Geophysicists, astronomers (especially those with an interest in history),

historians and orientalists should all find Richard Stephenson's work

fascinating. The culmination of many years of research, it discusses, in
depth, ancient and medieval eclipse observations and their importance
in studying Earth's past rotation. This is the first major book on this
subject to have appeared in the last 20 years. The author has specialised
in the interpretation of early astronomical records and their application
to problems in modern astronomy for many years. The book contains
an in-depth discussion of numerous eclipse records from Babylon, China,
Europe and the Arab lands. Translations of almost every record studied
are given. It is shown that although tides play a dominant long-term role
in producing variations in Earth's rate of rotation - causing a gradual
increase in the length of the day - there are significant, and variable,
non-tidal changes in opposition to the main trend.
HISTORICAL ECLIPSES AND EARTH'S ROTATION
HISTORICAL ECLIPSES AND
EARTH'S ROTATION

F. Richard Stephenson
University of Durham

CAMBRIDGE
UNIVERSITY PRESS
 CAMBRIDGE UNIVERSITY PRESS
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, Sao Paulo

Cambridge University Press

The Edinburgh Building, Cambridge CB2 8RU, UK

Published in the United States of America by Cambridge University Press, New York

www.cambridge.org
Information on this title: www.cambridge.org/9780521461948

© Cambridge University Press 1997

This publication is in copyright. Subject to statutory exception

and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.

First published 1997

This digitally printed version 2008

A catalogue record for this publication is available from the British Library

Library of Congress Cataloguing in Publication data

Stephenson, F. Richard (Francis Richard), 1941—
Historical eclipses and Earth's rotation / F. Richard Stephenson.
p. cm.
Includes bibliographical references and index.
ISBN 0 521 46194 4
1. Eclipses. 2. Earth-Rotation. I. Title.
QB541.S663 1997
523.3'8'09-dc20 96-20968 CIP

ISBN 978-0-521-46194-8 hardback

ISBN 978-0-521-05633-5 paperback
To my wife Ellen for her love, encouragement and patience
 Contents

Principal Symbols xv
1 Variations in the length of the day: a historical perspective 1
1.1 Introduction 1
1.2 The solar day 2
1.3 First speculations on variations in the Earth's rate of rotation 6
1.4 Discovery of the lunar secular acceleration 8
1.5 Early attempts to explain the Moon's secular acceleration 10
1.6 Discovery of the solar secular acceleration 15
1.7 Fluctuations in the mean motions of the Moon, Sun
and planets 17
1.8 The introduction of a theoretically invariant time-system 23
1.9 Short-term fluctuations in the Earth's rotation 26
1.10 The Julian and Gregorian calendars 29
2 Tidal friction and the ephemerides of the Sun and Moon 33
2.1 Introduction 33
2.2 Evaluation of the lunar tidal acceleration on TT 33
2.2.1 Transits of Mercury 34
2.2.2 Artificial satellite data 34
2.2.3 Lunar laser ranging 35
2.2.4 Selected value for n 35
2.3 The constancy of the lunar tidal acceleration 36
2.4 The tidal acceleration of the Earth's spin 37
2.5 Lunar and solar tidal dissipation 38
2.6 Ephemerides of the Sun and Moon 40
2.6.1 Solar ephemeris 40
2.6.2 Lunar ephemeris 41
2.7 Conclusion 42
3 Pre-telescopic eclipse observations and their analysis 43
3.1 Introduction 43
3.2 Observational requirements for determining AT in the
pre-telescopic period 43
3.3 Alternatives to eclipses in the pre-telescopic period 44

ix
x Contents

3.3.1 Equinox and related measurements 44

3.3.2 Close planetary conjunctions 45
3.3.3 Occultations of stars and planets by the Moon 46
3.3.4 Further remarks 47
3.4 Historical eclipses 47
3.5 Cause of eclipses 48
3.5.1 Solar eclipses 49
3.5.2 Lunar eclipses 55
3.6 Eclipse observations which are of value in the investigation
of AT 58
3.7 Sources of data 58
3.8 Solar eclipse observations and their reduction to AT 61
3.8.1 Total and annular solar eclipses 62
3.8.2 Solar eclipses which, although large, were not quite central 70
3.8.3 Timed solar eclipse contacts 71
3.8.4 Solar eclipse magnitudes 74
3.8.5 Observations that the Sun rose or set eclipsed 76
3.8.6 Records which merely note the occurrence of a solar eclipse 79
3.9 Lunar eclipse observations and their reduction to AT 82
3.9.1 Timed lunar eclipse contacts 83
3.9.2 Observations that the Moon rose or set eclipsed 86
3.9.3 Estimates of the degree of obscuration of the Moon at moon-
rise or moonset 89
3.10 Conclusion 91
4 Babylonian and Assyrian records of eclipses 93
4.1 Introduction 93
4.2 The Babylonian lunar eclipse records in the Almagest 95
4.3 Investigation of the Babylonian lunar eclipses cited by Ptolemy 97
4.4 Historical background to the astronomical cuneiform texts 105
4.5 Decipherment of the cuneiform texts 108
4.6 Origin and classification of the Late Babylonian astronomi-
cal texts 111
4.7 The Babylonian calendar 115
4.8 Techniques used in dating observational tablets 117
4.9 Units of time 118
4.10 'Lunar sixes' 119
4.11 Observation and prediction of eclipses 120
4.11.1 Observations 120
4.11.2 Predictions 122
4.12 Assyrian records of solar and lunar eclipses 124
4.13 Conclusion 127
 Contents xi

5 Investigation of Babylonian observations of solar eclipses 128

5.1 Introduction 128
5.2 The total solar eclipse of BC 136 Apr 15 129
5.3 Timed observations of solar eclipses 131
5.4 Estimates of solar eclipse magnitude 137
5.5 Solar eclipses occurring near sunrise or sunset 141
5.6 A possible allusion to a total solar eclipse in the Babylonian
Religious Chronicle 143
5.7 Conclusion 146
6 Timed Babylonian lunar eclipses 147
6.1 Introduction 147
6.2 Lunar eclipses for which only a single contact measurement
relative to sunrise or sunset is preserved 151
6.2.1 Measurements expressed to the nearest degree 151
6.2.2 Other less accurate measurements (to the nearest 5 or 10
degrees) 162
6.3 Eclipses for which two contact measurements relative to
sunrise or sunset are extant 167
6.4 Three or four timings relative to sunrise or sunset preserved
(total eclipses only) 173
6.5 Eclipse maxima timed relative to sunrise or sunset (partial
eclipses only) 178
6.6 First contacts timed relative to culmination of stars 184
6.7 Conclusion 190
7 Untimed Babylonian observations of lunar eclipses:
horizon phenomena 193
7.1 Introduction 193
7.2 Moon rising or setting eclipsed 194
7.3 Estimates of the proportion of the Moon obscured at its
rising or setting 208
7.4 Conclusion 211
8 Chinese and other East Asian observations of large
solar eclipses 213
8.1 Introduction 213
8.2 Records of solar eclipses from the Shang dynasty 215
8.3 The Chou dynasty and Warring States period
(c. 1050-221 BC) 219
8.4 The Ch'in and Han dynasties (221 BC to AD 220) 228
8.5 From the San-kuo to the Sui dynasty (AD 220-617) 238
8.6 The T'ang dynasty and Wu-tai period (AD 617-960) 245
8.7 The Sung, Kin and Yuan dynasties (AD 960-1368) 250
xii Contents

8.8 The Ming dynasty (AD 1368-1644) 260

8.9 Korea 262
8.10 Japan 266
8.11 Conclusion 269
9 Other East Asian observations of solar and lunar eclipses 273
9.1 Introduction 273
9.2 General remarks on recorded eclipse timings 274
9.2.1 Double hours and their subdivisions 274
9.2.2 Night-watches and their divisions 279
9.2.3 Estimates of lunar azimuth 283
9.2.4 Accuracy of early eclipse predictions 284
9.2.5 Sources of timed observations and terminology 284
9.3 Careful solar eclipse timings recorded in Chinese history 286
9.3.1 Solar eclipse timings recorded in sources other than
the Yuan-shih 288
9.3.2 Solar eclipse timings recorded in the Calendar Treatise
of the Yuan-shih 293
9.4 Lunar eclipse timings 295
9.4.1 Lunar eclipses timed in fifths of a night-watch 298
9.4.2 Lunar eclipses timed in double hours and marks 300
9.5 Azimuth measurements for lunar eclipses 307
9.6 Estimates of solar eclipse magnitude 310
9.7 Sun rising or setting eclipsed 315
9.8 Moon rising or setting eclipsed 318
9.9 Estimates of the proportion of the Sun or Moon obscured
at its rising or setting 325
9.10 Other records of lunar eclipses 329
9.11 Conclusion 333
10 Records of eclipses in ancient European history 334
10.1 Introduction 334
10.2 Solar eclipses analysed by Fotheringham 336
10.3 Other ancient reports of large solar eclipses 365
10.4 Greek lunar eclipse observations recorded in Ptolemy's
Almagest 368
10.5 The moonrise eclipse of BC 331 Sep 20/21 372
10.6 Conclusion 375
11 Eclipse records from medieval Europe 376
11.1 Introduction 376
11.2 Historical sources 376
11.3 Chronological discussion 378
11.4 Interpretation of observations of large solar eclipses 381
 Contents xiii

11.5 Place co-ordinates 386

11.6 Eclipses which were central at a known location 387
11.6.1 Detailed descriptions 387
11.6.2 Brief reports of total eclipses 413
11.7 Eclipses which were definitely partial, yet very large 416
11.8 Detailed records which are deficient in some way 422
11.9 Timed solar eclipse contacts 427
11.10 An eighth century occultation of Jupiter by the
eclipsed Moon 428
11.11 Conclusion 430
12 Solar and lunar eclipses recorded in medieval Arab chronicles 431
12.1 Introduction 431
12.2 Chronicles and chroniclers 432
12.3 Date conversion 434
12.4 Records of total and near-total solar eclipses 438
12.5 Other records of large solar eclipses 443
12.6 Sun eclipsed near sunrise or sunset 449
12.7 Moon eclipsed near moonrise or moonset 450
12.8 Selected records of total lunar eclipses 452
12.9 Conclusion 454
12.10 Appendix: Arab chronicles consulted 454
13 Observations of eclipses by medieval Arab astronomers 456
13.1 Introduction 456
13.2 Sources of data 456
13.3 Motives for observation 458
13.4 Calendrical remarks 461
13.5 Observational techniques 463
13.6 References to large solar eclipses in astronomical treatises 467
13.7 Solar eclipse times recorded by Ibn Yunus 468
13.8 Lunar eclipse times recorded by Ibn Yunus 476
13.9 Solar and lunar eclipse timings recorded by al-Battani
and al-Biruni 488
13.10 Solar magnitude estimates 494
13.11 Rising and setting phenomena 496
13.12 Conclusion 499
14 Determination of changes in the length of the day 501
14.1 Introduction 501
14.2 Timed and untimed observations 502
14.3 Lunar and solar tidal friction 506
14.4 Constant non-tidal component 506
14.5 Variable non-tidal component 509
xiv Contents

14.6 Further discussion of timed observations 511

14.7 Changes in the LOD 513
14.8 Geophysical discussion 516
14.9 Concluding remarks 517

Appendix A Timed data 518

Appendix B Untimed data 524

References 527

Acknowledgements 539

Index of eclipse records 540

Index of places of observation 549

Name index 551

Subject index 554

 Principal Symbols

C gravitational lunar acceleration on GMT/UT

C* lunar sidereal acceleration on GMT/UT
c lunar sidereal acceleration on GMT/UT with gravitational
term removed
d solar sidereal acceleration on GMT/UT
cy centuries (Julian)
GMT Greenwich mean time
E time in Julian centuries from the epoch 2000.0 (= J2000)
ET ephemeris time
h hours
J2000 epoch 2000.0
L lunar mean longitude on GMT/UT
Lr solar mean longitude on GMT/UT
LMT local mean time
LOD length of day
LT local apparent time
ms milliseconds
n lunar mean orbital angular velocity
nf solar mean orbital angular velocity (apparent)
n lunar orbital acceleration on TT
p.e. probable error (= 0.67 x s.e.)
q acceleration of Earth's spin (= AT/t2)
r lunar mean distance from Earth
s.e. standard error
SI Systeme Internationale (series of units)
T time in Julian centuries from the epoch 1900.0
t time in Julian centuries from the epoch 1800.0 (or 1820.0)
TAI international atomic time

xv
xvi Principal Symbols

TT terrestrial time
UT universal time
y years
z equation of time
AT Earth's rotational clock error (= TT — UT)
9 time in Julian centuries from an arbitrary epoch
X lunar mean longitude on TT
XI solar mean longitude on T T
A geographic longitude (negative to the east of Greenwich)
(/> geographic latitude
co terrestrial angular velocity of rotation
 Variations in the length of the day:
a historical perspective

1.1 Introduction
The main purpose of this book is to investigate in detail long-term varia-
tions in the length of the day, or equivalently, changes in the Earth's rate
of rotation, using pre-telescopic observations of eclipses. Such variations
are mainly produced by lunar and solar tides, but non-tidal mechanisms
are also significant. Despite its historical bias, this subject has become an
important topic in modern geophysics.
In studying changes in the length of the day which have occurred in
recent centuries (since the invention of the telescope), more accurate data
than eclipses are available. These include occultations of stars by the Moon
up to 1955 and systematic monitoring of the Earth's rotation relative to
the atomic time-scale since that date. However, observations of eclipses
provide by far the most consistent and reliable source of information
on variations in the terrestrial rate of rotation during the pre-telescopic
period.
Although numerous early accounts of eclipses are to be found in as-
tronomical works, many others are scattered in a variety of writings -
for example the Greek and Latin classics, imperial annals of China and
monastic chronicles of Europe. Observations recorded in these works are
often extremely crude by modern standards, but the lengthy time-scale
covered is highly suited to the detection of long-term trends which are
not apparent from more recent data. The investigation of this diverse
material adds a whole new dimension to what might otherwise have been
a somewhat narrow scientific discipline.
This book has as its nucleus an extensive paper by Stephenson and
Morrison (1995) entitled 'Long-term fluctuations in the Earth's rotation:
700 BC to AD 1990'. The present work is a considerable expansion of this
paper and contains detailed discussion of almost every observation used
- as well as an in-depth examination of calendar conversion and early
methods of measuring time. Although in general much the same ancient

1
2 1 Variations in the length of the day

and medieval eclipse observations are analysed, a significant number of

medieval Chinese data have now been added (see chapter 9).
In the present chapter, the gradual recognition of the existence of
changes in the length of the day will be outlined, commencing with
ancient writers and extending through recent centuries to the present. The
lunar acceleration will also be considered; from a historical viewpoint this
is closely linked with changes in the length of the day. Other related topics
include the introduction of modern time-systems which are independent
of changes in the Earth's rate of rotation, and the Julian and Gregorian
calendars.
It should be noted that in quoting the names of authors born before
about 1900, both surname and given name(s) will be cited throughout this
chapter. However, the publication of so many multi-authored papers in
recent decades would make continuation of this practice tedious.
NB throughout this book I have deliberately avoided the use of accents
above and below individual letters. This choice should considerably
simplify the text for the reader who is not a linguist, without - I hope
- causing too much inconvenience to the reader who has a linguistic
background.

1.2 The solar day

The period of rotation of the Earth has been regarded as one of the
fundamental units of time since remote antiquity. Although the full
realisation that our planet does not rotate at an absolutely uniform rate is
a relatively modern discovery, several preliminary stages in this recognition
can be discerned. These relate exclusively to apparent changes in the length
of the day which are caused by both the tilt of the Earth's axis and the
ellipticity of the terrestrial orbit around the Sun.
Seasonal changes in the length of daylight and darkness were probably
recognised at a very archaic stage in history. Even in relatively low
latitudes (c. 30-35 deg N) where the great ancient civilisations of the
Fertile Crescent, Indus Valley and China developed, the length of daylight
ranges over a ratio of as much as 1.4:1 from summer to winter. In Egypt,
the night was divided into 12 hours at least as early as 2150 BC, but a
similar division of the period of daylight cannot be traced until c. 1450
BC (Parker, 1974). By the twelfth century BC, evidence for the use of
24 equal hours in a full solar day can also be found. A papyrus of that
date lists the number of hours of daylight and darkness for each month
of the year, the total always being 24. This system spread to Greece and
ultimately to the modern world.
At least as early as the eighth century BC, the astronomers of Babylon
divided the solar day into 12 equal double hours, each of which was
 1.2 The solar day 3

subdivided into 30 equal parts; the basic unit thus corresponded to 4

minutes. Babylonian arithmetic was sexagesimal and the origin of our
division of the hour into 60 minutes and the minute into 60 seconds can
ultimately be traced to this system of reckoning.
The ancient Greeks employed both seasonal and equal hours (each 24
in number) side by side. On the former scheme, day and night were each
divided into 12 hours, each ranging from about 0.8 to 1.2 of our present
hours, depending on the season. Although seasonal hours were more
relevant to everyday life, equal hours were of greater value to astronomers
in studying the motion of the Moon, Sun and planets. This dual practice
was followed in the medieval Arab world and subsequently in Europe.
The great Alexandrian astronomer Claudius Ptolemy (c. AD 150) was
well aware that the length of the solar day - which may be defined as the
interval between two successive transits of the Sun across an observer's
meridian - varied slightly during the course of a year. In a section of
his Almagest entitled 'On the inequality in the days' (III, 9) he gave an
essentially correct explanation of the difference between the anomalistic
or irregular solar day and the mean solar day (although, of course, he
assumed that the Sun revolved around the Earth and that its motion was
epicyclic).
The annual variation which Ptolemy describes arises from two separate
causes: the ellipticity of the Earth's orbit and the tilt of the terrestrial axis
(the obliquity of the ecliptic). The former effect, which is also responsible
for the inequality of the seasons, produces a cyclical change in the apparent
daily solar motion in longitude (currently from 1.020 deg in January to
0.954 deg in July). Since the Sun (apparently) moves in the ecliptic
rather than along the celestial equator, the obliquity is responsible for an
additional variation of the daily motion of the Sun in right ascension.
As a result of the superposition of these two components, during the
course of the year the solar day can be up to about 30 sec longer or
shorter than the average value of 86 400 sec (the mean solar day) - see
figure 1.1. Ptolemy was unable to detect such a small variation, but he
could readily discern the cumulative effect. He remarked:
Neither of these effects causes a perceptible difference between the mean
and the anomalistic return (of the Sun) over a single solar day, but the
accumulated difference over a number of solar days is quite noticeable.
[Trans. Toomer (1984, p. 170).]
On account of the build-up of individual small changes in the length
of the solar day, a sundial (or other device keeping solar time - usually
known as apparent local time) may be appreciably fast or slow relative
to a clock measuring mean time (as defined by the average daily motion
of the Sun). The difference between local apparent time (subsequently
 1 Variations in the length of the day

35
30
25
20
15
10
5
0
-5
-10
-15
-20
-25

Jan Feb Mar April May June July Aug Sept Oct Nov Dec
Month
Fig. 1.1 Annual variation in the length of the apparent solar day (Alod) arising
from the tilt of the Earth's axis and the eccentricity of the orbit.

abbreviated to LT) and local mean time (LMT) at any moment is known
as the equation of time (z). Extreme values of z are currently attained
in mid-February (—0.24 h) and early November (+0.27 h). In February,
the Sun crosses the observer's meridian significantly after mean noon,
while in November the reverse is true. The form of this annual variation
has slowly altered down the centuries, largely on account of the gradual
advance of the Earth's perihelion caused by planetary action. Figures
1.2a and 1.2b compare the present curve with that calculated for 700 BC
(the approximate date of the earliest reliable astronomical observations
from any part of the world). For a series of equation of time curves
at 1000-year intervals from 4000 BC to AD 4000, see Hughes et al.
(1989).
Despite the early recognition of the apparent variation in the length
of the solar day, mean solar time remained an essentially theoretical
concept for more than 1500 years. As a result, all timings of eclipses (and
other celestial events) from any part of the ancient and medieval world are
expressed either directly or indirectly in terms of LT. Until the introduction
of pendulum clocks - the first by Christiaan Huyghens in 1656 - the
inaccuracy of primitive chronometers meant that the equation of time was
of little practical significance. However, this parameter gradually became
of more importance as attempts were made to use pendulum clocks to
mark LMT. This was first achieved by regulating the clocks with the
aid of a sundial and a conversion table from LT to LMT. In about AD
1700, such tables for setting clocks and watches began to be commercially
 1.2 The solar day
(a)
+0.30
+0.25 -
+0.20 -
+0.15
+0.10
+0.05
0
-0.05
-0.10
-0.15
-0.20
-0.25
-0.30
-0.35

Jan Feb Mar April May June July Aug Sept Oct Nov Dec
Month

(b)
+0.30 r
+0.25 -
+0.20
+0.15
+0.10
+0.05
0
-0.05
-0.10
-0.15
-0.20
-0.25

Jan Feb Mar April May June July Aug Sept Oct Nov Dec
Month

Fig. 1.2 Changes in the form of the equation of time curve between 700 BC (a)
and the present (b).

available (Macey, 1994, p. 443), while almost from its inception in 1767
the Nautical Almanac tabulated this quantity to the nearest 0.1 sec.
The data in the various national ephemerides did not begin to be spec-
ified on LMT until 1834. Commencing in 1840, the railways were largely
instrumental in the adoption of Greenwich mean time (GMT) throughout
6 1 Variations in the length of the day

Britain. GMT became the world standard following the International

Meridian Conference in 1884. Until 1924, GMT was reckoned from mean
noon at Greenwich, but in the following year it began to be measured
from Greenwich mean midnight instead. By 1960 the designation GMT
had been replaced internationally by universal time (UT). Today, three
separate definitions of UT are adopted: UTO, UT1 and UTC. Of these,
UTO is a local approximation to universal time, and is not corrected for
polar motion. UT1 is a more precise measure of UT, freed from the effects
of polar motion, and is used in navigation. Finally, UTC (co-ordinated
universal time) differs from international atomic time (defined in section
1.8 below) by an integral number of seconds and forms the basis of most
radio time-signals and legal time systems. The difference between the var-
ious designations of UT is only of the order of 1 sec, which is negligible
for historical studies. Hence throughout this book the general designation
UT will normally be used.
In order to compare with calculation an observation made at longitude
A deg and expressed in LT, it is necessary to convert the measured LT to
UT. This is given in hours by the equation
UT = LT + y l / 1 5 - z , (1.1)
where z is the equation of time. Here the adopted convention is that A is
positive to the west of the Greenwich meridian and negative to the east
of it (both through 180 deg).

1.3 First speculations on variations in the Earth's rate of rotation

Speculation that the rate of spin of the Earth - and hence the mean solar
day - might itself vary seems to have first occurred around the middle of
the eighteenth century. In 1754, the Royal Academy of Sciences at Berlin
offered a prize for solution of the following problem:
Whether the Earth in its rotation round its axis, by which it brings
about the alternation of day and night, has undergone any alteration since
the first period of its origin. What may be the cause of this, and what can
make us certain of it?
[Ley (1968, p. 157).]
According to Ley (1968, p. xi), Leonhard Euler may have first mentioned
the problem to the secretary of the Royal Academy in Berlin. Ley notes
that Euler, in a letter to the Bishop of Bergen (Henrik Pontoppidan),
written not long before the prize was offered, had stated that he suspected
that the day was growing shorter.
Immanuel Kant suggested, on the contrary, a cause that would lengthen
the day. In a paper whose title may be translated 'Whether the Earth
 1.3 First speculations on variations in the Earth's rate of rotation 1

has undergone any alteration of its axial rotation', Kant (1754) reasoned
that the combined attraction of the Moon and Sun, by moving the waters
of the Earth in a westerly direction, would oppose the diurnal rotation
of our planet. Estimating the velocity of ocean movement at '1 foot per
second', Kant considered the loss of momentum on the east coast of
North and South America - because of the large extension in latitude. He
calculated that in two million years the 'whole motion of the Earth' would
be exhausted, corresponding to a mean increase in the length of the day
by about 4 sec per century. Although this result is about 2000 times the
true figure, the basic principle was sound.
Kant further concluded that on account of the tidal retardation of its
spin, in the remote future the Earth would ultimately turn the same face
towards the Moon, the terrestrial surface then being at rest relative to
the source of disturbance. He also correctly explained the fact that the
Moon always turns the same face towards the Earth as being due to
the retardation of the lunar rotation by tides raised by the Earth. Tidal
friction is now recognised as the major long-term cause of variation in the
length of the mean solar day (usually abbreviated to LOD). Remarkably,
Kant later came to the conclusion that the Earth might be contracting
owing to gradual cooling, which would cause it to spin faster in opposition
to the effect of the tides.
Some 25 years after Kant formulated his hypothesis of the tidal retar-
dation of the Earth's spin, Sir William Herschel considered the question
of changes in the length of the day from an observational basis. In a letter
published in 1781, he remarked that no astronomer had then investigated
whether the Earth's diurnal rotation was constant or not, adding that this
was 'probably [on account of] the difficulty of finding a proper standard to
measure it by; since it is itself used as the standard by which we measure
all the other motions'.
Herschel made the further perceptive remark:
It is perhaps not altogether impossible but that inequalities may exist in
the (diurnal) motion which, in an age where observations are carried to
such a degree of refinement, may be of some consequence.
Because even the best of contemporary clocks were too imprecise to
demonstrate the existence of such 'inequalities', Herschel envisaged a long-
term plan. He recommended determination of the rotation periods of the
other planets - especially Mars - as accurately as possible relative to the
LOD, in the hope that future astronomers would be able to decide whether
or not these periods had maintained their relative values. This was to
be a vain hope. However, nearly a century beforehand, the analysis of
ancient observations of eclipses by Edmond Halley had provided the first
suggestion of a seemingly independent phenomenon: an acceleration of
8 1 Variations in the length of the day

the Moon. Ultimately, this proved to be partly real and partly apparent,
the cause of the apparent component being a gradual increase in the
adopted unit of time - i.e. the Earth's period of rotation.

1.4 Discovery of the lunar secular acceleration

As noted in section 1.1, although ancient observations are of low precision
compared with telescopic measurements, the greatly enhanced time-scale
renders them much more suitable for determining an accelerative term.
By Halley's epoch, the investigation of ancient eclipses to obtain im-
proved knowledge of the lunar motion was already a long-established
tradition. Both Hipparchus (c. 150 BC) and Ptolemy (c. AD 150) had
utilised Babylonian eclipse observations from several centuries previously
for this purpose (see chapter 4 below). Long afterwards, the great Chinese
astronomer of the thirteenth century Kuo Shou-ching had used a series
of eclipse observations extending over more than a thousand years to test
the accuracy of contemporary eclipse tables (see chapter 9). Nevertheless,
until Halley's researches, the mean motion of the Moon (i.e. with the
various short-term fluctuations averaged out) had been assumed to be
uniform.
In 1695, towards the end of a paper entitled 'Some account of the
ancient state of the city of Palmyra...', Halley recommended that the
longitudes of Baghdad, Aleppo and Alexandria be carefully determined
(using forthcoming lunar eclipses), explaining his motive in the following
words:
For in and near these Places were made all the Observations whereby
the Middle Motions of the Sun and Moon are limited: And I could then
pronounce in what Proportion the Moon's motion does Accelerate; which
that it does, I think I can demonstrate, and shall (God Willing) one day
make it appear to the Publick.

Two years earlier, Halley (1693) had investigated a series of celestial

observations made by the medieval Muslim astronomer al-Battani (AD
850-929). Al-Battani, whose latinised name was Albategnius, observed
both at Raqqa (in present-day Syria) and Antakya (Turkey) - see also
chapter 13. When Halley undertook his investigation, only an inaccurate
Latin translation of al-Battani's writings, made in the twelfth century,
was available. Halley's objective can best be summarised in the title of
his paper (also in Latin), which may be rendered as follows: 'Corrections
and comments on the ancient astronomical observations of Albategnius in
order to restore the solar and lunar tables of that author'. Although this
paper was purely an historical exercise, Halley's research may well have
inspired him to test contemporary lunar tables against the observations
 1.4 Discovery of the lunar secular acceleration 9

made by al-Battani and other early astronomers. However, apart from the
above brief reference in his 1695 paper his publications remain silent on
the question of the lunar acceleration.
In 1749, the Reverend Richard Dunthorne not only provided firm
evidence for the existence of a lunar acceleration but also obtained a
result which was not significantly improved upon until the present century.
Using lunar and solar tables which he had himself constructed, Dunthorne
analysed a selection of six ancient and medieval eclipse timings from the
following years: 721 and 313 BC (both from Babylon and recorded by
Ptolemy in his Almagest); 201 BC and AD 364 (both from Alexandria);
and AD 977 and 978 (both from Cairo). Details of these and other
roughly contemporary observations are given in chapters 4, 10 and 13
below. In each case Dunthorne found that the computed time was earlier
than the observed time, the discrepancy being greatest for the most ancient
observations. He inferred an acceleration of the Moon, and concluded:
If we take this Acceleration (of the Moon) to be uniform, as the
Observations whereupon it is grounded are not sufficient to prove the
contrary, the Aggregate of it will be as the Square of the Time: And if we
suppose it to be 10" in 100 years... it will best agree with the above
mentioned Observations.
Dunthorne here implied a coefficient of 92 of 10" in the Moon's longi-
tude (where 9 is measured in Julian centuries, each of 36 525 mean solar
days from an arbitrary epoch). Of course, this result, in common with
all similar deductions made until comparatively recent times, was based
on the assumption of a fixed length of day. It became customary to refer
to the coefficient of 82 in the lunar mean longitude - expressed purely
in arcseconds - as the Moon's secular acceleration - rather than quoting
twice this figure. Interestingly, the observation in 721 BC still remains one
of the very earliest reports of an eclipse from any part of the world, even
though numerous ancient records unknown to Dunthorne have since come
to light. Because the 721 BC eclipse occurred nearly 25 centuries before
Dunthorne's own time, a secular acceleration of 10" implied a discrepancy
between observation and theory of as much as 1.7 deg. The Moon takes
fully 3 hours to move through this angle, a time-interval which could
be measured with tolerable precision by the astronomers of Babylon (see
chapter 4).
In his paper, Dunthorne also stated that he had investigated several
eclipse observations made by Tycho Brahe in the late sixteenth century
and by Bernard Walther and Johann Muller (the latter better known
as Regiomontanus) in the late fifteenth century. Unfortunately he only
published very brief details. On comparing Tycho's measurements with
his own eclipse tables, Dunthorne had found that observation and theory
10 1 Variations in the length of the day

agreed 'full as well as could be expected considering the imperfection of his

(Tycho's) clocks...'. For the observations by Walther and Regiomontanus,
Dunthorne's computed lunar longitudes proved to be mainly some 5
arcmin ahead of their positions as derived from measurement. However,
in this latter case he considered the mutual discord between the individual
results too great for an acceleration to be inferred with certainty.
Not long after Dunthorne's work, Tobias Mayer (1753) made a similar
investigation of historical eclipses. He obtained a significantly lower result
for the lunar secular acceleration of 6".7, but subsequently revised it to
9". At about this same time, Joseph-Jerome Lalande (1757) obtained a
result of 10". Hence by early in the second half of the eighteenth century,
the existence of the secular acceleration of the Moon's mean motion was
firmly established on observational grounds.

1.5 Early attempts to explain the Moon's secular acceleration

The significant acceleration of the Moon derived from ancient data invited
theoretical explanation, and commencing in 1770 several prizes were
offered for its solution by the Academy of Sciences at Paris. Following
unsuccessful attempts by Leonhard Euler and Joseph-Louis Lagrange,
what in retrospect proved to be a partially successful explanation was
obtained by Pierre-Simon Laplace (1787a and b). Laplace attributed the
lunar acceleration to the effect of the gradual diminution in the eccentricity
of the Earth's orbit caused by planetary action. This produces a long-term
decrease in the mean action of the Sun on the Moon, resulting in a slow
increase in the lunar mean motion. The variation in the Earth's orbital
eccentricity - and hence the change in the average motion of the Moon -
is actually cyclic with a periodicity of the order of 100000 years (Berger,
1977). However, over the relatively short time-scale covered by historical
observations the lunar acceleration arising from this cause has consistently
been positive - see chapter 2.
Laplace approximated the change in eccentricity (ef) of the terrestrial
orbit by a linear equation of the form
e' = e'0-ay. (1.2)
Here e'o is the eccentricity at epoch 1850 and a is a constant; y is
reckoned in Julian years from 1850. Laplace deduced that the lunar
secular acceleration (C) is given by
C = lrn2e'on'a, (1.3)
where m is the ratio of the mean motions of the Sun and Moon (0.0747)
and vl the mean solar motion. Laplace derived a secular acceleration
of 11".135, which appeared to be in excellent accord with observation.
 7.5 Attempts to explain the Moon's secular acceleration 11

Consequently he rejected the possibility that variation in the LOD was

responsible. He remarked that since the time of Hipparchus, almost 2000
years previously, the LOD 'had not varied by ^Q of a second'. Some years
later, in 1792, Lagrange closely confirmed Laplace's value for the lunar
acceleration and it thus seemed that the problem was solved.
Laplace's apparently satisfactory result was not disputed until 1853
when John Couch Adams pointed out an error in his computations.
Laplace and Lagrange had integrated the necessary differential equations
as though the eccentricity of the Earth's orbit were constant, substituting
its variable value in the results. Adams showed that although such a
procedure was permissible when only a first approximation was required,
for higher precision it was necessary to introduce the variability of ef
into the differential equations. He at first replaced \m2 in equation (1.3)
by |ra 2 — (3771/64)m4, thus arriving at C = 8".3. Adams subsequently
reduced this result to 5".78 by evaluating the first few terms in a slowly
converging power series (to m7). The sum of the individual terms for C was
C = 10".66 - 2".34 - 1".58 - 0".71 - 0".25 = 5".78.
Adams' revised value, first announced in a letter to Charles-Eugene De-
launay, was published in 1859. It was disputed by such distinguished
mathematicians as Peter Andreas Hansen, Giovanni Plana and Philippe
de Pontecoulant, who each closely supported Laplace's figure. However,
following an extensive analysis, in which he evaluated as many as 42
terms, Delaunay (1859) obtained a secular acceleration of 6".ll - only a
slight revision of Adams' value. To quote Simon Newcomb (1898, p. 98):
Although their antagonists never formally surrendered, they tacitly
abandoned the field, leaving Adams and Delaunay in its undisturbed
possession.

Twentieth century investigations have closely confirmed Delaunay's de-

duction. In his exhaustive investigation of the lunar theory, Ernest W.
Brown (1915a) derived a value for C of 6//.03. More recently, Chapront-
Touze and Chapront (1983) made a new determination of this param-
eter. Their result comprises the following components: Earth figure
(0".1925); planetary perturbations (O".OO25); Earth's orbital eccentricity
(5".8864); lunar figure (-0".0151). The sum, 6".0463, is almost identical
to Brown's value.
William Ferrel (1865) and Delaunay (1865) independently revived the
hypothesis of tidal retardation of the Earth's spin to account for the
observed excess acceleration of the Moon. Ferrel derived a result for the
apparent acceleration of the Moon on the simplifying assumptions that the
surface of the Earth was covered by oceans and that the tidal wave took
the form of a prolate spheroid. He also made the further rather arbitrary
12 1 Variations in the length of the day

suppositions that the vertices of this spheroid were displaced by friction

through 30 deg (relative to the line of centres of the Earth and Moon) and
that the height of the tidal wave above low water was 2 feet (0.6 metres) at
the equator. Ferrel calculated that the effect of both lunar and solar tides
would be to retard the rotatory motion of the Earth at the equator by
'about forty-four miles in a century'. Since the Moon moves through 84"
in its orbit while the terrestrial equator turns through this distance, Ferrel
concluded that the apparent secular acceleration of the Moon would be
84". This was roughly an order of magnitude greater than the observed
acceleration and Ferrel concluded that the effect of tides must be largely
'counteracted by a corresponding contraction of the Earth's volume from
a gradual loss of heat' - essentially the explanation offered by Kant.
The equator rotates through 44 miles (70 km) in about 150 sec. As a
(Julian) century contains 36 525 days, it can be seen that Ferrel's result
would correspond to a rate of increase in the LOD by about 8 milliseconds
per century (ms/cy). A similar figure was reported by Sir William Thomp-
son - later Lord Kelvin - at a lecture delivered in 1866 (Kelvin, 1894, p.
68). Delaunay (1865) did not attempt to deduce the rate of lengthening of
the day due to tides. However, on the basis of the difference between the
observed secular acceleration of the Moon and its gravitational equivalent
(amounting to some 6"), he calculated that the equivalent secular retarda-
tion of the terrestrial rotation would be 27^ times as great or 164". Since
the Earth turns through this angle in about 11 sec, the corresponding rate
of increase in the LOD would be about 0.6 ms/cy. This is, in fact, about
one-quarter of the actual amount (see chapter 2).
Neither Ferrel nor Delaunay - nor Kelvin at this early date - seem to
have considered the reciprocal effect of lunar tidal friction on the Moon's
orbit. However, Emmanuel Liais (1866), who was of the opinion that
magnetism might play a key role in the interaction between the Earth
and Moon, stressed that the gain in angular momentum by the Moon
at the expense of the Earth would produce an orbital acceleration. Liais
concluded:

By the laws of mechanics, a reduction in the speed of a celestial body

causes a recession such that the 'angular movement' [mouvement angulaire]
is increased. Thus there exist (both) a real acceleration of the motion of
the Moon and an apparent acceleration arising from the increase in the
length of the day.

A satellite which gains angular momentum both recedes from its pri-
mary and loses angular velocity, as can be seen from the following equa-
tions, which are based on the approximation of circular motion. If A
represents the orbital angular momentum of the Moon, r the distance
 1.5 Attempts to explain the Moon's secular acceleration 13

from the Earth, and n the angular velocity,

dA/dr=Kr~1/2, (1.4)

dA/dn = -krC5l\ (1.5)

where both K and k are positive constants.
Liais effectively implied that the observed acceleration of the Moon
would be the sum of the apparent and real components, which is a clear
statement of the true situation.
Sir George Darwin (1880) and also Lord Kelvin and Peter Guthrie Tait
(1883) discussed the effect of tidal friction in detail. However, they were
well aware that the approximations which they made were necessarily
crude. Kelvin later (1894, p. 69n) stressed that direct evaluation was a
complex task. He remarked:

It seems hopeless, without waiting for some centuries, to arrive at any

approach to an exact determination of the the amount of the actual
retardation of the Earth's rotation by tidal friction, except by extensive
and accurate observation of the amounts and times of the tides on the
shores of continents and islands in all seas, and much assistance from true
dynamical theory to estimate these elements all over the sea. But supposing
them known for every part of the sea, the retardation of the Earth's
rotation could be calculated by quadratures.

Not until 1920 was the first reasonably successful calculation of this
kind made (by Sir Harold Jeffreys), but this empirical method has been
superseded by studies of the orbits of artificial satellites and lunar laser
ranging - see chapter 2.
Darwin emphasised that in the geological past the Moon must have
been much closer to the Earth than at present. He showed that since
the lunar orbital angular momentum is about 5 times that of the Earth
about its axis of rotation, if the Earth and Moon were once a single
fluid body, the rotation period would be about 4 hours. This led to his
'resonance theory' of the origin of the Moon, in which solar tides raised
on the primeval Earth were amplified by resonance to such an extent that
the planet broke into two. Although this theory is no longer regarded as
tenable, it was briefly revised in modified form some 25 years ago (Wise,
1969; O'Keefe, 1970).
The effects of tidal friction on the rotation of the Earth and motion
of the Moon may be briefly outlined as follows; for a more detailed
discussion, see Jeffreys (1976, pp. 316ff.). The gravitational attraction of
the Moon on the Earth will produce diametrically opposite tidal bulges
- as illustrated schematically in figure 1.3. At bulge X, which is slightly
nearer the Moon, the lunar gravitational force will be greater than at the
14 1 Variations in the length of the day

Tidal bulges,

Moon
Earth Phase lag
Fig. 1.3 Tidal interaction between the Moon and the Earth.

centre of the Earth, while at Y it will be weaker. On account of friction

in the oceans (and in the interior of the Earth owing to the imperfect
elasticity of the material) there will be a phase lag between the passage of
the Moon across any particular meridian of longitude and the associated
high tide. The two bulges will thus be out of alignment with the line of
centres of the Earth and Moon. Hence the gravitational attraction of the
Moon will have a component in the opposite sense to the direction of
rotation of the Earth - which will thus be retarded.
In the reciprocal case of the attraction of the bulges X and Y on the
Moon, the force produced by X is greater than that due to Y since the
distance XM is less than YM and the angle XME is greater than YME.
There will thus be a net force acting on the Moon with a component
in the direction of revolution. Following the principle of conservation
of angular momentum, the angular momentum lost by the Earth will
be gained by the Moon.
Despite the enormous mass of the Sun, solar tides are significantly
less effective in retarding the Earth's rotation - by a factor of about 4 -
because of the vastly greater mean distance.
Because the angular momentum in the Earth's orbit is so huge, the
Sun can have negligible real acceleration due to inverse tidal action.
Nevertheless, a gradual decrease in the length of the day caused by tides
would produce an apparent solar acceleration. Until early in the present
century there was no observational support for any solar acceleration and
this had somewhat weakened the arguments put forward in favour of the
tidal retardation of the Earth's spin. Eventually, evidence was to come
from a study of ancient eclipses - just as it had in the case of the lunar
acceleration.
 1.6 Discovery of the solar secular acceleration 15

1.6 Discovery of the solar secular acceleration

Interest in historical eclipses waned towards the end of the eighteenth
century, but it was revived by Sir George Airy (1853, 1857). Untimed
reports of total obscurations of the Sun now began to attract special
attention, the basic principle being that since the lunar umbra sweeps
out a very narrow path across the Earth's surface (see chapter 3), small
changes in the adopted value for the lunar acceleration would materially
alter the maximum degree of obscuration of the Sun at any particular
site. Other nineteenth century authors who analysed this type of material
included Giovanni Celoria (1877a and b), and Friedrich Ginzel (1884a
and b). In general, results for the lunar secular acceleration fairly close
to 10" were derived - essentially the value obtained by Dunthorne (1749)
and significantly greater than the theoretical figure.
By the mid-nineteenth century, a small secular acceleration of the
equinox (the reference point for measuring celestial co-ordinates) had
been discovered. Hansen (1857) calculated that the equinox had a secular
acceleration (coefficient of 92) due to planetary action of I'M, while
Newcomb (1895b) obtained l'Ml. (Recent computations by Laskar (1986)
yield an almost identical value of I'M 12.) Observational results for the
lunar acceleration inevitably included the effect of precession, and when
this was subtracted it became customary to refer to the remainder as the
sidereal acceleration of the Moon. Results for the sidereal acceleration
were a little closer to the theoretical result for C (approximately 6//.05),
which was calculated relative to an inertial frame. However, there was still
a significant discrepancy of several arcsec/cy2.
Throughout his career, Simon Newcomb considered historical allusions
to total solar eclipses as of very dubious worth. In his own studies,
starting in 1878, Newcomb relied instead on timed measurements. These
originated from Babylon and Greece between 721 BC and AD 136 (as
quoted in Ptolemy's Almagest) and from Baghdad and Cairo between AD
829 and 1004. (For details of these various observations, see chapters 4,
10 and 13.) Newcomb obtained a result for the lunar sidereal acceleration
of 8".8. As will be apparent from chapter 10 below, Newcomb was fully
justified in his rejection of the ancient European observations of total
solar eclipses which his contemporaries analysed. Often both the date and
place of observation were in doubt, while in most cases it was not even
clear whether the Sun was indeed completely obscured. However, medieval
observations of this kind have proved to be of considerable utility - see
chapter 11.
In 1905, Philip H. Cowell found evidence from ancient eclipses for a
significant solar acceleration. Only a minute secular acceleration of the
Sun (amounting to — 0".02) was expected on gravitational theory. This
16 1 Variations in the length of the day

was purely apparent and arose from a slow increase in the mean distance
of the Earth from the Sun caused by planetary perturbations. Cowell had
previously made several analyses of ancient eclipse records - both timed
and untimed - in order to derive a value for the lunar coefficient of 92.
He now concluded that to best satisfy the observations it was necessary
to assume a secular acceleration of the Sun as well as of the Moon. His
results for the sidereal lunar and solar accelerations (which will be denoted
by C* and d respectively) were C* = 10".9 and d = 4".l. At first Cowell
attributed the observed solar acceleration to the resistance of the Earth's
orbital motion by the 'ether'. Soon afterwards (1906a) he reverted to the
hypothesis of a steady increase in the unit of time - the mean solar day -
produced by tidal action.
Cowell noted that the ratio which he derived for C*/c1 (i.e. 2.7:1)
was much less than that of the mean motions of the Moon and Sun
(13.37:1). He correctly explained this difference as arising from the gradual
diminution in the angular velocity of the Moon as its orbit expanded under
the reciprocal action of the tides. In essence, Cowell showed that if co
represents the angular velocity of the Earth's rotation and n and nr the
mean angular velocity of revolution of the Moon and Sun, then, with 9
in centuries as above,

(£)
In equation (1.7), dn/d6 represents the real tidal acceleration of the
Moon (negative) and (n/co)(dco/d8) its apparent acceleration due to the
increase in the adopted unit of time. As noted earlier (section 1.5), tides
have negligible effect on the motion of the Sun so that any acceleration
must be purely apparent. The factor 0.5 is introduced in both equations
because the secular accelerations C* and d were customarily regarded as
coefficients of 62.
In general, if one takes the mean solar motion as 0^.0411 per second
of time, it is readily shown that the rate of lengthening of the day
corresponding to a selected value for d is given by 1.33 d ms/cy. From his
value for d of 4".l, Cowell derived a result for the rate of increase in the
LOD of 5 ms/cy, which is rather more than twice the current estimate.
Cowell's detection of the apparent solar acceleration, although based
on rather flimsy historical evidence, represented a pioneering step. His
conclusion that the apparent acceleration of the Sun arose from a diminu-
tion of the Earth's rate of rotation was rejected by Newcomb and other
authors (see Cowell, 1907). However, Cowell's work inspired other as-
tronomers - notably John K. Fotheringham - to solve directly for both
 1.7 The mean motions of the Moon, Sun and planets 17

C* and d when analysing ancient observations. In a protracted series of

papers, published between 1908 and 1935, Fotheringham analysed numer-
ous early observations of eclipses, occultations and equinoxes. Most of his
timed data were taken from the Almagest, but he also investigated several
allusions to eclipses in ancient historical and poetical works. He derived
(1920b) values for C* and d of respectively 10".6 and 1".5, remarking
that 'all classes of ancient observations' were adequately satisfied by these
figures. Fotheringham's result for the solar acceleration, corresponding
to a rate of increase in the LOD of 2.0 ms/cy, is remarkably close to
the figure deduced from present-day researches. Although both he and
Cowell used a miscellany of observations of varying degrees of reliability,
Fotheringham placed greater emphasis on the more viable timed data.
Willhelm de Sitter (1927) revised Fotheringham's analysis of ancient
observations. In quoting the lunar acceleration he subtracted the gravita-
tional contribution derived by Brown (6".03), obtaining a result which we
shall term c. De Sitter obtained values for c and d (the solar acceleration)
of respectively 5".22 + 0".30 (p.e.) and l".80 + 0.16 (p.e.). Fotheringham's
results, reinterpreted in this way, were to play an important part in the
recognition of short-term fluctuations in the Earth's rate of rotation as
revealed by telescopic observations. It is to this issue that we now turn.

1.7 Fluctuations in the mean motions of the Moon, Sun and planets
During his extensive researches on the lunar theory, Hansen discovered
two inequalities of long period in the Moon's mean longitude which he
attributed to the action of Venus. These terms, whose derived amplitudes
were respectively 15".3 and 21".5 and periods 273 years and 239 years,
were included in his Tables de la Lune, published in 1857. At the time,
Hansen's tables were believed to represent closely all observations of the
Moon since 1750. In particular, they were adopted for computing the
lunar ephemerides printed in the Nautical Almanac from 1862.
Of the two inequalities, the amplitude and period of the former were
closely confirmed by Delaunay (1860a and b). The term became known as
the 'Great Venus Term'. According to Brown (1919, p. 27), its amplitude
is 14".27 while its period is 271 years. However, the latter inequality was
shown by Delaunay (1860a and b) to have no theoretical justification. This
was the first reliable indication of major irregularities in the lunar motion
which could not be satisfactorily explained by gravitational theory. By
1870, Newcomb had noted that comparisons between observations of the
position of the Moon and predictions based on Hansen's theory revealed
significant deviations. As these discrepancies could not be explained by
observational errors, it was evident that they either arose from deficiencies
18 1 Variations in the length of the day

in Hansen's theory or fluctuations in the motion of the Moon of non-

gravitational origin.
In 1878, Newcomb analysed numerous telescopic observations of the
Moon, some - uncovered as the result of his own literature searches -
dating as far back as 1645. He demonstrated that the fluctuations in
the lunar motion could be represented approximately by an empirical
oscillatory term of period some 260 years and amplitude 12" on which
were superposed additional fluctuations on a time-scale of decades and
of typical amplitude 3 or 4 sec. Eventually Newcomb (1909) mapped
these fluctuations in great detail - see figure 1.4. He named the principal
fluctuation the 'Great Empirical Term', assigning it a period of 275 years
and amplitude 12".95.
In 1878 Newcomb made the following discerning comment:

Now it is most remarkable that observations of transits of Mercury

agree with those of the Moon, and with those of the first satellite of
Jupiter, in indicating that this apparent inequality (in the Moon's motion)
was in part at least due to the Earth's rotation.

Lord Kelvin, at a lecture delivered in 1876, had also ascribed the

fluctuations to variations in the Earth's rate of rotation (Kelvin, 1894,
pp. 271-2). However, it was many years before this conclusion was
generally accepted. Not long before his death, Newcomb (1909) was more
cautious, asserting:

I regard these fluctuations as the most enigmatic phenomenon presented

by the celestial motions, being so difficult to account for by the action of
any known causes, that we cannot but suspect them to arise from some
action in nature hitherto unknown.

Although Newcomb again considered the possibility that 'the inequal-

ities are only apparent, being perhaps due to fluctuations in the Earth's
speed of rotation, and therefore in our measure of time', this remained
no more than a hypothesis to him. In the concluding remarks to his
Researches on the Motion of the Moon and Related Astronomical Elements,
published posthumously in 1912, he stated:

The most unsatisfactory feature of the conclusion of the entire work as

carried through by the author is that, until the matter is cleared up, it will
be impossible to predict the Moon's longitude with the precision required
for astronomical purposes. We shall be obliged to correct the Moon's
mean longitude from time to time, perhaps at intervals of 10 or 20 years,
from observations.

This was an unhappy conclusion to many years of research.

 1.7 The mean motions of the Moon, Sun and planets 19

«~1620

11920 -M820
Fig. 1.4 Fluctuations in the lunar mean longitude from AD 1630 to 1909 as
derived by Newcomb (1909).
20 1 Variations in the length of the day

Towards the end of the nineteenth century, Ernest Brown, on the

suggestion of Sir George Darwin, undertook the derivation of a new
theory of the motion of the Moon. One of the main objectives of this
major task - which was to occupy many years - was to help elucidate
the nature of the fluctuations in the lunar mean longitude. The eventual
outcome was the publication of Brown's Tables of the Motion of the Moon
in 1919. These tables were used for calculating the lunar ephemeris in
the Nautical Almanac between 1923 and 1959 - replacing Hansen's tables
which since 1883 had incorporated certain corrections by Newcomb. In his
theory, Brown included a modified version of Newcomb's Great Empirical
Term (of period 257 years and amplitude 10".71) and also the gravitational
portion of the Moon's secular acceleration (which he calculated to be 6".03
at 1900).
Despite the exhaustive theoretical investigation by Brown, there was
still discord with observation, whether or not the Great Empirical Term
was included. It now seemed clear that the fluctuations in the Moon's
motion did not arise from any defect in the lunar theory. Brown (1915b)
listed various hypotheses which had been proposed to account for these
irregularities, such as the effect of undiscovered matter in the solar system,
non-gravitational forces, oblateness of the Sun, magnetic forces and peri-
odic disturbance of the motion of the Moon by a swarm of meteorites.
Although he regarded none of these ideas as viable, Brown did not at this
stage reconsider the possibility of fluctuations in the Earth's rate of rota-
tion. In this same paper he noted similarities between the fluctuations in
longitude for the Moon, Sun and Mercury since AD 1750 and suggested:

We must therefore look for some kind of a surge spreading through the
solar system and affecting planets and satellites the same way but to
different degrees.

However, publication of his paper was to inspire Hermann Glauert -

a Cambridge postgraduate student - to reopen the question of variations
in the LOD.
Glauert (1915) undertook an analysis of what were then recent obser-
vations of the Moon and Sun - and also of the faster-moving planets
Mercury and Venus. His investigation covered the period from 1865 to
1914, during which - fortuitously - the mean longitude of the Moon had
undergone unusually large and rapid fluctuations. Glauert was able to
show that the discrepancies in mean longitude for the Sun, Mercury and
Venus fairly closely resembled those for the Moon. He concluded:

It appears that the errors in the longitude of the Moon, and the three
bodies considered in this paper may be accounted for by a rather irregular
variation in the rotation of the Earth...
 1.7 The mean motions of the Moon, Sun and planets 21

Although Glauert's result was inconclusive, since he was not able to

prove that the deviations were in the ratio of the mean motions of the
celestial bodies, he had paved the way for a return to the idea that the
fluctuations were only apparent and had their origin in the variations
of the adopted unit of time. Other partially successful investigations of
telescopic observations were carried out by Robert T. A. Innes (1925),
Brown (1926), Sir Harold Spencer Jones (1926, 1932) and de Sitter (1927).
However, not until a later paper by Jones (1939) was it clearly demon-
strated that changes in the Earth's rate of rotation were fully responsible
for the observed anomalies. De Sitter, and subsequently Jones, expressed
the problem as follows. Let B be defined as the fluctuation in the Moon's
mean longitude. Then the difference (5L) between observational and grav-
itational theory (Brown's theory with the Great Empirical Term removed)
for the lunar mean longitude can be represented by an equation of the
form

5L = a + bT + cT2 + B, (1.8)

where T is expressed in Julian centuries from the epoch 1900.0. For the
Sun, the corresponding expression - the difference between observation
and Newcomb's (1895a) theory - is:

51/ = a! + b'T + c'T2 + QBri/n. (1.9)

Here nf/n is the ratio of the mean solar and lunar motions, while Q
is a factor to be determined (as are the coefficients a,b,a\bf and cf).
Similar expressions apply to Mercury and Venus. In the case of the Sun,
nf /n = 0.0747, while for Mercury and Venus the corresponding figures are
0.310 and 0.112. It was necessary to use a value for c, the lunar coefficient
of T 2 , derived from ancient observations in order to solve for Q. As noted
above, de Sitter revised Fotheringham's result for c, obtaining 5".22. If the
observed fluctuations in the mean longitude were entirely Earth-based, Q
would have the value unity. Although de Sitter was able to show that,
within the limits of error, Q had the same value for the Sun, Mercury and
Venus, his result was considerably greater then unity: 1.25 + 0.03 (p.e.).
Shortly afterwards, Jones (1932) reinvestigated the problem and found Q
= 1.19 + 0.05 (p.e.). Hence neither solution excluded the possibility that
part of the observed lunar fluctuations was real.
In 1939, Jones made an exhaustive analysis of the following series of
observations: occultations of stars by the Moon and other lunar observa-
tions since AD 1680; solar declinations since 1760; solar right ascensions
(RA) since 1835; Mercury transits since 1677; and measurements of the
RA of Venus since 1835. His principal conclusions can be divided into
two parts:
22 1 Variations in the length of the day

I i I I I l J I I I I i i

+ 15" - • ^ £\ -

+ 10" - / .• • r\.. —

° \°
+5"

/
-5" -
" * /

-10" ® Mercury (November Transits) °\

o Mercury (May Transits) ?\
-15" • Sun Declinations
o
I l I I I I I I I I i I I
1680 1700 20 40 60 80 1800 20 40 60 80 1900 20
Year
Fig. 1.5 Fluctuations in the lunar mean longitude from AD 1680 to 1939 as
derived by Jones (1939).

(i) For the Sun, Moon, Mercury and Venus, Q = 1.025 + 0.050 (p.e.).
Hence:

The fluctuations in the mean longitudes of the Sun, Moon,

Mercury and Venus are proportional to the respective mean motions
of these bodies and can be attributed to variations in the rate of
rotation of the Earth caused by changes in the moment of inertia
[sic].

(ii) On the assumption of a lunar secular acceleration of 5".22, the

secular acceleration of the Sun was found to be 1".23 + 0".04 (p.e.),
while for Mercury and Venus the accelerations were respectively
5". 10 and 2".00. Thus:

The secular accelerations of the Sun, Mercury and Venus are

proportional to their mean motions and can be accounted for by
retardation of the Earth's rotation by tidal friction.

Figure 1.5, which is taken from figure 1 of Jones' (1939) paper, shows
the fluctuations in the mean longitude of the Moon which he determined
from observations of the Moon (solid line), Sun and Mercury. The results
for the Sun are scaled up by a factor of 13.37 and for Mercury by 3.23 -
the reciprocal ratios of the mean motions relative to the Moon. Although
 1.8 Introduction of a theoretically invariant time-system 23

the scatter among the data for the Sun and Mercury is fairly large, there
is good general accord with the lunar curve.
Jones made the comment that observations of the Sun and Mercury
agree in indicating that the average effects of tidal friction during the
past 250 years are smaller than the average effects over the past 2000
years. He particularly had in mind the difference between his result
for the solar acceleration derived from telescopic data (1".23) and that
obtained by de Sitter from ancient observations (l".80). Tidal friction has
probably remained virtually constant over the last two or three millennia
(see chapter 2). Although Jones assumed that the variations in the rate of
rotation of the Earth as observed from the telescopic observations were
caused by changes in the terrestrial moment of inertia, this is only one of
several possible causes. Today, fluctuations in the Earth's rate of rotation
at the millisecond level have been detected on a wide range of time-scales
from days to millennia. These are superposed on the main tidal slowing
down and are ascribed to a variety of mechanisms: e.g. varying trade wind
patterns, core-mantle coupling, sea-level changes and post-glacial uplift -
see also section 1.9.

1.8 The introduction of a theoretically invariant time-system

As early as 1929, Andre-Louis Danjon recognised that mean solar time
(i.e. GMT) was unsatisfactory for astronomical purposes. A new time-
system which was independent of variations in the Earth's rate of rotation
was required instead. Danjon proposed the adoption of a theoretically
uniform time-system, based on Newtonian dynamics. He wrote:

We should not delay longer in adopting a new practical definition of

time... in order to fix the time by the position of the planets in their orbits
it is necessary to make the calculations in celestial mechanics accounting to
the laws of nature.
[Trans. Clemence (1971).]
Little interest was shown in Danjon's proposal for nearly 20 years.
However, the corrections to the mean longitudes of the Moon and Sun
derived by Jones (1939) were to form the basis for both the definition of a
Newtonian time-system and its measurement (using lunar observations).
When Jones produced his paper in 1939, the standard expressions for
the mean longitudes of the Moon (L) and Sun (I/) relative to the mean
equinox of date were as follows:

L = 973 571^71 + 1732 564406^06T+7^14T 2 +0".0068T 3 +GET, (1.10)

Lf = 1006908".04 + 129 602 768".13T + l".089T 2 . (1.11)

24 1 Variations in the length of the day

Here (as above) T is measured in Julian centuries of 36 525 mean solar

days from the epoch 1900.0 (Julian day number 2415020.0) and GET
denotes the Great Empirical Term. These equations had been derived
respectively by Brown (1915a) and Newcomb (1895b). Both of the above
expressions include linear and quadratic terms deduced by Newcomb
(1895b) to allow for precession - the latter of coefficient l'Ml.
As noted above, Jones (1939) adopted de Sitter's result for the lunar
acceleration c of 5".22. In order to secure close agreement with modern
observations he also added a constant (+4".65) and a linear term (+12".96
T). The resulting correction to Brown's expression is therefore
5L = +4".65 + 12".96T + 5".22T2 - GET. (1.12)
In the case of the Sun, Jones derived the following amendment to
Newcomb's expression for L1:
51/ = +l".00 + 2".97T + 1".23T2. (1.13)
In 1948, Gerald M. Clemence - who at the time was unaware of
Danjon's suggestion - used Jones' correction to Newcomb's mean solar
longitude (equation (1.13)) to define 'Newtonian time'. This definition was
based on the mean motion of the Sun in longitude: 1 arcsec in 24.349 sec
of time. Denoting AT as the difference between GMT and the proposed
new time-system (in the sense Newtonian time - GMT), Clemence derived
the equation
AT = 24.349 + 72.3165T + 29.949T2 + 1.8215 sec. (1.14)
Subsequently at the International Astronomical Union (IAU) General
Assembly held at Rome in 1952 it was recommended that:
In all cases where the mean solar second is unsatisfactory as a unit of
time by reason of its variability, the unit adopted should be the sidereal
year at 1900.0; that the time reckoned in these units be designated
Ephemeris Time; that the change of mean solar time to ephemeris time be
accomplished by the following correction:
AT = 24.349 + 72.3165T + 29.949T2 + 1.8215 sec,
where T is reckoned in Julian centuries from 1900.0 January 0 Greenwich
Mean Noon and B has the meaning given by SPENCER JONES in
Monthly Notices, R.A.S 99, 541 (1939) and that the above formula also
define the second.
[IAU (1954).]
The name 'ephemeris time' (usually abbreviated to ET) had been sug-
gested by Dirk Brouwer.
The unit of ET was thus chosen to be the same as that of Newcomb's
solar ephemeris. In this ephemeris, the mean motion of the Sun was
 1.8 Introduction of a theoretically invariant time-system 25

expressed in terms of the average length of the second - regarded as

1/86400 of the mean solar day - during the latter half of the eighteenth
century and most of the nineteenth. This resulted from the fitting of
his theory to observations ranging in date from 1750 to 1892 (Newcomb,
1895b). Thus the average LOD on the GMT/UT scale during this interval
(mean epoch c. 1820) became the standard LOD on the ET scale. At the
IAU General Assembly at Dublin in 1954, the duration of the second was
fixed on the basis of the above definition as 1/31 556925.975 of the length
of the tropical year for 1900.
On account of its slow motion and extreme brilliance, the Sun was by
no means a suitable object for measuring ET. The much more rapidly
moving Moon was well suited for this purpose. A revised formula was
thus needed for the lunar mean longitude. Clemence (1948) showed that
since the mean motion of the Moon is 13.37 times as rapid as that of the
Sun, equation (1.13) is equivalent to the following correction to the mean
lunar longitude:
AL = 13".37 + 39".71T + 16".41T2. (1.15)
In order to convert Brown's expression for L to Newtonian Time (i.e. ET),
the correction (5L — AL) , as obtained from equations (1.12) and (1.15),
was required. Representing the lunar mean longitude on the new system
by K
A = L + 5L-AL, (1.16)
or, inserting numerical values,
k = L- 8".72 - 26".75T - 11".22T2 - GET, (1.17)
one could thus derive AT directly by comparing longitudes of the Moon
computed from the lunar theory with observation.
The above amendment to the lunar mean longitude (equation (1.17))
was also recommended at the 1952 IAU Rome meeting 'in order to bring
the lunar ephemeris into accordance with the solar ephemeris'. With a
minor alteration to the coefficient of T (to 26".74), it was incorporated
in the Improved Lunar Ephemeris (ILE, 1954) and subsequently the lunar
ephemeris j = 2 (IAU, 1968).
From section 1.6, twice the coefficient of T 2 in the above equation (i.e.
—22.44 arcsec/cy2) represents the estimated non-gravitational acceleration
of the Moon (dn/dt or n) caused by tides. This gradual retardation is
accompanied by a slow recession of the Moon from the Earth. As will
be discussed in chapter 2, the most reliable modern investigations, using
a variety of techniques, indicate a value for n close to —26 arcsec/cy2.
In 1960, the International Conference on Weights and Measures ap-
proved the definition of the second as:
26 1 Variations in the length of the day

the fraction 1/31 556925.9747 of the tropical year 1900 January 0 at 12

hours ephemeris time.
In the same year, ET became the standard time-system adopted in the
Astronomical Ephemeris. By this date the numerical coefficients for AT
had been very slightly altered and the finally adopted formula was
AT = 24.349 + 72.318T + 29.950T2 + 1.821445 sec. (1.18)
Ephemeris time was to have a relatively short active life. As early as
1945, Isidor Isaac Rabi had suggested the idea of using atomic resonances
in a chronometer. In 1955, the first atomic clock - the caesium beam
resonator - was developed by L. Essen and J. V. L. Parry (Essen and
Parry, 1957). This device had a stability of the order of 1 part in 1013,
losing or gaining less than 10~8 sec in a day. It was thus far superior to any
solar system clock and - in particular - was ideally suited to the accurate
determination of fluctuations in the LOD. By July 1955, atomic time-scales
became available in several countries. At the International Conference on
Weights and Measures in 1967, the second was redefined as:
The duration of 9192631 770 periods of the radiation corresponding to
the transition between the two hyperfine levels of the ground state of the
caesium atom 133.
This unit, which is the fundamental unit of time in the international
system (SI), was chosen to be as close to the length of the second of the
ET scale as measurement would allow (Markowitz et al., 1958).
The new standard became known as international atomic time (TAI)
in 1971. As there had been no discontinuity in the adopted time-scale
since 1955, the designation TAI can be effectively used as far back as that
date. The origin of TAI was arbitrarily chosen so that the TAI and UT1
readings were identical at 0 h on Jan 1 in 1958. Since the value of AT
was 32.184 sec at this moment, the difference ET-TAI was also 32.184
sec. Currently (1997) AT has a value of close to 60 sec.
In 1984, ET was replaced by terrestrial dynamical time (TDT) for
astronomical purposes and the fundamental geocentric ephemerides were
defined in terms of TDT. In order to provide continuity with ET, the
difference between TDT and TAI was set as 32.184 sec. No significant
variations between TDT - now simply known as terrestrial time (TT) -
and TAI have been detected. The designation TT will be used in place of
both TDT and ET throughout most of the remainder of this book.

1.9 Short-term fluctuations in the Earth's rotation

Although the fundamental aim of this book is the investigation of vari-
ations in the Earth's rate of rotation on a millennial time-scale, some
 1.9 Short-term fluctuations in the Earth's rotation 27

brief remarks on shorter-term fluctuations are also appropriate. These

fluctuations - which cannot be resolved using pre-telescopic observations
- may be divided into two main groups, depending on their periodicity.
The categories are: (i) sub-annual and inter-annual variations; and (ii)
decade fluctuations.
Variations in category (i), which are of approximate amplitude 1 milli-
second (1 ms), are of too short a periodicity to be in any way relevant to
the theme of the present book. It is well established that they are produced
by exchange of angular momentum between the atmosphere and the sur-
face of the Earth, alterations in the force of the winds against mountain
ranges being largely responsible. (For reference, the principal moment
of inertia of the atmosphere is about 1.8 x 10~6 of that of the Earth
itself.) Significant peaks at about 6 months, 1 year, 2.2 years and 4 years
have been detected (e.g. Hide et a/., 1980; Jordi et al, 1994). The higher
frequency oscillations can only be mapped in detail (using techniques such
as very long base-line interferometry) since 1955, when the TAI time-scale
was introduced. However, the inter-annual variations can be detected as
far back as 1890 by analysing timings of lunar occultations of stars.
The fluctuations which occur on a time-scale of decades - category (ii) -
are of typical amplitude 3 ms. These variations can be traced over the last
350 years or so, mainly using recorded timings of occultations of stars by
the Moon (Stephenson and Morrison, 1984). The principal causal mech-
anism would appear to be angular momentum transfer between the outer
fluid core of the Earth and the surrounding solid mantle by the process
of electromagnetic coupling. However, small alterations in global sea-level
owing to freezing or melting of polar ice may also be partly responsible.
Over a period of several decades, the cumulative effect of systematic
changes in the LOD at the millisecond level is very significant. To give
a fairly typical example, over the past 40 years (some 15 000 days) the
average LOD has been approximately 2 ms longer than the standard
value of 86400 SI sec. Hence the accumulated clock error TT — UT
(i.e. AT) during that interval has amounted to some 30 sec. Even fairly
crude telescopic timings of occultations of stars (which are effectively
instantaneous events) can be used to trace similar decade fluctuations in
the past in some detail. Resolution of these fluctuations is very good as far
back as about AD 1860, but progressively deteriorates in earlier centuries
(Stephenson and Morrison, 1984).
Figure 1.6, which is based on figure 1 of Stephenson and Morrison
(1995), depicts the form of the AT curve since AD 1620 by a dotted line.
If only tides and other long-term effects producing a fairly steady change
in the LOD were significant, the basic form of this curve would have been
parabolic. However, on a time-scale of centuries the decade fluctuations
have tended to obscure the long-term trend (which is represented by a
28 1 Variations in the length of the day

100 I I I I I I I I I I I I I I I
I ' '
" A
I \\ /

_ \ /
— W - — Long-term trend
-* 50
o
_ ^ Actual / /

Z>
_ \
\
I \

- \
\ -

I I | I I I I I I I I I I I I I I I I
1600 1700 1800 1900 2000
Year
Fig. 1.6 The form of the AT curve since AD 1620 (shown by a dotted line),
as depicted by Stephenson and Morrison (1995). The long-term parabolic trend
(represented by a dashed line) is also depicted. The cusp of this mean parabola
is placed close to AD 1820.

dashed line in the diagram). The cusp of this mean parabola is placed
close to AD 1820 in figure 1.6. As discussed in section 1.8, this is
the approximate mean epoch of the observations analysed by Newcomb
(1895b) which were subsequently used to define the standard LOD on
the ET scale. From Stephenson and Morrison (1995), the approximate
equation of this parabola is —20 + 3It 2 , where t is expressed in centuries
relative to AD 1820 (see also chapter 14).
Figure 1.7 shows the first time-derivative of the modern AT curve,
as derived by Morrison and Stephenson (1981). This diagram delineates
actual changes in the LOD relative to the standard length of 86 400 SI sec.
Power spectral analysis reveals a broad peak around a periodicity of 30
years. Calculations indicate that torques of the order of 1018 Nm would
be required to produce the observed variations. Such torques are much
greater than those produced by lunar and solar tides - which amount to
about 5 x 1016 Nm (see chapter 2).
In analysing the various eclipse observations later in this book, the
primary objective will be to determine a series of AT results by comparing
individual values of UT derived from observation with their computed
equivalents on TT. For each measurement,
= TT-UT. (1.19)
 1.10 The Julian and Gregorian calendars 29

1600 2000

Fig. 1.7 The first time-derivative of the modern AT curve, as derived by

Morrison and Stephenson (1981); changes in the LOD relative to the standard
mean solar day of length of 86 400 SI sec are shown.

1.10 The Julian and Gregorian calendars

It seems appropriate to conclude this introductory chapter with some
remarks on the Julian and Gregorian calendars, to which all eclipse dates
are reduced throughout this book.
The fundamental calendrical periods are the tropical year of 365.2422
mean solar days (which fixes the seasons) and the synodic month of
29.5306 days (which determines the availability of moonlight). Early
calendars followed several basic forms. Some were purely solar (e.g. that
of ancient Egypt or the Julian system), others were exclusively lunar (e.g.
the Muslim calendar), while a number of calendars were luni-solar (e.g.
those of Babylon and China). Over the centuries, a variety of epochs have
been adopted for counting years - for example the epoch of the Olympiads
(776 BC), the traditional date of the foundation of Rome (753 BC), the
Seleucid era (311 BC), the era of Diocletian (AD 284), and the era of the
Hijra (AD 622) - in addition to the Christian era. Conversion between
the various early calendars and their Julian or Gregorian equivalents will
be discussed in the relevant chapters of this book.
The Julian calendar was instituted by Julius Caesar in 45 BC with the
advice of the Alexandrian astronomer Sosigenes. This calendar replaced
the luni-solar system of the Roman republic which, owing to faulty inter-
calation, became increasingly out of phase with the seasons. For example,
by 50 BC the vernal equinox was falling in mid-May rather than in late
March. Caesar directed that the year which is now designated 46 BC
should contain an extra 90 days; starting with the following year a reg-
ular four-year cycle was to be adopted. In each cycle, the first year was
30 1 Variations in the length of the day

assigned 366 days, while the remaining three years each contained 365
days. The average length of a Julian year was thus exactly 365.25 days.
Initially, the extra day was intercalated after Feb 24; use of Feb 29 is a
later development.
After the death of Caesar, an extra day was erroneously inserted every
three years. This practice continued until 9 BC, following which Augustus
restored the status quo by decreeing that further intercalation be suspended
for 16 years, all years in this interval having 365 days. Finally, in AD 8
the regular cycle initiated by Julius Caesar was once more adopted. This
continued without further modification until AD 1582.
It was not until AD 525 that years actually began to be counted
relative to the Christian era. The Anno Domini system was first introduced
by the monk Dionysius Exiguus, who employed it in his Easter tables.
In England, these tables, and with them the Christian era, were adopted
at the Synod of Whitby in AD 664. Not long afterwards, the Venerable
Bede (AD 672-735) systematically employed Anno Domini in his important
chronological writings, which gradually led to its popularisation. Use of
the Christian Era became widespread in Europe by the eleventh century.
Dionysius had derived an incorrect date of 753 AUC (from the foun-
dation of the city of Rome) for the birth of Christ; this event probably
took place about 4 or 5 years previously. Although Dionysius' result was
questioned by Bede, it was not altered by him. In the late ninth century,
the German monk Regino rejected the equation 753 AUC = AD 1, but by
then Dionysius' choice was too well established for any further alteration
to be made. The use of BC dates (in addition to AD) did not find general
favour until AD 1650, following the extensive chronological researches of
Dionysius Petavius (AD 1583-1652).
Today, historians adopt a standardised Julian calendar which follows
the same basic rules for all BC and AD dates up to AD 1582. For any
date after the Christian era (up to AD 1582), all year numbers which are
divisible by 4 - including century years - are bissextile or leap years. Since
there is no year zero on this system (1 BC is immediately followed by AD
1), 4 BC is not a leap year, but 5 BC, 9 BC and so on are bissextile. In
order to avoid a discontinuity of a year, historians of astronomy frequently
adopt a variant of the BC/AD scheme in which years are numbered as
positive or negative. On this system (which includes a year zero), years
represented by positive numbers (e.g. +1133) correspond exactly with
their AD equivalents. However, years denoted by negative numbers differ
by unity from their BC alternatives; thus —430 corresponds to 431 BC. In
particular, the year 0 is equivalent to to 1 BC. This choice simplifies the
enumeration of intervals between selected years. Throughout the present
book, all dates derived from historical researches will be expressed on
the BC/AD system. However, when AT results are tabulated - as in the
 1.10 The Julian and Gregorian calendars 31

Appendix - positive and negative year numbers will normally be used

instead since a continuous system is necessary when analysing the results.
As an average Julian year is slightly longer than a tropical year (365.25
days compared with 365.2422 days), the start of each year on the Julian
calendar became progressively delayed relative to the seasons. The differ-
ence amounted to one day every 128 years. Over the centuries, this defect
produced a marked change in the vernal equinox (and thus in the date
of Easter). Eventually it led to the reformation of the calendar by Pope
Gregory XIII in AD 1582. For this task, Pope Gregory was advised by
the Jesuit astronomer Christopher Clavius (AD 1537-1612).
In essence, there were two main steps in the adoption of the Gregorian
calendar:
(i) Ten days in AD 1582 were omitted from calendar reckoning. The
date 1582 Oct 4 (a Thursday), the last day of the Julian system, was
immediately followed by 1582 Oct 15 (a Friday), the first day of the
Gregorian scheme. In consequence, the date of the vernal equinox
was restored to Mar 21, as it had been at the time of the Council
of Nicea in AD 325. The sequence of weekdays was not interrupted,
however.
(ii) In century years which were not divisible by 400 (e.g. AD 1700,
1800 and 1900) the intercalary day was now omitted. All other years
which were divisible by 4 (including 1600 and 2000) remained as
leap years.
Although the Gregorian calendar was at once officially adopted by
most Roman Catholic countries, other countries delayed in its acceptance.
For example in England the change was not made until as late as 1752;
by then, since 1700 was not a leap year on the Gregorian reckoning,
it was necessary to omit 11 days. Thus the day after 1752 Sep 3 (a
Wednesday) was Sep 14 (a Thursday). The mean length of the Gregorian
year is 365.2425 days, only 0.0003 days longer than the tropical year. The
discrepancy amounts to no more than one day in about 3300 years.
On a systematic solar calendar - such as the Julian or Gregorian
system - it is a relatively straightforward matter to calculate the exact
number of days between any two selected epochs. Most computations of
this sort make use of the Julian day number (JDN), devised by Joseph
Justus Scaliger (AD 1540-1609). This scheme is based on a cycle of 7980
Julian years, commencing in 4713 BC. The precise epoch from which
Julian days are numbered is Greenwich noon on Jan 1 (Julian calendar)
in that year. A major reference epoch is 1900.0 (i.e. Greenwich noon
on 1899 Dec 31), for which the JDN is 2415020.0. Since most modern
astronomical computations make use of terrestrial time, it has become
32 1 Variations in the length of the day

customary to use the Julian ephemeris date (JED) where necessary; this
begins at 12h TT.
Various algorithms have been developed to convert from either the
Julian or Gregorian calendar to the JDN, among the simplest being those
deduced by Muller (1975), which make full use of the rules of FORTRAN.
These are given in equations (1.20) and (1.21). Here J is the JDN, Y the
year, M the month and D the day of the month; all variables are treated
as integer. For any date AD on the Gregorian calendar:
J=367*Y - 7*(Y+(M+Y)/12)/4 - 3*((Y+(M-9)/7)/100+l)/4
+275*M/9+D+1721029. (1.20)
For any date on the Julian calendar the following simpler formula
applies:
J=367*Y-7(Y+5001+(M-9)/7)/4+275*M/9+D+1729777. (1.21)
I have employed these formulae in many of the computer programs
which I have developed.
 Tidal friction and the ephemerides
of the Sun and Moon

2.1 Introduction
In order to determine the value of AT from an eclipse observation, it
is necessary to be able to calculate accurately the positions of the Sun
and Moon at any selected epoch. By definition, the Sun has negligible
acceleration on TT. However, the longitude of the Moon contains an
appreciable quadratic term - part of which (owing to the reciprocal action
of the tides) can only be determined empirically. Since many of the eclipses
recorded in history are remote from the present-day, the effect of the lunar
accelerative term on their calculated visibility is substantial. Furthermore,
knowledge of the tidal component of this acceleration (usually denoted by
n), leads directly to a determination of the effect of the tides on the Earth's
spin - see section 2.4. Consequently, it is important to investigate both
the numerical value of n and its constancy during the historical period.
Each of these questions will be considered in the immediately following
sections (2.2 and 2.3).

2.2 Evaluation of the lunar tidal acceleration on TT

As discussed in chapter 1, the gravitational component of the lunar
acceleration (coefficient of T2 equal to 6".05) is well established. Up to
about 1970, all estimates of the non-gravitational lunar acceleration were
based on analyses in a UT framework. Results for the secular acceleration
of the Moon (c) and Sun (cr) on UT can be easily converted to n using
the formula:
n = 2(c — (n/n')cr) arcsec/cy2, (2.1)
where n/nr is the ratio of the mean lunar and solar motions. The factor of
2 is introduced because whereas c and d are coefficients of T 2 , h is a true
acceleration. Substituting the numerical value for n/nf of 13.37, we obtain
n = 2c - 26.1 Ad arcsec/cy2. (2.2)

33
34 2 Tidal friction and the ephemerides of the Sun and Moon

Thus for example using de Sitter's (1927) results for c of 5".22 + 0".30
(p.e.) and d of l".8O + 0".16 (p.e.) leads to n = -37.7 + 6.4 (s.e.)
arcsec/cy2. Replacing the above figure for d by Jones' (1939) value of
T.23 + 0".04 (p.e.) gives instead n = -22.44 + 1.8 (s.e.) arcsec/cy2. As
noted in chapter 1, this latter value for n was adopted in both the Improved
Lunar Ephemeris (ILE, 1954) and the lunar ephemeris j = 2 (IAU, 1968).
Commencing in 1970, several investigators attempted to utilise ancient
eclipse observations to solve simultaneously for both h and the mean
acceleration of the Earth's spin (cb/co). These authors included Newton
(1970 and 1972b), Muller and Stephenson (1975) and Muller (1976), results
for h ranging from —42 to —30 arcsec/cy 2. Such efforts proved to be of
limited usefulness because in the analysis of eclipses the two parameters
n and (cb/co) are highly correlated. In practice, only a linear combination
of these quantities can be satisfactorily determined from eclipse data.
In recent years, investigators of early eclipse observations have assumed
an independent value for n in order to solve exclusively for AT. Fortu-
nately for Earth rotation studies, several precise determinations of n using
a variety of other means have been made in recent years. Techniques
include the following: transits of the planet Mercury across the solar disk,
artificial satellite data, and lunar laser ranging. Each of these methods
will be discussed briefly below.

2.2.7 Transits of Mercury

In 1975, Morrison and Ward determined n from a comparison over most
of the telescopic period between the AT curves obtained from (i) transits
of the planet Mercury across the solar disk and (ii) occultations of stars
by the Moon. Both Mercury and the Sun have negligible acceleration on
TT. The investigation by Morrison and Ward, which covered the period
from 1677 to 1973, yielded a result for n of —26 + 2 arcsec/cy2. This value
has been closely confirmed in recent years using both artificial satellites
and lunar laser ranging.

2.2.2 Artificial satellite data

The trajectory of a close artificial satellite is significantly perturbed by
the tidal deformation of the solid Earth and oceans. Analysis of orbital
measurements yield results for the tidal phase lag, leading to a value for n.
Early results for n obtained using this technique include: —27.3 + 5.2
arcsec/cy2 (Lambeck, 1977) and -27.4 + 3 (Goad and Douglas, 1978).
Most recently, Christodoulidis et al (1988) deduced -25.27 + 0.61. In
particular, these last authors determined the individual contributions to
n from long-period, diurnal and semi-diurnal tides to be as follows (in
 2.2 Evaluation of the lunar tidal acceleration on TT 35

arcsec/cy 2 ): -0.69 + 0.36 (long-period), -3.18 + 0.25 (diurnal), and

-21.40 + 0.43 (semi-diurnal). By far the largest contribution (-20.00 +
0.40) comes from a single tide - the M2 (principal lunar semi-diurnal).

2.2.3 Lunar laser ranging

The first laser retro-reflector array was installed on the lunar surface
by the Apollo 11 astronaut Neil Armstrong in 1969. There are now five
instruments of this kind on the Moon - the last deployed by Lunakhod 2 in
1973; the signals have been mainly transmitted and received at McDonald
Observatory in Texas.
The first published measurements in 1978 by Calame and Mulholland
and separately by Williams et al. gave respectively n = —24.6 + 5 and
—23.8 + 4 arcsec/cy 2. (Calame and Mulholland estimated the formal
standard deviation to be + 1.6 arcsec/cy2, but suggested that + 5 was
a more realistic error estimate.) Soon afterwards, Ferrari et al (1980)
obtained -23.8 + 2.6, while Dickey et al (1982) deduced -23.8 + 1.5.
Since 1992, results for h close to —26 arcsec/cy2 have been derived from
lunar laser ranging, the most recent figure being —25.88+0.5 (Dickey et al,
1994). James G. Williams, one of the laser ranging team at Jet Propulsion
Laboratory, Pasadena, has informed me (personal communication: April
1994) that lately values for h in the range —25.8 to —26.0 arcsec/cy2 'have
consistently been obtained'.
The rather low results obtained in the early 1980s were to have ad-
verse consequences for contemporary ephemerides; a value for h of close
to — 23".9/cy2 was incorporated in both the Jet Propulsion Laboratory
numerically integrated ephemeris DE200/LE200 (Standish, 1982) and
the semi-analytical ephemeris ELP 2000 (Chapront-Touze and Chapront,
1983). Fortunately, a much more viable value for h of—26.21 arcsec/cy2 is
implicit in the long-term Jet Propulsion Laboratory numerical integration
known as DE102/LE51 (Newhall et al., 1983). This ephemeris, which is
very useful for historical researches, covers the entire period from 1411
BC to AD 3002.

2.2.4 Selected value for h

The weighted mean of the latest published results for h obtained from
artificial satellite data (-25.27+0.61) and lunar laser ranging (-25.88+0.5)
is 25.63 + 0.39 arcsec/cy2. However, bearing in mind the comment by
Williams quoted above, I have adopted a figure for h of —26.0 arcsec/cy2.
This was the choice of Stephenson and Morrison (1995).
36 2 Tidal friction and the ephemerides of the Sun and Moon

It is readily shown that as the lunar orbit expands,

where r is the mean lunar distance. Inserting the appropriate values for
r (= 3.84 x 108 m) and n (= 1.732 x 10 9 7cy) gives r = 0.148n m/cy.
A lunar acceleration of —26/cy2 is equivalent to a rate of retreat of the
Moon (r) of 3.86 m/cy. Of more direct relevance in the present context,
the length of the synodic month can be calculated to be increasing at an
average rate of 0.038 sec/cy. Hence since 700 BC, the mean length of the
month has increased by only 1.0 sec.

2.3 The constancy of the lunar tidal acceleration

The degree of constancy of ri, the lunar tidal acceleration, during the
interval of approximately 2700 years covered by reliable historical obser-
vations is clearly a matter of some concern. This acceleration is directly
proportional to the lunar tidal torque and hence to the rate of dissipation
of the Earth's rotational energy (see section 2.5).
Lunar tidal dissipation takes place mainly in the hydrosphere; bodily
friction represents only a small proportion of the total - perhaps 10 per
cent (e.g. Lambeck, 1980, p. 329). Hydrospheric dissipation occurs both
in the deep oceans and the continental shelves, but the precise ratio of the
contributions from these two sources is uncertain. On the time-scale cov-
ered by the present investigation, dissipation in the deep oceans ought to
be sensibly constant since global parameter changes (e.g. ocean-continent
distribution) should be negligible over a very few millennia. However,
there is evidence that even on the shelf areas changes should be very
small. There has not been any appreciable change in sea-floor topogra-
phy and although significant localised sea-level variations have probably
occurred, global alterations in sea-level have been minor. Studies of
tide-gauge records by Nakiboglu and Lambeck (1991) indicate a global
sea-level rise in recent decades by 1.15 + 0.3 mm/y. Measurements of
global sea-level variations on a time-scale of millennia, although less reli-
able, suggest roughly the present rate over the past 2700 years (Pirazzoli,
1991). A significantly more marked change would appear to be unlikely
since the entire period is long after the last ice-age (which ended some
10000 years ago). On this basis, it may be estimated that since 700 BC
mean sea-level has probably not changed by more than about 3 m.
In general, the average depth of water on the continental shelves is
about 100 m; a change in the global sea-level by 3 per cent (i.e. by 3 m)
would probably alter the area of the shelves by a similar ratio. No model
is available which can predict just how tidal dissipation would respond
 2.4 The tidal acceleration of the Earth's spin 37

to variations in both area and water depth on the continental margins.

However, the very small magnitude of these variations suggests that the
effect on global dissipation will be negligible. In summary, the assumption
of a sensibly constant value for n over the historical period seems justified.
(I am grateful to Dr J. M. Huthnance of the Proudman Oceanographic
Laboratory, Bidston, Merseyside for helpful discussion on this question.)

2.4 The tidal acceleration of the Earth's spin

Analysis of ancient and medieval eclipse observations leads to a result
for the total rate of change in the LOD owing to both tides and other
causes. Hence any residual non-tidal lengthening of the day (for example
produced by post-glacial uplift, sea-level variations, etc.) may be obtained
by subtracting the tidal component which is produced by the action of
both the Moon and Sun.
Christodoulidis et al (1988) analysed the perturbations of the orbits of
near-Earth artificial satellites by tidal effects. On the basis of conservation
of angular momentum in the Earth-Moon system, these authors obtained
the following empirical relation between the observed tidal acceleration of
the Moon and the acceleration of the Earth's spin caused by lunar and
solar tides (coj):
cbT = (+49 + 3) n arcsec/cy2. (2.4)
This result was only marginally smaller than that obtained by Lam-
beck (1980, p. 337) by combining oceanic estimates, tidal solutions and
astronomical estimates for d>j - i.e. (+51 + 4) n arcsec/cy2.
Converting from arcsec/cy2 to rad/sec 2 (the customary units for d>j),
by multiplying by the factor 4.868 x 10~25 gives
d>T = (2.39 + 0.15) x 10" 23 n rad/sec 2 . (2.5)
Inserting n = —26.0 arcsec/cy2 yields
cbT = (-6.20 + 0.38) x 10~ 22 rad/sec 2.
This last result may be reduced to the rate of change in the LOD (Zj)
in ms/cy by multiplying by a factor 3.745 x 10 21:
Z T = 2.3 + 0.1ms/cy.
This result will be assumed in subsequent investigations; see chapter 14.
If the LOD were to increase at a constant rate of 2.3 ms/cy, there
would be a parabolic divergence between UT (as measured by the Earth's
rotation) and TT. This would take the form
(2.6)
38 2 Tidal friction and the ephemerides of the Sun and Moon

-800 -400 0 +400 +800 +1200 +1600 +2000

-600 -200 +200 +600 +1000 +1400 +1800
Year

Fig. 2.1 The estimated range in ATtidai in sec as far back as 800 BC according
to the results of Christodoulidis et al (1988).

where qj is a constant and t is reckoned from the epoch at which the unit
of UT is equal to the unit of TT. From chapter 1, this epoch lies close to
AD 1820. If AT is measured in seconds and t in Julian centuries of 36 525
ephemeris days, then qj can be expressed in terms of Zj (ms/cy) by
qT = 36.525ZT/2. (2.7)
With Zj = 2.3 + 0.1 ms/cy, then from equations (2.6) and (2.7)
(42 + 2)£2. (2.8)
On this basis, the estimated range in ATtidai in sec as far back as 800 BC
is depicted in figure 2.1. Since many ancient and medieval observations
were timed to a small fraction of an hour, it can be seen that especially
in earlier centuries any significant non-tidal variations from the mean AT
parabola should be readily apparent.

2.5 Lunar and solar tidal dissipation

The torques which the Moon and Sun produce on the rotating Earth
may be readily estimated. From Kepler's third law and the principle
of conservation of angular momentum, the lunar tidal torque (F) is
approximately given in terms of n by
F = -EMr2 n/3(E + M), (2.9)
 2.5 Lunar and solar tidal dissipation 39

where E and M are the masses of the Earth and Moon and r is the mean
separation. Expressing n in terms of arcsec/cy2 in the above equation and
inserting numerical values for E (= 5.58 x 1024 kg), E/M (= 81.3) and r
(= 3.84 x 108 m) gives the mean lunar tidal torque F as
F 1.74 x lO15n Nm. (2.10)
With h = -26.0 arcsec/cy2, F ~ +4.5 x 1016 Nm.
Jeffreys (e.g. 1976, p. 319) showed that if friction in the oceans is
assumed to be linear, the ratio of the solar tidal torque (Ff) to that of the
Moon is equal to 1/4.9. At the other extreme, for quadratic friction, this
ratio increases to 1/3.8 (Newton, 1968). Hence if one adopts a mean ratio
for F1 IF of 0.23, the combined lunar and solar torque amounts to some
+5.6 x 1016 Nm. In opposition to the main deceleration, the semi-diurnal
atmospheric tide is responsible for a small accelerative term. The torque
produced by this mechanism is estimated by Volland (1990) as (—3.2 +
0.3) x 1015 Nm. Adding this to the above result of +5.6 x 1016 Nm
gives +5.3 x 1016 Nm as a fair estimate of the tidal torque (FT) causing
a retardation of the Earth's spin.
The rate at which the tidal torque does work (Pj) is given by
PT = FT(co-n). (2.11)
Inserting the appropriate values for co (= 7.29 x 10~5 rad/sec) and n
(= 0.27 x 10" 5 rad/sec) yields
PT ~ 3.7 x 1012 W.
The first estimate of global tidal dissipation based on oceanic studies
was made by Jeffreys in 1920. Jeffreys considered that the contribution
from the open oceans would be negligible, but he was of the opinion that
shallow seas would be a major source of energy loss. Following a paper
by Taylor (1919) which contained estimates of the rate of dissipation
of energy by tides in the Irish Sea, Jeffreys (1920) obtained data on
tidal ranges and currents for shallow seas in various parts of the Earth
(mainly from the Admiralty Pilot). On this basis, he calculated a value
for Pj of 1.1 x 1012 W. Soon afterwards, Heiskanen (1921) made similar
computations, deriving Pj = 1.9 x 1012 W. Although both of these results
were considerably smaller than the value derived above (i.e. 3.7 x 1012 W),
they were at least of the correct order of magnitude.
In relatively recent years, several numerical models of global ocean
tides have been published. These incorporate terms for ocean loading
and gravitational self-attraction into the Laplace tidal equations. Perhaps
the most comprehensive solution for Pj obtained in this way is that by
Lambeck (1980, p. 330). By summing the contributions of the individual
oceanic tides (both lunar and solar), Lambeck derived a value for Pj of
40 2 Tidal friction and the ephemerides of the Sun and Moon

4.42 x 1012 W. This result is in adequately close accord with that obtained
above from astronomical data.

2.6 Ephemerides of the Sun and Moon

All eclipse computations in this book have been made using specially
designed programs - which are adaptations of standard ephemerides for
the Sun and Moon. In computing solar positions I have used Newcomb's
analytical ephemeris (Newcomb, 1895a). All Fourier terms of ampli-
tude greater than 0.025 arcsec in longitude and 0.005 arcsec in latitude
have been included. For lunar positions a modification of the analytical
ephemeris j — 2 (IAU, 1968), in which n = —26 arcsec/cy2, has been
adopted. In this case, all Fourier terms in longitude greater than 0.05 arc-
sec, and in latitude greater than 0.01 arcsec have been included. In testing
the accuracy of these ephemerides, I am grateful to Dr E. M. Standish of
Jet Propulsion Laboratory, Pasadena, for supplying comparison solar and
lunar positions based on DE102/LE51. Further details are given below.

2.6.1 Solar ephemeris

Newcomb's ephemeris for the Sun - which effectively defines the TT
scale - incorporates the following expression for the solar mean longitude
relative to the mean equinox of date:
X[900 = 1006908^.04 + 129 602768".13T + l".089T 2 , (2.12)
where T is in Julian centuries of 36 525 days measured from the epoch
1900.0 (Julian day number 2415020.0). The corresponding expression for
the epoch J2000 (i.e. 2000.0: Julian day number 2451545) is
4)oo = 1009 677".26 + 129 602 770".31£ + l".089£ 2 , (2.13)
where E is also in Julian centuries.
For comparison, the linear term in the solar mean longitude derived
by Bretagnon (1982) relative to J2000 has the coefficient 129 602771".36,
while the parabolic term has coefficient +l // .093. (A minute cubic term
is ignored.) The difference between the two parabolic terms (0".004£2) is
negligible, amounting to no more than 2" at 700 BC - the approximate
date of the oldest reliable eclipse observations. Since the disparity between
the linear terms (l".05E) reaches some 30 arcsec at 700 BC, this would
alter the computed times of eclipses or occultations by about 1 minute.
Such a discrepancy is insignificant at this early period.
A comparison which I have made between solar longitudes obtained
using Newcomb's ephemeris and those from the numerically integrated
ephemeris DE102 for dates of historical eclipses yields deviations less
 2.6 Ephemerides of the Sun and Moon 41

than about 30" throughout the entire period since 700 BC; these show a
roughly linear trend.

2.6.2 Lunar ephemeris

The mean lunar longitude derived by Brown (1915a) relative to the mean
equinox of date is
L = 973571".71 + 1732 564 406".06T + 7".14T2 + 0".0068T3. (2.14)
As above, T is measured in Julian centuries from the epoch 1900.0.
From chapter 1, the correction to the above expression necessary to
convert from a UT to a TT framework - as in j = 2 - is
X = L - 8".72 - 26".74T - 11".22T2, (2.15)
where X is the mean longitude of the Moon on TT.
The equivalent value of n is —22.44 arcsec/cy2. A further amendment
to j = 2 is necessary to incorporate a value for h of —26.0 arcsec/cy2, as
derived by Morrison (1979). From an analysis of numerous occultations
reported between 1943 and 1974, Morrison showed that the following
addition should be made in order to obtain the optimum accord with
modern observations:
5^ = -l / / .544 + 2".330T - 1".78T2. (2.16)
The sum of equations (2.15) and (2.16), i.e. -10".26-24".41T-13".00T 2 ,
is the total correction which has been applied to Brown's expression for
the lunar mean longitude. This yields a revised result at the epoch 1900.0
of
A19Oo = 973 561^.45 + 1732 564 381".65T - 5".86T2 + 0".0068T3. (2.17)
When expressed in terms of the epoch J2000 this becomes:
X2000 = 785 937".25 + 1732 564 369".95£ - 5".88£2 + 0".0068£3. (2.18)
For comparison, the semi-analytical lunar ephemeris devised by
Chapront-Touze and Chapront (1983), when adjusted for precession of
5029". 10E + l ' U l E 2 (Laskar, 1986) incorporates a coefficient of E of
1732 564 372".84 and a quadratic term of -4".79£ 2 . The latter corre-
sponds to a value for n of —23.9 arcsec/cy2. As noted earlier (section
2.2.3), this has in retrospect proved to be an unfortunate choice.
Over the period covered by historical observations, the main discrepancy
between the adopted expression for the lunar mean longitude and that
due to Chapront-Touze and Chapront (1983) lies in the difference between
the quadratic terms - i.e. l".03£ 2 . This amounts to as much as some 750"
42 2 Tidal friction and the ephemerides of the Sun and Moon

around 700 BC. By comparison the linear terms differ by only about 3"£,
or less than 80" at the above epoch.
A long-term comparison between the modified version of j = 2 used
in this study - based on equation (2.18) - and DE102/LE51 (which
incorporates h = —26.21 arcsec/cy2) yields much smaller deviations. This
conclusion is expected since the difference between the quadratic terms
in X is only about OMIT 2 . Discrepancies in longitude between the two
ephemerides reach about 100" at 700 BC, while the maximum discrepancy
in latitude is about 10".
Most of the deviations in the lunar position between the adopted
ephemeris and DE102/LE51 can be accounted for by the slight discord
between the selected value for n (—26.0 arcsec/cy2) and that implicit in
DE102/LE51. When this parabolic difference is allowed for, the discrep-
ancies in longitude at 700 BC reduce to no more than 25" while latitude
errors do not exceed 2". Both of these deviations - which show a gradual
decrease as E diminishes - are negligible for the present purpose.

2.7 Conclusion
Having considered the accuracy of the adopted solar and lunar
ephemerides, we are now in a position to discuss the various techniques
available for the analysis of solar and lunar eclipse observations. This
will form the subject of chapter 3, after a preliminary discussion of the
viability of alternative data.
 Pre-telescopic eclipse
observations and their analysis

3.1 Introduction
Despite their relatively low precision, pre-telescopic observations cover
a sufficient time-span for long-term trends in the length of day (LOD)
to become apparent. These trends cannot be discerned from modern
measurements. Here we have the main reason why archaic observations
are so important in the study of the Earth's past rotation.
The analysis of ancient and medieval eclipse records is just one of
several techniques which have been utilised in recent years to investigate
variations in the LOD over the historical past. Before discussing in detail
the application of eclipse observations to this problem, it is necessary to
briefly consider other available methods and to explain why eclipses are
to be preferred to other types of data.

3.2 Observational requirements for determining A T in the

pre-telescopic period
Any early astronomical observation which is of value in studying changes
in the LOD in the past must satisfy a wide variety of criteria. These may
be listed as follows:

(i) The observation must involve at least one of the brighter and more
rapidly moving objects in the solar system (i.e. the Moon, Sun or
one of the inner planets Mercury, Venus and Mars).

(ii) The exact Julian or Gregorian date of the observation must either
be specified directly or be able to be determined unambiguously.

(iii) A reasonably accurate value of the UT must be reducible from the

reported circumstances, or - as in the case of total solar eclipses -
the rotational phase of the Earth must be able to be derived.

43
44 3 Pre-telescopic eclipse observations and their analysis

(iv) The corresponding TT must be able to be precisely calculated from

the dynamical equations of motion of the celestial body or bodies
concerned.
(v) In most cases the geographical position of the observer must be
known with some precision - preferably to within a small fraction
of a degree in both latitude and longitude.
Since careful measurements of the co-ordinates of the Sun, Moon and
planets are fairly rare throughout the pre-telescopic period, studies of
the Earth's past rotation have mainly concentrated on the observation of
specific events, such as conjunctions between two celestial bodies. As will
be discussed in detail below, numerous early records of both solar and
lunar eclipses satisfy all of the above conditions and form an impressive
body of data. Other types of observation are of much more limited utility.

3.3 Alternatives to eclipses in the pre-telescopic period

Potentially viable observations other than eclipses include: (i) equinox
and related measurements; (ii) close planetary conjunctions; and (iii)
occultations of stars and planets by the Moon. These will be considered
in turn below.

3.3.1 Equinox and related measurements

Early astronomers not only recorded the dates of the vernal and autumnal
equinoxes but also occasionally made careful estimates of the local times
when the Sun crossed the celestial equator - for example by measuring the
meridian altitude of the Sun on a pair of dates adjacent to either equinox.
At the equinoxes, the declination of the Sun is changing by about 1 arcmin
in an hour. This is close to the limit of resolution of the unaided eye
so that in principle individual measurements are capable of fixing AT
to the order of an hour. However, on account of instrumental defects
and other factors, most early measurements do not attain this ideal.
Preserved equinox and related observations (e.g. determinations of the
meridian altitude of the Sun at other times of year) from the pre-telescopic
period fall into two main groups: (i) ancient Greek measurements and
(ii) medieval Arab records.
A series of 20 Greek equinox observations from the time of Hipparchus
is reported by Ptolemy in his Almagest (III, 2). These range in date from
162 to 128 BC and most observations were probably made by Hipparchus
himself. In each case the time when the Sun crossed the celestial equator
is estimated to the nearest 6 hours. The most recent investigations of these
measurements in the study of the Earth's past rotation are by Newton
 3.3 Alternatives to eclipses 45

(1970, pp. 9-16) and Muller (1976). Newton obtained a best estimate for
AT of +4.2 + 1.0 h (i.e. +15 100 + 3600 sec) at epoch -143 (i.e. 144 BC),
while Muller deduced +4.2 + 0.8 h (+15 100 + 2900 sec) at essentially the
same epoch. As will be evident from subsequent chapters of this book,
individual ancient eclipse observations frequently lead to uncertainties in
AT smaller than those obtained from the whole series of equinox data.
In his Kitab Tahdid Nikayat al-Amakin Li-Tashih Masafat al-Makasin
(The Determination of the Co-ordinates of Positions for the Correction
of Distances between Cities), the great Muslim astronomer al-Biruni (AD
973-1048) recorded 12 equinox times determined between AD 829 and
1019. This treatise has been translated into English by Ali (1967). Most
of these equinox times are quoted to the nearest hour. Al-Biruni also
listed more than 40 measurements of the meridian altitude of the Sun
between AD 832 and 1018, as well as a few miscellaneous observations.
The equinox data will be referred to below as set (a), while the other
observations will be denoted set (b).
The most comprehensive study of the medieval measurements is by
Newton (1972a), who expressed his results in terms of co, the rate of spin
of the Earth. From set (a), Newton derived a value for 109 (oo/co) of
—18.3 + 7.0 cy" 1 . The corresponding value for AT at the mean epoch of
AD 890 may be deduced as approximately +2500 + 1000 sec. Newton
did not quote a separate result for (d>/co) from set (b) but he stated that
his combined solution from (a) and (b) was 109 (oo/co) = —26.5 + 5.8
cy" 1 . This implies a result for set (b) of —44.6 + 10.4 cy" 1 . At the mean
epoch of AD 940, the equivalent value of AT would be about +5400
+ 1300 sec. It is thus clear that there is considerable discord between
the individual solutions, the second being more than twice as large as
the first. In particular, the standard errors would appear to have been
considerably underestimated. It may be concluded that by the medieval
period, equinox observations and meridian altitude measurements give
only a general indication of the value of AT.

3.3.2 Close planetary conjunctions

Many close conjunctions between two planets or between one planet and
a bright star are recorded in Babylonian, East Asian and Arab history.
In some cases, it is specifically stated that a planet appeared to conceal
another planet or star. Most recorded occurrences of this kind are untimed,
but a number of such events observed by Arab astronomers between AD
858 and 1003 were timed to the nearest hour or so. These observations
are reported by Ibn Yunus (d. AD 1009) in his al-Zij al-Kabir al-
Hakimi (Astronomical Handbook dedicated to Caliph al-Hakim), sections
of which have been translated into French by Caussin (1804). Most
46 3 Pre-telescopic eclipse observations and their analysis

Immersion of star
at dark limb

Emersion of star
at bright limb

Fig. 3.1 Schematic representation of a lunar occultation.

conjunctions reported by Ibn Yunus involve the planet Venus. Although

Venus moves fairly rapidly (mean motion 1.6 times as fast as that of the
Sun), it is such a brilliant object that its light tends to overpower other
planets or stars which are located in its immediate vicinity.
From an analysis of the medieval Arab data, Newton (1970, pp. 164-
174 and 200-210) derived 109 (co/co) = - 2 0 + 24 cy" 1 . At the mean
epoch of AD 970 - roughly the same average date as for the Arab
equinox observations - the corresponding value of AT is +2300 + 2700
sec. Newton expressed some disappointment at the large uncertainty in his
result. Like the medieval equinox and related data, it would appear that
close planetary conjunctions have little to offer in the determination of AT.

3.3.3 Occultations of stars and planets by the Moon

Figure 3.1 shows a schematic representation of a lunar occultation. First
contact (immersion) occurs at the eastern limb of the Moon and last
contact (typically about an hour later) at the western limb.
Although telescopic observations of occultations provide the best dataset
for measurement of AT from soon after the introduction of the telescope
to the advent of the TAI scale in 1955 (Stephenson and Morrison, 1984),
most pre-telescopic reports of these events are of little use. Owing to the
brilliance of the Moon - especially at large elongations from the Sun -
it is difficult for the unaided eye to decide whether or not an occulta-
tion is actually taking place unless the occulted object is extremely bright
 3.4 Historical eclipses 47

(e.g. Venus or perhaps Jupiter). Further, the terminology in early texts

is often vague.
Among ancient and medieval civilisations, only the Babylonians and
the inhabitants of East Asia (Chinese, Koreans and Japanese) recorded
occultations on a regular basis. Several tens of observations of this kind
are described in Babylonian history, but East Asian history is replete with
such reports. Nevertheless, very few times are noted; in most cases it
is simply stated that an occultation of a certain star or planet occurred
on a specified day. The occulted object is often described as 'entering'
the Moon. Ancient Greek astronomers occasionally timed occultations of
stars by the Moon but only two events of this kind involving a bright star
are preserved (in the Almagest, VII, 3).
Hilton et al. (1992) analysed a series of 58 untimed Chinese observations
of occultations of planets by the Moon between 68 BC and AD 575.
Expressing AT in the form qt2, where t is in Julian centuries measured
from an epoch around AD 1800, they derived values for q by minimising
the sum of the squares of the least separations between Moon and planet
as a function of q. By applying equal weights to all observations, Hilton et
al. obtained a result for q of +12.6 + 10.2 arcsec/cy2. At the mean epoch
- which was close to AD 450 - the approximate value for AT would
thus be +2400 + 1900 sec. By adopting different weighting schemes, they
derived results for q of around +35 arcsec/cy2 (AT ^ +6600 sec) but the
uncertainties were very high - of the order of 20 or 30 arcsec/cy2. Hilton
et al. concluded that:

Overall the check on the change in the rotation rate (of the Earth) is
very weak, but it represents the limit of what can be done with known,
untimed occultation records.

3.3.4 Further remarks

Ancient and medieval observations of equinoxes, close planetary conjunc-
tions and occultations yield disappointing results for AT. It is indeed
fortunate that a promising array of both timed and untimed observations
of eclipses is available. Reliable records of these events commence around
700 BC and continue throughout most of the pre-telescopic period.

3.4 Historical eclipses

Both solar and lunar eclipses can often be very impressive events. As
a result they have attracted considerable attention throughout much of
recorded history. Indeed, several thousand individual records of eclipses
48 3 Pre-telescopic eclipse observations and their analysis

are preserved from the pre-telescopic period in a variety of sources. How-

ever, only a small proportion of these early data proves to be of value in
the study of the Earth's past rotation. Most useful observations, notably
timings of individual phases, are the work of astronomers. Neverthe-
less, many useful accounts of eclipses - especially total obscurations of
the Sun - were written by chroniclers and others who had no special
interest in astronomy.
One of the principal reasons for early astronomers observing eclipses
was to test the accuracy of existing almanacs and improve future pre-
dictions. In addition, medieval Arab stargazers measured the times of
lunar eclipses in order to determine longitude on the Earth's surface.
Especially in Babylon and China, the ultimate goal in watching for
eclipses and other celestial phenomena was astrological. There can be
little doubt that without the influence of astrology many of the astro-
nomical observations which exist from the ancient and medieval world
would have been neither made nor preserved. Here we have perhaps
the only positive contribution which this pseudo-science has made to
world culture!
In reporting eclipses, the main concern of chroniclers seems to have
been to describe the phenomenon as a spectacle. Many of these events
evidently produced a profound impression on eyewitnesses, whether or
not they had any special interest in astronomy. Eclipses of both Sun and
Moon are frequently noted in historical writings, but in such works only
solar obscurations are of much value in the study of changes in the LOD.
Before discussing the different types of eclipse observation which are
available from various parts of the ancient and medieval world, some
remarks on the cause of eclipses are appropriate.

3.5 Cause of eclipses

From the point of view of Earth rotation studies, solar and lunar
eclipses have a number of mutually distinct features. In the case of
a solar obscuration, a shadow is cast by the Moon on the rotating
Earth. As a result, the visibility of the various phases at any par-
ticular site depends very much on the value of the accumulated clock
error AT at that date. A lunar eclipse occurs when the Moon enters
the terrestrial shadow so that only the local time of each individual
stage is a function of AT; parameters such as magnitude and dura-
tion are not usually affected by Earth's rotation. Nevertheless, early
solar and lunar observations are sufficiently diverse as to complement
one another in the investigation of long-term changes in the length of
the day.
 3.5 Cause of eclipses 49

Partial
(penumbra)

Annular

Partial
(penumbra)

Fig. 3.2 Basic mechanism of solar eclipse formation.

5.5.7 Solar eclipses

Solar eclipses are produced by the direct interposition of the Moon be-
tween the Sun and Earth. As seen from our planet, the Moon is in
conjunction with the Sun (astronomical new Moon) every synodic month
- i.e. at mean intervals of 29.5306 days. Hence if the lunar and terrestrial
orbits were in the same plane, there would be 12 or 13 solar eclipses each
year. In fact, the two orbits are inclined at an angle of about 5.1 deg. Thus
from the Earth's viewpoint, at most conjunctions the Moon passes either
above or below the Sun and no eclipse is produced. A solar eclipse can
only occur when at the time of conjunction the Moon is close to one of
its nodes - the two points 180 deg apart where its orbital plane intersects
that of the Earth. During an average century, there are approximately
238 solar eclipses visible on (at least some part of) the Earth's surface.
This figure represents only about 19 per cent of the new Moons occurring
during this same time. Nevertheless, on the Earth as a whole, eclipses of
the Sun are still a fairly frequent phenomenon. In most years two or three
of these events take place, but rarely this number can be as large as four
or even five (as last took place in 1935).
The basic mechanism of solar eclipse formation is shown in figure 3.2.
Within the penumbral region, the Sun is partially obscured by the Moon,
but in the full umbral shadow the entire solar disk is hidden. Although
the Moon is much smaller than the Earth (ratio of diameters 0.272), the
penumbral shadow has a diameter rather more than half (approximately
0.54) of the Earth at mean distance. Hence a partial solar eclipse is usually
visible over a wide area. Within the penumbra, variations in the distance
of the Moon and Sun from the Earth owing to the ellipticity of the orbits
50 3 Pre-telescopic eclipse observations and their analysis

have little effect on the appearance of an eclipse to the unaided eye.

However, the diameter of the umbra is so small (seldom exceeding 0.02 of
the terrestrial diameter) that inside this region even minor variations in
distance can have a critical effect on visibility.
The rather obvious fact that the apparent sizes of the Sun and Moon
are very similar (mean semi-diameters 960" and 931" respectively) is an
indication that the apex of the umbral cone can just about reach the
Earth on average. The lunar orbit is sufficiently elliptical (eccentricity
0.055, compared with 0.017 for the Earth around the Sun) to produce a
range of both total and annular eclipses. Under optimum conditions -
when the Moon is nearest the Earth (i.e. at perigee), the Sun is furthest
away, and both bodies are in the observer's zenith - the lunar disk may
appear fully 1.08 times as large as that of the Sun (semi-diameter 1023",
compared with 946"). However, at the other extreme - when the Moon
is furthest from the Earth (at apogee), the Sun is closest to us, and
both bodies are near the observer's horizon - the Moon may cover no
more than about 0.90 of the Sun (semi-diameter 881", compared with
978"). These two extremes are illustrated in figures 3.3a and 3.3b. Of the
approximately 238 eclipses visible on the Earth during an average century,
some 84 are partial, 77 are annular, 66 are total and 11 are combinations
of annular and total (von Oppolzer, 1887). These frequencies are only
slightly variable from one century to the next.
When the umbra extends fully to the Earth, the entire solar photosphere
is screened from a small section of the surface of our planet for a short
time. On account of the orbital motion of the Moon and rotation of
the Earth, the umbral shadow sweeps progressively across the terrestrial
surface in a narrow band; this is a section of a small circle - see figure
3.4, which shows the path of totality at the eclipse of AD 968 Dec 22
as calculated by Schroeter (1923). For observers in this zone, the sky
suddenly grows very dark and often several stars become visible. The
corona (the outer atmosphere of the Sun) also appears, but for some
reason this is very rarely mentioned in early records. Under a vertical Sun
the umbra can never exceed about 270 km in diameter. However, at low
solar altitudes the elongated shadow of the Moon may be much wider
than this, occasionally exceeding 500 km.
The change in light intensity during the various stages of a total solar
eclipse is shown schematically in figure 3.5 (lower curve). During totality,
most of the available light comes from scattered sunlight near the horizon
where the eclipse is still only partial. When the atmosphere is relatively
dust-free, the illumination may decrease by more than six orders of
magnitude between the start of the eclipse and the onset of totality; most
of the diminution in light level occurs in the last few seconds before the
Sun is completely obscured.
 3.5 Cause of eclipses 51

\ \ Maximum ratio
of lunansolar
semi-diameter
= 1.08

(b)
So\ar limb

Minimum ratio
of lunansolar
semi-diameter
= 0.90

Fig. 3.3 Extreme total (a) and annular (b) obscurations of the Sun.

If the apex of the umbral cone stops short of the Earth, an annular
or ring eclipse is formed instead. In most events of this kind, the loss of
daylight is fairly small. This is evident from figure 3.5 (upper curve), where
the light variation is calculated for a typical annular eclipse in which 0.95
of the solar diameter is obscured. In this diagram, approximate allowance
is made for limb darkening around maximal phase.
When the apex of the umbral cone just touches the Earth, an eclipse
may be marginally annular near the sunrise and sunset locations and
marginally total near the noon position; at the former places an observer
is significantly further from the Moon (by up to one Earth radius).
The magnitude (JH) of a partial solar eclipse is defined as the fraction
of the solar diameter obscured by the Moon at the moment of greatest
phase. In the case of a central eclipse, the magnitude (/ic) is given by the
52 3 Pre-telescopic eclipse observations and their analysis

30° W 15° W 0° 15° E 30° E 75° E

60° N
60° N

55° N
55° N

50° N
50° N

45° N
45° N

40° N
40° N

35° N
35° N

30° N 30° N

45° E

Fig. 3.4 Path of totality at the eclipse of AD 968 Dec 22 as calculated by

Schroeter (1923).

following expression:
= (P - U)/(P + 17), (3.1)
where P is the penumbral radius and U the umbral radius on a plane
passing through the place of observation and perpendicular to the shadow
axis (effectively the continuation of the line joining the centres of the Sun
and Moon). Both P and U are customarily expressed in terms of the
Earth's equatorial radius. The adopted convention is that U < 0 for total
eclipses and U > 0 for annular obscurations; P is always positive. Since
(P — U) is constant, having a numerical value of 0.5464, equation (3.1)
reduces to
= 0.5464/(P + U). (3.2)

Although the magnitude of a central annular eclipse is always less than

unity, in the case of a total eclipse it exceeds unity. It should be noted that
(ignoring minor local variations) the magnitude of a central eclipse is the
same whether the place of observation lies near the middle of the shadow
or close to its edge. As nc is also equal to the ratio of the apparent lunar
 3.5 Cause of eclipses 53

Start

10 2
-
"(75
S io- 3
1

10 -5

10" 6

0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2
Hours
Fig. 3.5 Changes in light intensity during the various stages of typical annular
and total solar eclipses.

and solar diameters, the extreme range in magnitude for a central eclipse
extends from 0.90 to 1.08 (see above).
The Moon orbits the Earth at about 3400 km per hour. However, on
account of the terrestrial rotation, the speed of the lunar shadow relative
to a point on the ground is quite variable and is frequently much less
than the above figure. As a result, the times of onset and the durations
of both partial and central (i.e. total and annular) eclipses are much
affected by geographical circumstances. When an eclipse occurs with the
Moon overhead at the equator, the effect of the Earth's speed of rotation
(some 1700 km per hour) is to roughly halve the ground velocity of the
shadow. Near rising or setting, the corresponding contribution from the
Earth's rotation is minimal so that the shadow travels much faster along
the ground. Under ideal conditions, the interval from first to last contact
(i.e. from the very beginning of the visible eclipse to its end) can exceed 4
hours. However, a duration as long as this is fairly rare.
For totality, the maximum duration (from second to third contact) is
rather less than 8 min while for annularity the duration may exceed 12
min. The relatively long durations of annular eclipses arise largely from
the low orbital velocity of the Moon when furthest from the Earth. Two
solar eclipses of remarkable duration occurred in 1973. Totality on Jun 30
54 3 Pre-telescopic eclipse observations and their analysis

lasted for 7.1 min, while annularity on Dec 24 extended for 12.0 min. This
combination of circumstances will remain unparalleled for more than a
century. Although ancient and medieval astronomers often measured the
durations of eclipses from first to last contact, no careful determinations
of the brief span of totality or the ring phase are preserved until as late as
the eighteenth century. Primitive clocks (clepsydras, etc.) were probably
too imprecise for this purpose.
During an average eclipse of the Sun, only about one-sixth of the
terrestrial surface encounters some part of the lunar shadow. As a result,
at any one location only about 38 solar eclipses (roughly one-sixth of
238 - the total for the Earth as a whole) are visible in a typical century.
On some occasions, only a small fraction of the Sun will be obscured by
the Moon and to the unsuspecting bystander the phenomenon may well
pass unnoticed. At other times, the loss of daylight may be pronounced.
Central eclipses are extremely rare at a given site. According to the
statistical estimate of Meeus (1982), the mean interval between total
eclipses at any given point on the Earth's surface is about 375 years, while
for annular obscurations it is approximately 224 years. However, the actual
interval between successive central eclipses occurring at a particular place
is very variable. Often, many centuries can elapse, while very occasionally
two such events can take place in not much more than a year. For
instance, both the eclipses of 1983 Jun 11 and 1984 Nov 22 were total
at Port Moresby (Papua New Guinea) - a circumstance which aroused
considerable local interest. One of the most remarkable examples in
ancient and medieval history occurred at Split (Croatia) in the early
thirteenth century. Both the eclipses of AD 1239 Jun 3 and 1241 Oct 6
were described by a chronicler of that city as causing intense darkness, and
producing great terror among the populace (see chapter 11 for details).
So many careful observations of partial solar eclipses are preserved
from antiquity that it seems likely that early astronomers must have often
used some means to dim the brilliant disk of the Sun. It is known that
medieval Arab astronomers were in the habit of viewing the eclipsed Sun
by reflection in water (see chapter 13); it could then be scrutinised more
easily. However, under these circumstances even a slight breeze can distort
the solar image. The first century AD Roman author Seneca tells us that
in his time more viscous liquids were utilised for this purpose:
Whenever we want to watch an eclipse of the Sun we set out basins
filled with oil or pitch, because the heavy liquid is not easily disturbed and
so preserves the images it receives...
[Naturales Quaestiones, I, 11; trans. Corcoran (1971, pp. 69-71).]
Possibly similar techniques were adopted by early astronomers in other
parts of the world, but in general little is known about the methods used.
 3.5 Cause of eclipses 55

M1 = Moon partially in umbral shadow: partial eclipse

M2 = Moon fully in umbral shadow: total eclipse
M3 = Moon partially in penumbral shadow: penumbral eclipse
Fig. 3.6 Schematic representation of a lunar eclipse.

3.5.2 Lunar eclipses

A lunar eclipse takes place whenever the Moon encounters the terrestrial
shadow, as depicted schematically in figure 3.6. At the Moon's distance,
the diameter of the penumbral cone is about 4.8 times the lunar diameter,
while the corresponding ratio for the umbral cone is approximately 2.7:1.
Figure 3.7, which is from the book Astronomicum Caesareum, pub-
lished in 1540 by Petrus Apianus, shows the Moon, umbral shadow and
penumbral shadow roughly to scale.
During a penumbral eclipse, there is a slight reduction in the amount of
sunlight reaching the Moon, but this is scarcely noticeable to the unaided
eye. As a result, records of these events from the pre-telescopic period
are virtually unknown. An almost unique exception is the marginally
penumbral eclipse of BC 188 Aug 1/2, which was observed in Babylon
(see chapter 6). By contrast, umbral eclipses are readily visible and in
general only these will be considered below.
For the Moon to enter the umbra either partially or totally, it must be
close to one of the the two nodes of its orbit when at opposition (i.e. full
Moon). In a typical century there are 154 lunar eclipses of which 70 are
total and 84 are partial (Liu and Fiala, 1992). The combined figure repre-
sents only about 12 per cent of the number of full Moons during the same
interval. Hence on the Earth as a whole, eclipses of the Moon occur signif-
icantly less frequently than their solar counterparts (only about two-thirds
as often). Occasionally, there can be as many as three lunar obscurations
in a single year. Very rarely, as last happened in 1982 and will not recur
for more than 500 years!, three total eclipses can occur in one year.
When the Moon passes fairly centrally through the terrestrial umbra,
the duration from start to finish (first contact to last contact), can reach
about 3.7 hours, while totality (second contact to third contact) can last
up to about 1.7 hours. As already noted, the rotation of the Earth has
56 3 Pre-telescopic eclipse observations and their analysis

Fig. 3.7 The Moon, umbral shadow and penumbral shadow as depicted in the
book Astronomicum Caesareum by Petrus Apianus (Ingolstadt, 1540).

no effect on these durations under normal circumstances. However, if the

Moon happens to be near the observer's horizon, part of an eclipse may
occur before local moonrise or after moonset, thus apparently reducing
the durations.
For a partial eclipse, the magnitude is simply the fraction of the lunar
diameter in shadow at maximal phase. In general, the magnitude may be
derived from the relation:

= (s + S - d)/2s. (3.3)
 3.5 Cause of eclipses 57

Here s is the lunar semi-diameter, S the semi-diameter of the Earth's

shadow at the Moon's distance and d the minimal angular distance of
the Moon from the centre of the shadow. Mean values for s and S are
respectively 933" and 2472". In the case of a total eclipse, \i can range
from unity up to a maximum of 1.89, the latter value applying only when
the Moon passes centrally through the umbra. However, only rarely does
the magnitude exceed 1.80.
As the Moon is so much dimmer than the Sun, even small partial
eclipses are relatively easy to observe with the unaided eye. For instance,
a survey of lunar eclipses recorded by medieval Arab chroniclers has
shown that although most observations relate to total eclipses, several
obscurations of magnitude less than about 0.5 (including two as small as
0.28 and 0.12) were noted (Said and Stephenson, 1991).
Since the lunar parallax is fairly small (approximately 1.0 deg), no mat-
ter where an observer is situated on the Earth's surface the proportion
of the lunar disk covered at any given instant is virtually the same pro-
vided that the Moon is above the horizon at the time. Similar remarks
apply to the UT of the various phases; the local time of each contact
depends directly on the geographical longitude of the place of observa-
tion. As a result, lunar eclipses have a lengthy history of application
to longitude determination - extending at least as far back as the ninth
century AD (see chapter 13). Solar eclipses are of little utility for this
pursuit owing to the irregular movement of the Moon's shadow across
the Earth's surface.
Although lunar obscurations are less frequent than solar eclipses on the
Earth as a whole, the reverse is true at any specific place. This is because
at any moment (weather permitting), the phenomenon is visible over the
entire hemisphere of the Earth at which the Moon is above the horizon. In
an average century as many as 95 lunar eclipses are visible at a given site,
including those in which the Moon rises or sets whilst the phenomenon
is is progress. This number, which represents more than 60 per cent of
the quota for the entire Earth, is roughly l\ times the number of solar
obscurations in the same time.
The Earth's atmosphere materially influences the appearance of a lunar
eclipse. It effectively increases the diameter of the terrestrial umbra by
about two per cent and also imparts a diffuse edge to the shadow. On
account of refraction, the atmosphere allows some sunlight to reach the
lunar disk even during totality. When there is minimal volcanic dust,
etc. in the upper air, the totally eclipsed Moon takes on a deep red hue
due to the preferential transmission of long-wavelength light. However,
following a major volcanic eruption, there may be so much dust that the
Moon may be virtually invisible during totality. Both red and very dark
eclipses are recorded from time to time in history.
58 3 Pre-telescopic eclipse observations and their analysis

3.6 Eclipse observations which are of value in the investigation of A T

Observations of solar eclipses which are potentially viable for the study
of Earth's past rotation include the following: (i) untimed total and
annular eclipses; (ii) instances (also untimed) where totality or annularity
is specifically denied; (iii) measurements of the local time of individual
phases; (iv) estimates of eclipse magnitude (at greatest phase); and (v)
cases where it is noted that the Sun rose or set whilst eclipsed. As will be
discussed in section 3.8.6, reports that merely note the occurrence of an
eclipse without further details are of minimal value.
To the above list may be added the following types of lunar eclipse
observation: (vi) measurements of the local time of the various phases;
(vii) examples where it is merely noted that the Moon rose or set whilst
eclipsed; and (viii) specific estimates of the degree of obscuration of the
Moon at moonrise or moonset. Since the peak magnitude of a lunar eclipse
and the durations of individual phases are independent of AT, estimates
of these quantities - or statements that the Moon was either totally or
partially obscured - do not make any direct contribution to the study of
the Earth's past rotation. However, as will be shown in sections 3.8.4,
3.9.1 and 3.9.3, analysis of recorded magnitudes and durations enables the
material in some of the above categories to be better assessed.
Before considering the derivation of AT from observations in each
of the above categories, the various historical sources will be briefly
outlined. Fuller details, together with a discussion of the relevant calendar
conversion, will be given in chapters 4 to 13.

3.7 Sources of data

Practically all of the ancient and medieval eclipse observations which are
of value in the investigation of long-term changes in the length of the day
originate from only four civilisations. These are: Babylon (c. 700 BC to
50 BC); China (c. 700 BC to AD 1500); Europe (c. 500 BC to AD 1600);
and the Arab dominions (c. AD 800 to 1300). Somewhat surprisingly,
there do not appear to be any useful records from ancient Egypt (apart
from the Greek city of Alexandria), India or Central America, for exam-
ple. According to the first century BC Greek historian Diodorus of Sicily
(Library of History, book I, chapter 81), the Egyptians 'keenly observed
the movements, orbits and stationary points of the planets'. However,
this somewhat ambitious assertion is not supported by modern research
into ancient Egyptian astronomy (Parker, 1974). Indian astronomers were
much concerned with the prediction of eclipses and other aspects of
mathematical astronomy (cf. Yabuuchi, 1979), but no early observations
appear to be extant. Although the Mayans developed a highly sophisti-
 3.7 Sources of data 59

cated astronomy and were very interested in eclipse prediction (Thompson,

1974), no actual observations are known to survive. Many Mayan codices
(possibly containing valuable astronomical records) were destroyed by the
conquistadors in AD 1540.
Babylonian astronomers systematically observed eclipses along with
many other celestial phenomena. Their records, inscribed on clay tablets,
first came to light rather more than a century ago and are now largely
in the British Museum. Translations of numerous astronomical texts have
been published by Sachs and Hunger (1988, 1989, 1996). Only about ten
per cent of the original material is extant, and nearly all surviving tablets
are badly damaged. Nevertheless, nearly 150 important observations of
eclipses - mainly of the Moon - are available; these are discussed in
chapters 4 to 7 below. In arriving at this total, I have treated each
individual measurement etc. as a separate observation, even if the same
eclipse is counted several times. When dates are fully preserved, they
are invariably accurately expressed in terms of the Babylonian luni-solar
calendar. Precise conversion to the Julian calendar is straightforward (for
details, see chapter 4). The place of observation is also well established as
Babylon. There is nothing in the available records to indicate just when the
cause of eclipses was known to the Babylonians. They probably obtained
this knowledge from the Greeks during the Hellenistic period (late fourth
century BC onwards). The first true explanation of eclipses appears to
have been given by Anaxagoras of Athens c. 450 BC - see also chapter 10.
It was the practice of the astronomers of Babylon to estimate the mag-
nitude of an eclipse (usually to the nearest twelfth of the lunar or solar
diameter) and also to carefully time the various phases. Times were usu-
ally expressed to the nearest US (time-degree) equal to four minutes; these
were probably determined with the aid of a clepsydra (water clock), but the
records are silent on this question. Apart from the texts devoted to astron-
omy, Babylonian historical records make virtually no references to eclipses.
Chinese eclipse records (see chapters 8 and 9) are also dominated by
the observations of astronomers, who maintained a regular watch of the
sky for all kinds of celestial events. More than a thousand individual
sightings of eclipses are reported in Chinese history, but most accounts
simply note the occurrence of such an event without any other details.
Only a relatively small proportion of these observations is of value in
the investigation of AT, but this still represents a significant number of
data (more than 100 in total). Practically all dates are precisely recorded
(relative to the Chinese luni-solar calendar - see chapter 8) and the place
of observation can usually be taken as the dynastic capital of the time.
The correct explanation of eclipses was not given in China until as late as
the first century AD - long after its realisation by the Greeks.
Chinese eclipse reports from before AD 400 almost exclusively relate to
60 3 Pre-telescopic eclipse observations and their analysis

the Sun and there are several allusions to total or very large obscurations
from this period. Between about AD 400 and 1300, many timings of both
solar and lunar eclipses - determined with the aid of a water clock - are
preserved. Most solar and lunar measurements were quoted to the nearest
k'o ('mark'); one hundred of these units equalled a full day and night,
so that each was equivalent to 0.24 h. However, some lunar times were
expressed to the nearest fifth of a keng ('night watch') - a unit whose
length varied with the seasons but was roughly equal to half an hour.
After AD 1300, few careful observations are extant until the Jesuit era
(seventeenth century). Chinese records of eclipses are largely found in the
official dynastic histories (cheng-shih), all of which have been printed (and
reprinted in recent years) and are thus readily accessible. An extensive
and valuable compilation of eclipse and other astronomical records from
China has been co-ordinated by Beijing Observatory (1988).
From about the middle of the first millennium AD, Korean and Japanese
astronomers followed much the same observing pattern as their Chinese
counterparts. However, few eclipse records from these countries appear to
be sufficiently detailed and reliable to contribute effectively to the study
of the Earth's past rotation.
Nearly all early European accounts of eclipses are the work of historians,
chroniclers, etc. - rather than astronomers. With the exception of some ten
observations made by Greek astronomers between about 200 BC and AD
130 (all contained in Ptolemy's Almagest), most ancient records of eclipses
are of dubious reliability. These are largely found in the Greek and Latin
Classics and are the subject of chapter 10 below. Ginzel (1899) made
an extensive compilation of this material; see also Newton (1970). After
about AD 800, some 50 qualitative but careful descriptions of total or very
large solar obscurations are preserved in monastic and town chronicles
(see chapter 11). Several independent accounts may often relate to the
same event; for instance some ten separate reports of totality in AD
1239 are preserved from a variety of sites in the Mediterranean region.
Despite the lack of measurements, and often only minimal understanding
of the cause of eclipses, medieval European records are frequently of great
value in the determination of AT. Dates are normally accurately specified
(mostly in terms of the Julian calendar), while usually the observation
can be assumed to have been made at the place where the chronicle was
composed. Large numbers of medieval annals have been published in their
original language (generally Latin) by editors such as Muratori (1723-)
and Pertz (1826-), and are thus readily accessible. Celoria (1877a and
1877b) and Ginzel (1884a, 1884b and 1918) assembled numerous eclipse
records from these sources, while many translations have been published
by Newton (1972b). A few careful timings made by fourteenth and
fifteenth century European astronomers are also extant - see chapter 11.
 S.8 Solar eclipse observations and their reduction to AT 61

From about AD 800 to 1500, Arab chroniclers frequently documented

the occurrence of both solar and lunar eclipses - as well as other striking
celestial phenomena such as comets and meteors. Although times are
only crudely estimated, there are several vivid accounts of total eclipses
of the Sun between AD 900 and 1300. These are discussed in chapter
12. Over a relatively short period - between about AD 830 and 1020 -
Arab astronomers recorded careful timings and magnitude estimates for
both solar and lunar eclipses. More than 40 separate observations of
this kind are preserved (see chapter 13). Individual eclipse phases were
timed indirectly by measuring the altitude of the Sun, Moon or selected
bright stars. In each account the place of observation is clearly stated and
nearly all dates are carefully noted in terms of the Islamic lunar calendar.
The major chronicles and astronomical treatises have been published in
Arabic. Translations of several astronomical works are also available,
notably the treatise of Ibn Yunus (d. AD 1009), which contains many
careful observations of both solar and lunar eclipses (Caussin, 1804). Said
et al. (1989) have published translations of a wide variety of solar eclipse
records compiled from Arab chronicles.
Combining the above four sets of data, we have a remarkable series
of more than 300 useful observations of both solar and lunar eclipses,
extending from about 700 BC to the telescopic era. (As noted above, re-
ports of different phases of the same eclipse are counted as separate data.)
There are a few earlier eclipse records, notably from China c. 1000 BC (see
chapter 8). However, the interpretation of these archaic texts is open to
speculation, while the dates are uncertain. Most of the accessible sources
from approximately 700 BC onwards have been thoroughly searched for
references to eclipses. Although consultation of material as diverse as (i)
unpublished Babylonian tablets, (ii) medieval European printed chroni-
cles, and (iii) Arab manuscripts is currently in progress, the acquisition of
further useful eclipse data seems likely to be fairly slow unless some fresh
archive comes to light.

3.8 Solar eclipse observations and their reduction to AT

In section 3.6 above, five different categories of solar eclipse observation
which are in principle of value in the calculation of AT are listed. The
determination of AT from each type of observation will now be discussed,
with examples. In each case it will be assumed that the only unknown
quantity is AT itself. As noted in chapter 2, the various orbital parameters,
including the lunar acceleration h, are known with more than adequate
precision.
62 3 Pre-telescopic eclipse observations and their analysis

3.8.1 Total and annular solar eclipses

Throughout this investigation it will be assumed that an eclipse was only
fully total or annular if a record clearly describes either the complete
disappearance of the Sun or the reduction of the solar disk to a ring of
light. As discussed in chapter 11, other effects - such as darkness or the
visibility of stars - are only general indications of a very large eclipse. If
the sky is clear, the onset of totality should be extremely well defined, as
is evident from figure 3.5. This is confirmed by the impressions of modern
observers (e.g. Muller, 1975; see also the many eighteenth and nineteenth
century reports compiled by Ranyard, 1879). An interesting illustration is
provided by the eclipse of AD 1715 May 3, which was total in England.
Edmond Halley circulated advance notification of this eclipse throughout
much of England, specifically asking those who happened to be near the
edges of the belt of totality to keep a special watch. As a result, Halley
(1715) was able to obtain several careful unaided-eye observations from
near the northern and southern limits of totality. These reports proved to
be so self-consistent that it has proved possible to use them to determine
the mean solar semi-diameter within 0.1 arcsec (Parkinson et a/., 1988).
The following account, relating to the eclipse of AD 1241 Oct 6, is fairly
typical of medieval descriptions of the total phase. It originates from the
monastery of Stade in north Germany (see chapter 11):

1241 ... There was an eclipse of the Sun ... on the day before the Nones
of October (Oct 6), on Sunday some time after midday. Stars appeared
and the Sun was completely hidden from our sight ...

Use of the first person plural affirms that totality was witnessed in Stade.
In the case of an annular eclipse, the ring phase is less marked since
a portion of the Sun stays unobscured at all times. Hence the loss
of daylight is usually relatively small. Although more than 40 careful
accounts of totality are preserved from various parts of the world in the
pre-telescopic era, no more than four direct allusions to annularity can
be traced in this period (AD 873, 1147, 1292 and 1601). Evidently, most
events of this kind passed unnoticed. The ring phase is clearly described
in the following brief report from Ta-tu (the former name for Beijing) on
a date corresponding to AD 1292 Jan 21 (see also chapter 8):

Chih-yuan reign period, 29th year, first month, day chia-wu [31]. The
Sun was eclipsed. A darkness invaded the Sun, which was not totally
covered. It was like a golden ring ...

Even if the Sun is only marginally reduced to a complete circle of

light, the effect of irradiation is to greatly enhance visibility of this ring,
as has been noted by modern observers. Immediately outside the zone
 3.8 Solar eclipse observations and their reduction to AT 63

la)

Ib)

Fig. 3.8 Form of the crescent as seen just outside the central zone for a
generally annular eclipse of typical central magnitude 0.95 (a) and a generally
total obscuration of typical central magnitude 1.05 (b).

of annularity, the angle between the cusps of the crescent is small and
remains less than 90 deg (i.e. making at least three-quarters of a full
circle) even at some considerable distance from the central zone. This is
illustrated in figure 3.8a, which is based on a typical central magnitude
of 0.95. (By contrast, when a generally total obscuration is viewed from
outside the umbral region this angle always exceeds 180 deg - see figure
3.8b.) Because of this feature, early descriptions of eclipses which were
generally annular on the Earth's surface are often vague and difficult to
interpret (e.g. 'a human head was seen in the Sun' - as reported from both
Cologne and Hirschau in Germany in AD 1207 - see chapter 11). Hence
unless a text clearly asserts that a full ring of sunlight was seen (or, on
the contrary, likens the Sun to a crescent), the record is best rejected.
Provided it can be confidently established that a certain solar eclipse
was fully total or annular on a definite date and at a specified place,
measurement of the time of day when it occurred is usually superfluous.
This results from the narrowness of the central zone on the Earth's surface.
The effect of increasing AT is to displace the whole of the calculated zone
of totality or annularity towards more easterly longitudes, the converse
being true if AT is diminished. For any particular observation, made at
a site with latitude <\> deg and longitude A deg, an approximate solution
for AT may be obtained in the following way. (NB the adopted sign
conventions are: positive for latitudes north of the equator and positive
for longitudes west of the Greenwich meridian up to 180 deg.)
64 3 Pre-telescopic eclipse observations and their analysis

On the assumption that the Earth has always rotated at its present
rate (i.e. AT = 0 at any epoch), the longitude A' at which the central
line intersects the parallel of latitude (0) passing through the site is first
derived. The required value for AT is then given by the formula
AT = 240(Af -A) sec. (3.4)
To give an example of this method of investigation, two separate late
Babylonian texts - which are now in the British Museum - describe a
complete eclipse of the Sun on a date corresponding to BC 136 Apr 15.
(Photographs of both tablets are shown in figures 3.9a and 3.9b.) During
totality, four planets and several stars were said to be visible (for full
details see chapter 5). This is by far the most reliable ancient record of
such a phenomenon.
The zone of totality in 136 BC - computed on the assumption that AT
= 0 - is shown in figure 3.10a. This zone would pass far to the west
of Babylon ((/> = +32.55 deg, A = —44.42 deg), the central line crossing
the parallel of latitude for the city at A' = +4.3 deg. Hence in order to
obtain satisfactory agreement with the record, a correction in longitude of
approximately +48.8 deg would be required. Such a correction implies a
result for AT of around +11700 sec (3.25 h). The computed central line
with AT = +11700 sec is depicted in figure 3.10b.
In practice, it is possible to set firm upper and lower bounds to the
value of AT at the appropriate epoch by taking into account the width
of the zone of totality or annularity. Thus, as shown in figure 3.11, for the
southern edge of the umbral shadow to just reach Babylon, a value for AT
of +11210 sec (3.11 h) would be needed. For the northern edge to reach
the site, the corresponding figure would be +12140 sec (3.37 h). Hence to
fully satisfy the observation, only values of AT between these two limits -
i.e. +11210 < AT < +12140 sec - are acceptable. The true value of AT
in 136 BC has an equal probability of lying anywhere in this interval, but
not outside it; this represents a tolerance of no more than about four per
cent relative to the mean figure of +11680 sec. It should be mentioned
here that I have usually estimated AT limits obtained from observations
of total and annular obscurations - and also for careful reports which
specifically deny a central eclipse - to the nearest 10 sec. All other AT
values or limits (e.g. obtained from contact timings or rising and setting
phenomena) will be quoted to the nearest 50 sec - a precision which is
still more than adequate for investigating long-term trends in AT.
Since for any given value of AT the computed central line in 136 BC
crosses the parallel of latitude for Babylon at only one point along the
entire track, no other solution range than the above will suffice. However,
if a central line crosses the observer's latitude twice, as was true for places
between latitudes 40 and 50 deg in AD 968 (see figure 3.4), an observation
 3.8 Solar eclipse observations and their reduction to AT 65

(a)

jf *

• •• •

1
-
It: 1
•
%

mlillHfl
m
MS/PM/O6J No 1
|BHITiSH llMUSF U M I 2| inch" 4
^ I ^ ^ ^ H

Fig. 3.9 Two Late Babylonian tablets describing a complete eclipse of the Sun
on a date corresponding to BC 136 Apr 15. (Courtesy: British Museum.)
66 3 Pre-telescopic eclipse observations and their analysis

(a) A = 4.3° W

= 32.5° N

(b) A = 44.5° E

= 32.5° N

Fig. 3.10 Zone of totality in 136 BC, computed on the assumption that (a) AT
= 0 and (b) AT = +11700 sec. The eclipse was observed to be complete at
Babylon. (Courtesy: Dr S. Bell.)
 3.8 Solar eclipse observations and their reduction to AT 67

A7= 12140 sec

AT=11210sec

Lat. = 32°.55 Lat. = 32°.55

(a) (b)

Fig. 3.11 Lower (a) and upper (b) bounds to the value of AT as derived from
observations of the total solar eclipse of 136 BC at Babylon.

of a central eclipse will lead to two distinct ranges for AT. In principle,
these can only be separated if a measurement (or at least an estimate)
of the local time is recorded. However, in practice, the two derived AT
ranges are often so far apart that only one solution can possibly be viable
- as indicated by comparison with AT results obtained from roughly
contemporaneous observations. An example in which the two sets of AT
limits are unusually close is provided by the total solar eclipse of AD
454 Aug 10. As viewed from the capital of Chien-k'ang (Nanjing: cf) =
+32.03 deg, A = —118.78 deg), this was said to be complete, and 'all the
constellations were brightly lit' (see chapter 8). Unfortunately, the time of
day is not specified.
As can be seen from figure 3.12, which is calculated using AT = 0, the
central line in AD 454 would cross the latitude of Nanjing twice (at A!
= —115 deg and A! = —90 deg), reaching its maximum distance north
of the equator (cj) = +33.0 deg) between these positions. Since these two
longitudes are respectively some 4 and 29 deg to the west of Nanjing, the
observation can thus be satisfied by two discrete ranges of AT: either
around +1000 sec (for an early morning observation) or +7000 sec (for a
mid-morning observation). Allowing for the width of the track of totality,
the precise limits may be deduced; these are (i) +70 < AT < +1800 sec
and (ii) +6130 < AT < +7900 sec. As will be discussed below (chapter
8), only the latter choice is supported by observations of similar date,
so that set (i) can effectively be discounted. In most cases where two
 3 Pre-telescopic eclipse observations and their analysis

45

-90' -105° -120° -135°

Longitude

Fig. 3.12 Part of the central zone for the total solar eclipse of AD 454 Aug 10
computed using AT = 0. Totality was observed at the Chinese capital of Nanjing.

AT ranges are indicated they are much further apart than this and an
unambiguous solution can thus be readily obtained. A dual result will
only be considered if both alternatives seem viable.
In 136 BC the zone of central eclipse was inclined at a rather large
angle to the equator in the vicinity of Babylon. Hence although the
track of totality was fairly wide (some 270 km) the resultant range of
AT is reasonably narrow. However, if the path of the umbra ran almost
parallel to the equator near the place of observation - as in AD 454 -
the corresponding range of AT would be considerably increased. In AD
454 the breadth of the umbral zone was only 110 km but the derived AT
range (i.e. +6130 < AT < +7900 sec) is roughly twice as wide as that
obtained in 136 BC In extreme cases the AT limits may be so wide apart
as to render the observation redundant, even when a careful description of
totality or the ring phase is preserved. This is well illustrated by the eclipse
of AD 1185 May 1 (see figure 3.13, which is taken from Schroeter, 1923).
On this date, the central zone (260 km in width) ran almost parallel to the
equator across northern Russia. Detailed accounts of totality, mentioning
visibility of the chromosphere, are preserved from Novgorod (4> = +58.50
deg, A = —31.33 deg) - chapter 11. Although only a single range of
AT can satisfy the observations made at this site, the limits are extremely
 3.8 Solar eclipse observations and their reduction to AT 69

30°W 15°W 0 1b°b 30°E 45°E \ -^ AJ!*—^ 75°E

K
/
/ 60°N
60°N

/ \

'/Y
/ /
WAVVV
if 3\\Vvv>
55°N ' /
V / 55°N

iK^:::::*:v-^il *
mm \ \ \ \ ^\ \ \Novgorod J v ^ \ \ \ 50°N
50°N
iJprA \ \y
r
t
45°N
4R°N . / ~4 .
•

40°N 40°N

35°N 35°N
S /
30°N 7-Z / T4 0 15°E
30°N

45°E

Fig. 3.13 Track of totality for the eclipse of AD 1185 May 1 according to
Schroeter (1923). The eclipse was recorded as complete at Novgorod, the Russian
capital.

wide: between —2200 and +10 500 sec. Roughly contemporaneous data
indicate a true result for AT of close to +1200 sec at this fairly late
date. Hence calculation can no more than confirm the reports of totality
from Novgorod. Several similar examples could be given. Clearly the local
geographical circumstances are of considerable importance in determining
the utility of a particular observation.
Rather than map the belt of totality or annularity on each occasion,
it is preferable to use a technique involving projection of the shadow on
the 'fundamental plane' - the geocentric plane perpendicular to the axis
of shadow and passing through the centre of the Earth (see figure 3.14).
In this procedure, which has been adopted throughout this book, AT is
varied until the modulus of the minimal distance ( |D| ) of the place of
observation from the shadow axis equals that of the umbral radius ( \U\ )
on the fundamental plane. The usual convention is to express both D and
U in units of the Earth's equatorial radius. Limits to AT are obtained
70 3 Pre-telescopic eclipse observations and their analysis

Fig. 3.14 Projection of the lunar shadow on the 'fundamental plane'.

corresponding to each edge of the central zone separately touching the

place of observation. An iterative solution is necessary since D does not
vary linearly with changes in AT.

3.8.2 Solar eclipses which, although large, were not quite central
Many solar eclipse reports deny that the central phase was witnessed,
affirming instead that the Sun was reduced to a crescent (e.g. like a
hook'). If a record states that the cusps of the crescent were pointing
upwards, then it is clear that the eclipse was viewed from the south of the
central zone and vice versa. Unfortunately, it is very rare for a text to be
so specific and usually it is necessary to make two separate solutions -
one on the assumption that the umbral region passed to the north of the
observing site and the other to the south of it.
 3.8 Solar eclipse observations and their reduction to AT 71

Computation proceeds very much as if the central phase was witnessed,

but instead a range of AT - corresponding to the unobserved umbra -
is excluded. Almost any reasonable value of AT on either side of the
excluded range is usually acceptable since estimates of the magnitude of
a large partial eclipse - if recorded at all - are often very crude. If the
prohibited range of AT is very narrow (i.e. for an eclipse in which the
apparent diameters of the Sun and Moon were nearly equal), almost any
feasible solution space is permitted; the observation may thus prove to
be redundant. An example of a more useful record is provided by the
eclipse of AD 1004 Jan 24, in which the belt of annularity was unusually
narrow. On this occasion, a very large partial obscuration of the Sun was
reported from Cairo (cj) = 30.05 deg, A = —31.25 deg) by the astronomer
Ibn Yunus in the following words (see chapter 13):
The Sun was eclipsed until what remained of it resembled the crescent
Moon on the first day of the month. I estimated the magnitude of the
eclipse as 11 digits ...

In the vicinity of Cairo, the width of the central zone was only about 30
km. Calculations show that for any value of AT in the narrow range be-
tween +1770 and +1940 sec, the phase would have been annular at Cairo.
Hence to satisfy the observation of a partial eclipse, either AT < +1750
sec or > +1920 sec. Values of AT far beyond either of these limits would
still produce a large partial eclipse at Cairo. As both estimates of magni-
tude can be considered merely approximate (and not even in good mutual
accord - see section 3.8.4) the main conclusion from the above careful de-
scription is that the value of AT did not lie between +1750 and +1920 sec.

3.8.3 Timed solar eclipse contacts

Many measurements of solar eclipse timings are preserved in ancient and
medieval history, notably from Babylon, China and the medieval Arab
dominions. At each eclipse the astronomers typically determined three
instants - the start (i.e. first contact), greatest phase and end (last contact)
- unless the Sun was below the horizon or obscured by cloud for part of
the eclipse duration. The various measurements, whether made directly
(for example, with the aid of a water clock) or indirectly (in terms of solar
altitude) are readily reducible to local apparent time and hence to UT.
On only two occasions during the whole of the pre-telescopic period (at
Babylon in 136 BC and at the Chinese capital of Ch'ang-an in AD 761)
was the onset of totality carefully timed, but in neither case is there an
estimate of the duration of the total phase.
Babylonian astronomers measured the time of first contact relative to
the moment of sunrise or sunset, depending on which was nearer. Times
72 3 Pre-telescopic eclipse observations and their analysis

of the later phases were then determined relative to the start of the
eclipse. In computing the LT of rising or setting of the Sun or Moon,
I have adopted the standard figure for mean horizontal refraction (0.57
deg). Except in extreme circumstances, corrections for temperature and
pressure are unlikely to exceed 0.05 deg (Newton, 1972a) - a negligible
amount. When the time of an eclipse contact was measured relative to
sunrise or sunset, I have assumed that the upper limb of the Sun or
Moon (semi-diameter roughly 0.26 deg) was on the visible horizon at
these moments (a depression of the Sun's centre relative to the theoretical
horizon of approximately 0.8 deg). There is some evidence that the
Babylonians adopted this definition, which corresponds to either the very
first (morning) or last (evening) gleam of sunlight (Stephenson, 1974).
In the vicinity of Babylon the terrain is so flat that the horizon is
remarkably level. Hence in computing the LT of sunrise or sunset, no
allowance need be made for horizon profile. However, since the walls
of Babylon were some 15 m in height (Ravn, 1942, pp. 20 and 28), the
observatory would have to be at least above this level. The depression of
the horizon from this height would be 0.1 deg. According to Diodorus of
Sicily (II, 9), the Babylonian astronomers used to survey the sky from the
top of the ziggurat, a tower fully 90 m in height. From here the depression
of the horizon would be as much 0.27 deg. Since Diodorus' story may
be no more than hearsay, the figure of 0.1 deg (corresponding to some
15 m above ground level) will be adopted instead. When refraction and
semi-diameter are included, this is equivalent to a solar depression of 0.9
deg at the moment of sunrise or sunset.
To give an example of a Babylonian timed observation, the solar eclipse
of BC 254 Jan 31 was stated to begin 56 deg (= 3.73 h) before sunset,
reaching maximal phase after a further 12 deg (0.80 h) and clearing after
an additional 11 deg (0.73 h) - see chapter 5. In order to convert these
measurements to LT, it is first necessary to compute the LT of sunset, then
subtract the appropriate intervals. On this occasion, the Sun would attain
a depression of 0.9 deg (i.e. assumed sunset) at 17.26 h; hence the LTs
of the three observations would respectively be 13.53, 14.33 and 15.06 h.
These times are all in hours and decimals - a system which is standard
throughout this book.
It was the practice of Chinese astronomers to measure solar eclipse times
in double hours and marks, the first of the 12 double hours commencing
at an LT of 23 h. For instance, the eclipse of AD 1068 was reported to
have begun at 8 marks in the hour of szu (9-11 h), reached its maximum
after 5 marks in the hour of wu (11-13 h) and ended at 3 marks in the
hour of wei (13-15 h). These measurements correspond to LTs of 10.96,
12.44 and 13.84 h (see chapter 9).
In order to fix the times of the various contacts for eclipses of the
 3.8 Solar eclipse observations and their reduction to AT 73

Sun, medieval Arab astronomers systematically measured solar altitudes,

afterwards reducing their results to local time with the aid of an astrolabe
or tables. For example, at the eclipse of AD 923 Nov 11, the altitude of
the Sun at mid-eclipse was estimated as 8 deg in the east, and at the end
as 20 deg in the east (see chapter 13). The equivalent LTs are respectively
7.56 and 8.72 h. There is no mention of first contact in the text, but it may
be inferred from the measured altitudes at mid-eclipse and last contact
that the Sun probably rose eclipsed.
Once reduced to LT, the various measurements may be converted to
UT by allowing for the equation of time (z) and the geographic longitude
(/I), as discussed in chapter 1 (equation (1.1)). In order to derive a figure
for AT from any particular observation, it is necessary to use an iterative
solution since the computed value of the TT for any phase of a solar
eclipse is a function of AT itself. The Chinese observations of the eclipse
of AD 1068 Feb 6 have already been discussed. This was recorded at the
capital of Pien (</> = +34.78 deg, A = —114.33 deg), measured times of the
various phases leading to LTs of 10.96, 12.44 and 13.84 h. Adjusting for
the equation of time (z = —0.24 h) and the longitude of Pien (equivalent
to —7.62 h) the corresponding UTs are respectively 3.60, 5.07 and 6.48 h.
Using iteration, the computed TTs are 4.05, 5.40 and 6.71 h, leading to
AT values (to the nearest 50 sec) of +1600, +1200 and +850 sec. These
three results are in fair accord with one another, bearing in mind that the
measured times were only expressed to the nearest mark (i.e. 0.24 h or
some 900 sec).
Despite the limited resolution of the unaided eye, I have normally
assumed geometrical contacts for both the beginning and end of a solar
eclipse. Since ancient and medieval methods of timing were fairly crude,
determination of the LT of an actual contact between the lunar and
solar limbs should only be marginally more precise than for the less
well defined moment of maximal phase. In principle, apart from the rare
measurements of the onset of totality, timings of last contact should be the
most accurate since the observer merely needs to watch the indentation at
the western limb of the Sun gradually disappear. However, among ancient
and medieval astronomers it was common practice to roughly calculate in
advance the time of onset for an eclipse so that observers would have some
idea of when to look for first contact. In any given series of observations,
I have thus not weighted observations according to phase.
On very rare occasions, a medieval Arab observer stated that he was late
in detecting the start of an eclipse and had thus applied a small empirical
correction to the measured time in order to estimate the moment of true
first contact (see chapter 13). Under these circumstances, I have normally
accepted the amended result.
Early astronomers occasionally reported the moment of mid-eclipse.
74 3 Pre-telescopic eclipse observations and their analysis

Unless it is clear from the text that this result was obtained by averaging
the times of first and last contact, I have assumed that maximal phase
was intended. For a solar obscuration, mid-eclipse cannot be identified
directly; this moment can often be several minutes earlier or later than
that of greatest phase owing to the uneven speed of the shadow across
the ground. (In the case of a lunar eclipse, the moments of greatest phase
and mid-eclipse are, of course, effectively identical.)

3.8.4 Solar eclipse magnitudes

It was the usual practice of Babylonian, Greek and Arab astronomers to
estimate the degree of obscuration of the Sun or Moon at maximal phase
to the nearest 'finger' or 'digit'. These units were each equal to one-twelfth
of the apparent solar or lunar diameter. Chinese observers adopted a
different system in which the basic unit of magnitude was originally one-
fifteenth of the diameter of the Sun or Moon but after about AD 1000
was increased to one-tenth of the appropriate diameter. In most records
of eclipse magnitude it is clear from the text that the luminary was above
the horizon at greatest phase; if there is any doubt on this question, the
observation will be rejected.
Each magnitude observation can lead to a discrete value for AT without
the need for a measurement of time. However, there are several drawbacks
in using this method. Preserved data from any historical source are
relatively few in number. Further, not only do they represent crude
unaided eye estimates, the computed magnitude of a solar eclipse is only
a weak function of AT.
Partly to assess the potential of this method, Stephenson and Fatoohi
(1994a) investigated the accuracy of a series of recorded estimates of
lunar eclipse magnitudes from different parts of the world. For each
individual observation, they compared the estimated magnitude with its
computed equivalent (which is independent of AT). Stephenson and
Fatoohi concluded that the mean uncertainties in such estimates are close
to ten per cent. In the case of a solar eclipse, an error in the magnitude
by this amount would lead to an uncertainty in the value of AT by some
2000 to 2500 sec. Hence compared with other eclipse data, observations
of this type should give only poor resolution. Since it is seldom stated
whether the upper or lower portion of the Sun was covered, each estimate
of magnitude can - in principle - lead to up to four discrete values for
AT. However, in practice more than one viable result is seldom obtained
since, to obtain full accord with observation, calculation must place the
Sun above the horizon at maximal phase.
An example is provided by an observation made at the Chinese capital
of Lo-yang ((/> = +34.75 deg, A= —112.47 deg) on a date equivalent to
 3.8 Solar eclipse observations and their reduction to AT 75

b b

-70° _ 8 o° -90°
Longitude
Fig. 3.15 Central line and two curves corresponding to magnitude 0.53 (each
computed using AT = 0) for the eclipse of AD 489 Mar 18. The Sun was
estimated to be A (i.e. 0.53) covered at the Chinese capital of Lo-yang.

AD 489 Mar 18. The record indicates that ^ (i.e. 0.53) of the solar
diameter was covered at maximum (see chapter 9). For this eclipse, the
central line and the two curves corresponding to magnitude 0.53 (each
computed using AT = 0) are charted in figure 3.15. It can be seen that
only the more southerly curve for magnitude 0.53 intersects the parallel of
latitude for Lo-yang - and at a single longitude (A! = —98.1 deg). Hence
using equation (3.4), the equivalent value of AT is approximately +3450
sec, although with a very large uncertainty. As in the case of central and
near-central solar eclipses, I have preferred to use an iterative technique
to solve for AT. This more convenient procedure selects those values of
AT which yield the observed magnitude at the required location with the
Sun above the horizon.
The Arab report of a very large partial eclipse in AD 1004 has been
considered in section 3.8.2. Observers in Cairo expressed the magnitude
in two different ways:
... what remained of it resembled the crescent Moon on the first day of
the month. I estimated the magnitude of the eclipse as 11 digits ...
These two estimates are in poor accord with one another. When the
crescent Moon is about 24 hours old, only some two per cent of the
lunar diameter is sunlit, which represents a phase considerably more than
the estimated magnitude of 11 digits (0.92). However, the allusion to the
crescent Moon may be purely descriptive. Accepting the latter value (0.92)
76 3 Pre-telescopic eclipse observations and their analysis

leads to two results for AT: +150 sec if the lower limb of the Sun was
obscured (Cairo to the north of the central line) and +3400 sec if the
upper limb was covered (Cairo to the south of the central line). Because
of the very large magnitude, this observation is unique in yielding two AT
results which are fairly close together and thus both reasonably viable.
Usually if two values for AT do emerge, they are of the order of 20000 sec
apart - a separation which is so great that one of the results is obviously
discordant.
Very occasionally, Arab astronomers expressed eclipse magnitudes in
terms of the area of the solar disk rather than its diameter. Centuries
before, Ptolemy (Almagest, VI, 7) was aware of similar practices, although
there is no evidence that any of the observations which he cites were in
this form. An example of an Arab record - dating from AD 928 - is as
follows (see chapter 13):

The Sun rose eclipsed by (a little) less than one quarter of its surface and
the eclipse continued to increase until one-quarter of it was eclipsed ...

Such observations are readily reduced to linear magnitude by simple

geometry. In table 3.1, I have listed the area and linear equivalents for
solar and lunar eclipses. For solar obscurations, I have assumed equal
semi-diameters for the Moon and Sun, while for lunar obscurations I have
adopted a typical umbral radius of 2.70 times that of the Moon. It will be
noted that discrepancies (all of the same sign) are greatest for moderately
small eclipses. An eclipse in which half of the area of the solar disk (but
as much as much as 0.60 of the diameter) is covered by the Moon is
represented diagrammatically in figure 3.16.
Ptolemy (Almagest, VI, 8) gives a similar table to table 3.1, all magni-
tudes being, however, expressed in digits. In general, there is close accord
between his results and those given in table 3.1.

3.8.5 Observations that the Sun rose or set eclipsed

Several observations indicating that the Sun was eclipsed when it was on
the horizon are to be found among the records of Babylonian, Chinese
and Arab astronomers. In most instances, the fraction of the Sun which
was obscured at these moments is not specified; it is merely implied that
the Sun was visibly eclipsed when it rose or set. Observations of this kind
set limits to AT since at least some part of the solar disk must be eclipsed
when the Sun is on the appropriate horizon. For a recent photograph
showing the partially eclipsed Sun on the point of setting, see figure 3.17.
If it can only be established from the record that the Sun reached the
horizon between first and last contact, AT can only be fixed within very
 3.8 Solar eclipse observations and their reduction to AT 11

Table 3.1 Conversion from area to linear magnitude for both

solar and lunar eclipses.

Solar Lunar Solar Lunar

Area mag. Lin. mag. Lin. mag. Area mag. Lin. mag. Lin. mag.

0.05 0.12 0.11 0.55 0.64 0.57

0.10 0.19 0.17 0.60 0.68 0.61
0.15 0.26 0.23 0.65 0.72 0.65
0.20 0.31 0.28 0.70 0.76 0.69
0.25 0.36 0.32 0.75 0.80 0.73

0.30 0.41 0.37 0.80 0.84 0.77

0.35 0.46 0.41 0.85 0.88 0.81
0.40 0.51 0.45 0.90 0.92 0.86
0.45 0.55 0.49 0.95 0.96 0.92
0.50 0.60 0.53 1.00 1.00 1.00

Solar IJmb

\
Area mag
\ = 0.50

Fig. 3.16 Solar eclipse in which half of the area of the Sun's disk (but as much
as much as 0.60 of the diameter) is covered by the Moon.

wide limits. However, it is sometimes possible to decide from the text if

the phase was growing or diminishing at the time - thus roughly halving
the indicated range of AT. Further refinements cannot usually be made.
In calculating the LT of sunrise or sunset under these circumstances,
I have assumed that the centre (rather than the upper limb) of the Sun
78 3 Pre-telescopic eclipse observations and their analysis

Fig. 3.17 Photograph by Giancarlo Gengaroli showing the partially eclipsed Sun
in 1994 May 10 just before setting. (Courtesy: Vastronomia.)

was on the visible horizon when each observation was made. Ideally,
some correction should be made for the position angle of contact between
the lunar limb and the disk of the Sun at the appropriate phase of the
eclipse. However, such refinements seem unnecessary since only crude
allowance for topographic effects can be made (height of the observer
above the ground and horizon profile). Most early observatories were
located within city walls. Hence where possible astronomers usually
selected an elevated site in order to obtain an uninterrupted view of
the horizon. For example the so-called 'Ancient Observatory' at Beijing,
established by Jesuit astronomers in AD 1670, is located on a raised
platform some 10 m above ground level. As noted above (section 3.8.3),
the Babylonian observatory was probably at least 15 m above the ground.
Adopting a mean observer height of some 10 or 15 m, leads to an assumed
solar depression of roughly 0.65 deg under the above circumstances.
In the vicinity of Babylon and several other ancient cities which were
equipped with observatories, the horizon is extremely level. However, at
certain other important sites the surrounding countryside is quite hilly and
careful allowance for horizon profile would be needed in order to deduce
the LT of sunrise or sunset accurately. To give an arbitrary example,
200 m hills on the eastern or western horizon at a distance of 10 km
would delay sunrise or advance sunset by at least 5 min. Such an amount
 3.8 Solar eclipse observations and their reduction to AT 79

would materially affect the derived AT limits - especially in medieval

times when the value of this parameter was of the order of 1000 sec. In
summary, if the necessary correction is likely to be significantly less than
this amount, I have retained the observation but have made no allowance
for horizon profile. Otherwise, I have rejected it. In chapter 14, where the
various results obtained in this book are analysed, I have reconsidered the
more critical AT limits.
A useful illustration is provided by the solar eclipse of BC 322 Sep 26.
This was observed at Babylon - see chapter 5. The eclipse was recorded
in the following words:

Year 2 of (king) Philip, (month VI). The 28th, around 3 deg before
sunset, solar eclipse ... It set eclipsed.

The recorded time before of only 3 deg (i.e. 12 minutes) before sunset
leads to a discrete value for AT, as will be discussed in chapter 5. However,
the mere fact that some part of the eclipse was visible before sunset can
set a useful lower limit to AT. The calculated LT of sunset at Babylon on
the day in question was 18.08 h, a UT of 15.01 h. It may thus be inferred
that the UT of first contact was earlier than 15.01 h; otherwise no part
of the eclipse would be visible before the Sun set. Computing iteratively,
the TT of first contact is 18.74 h, leading to the result AT > +13 400 sec.
Although the text implies that the eclipse began only a few minutes before
sunset, there is always the possibility of a scribal error in the time interval.
Hence to obtain an upper limit to AT, it seems best to assume only that
last contact took place after the Sun reached the western horizon - i.e.
that some part of the eclipse remained visible until sunset. In this case,
the LT would be > 18.08 h, and the UT > 15.01 h. Comparing with the
computed TT for this phase of 20.02 h gives AT < +18 000 sec. When
the two separate limits are combined, it may be concluded that +13 400
< AT < +18 000 sec, both of which represent fairly firm bounds.
Since estimates of the degree of obscuration of the Sun at sunrise or
sunset are quite rare, these will be considered in section 3.9.3 along with
the more frequent lunar observations of this kind.

3.8.6 Records which merely note the occurrence of a solar eclipse

The vast majority of ancient and medieval eclipse records mention no
more than that an eclipse of the Sun or Moon occurred on a specified
date. Lunar reports of this kind are clearly valueless in the determination
of AT. In principle, solar observations can be used to determine AT, but
there are several problems inherent in the analysis of such data. If it is
assumed that a bare record of a solar eclipse can imply any magnitude
80 3 Pre-telescopic eclipse observations and their analysis

(from scarcely visible to central), the range in AT obtained from a single

observation is huge - some 40000 sec. Even in ancient times - when
in any case such data are less frequent - results of this sort tell us very
little. Additionally, experience shows that because of the non-random
distribution of population centres the AT values obtained in this way
may well seem to imply large-scale variations in the LOD which prove to
be wholly spurious. Apart from a brief discussion in the present section,
such observations will not be considered in this book. They are included
here mainly because Newton (1970 and 1972b) attached much importance
to them.
The difficulties associated with the analysis of laconic records of solar
eclipses are perhaps best illustrated by consideration of the almost con-
temporaneous total solar eclipses of AD 1178 Sep 13 and 1187 Sep 4. In
the interval of only nine years between them, AT is unlikely to have varied
by more than a few tens of seconds. As a result of an extensive search
through published compilations of European chronicles, Ginzel (1884a)
was able to assemble 14 separate records of the first event and as many
as 27 for the second. No progress in uncovering further records of these
two eclipses has since been made. Most of the accounts which Ginzel
compiled were from local annals (e.g. of monasteries) and hence the place
of observation may be confidently established. However, a few other
reports were contained in chronicles which drew on information from a
wide area. In almost every case, it is merely reported that an eclipse of
the Sun occurred on a specified day without giving any descriptive details.
For instance, the following brief report is from Italy in AD 1178:
1178 ...On the Ides of September (Sep 13), the Sun was obscured [sol
obscuratus est].
[Chronicon Fossae Nuovae]
Figure 3.18 marks the locations of each observing site for the eclipse of
AD 1178 and figure 3.19 for the eclipse of 1187. Both charts are drawn on
a rectangular grid in order to best demonstrate the effect of altering AT.
The computed tracks of totality on each date, based on the assumption of
a constant length of day of 86400 SI seconds (AT = 0), are shown. On the
seemingly plausible assumption that the individual observations should be
equally weighted, it is possible by displacing the calculated central lines
(i) in 1178 towards the east by about 20 deg and (ii) in 1187 towards the
west by some 10 deg to derive what appear to be their most likely true
positions. The implied values for AT are roughly +5000 sec in 1178 and
—2500 sec in 1187. These results are at considerable variance with one
another - and also with those obtained from roughly contemporaneous
Chinese and Arab timed measurements, which indicate a value for AT at
this date of around +1000 sec (see chapters 9 and 13).
 3.8 Solar eclipse observations and their reduction to AT 81

10° W 30° E

Fig. 3.18 Locations of the various European sites where the solar eclipse of A D
1178 Sep 13 was observed. Equally weighting the observations, the 'best •fitting'
central line lies some 20 deg to the east of the computed central line.

Rather than assume huge variations in AT in such a short time,

it is much more reasonable to conclude that analysis of this sort is
very much influenced by 'population bias' - the non-random distribu-
tion of population centres (or at least those places from which pre-
served eclipse records are accessible) in medieval Europe. A similar
illustration, involving the eclipse of AD 1039, is given by Muller and
Stephenson (1975). For independent comments, see Lambeck (1980),
p. 314.
In summary, the use of careful descriptions of eclipses - whether quali-
tative or quantitative records - is much to be preferred.
82 3 Pre-telescopic eclipse observations and their analysis

10° W

Fig. 3.19 Locations of the various European sites where the solar eclipse of AD
1187 Sep 4 was observed. Equally weighting the observations, the 'best-fitting'
central line lies some 10 deg to the west of the computed central line.

3.9 Lunar eclipse observations and their reduction to AT

As noted earlier, lunar eclipses have more restricted usage than their solar
counterparts in the study of the Earth's past rotation. However, the fact
that relatively more timings and rising and setting observations of lunar
eclipses have survived does much to redress this balance.
Before I discuss the various ways in which lunar eclipse records can be
used to determine AT, some comments on the adopted dating scheme are
necessary. Throughout this book I have assigned a double date on the
Julian calendar - e.g. Jun 13/14 - for all eclipses of the Moon. This is
partly because these events, which can last for several hours, often begin
 3.9 Lunar eclipse observations and their reduction to AT 83

before local midnight and end after it. Furthermore, since most observing
sites are much to the east of the Greenwich meridian, the LT tends to
be several hours ahead of the UT. An additional reason for this choice is
that several cultures began the new day at times other than midnight. For
instance, the Babylonian day began at sunset (i.e. roughly 6 hours before
the start of the civil day) and this practice was later followed by Muslims.
On the other hand, it seems that Chinese astronomers did not change the
calendar date until sunrise (some 6 hours after the start of the civil day).
A double date thus helps to avoid confusion.

3.9.1 Timed lunar eclipse contacts

Ancient and medieval timings of partial lunar eclipses generally relate
to three phases: first contact, maximal phase (i.e. mid-eclipse) and last
contact. In the case of total obscurations, four separate times are usually
recorded; these are the two external contacts and also the beginning
and end of totality (second and third contact). Although most early
astronomers determined times directly (e.g. with the aid of a clepsydra),
Arab observers were in the habit of measuring altitudes of the Moon or
a suitably placed bright star instead.
In analysing all observations of the umbral contacts for lunar eclipses,
I have applied the standard increment of two per cent to the radius of the
Earth's shadow to approximately allow for the effect of the atmosphere.
There is also some evidence that an observer using the unaided eye may
mistake the deep penumbral shadow for the umbra itself. For example,
at the very beginning of the seventeenth century (a little before the in-
troduction of the telescope), the Belgian astronomer Wendelin criticised
a contemporary for failing to make this distinction (Pingre, 1901, p. 20).
However, recent investigation of recorded lunar eclipse durations (from
first to last contact) by unaided-eye observers shows that the mean ratio
of the measured and computed values is very close to unity, although
there is significant scatter among individual results. A survey of such
measurements by Stephenson and Fatoohi (1994b) revealed that the ap-
propriate mean ratios for Babylonian (after 300 BC), Chinese and early
seventeenth century European astronomers were respectively 1.02 + 0.03,
0.99 + 0.04 and 1.01 + 0.01. Hence, when a series of contact timings is
used to determine AT, penumbral effects should be negligible.
Once a lunar eclipse timing is reduced to UT, computation of the
corresponding TT leads to a result for AT by direct subtraction. There
is no need for an iterative solution (unlike in the case of a solar eclipse)
since at any given moment the local circumstances of a lunar eclipse are
virtually identical over the entire hemisphere of the Earth's surface from
which the Moon is visible at the time.
84 3 Pre-telescopic eclipse observations and their analysis

The total lunar eclipse of BC 226 Aug 1/2 was one of the many
carefully timed by Babylonian astronomers. The record contains the
following details (see chapter 5):
Night of the 14th ... At 52 deg after sunset, when a Cyg culminated,
lunar eclipse; when it began on the east side, in 17 deg (of) night time it
covered it completely; 10 deg night time maximal phase; when it began to
clear, it cleared in 15 deg night time from south to north ...42 deg onset,
maximal phase and clearing ... .
[Trans. Sachs and Hunger (1989, p. 141).]
The reference to the culmination of the star a Cyg provides an inde-
pendent check on the time of first contact - in addition to the sunset
measurement. It will be noted that the reported duration of 42 deg is
indeed the sum of the durations of the individual phases. The four in-
tervals after sunset correspond to 3.47, 4.60, 5.27 and 6.27 h respectively.
Since sunset occurred at a LT of 18.94 h, the corresponding LTs for the
contacts are 22.41, 23.54, 0.21 h and 1.21 h. Although the Babylonian date
(commencing at sunset) of all four contacts was Aug 2, the eclipse began
on a civil date (starting at local midnight) of Aug 1 and ended on Aug 2.
The various results are shown diagrammatically in figure 3.20. Converting
the LTs to UT (all on Aug 1) and comparing with the computed TTs (all
on Aug 2), we have the following results (to the nearest 50 sec):
• First contact: UT = 19.51 h, TT = 0.38 h, AT = +17 550 sec.
• Second contact: UT = 20.63 h, TT = 1.50 h, AT = +17 550 sec.
• Third contact: UT = 21.30 h, TT = 2.57 h, AT = +18 950 sec.
• Last contact: UT = 22.30 h, TT = 3.68 h, AT = +19350 sec.
During the eclipse, a Cyg (RA 19.43 h) was 11.00 h to the east of the
Sun. Hence the LT of culmination (identifying first contact) was 12.00 +
11.00 = 23.00 h, a UT of 20.09 h on Aug 1. This is as much as 0.68 h after
the corresponding sunset measurement and leads to a result for AT of
+15 450 sec. The relation between the UT, TT and AT values for each of
the five measurements is illustrated in figure 3.21. All of the AT results are
in fair (but by no means excellent) accord with one another. Fortunately
there are many further observations from around this period.
To give an additional example, astronomers in Baghdad ((/> = +33.33
deg, A=—44.42 deg) determined the time of first and second contact at
the eclipse of AD 854 Feb 16/17 by measuring star altitudes (see chapter
13). The altitude of a Tau at first contact was found to be 45.5 deg in the
east, while that of a CMi at the beginning of totality was between 22 and
23 deg in the east. At the date in question, the celestial co-ordinates (RA
 3.9 Lunar eclipse observations and their reduction to AT 85

Equivalent
Babylonian dates

Aug 1 Aug2 2nd Ct Last Ct

1stCt 3rdCt
Sunset
-52°

17h 18h 19h 20h 21 h 22h ! 23h ! 0 2h 3h 4h

I I

18h.94 22h.41 23h.54 0h.21 1h.21

Aug 1 Aug 2
Civil dates
Fig. 3.20 LT measurements made by Babylonian astronomers for the lunar
eclipse of BC 226 Aug 1/2.

> 19h.51 (First Ct)

CQ

2Oh.O9 (First Ct)(a Cyg)

(First Ct) 0h.38 - - -

J - 20h.63 (Second Ct)
>
CQ

(Second Ct) 1h.5O 21h.3O (Third Ct)

^o-

22h.30(Last Ct)
(Third Ct) 2h.57 - ~
<&<
>
CQ

(Last Ct) 3h.68

>
C
CQ

Fig. 3.21 Relation between the UT, TT and equivalent AT values for each of
the five measurements made at the lunar eclipse of 226 BC.
86 3 Pre-telescopic eclipse observations and their analysis

and dec) of a Tau were 3.52 h, +13.5 deg; for a CMi the corresponding
figures were 6.65 h, +7.46 deg. For comparison, the solar RA was 9.67 h.
Calculations show that a Tau would reach the measured altitude (45.5
deg) 2.93 h before crossing the meridian. Since the star was 6.15 h to the
west of the Sun in hour angle, the corresponding LT of first contact was
12.00 - 2.93 - 6.15 = 2.92 h. Similarly, a CMi would attain an elevation
of 22.5 deg at 4.53 h before meridian transit — an LT of 4.45 h for second
contact. The equivalent UTs are respectively 0.01 and 1.53 h. Comparing
with the computed TT of first and second contact (0.71 and 1.91 h) gives
AT results of +2500 and +1350 sec. At this comparatively recent epoch,
AT was fairly small. Hence the discrepancy between the individual values
(1150 sec), although rather less than that noted among the Babylonian
observations in 226 BC is relatively much more significant.

3.9.2 Observations that the Moon rose or set eclipsed

Observations of this type fall into two distinct groups: (i) those for which
it is merely stated that the Moon rose or set eclipsed; and (ii) instances
where the degree of obscuration of the lunar disk at moonrise or moonset
is estimated. Reports in the former category yield limits to the value of
AT and will be discussed here. Those in the latter group enable specific
values for AT to be determined; they will be considered in section 3.9.3.
The method of investigation, which basically resembles that used for
a solar eclipse observed at sunrise or sunset, is illustrated in figure 3.22.
Here it is assumed that at place A the Moon was seen to rise eclipsed
to some unknown degree. The diagram shows the Moon entering the
terrestrial umbra (first contact) and also leaving it (last contact). For a
certain clock error ATi, moonrise would occur at A just at first contact
so that the whole of the eclipse would be visible. However, for a different
clock error AT2, the effect of the Earth's rotation would be to move A
to Af where moonrise occurs just at last contact. In this latter case, no
part of the eclipse would be seen. Hence in order to obtain accord with
observation, the required value of AT can be anywhere in the indicated
range S AT (i.e. ATi < AT < AT2), but not outside it.
In applying this technique, I have assumed that the centre of the Moon
was on the visible horizon at moonrise or moonset and thus 0.65 deg
below the theoretical horizon (see section 3.8.5). When calculating the
local time of moonrise and moonset it is also necessary to make an
allowance for parallax, which can frequently reach a large fraction of a
degree. The LT of rising or setting of the Sun or Moon at any selected
place is a function of the declination of the luminary - and hence of AT.
However, on account of the very slow change in the solar declination
(never more than 1 arcmin in an hour), in computing the local time of
 3.9 Lunar eclipse observations and their reduction to AT 87

Moon at
first Ct

Mqonrise at first conta

Edge of umbra

Sun

Fig. 3.22 Method of investigation for an eclipse in which the Moon was observed
to rise eclipsed to some unknown degree. A discrete range of AT (i.e. SAT) is
indicated.

sunrise or sunset, it is quite unnecessary to make any allowance for AT,

corrections amounting to less than 0.01 h. This very minor variation has
been ignored in previous discussion. On the other hand, since the Moon
moves so much more rapidly than the Sun (by a factor of approximately
13), it is necessary to deduce the LT of moonrise or moonset iteratively
using successive approximations to the final choice of AT. The response to
changes in AT is particularly significant for spring moonrise and autumn
moonset.
If an eclipse was generally total, yet the record states that the Moon
rose or set whilst partially covered, it is usually possible to decide from the
text whether the Moon reached the horizon before or after the occurrence
of the total phase. Otherwise, two discrete ranges of AT are obtained, one
corresponding to moonrise or moonset between first and second contact
and the other between third and last contact. When the totally eclipsed
Moon is on the horizon, it is probably always invisible to the unaided
eye. In principle, if the sky was clear, experienced observers, who would
be well aware that the eclipsed Moon was accurately in opposition to the
Sun, would probably be able to determine whether any trace of the lunar
partial phase was discernible when the Sun reached the opposite horizon.
This is illustrated by a Chinese account from Lin-an (0 = +30.25 deg, A
= -120.17 deg) of the total lunar eclipse of AD 1168 Mar 25/26 (see also
chapter 9):

On this evening, when the Moon rose there was light cloud. Until the
fall of darkness it could not be seen that the Moon had been totally
eclipsed. When the 3rd mark of the initial half of the hour hsu was
88 3 Pre-telescopic eclipse observations and their analysis

reached, as expected it was shining and so it could be known that this

eclipse had been total on rising ...
Evidently the light cloud did not hinder observations on this occasion.
However, interpretation of a negative sighting must necessarily be some-
what dubious. Hence any observation which alleges that the Moon rose
or set totally eclipsed will be discounted throughout this book.
In order to further illustrate the method of analysis for moonrise
and moonset eclipses, I have selected two separate examples, both from
Babylon (see chapter 7): a record of a total eclipse from 667 BC and a
report of a partial eclipse from 66 BC which states that the Moon rose
before maximal phase.

(1) BC 667 Oct 14/15 '...It set eclipsed. (Began) at 20[+x] deg before
sunrise'.
This record is found on a tablet which lists many reports of lunar eclipses.
Although the eclipse was actually total, the text is so badly damaged that
it can only be inferred that the Moon set between first and last contact.
(i) For first contact before moonset, LT of moonset = 6.15 h.
Hence LT of contact < 6.15 h, UT < 3.02 h. Comparing the
UT with the computed TT of 6.19 h, one gets AT > 11400 sec.
(ii) For last contact after moonset, LT of moonset = 6.31 h. Hence
LT of contact > 6.31 h, UT > 3.20 h. Computed TT of 9.73 h,
thus AT < 23 500 sec.
When these two AT limits are combined, 11400 < AT < 23 500 sec.
It will be noted that the computed LTs of moonset differ by as much
as 0.16 h, mainly on account of the large AT range.

(2) BC 66 Dec 28/29 '... When the Moon rose, 2 fingers on the south side
[were eclipsed]. In 9 deg night, over a third of the disk [was eclipsed]; 8
deg duration of maximal phase, until it began to become bright ...'
It is evident that the Moon rose between first contact and maximum.
(i) For first contact before moonrise, LT of moonrise = 16.87 h. Hence
LT of contact < 16.87 h, UT < 14.00 h. Computed TT = 16.64 h;
thus AT > 9500 sec.
(ii) For maximum phase after moonrise, LT of moonrise = 16.90 h.
Hence LT of contact > 16.90 h, UT > 14.14 h. Computed TT =
17.75 h; thus AT < 13 350 sec.
Combining these limits, one obtains 9500 < AT < 13 350 sec.
Here the two calculated LTs of moonrise are not significantly different,
mainly because the AT range is relatively small.
 3.9 Lunar eclipse observations and their reduction to AT 89

3.93 Estimates of the degree of obscuration of the Moon

at moonrise or moonset
Several careful estimates of the fraction of the Moon which was enveloped
in the Earth's shadow when it rose or set are preserved. As for both solar
and lunar eclipse magnitudes generally, Babylonian and Arab estimates
of this ratio are usually expressed to the nearest twelfth of the disk and
Chinese estimates to the nearest tenth or fifteenth. Since the phase of a
central lunar eclipse changes by such an amount in only about 5 min, it
may be inferred that an observation of this type might compete fairly well
with a carefully timed contact.
When the Moon is rising or setting, atmospheric refraction is maximal.
Nevertheless, distortion of the lunar limb is not necessarily critical. Several
good quality photographs which have been published in various journals
etc. show the eclipsed Sun on the horizon - see figure 3.17 - but few seem
available for the eclipsed Moon in a similar situation. Photographs such
as these lead to the inference that on most occasions a reasonable estimate
of the proportion of the disk in shadow would be possible, whether for
the Sun or the Moon. In any case, if the limb of the luminary were very
distorted, an observer in the ancient or medieval world might well feel that
such an exercise was pointless. As will become apparent in subsequent
chapters, individual results for AT obtained from these observations prove
to be remarkably self-consistent.
If an estimate of the extent to which the Moon was covered at moonrise
or moonset is recorded, it would seem best to suppose that the whole of
the disk was just clear of the horizon when the observation was made.
Allowing for horizontal refraction and assumed observer height (10-15
m), this corresponds to a depression for the Moon's centre of about 0.4
deg below the theoretical horizon. Using an iterative solution for AT, the
calculated phase at moonrise or moonset may be varied until it equals the
observed fraction. Each estimate of the proportion of the Moon covered
at its rising or setting normally leads to a single value for AT (unless it is
uncertain whether the phase was increasing or declining at the time).
The profile of the horizon is less important for this type of observation
since specific limits to AT are not set. Horizon irregularities merely
increase the scatter in the AT results which are obtained. Unless very
significant, they can thus be ignored.
As noted above (section 3.8.4) Stephenson and Fatoohi (1994a) inves-
tigated the accuracy with which the unaided eye can determine the peak
magnitude of a lunar eclipse. They analysed many lunar magnitude es-
timates from the pre-telescopic period originating from Babylon, China,
the Arab lands and Europe. In each case there was nothing in the various
records to suggest that the observer was prevented from viewing the Moon
90 3 Pre-telescopic eclipse observations and their analysis

around maximal phase - either by cloud or the Moon being below the
horizon at the time.
Comparison between observation and computation for these reference
data showed that the median error in estimating magnitude (as a pro-
portion of the lunar diameter) was 0.08. More significantly, it was also
demonstrated that the various observers systematically overestimated the
magnitudes of small eclipses (less than half covered) by up to about 0.15
(in units of the lunar diameter). For larger obscurations much the re-
verse was true. The effect of the deep penumbral shadow may well lead
to overestimates of magnitude for small eclipses but physiological effects
may well be important at all phases.
The results of this investigation are shown in figure 3.23, which is taken
from the paper by Stephenson and Fatoohi (1994a). In this diagram,
the observational error in estimating magnitude (in the sense computed-
observed) is plotted as a function of the computed magnitude. A least
squares straight line fit (for want of a better solution) to the data em-
phasises the remarkable skewness of this distribution, the gradient being
as much as 0.19. Relative to this line, the median error of observation
is only 0.05.
Each estimate of the proportion of the Moon covered at its rising or
setting normally leads to a single value for AT (unless it is uncertain
whether the phase was increasing or declining at the time).
As an example, I have selected the eclipse of AD 438 Jun 23/24, which
was observed in the Chinese capital of Chien-k'ang (cj) = +32.03 deg,
A = —118.78 deg). The record reads as follows (see chapter 9):
When the Moon first rose, it was already eclipsed. Its brightness had
already regained 1/4 (of itself)...
The proportion of the lunar diameter covered at moonrise was thus
estimated as 0.75. The eclipse was actually total, but it seems clear from
the text that the phase was already declining when the Moon first came
up. Making a minor adjustment of +0.03 to the estimated fraction of 0.75
to allow for systematic errors (as derived from figure 3.23) gives a result
of 0.78.
Chien-k'ang (on the site of the modern city of Nanjing) was situated in
a fairly level plain beside the Yangtze river. Using an iterative procedure,
as in section 3.9.2, the computed LT of moonrise (assuming a depression
below the theoretical horizon of 0.4 deg) is 19.18 h, a UT of 11.20 h. Since
the computed TT when 0.78 of the lunar diameter was obscured after
maximum is 12.82 h, the equivalent value of AT is +5800 sec.
As already noted, it is rare to find an estimate of the proportion of the
Sun eclipsed when on the horizon. Investigation of the few observations of
this type follows very much the above pattern. Although no allowance need
 3.10 Conclusion 91

0.26—|
0.24 —
0.22 —
0.20 —
0.18 —
0.16 —
0.14 —
0.12 —
0.10 —
0.08 —
0.06 —
0.04 —
0.02 —
</)
CD -0.00 —
CD
£ -0.02 —
C

2 -0.04 —
LU
-0.06 —
-0.08 —
-0.10 —
-0.12 —
-0.14 —
-0.16 —
-0.18 —
-0.20 —
-0.22 —
-0.24 —
-0.26
i i i i i n i T T
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Computed magnitude of eclipse

Fig. 3.23 Results of an investigation of recorded lunar eclipse magnitudes

(Stephenson and Fatoohi, 1994a).

be made for AT in computing the LT of sunrise or sunset, an iterative so-

lution is necessary in order to match the calculated phase with observation.

3.10 Conclusion
The various techniques developed in this chapter will be used to analyse
the observations discussed in chapters 4 to 13. When it is felt that further
92 3 Pre-telescopic eclipse observations and their analysis

explanation is necessary, this will be given at the appropriate place. In

most cases a brief derivation of each individual AT result will be given
for direct reference, followed by tabulation later in the same chapter. All
results will be summarised in the Appendix and analysed in chapter 14.
(From here on, the abbreviations long, and lat. will be used instead of A
and 4> respectively.)
 Babylonian and Assyrian
records of eclipses

4.1 Introduction
More celestial observations are preserved from Babylon than from any
other contemporary civilisation. Yet until about a century ago, when
large numbers of clay tablets devoted to astronomy began to be un-
earthed at the site of Babylon, little was known about the achievements
of the skywatchers of this once great city. What could be established was
mainly based on ancient Greek texts and the Old Testament. Both the
Prophet Isaiah (e.g. 47:13) and the ancient Greek writer Strabo (Geog-
raphy, XVI, 1.6) stress the Babylonian preoccupation with astrology. As
noted in chapter 3, the ancient Greek historian Diodorus Siculus (Library
of History, II, 9) implies that the lofty ziggurat - built during the reign of
Nebuchadrezzar II (604-563 BC) was used as an observatory. Figure 4.1
shows a schematic view of Babylon in the days of Nebuchadrezzar II -
as visualised by Herbert Anger (Unger, 1931, figure 7).
Among writers of the ancient Greek and Roman world whose works
are still extant, only the great Alexandrian astronomer Claudius Ptolemy
(c. AD 150) hints at the true scale on which celestial observation was
practised at Babylon. In his Mathematike Syntaxis (Mathematical Sys-
tematic Treatise) - which later became known as the Almagest - Ptolemy
specifically mentions sets of Babylonian eclipse observations to which he
had access. Examples of his comments are as follows:
(i) First, the three ancient eclipses which are selected from those
observed in Babylon...
[Almagest, IV, 6; trans. Toomer (1984, p. 191).]

(ii) First then, to correct the actual mean motion in latitude, we looked
for [pairs of] lunar eclipses (among those securely recorded)
separated by as great an interval as possible...
[Almagest, IV, 9; trans. Toomer (1984, p. 206).]

93
94 4 Babylonian and Assyrian records of eclipses

Fig. 4.1 A schematic view of Babylon in the days of Nebuchadrezzar II - as

visualised by Herbert Anger (Unger, 1931).

(iii) He (Hipparchus) says that these three eclipses which he adduces

are from the series brought over from Babylon and were observed
there...
[Almagest, IV, 11; trans. Toomer (1984, p. 211).]

It is regrettable that Ptolemy actually cites no more than ten Babylonian

records of lunar eclipses from the apparently large number available to
him. Furthermore, no observations of solar eclipses from Babylon are
preserved in the Almagest, although there are a few references to other
 4.2 Babylonian lunar eclipse records in the Almagest 95

celestial phenomena reported from this site, such as conjunctions of planets

with stars. These various observations range in date from 721 to 229 BC,
lunar eclipses covering the period from 721 to 382 BC.
The discovery of vast numbers of astronomical cuneiform texts at the
site of Babylon during the 1870s and 1880s was eventually to revolutionise
knowledge of Babylonian astronomy. These texts, which are in the form of
inscribed clay tablets, range in date from about 730 BC to AD 75. Many
of the eclipse records which they contain have proved so important in
studies of the Earth's past rotation that the investigation of this material
will occupy fully three chapters of the present book (5, 6 and 7). In
particular, the earlier observations must have formed the ultimate source
of the material used by Hipparchus and Ptolemy. Despite the existence
of this huge archive, the ten Babylonian lunar eclipse records cited in the
Almagest form an interesting set of data and deserve to be analysed in their
own right. These observations will be investigated in the present chapter,
followed by a discussion of the historical background to the astronomical
cuneiform texts. At the close of this chapter, the eclipse records found on
the Assyrian cuneiform tablets of the eighth and seventh centuries BC will
be briefly considered. However, it should be stressed that these records
are mainly in the form of astrological texts and - unlike the Babylonian
material - contain little quantitative information.

4.2 The Babylonian lunar eclipse records in the Almagest

The very earliest astronomical records quoted in the Almagest are ex-
clusively of Babylonian origin. (For a discussion of the Greek eclipse
observations which Ptolemy cites, see chapter 10.) Ptolemy evidently did
not have access to material much older than 721 BC - the earliest Baby-
lonian eclipse record which he discusses. He states (III, 7) that beginning
with the reign of King Nabonassar (correct name: Nabu-nasir) of Baby-
lon (747-733 BC) 'the ancient observations are, on the whole, preserved
down to our own time'. This is apparently why Ptolemy chose the era of
Nabonassar (747 BC) for numbering years. There is a Hellenistic tradi-
tion that King Nabonassar (like Emperor Ch'in Shih-huang in China - see
chapter 8) destroyed the records of his predecessors. His alleged purpose
was to ensure 'that the reckoning of the Chaldean kings might start with
himself (Brinkman, 1968). Although this tradition may lack firm grounds,
it is a fact that Babylonian history and chronology are extremely weak
in the two centuries immediately prior to Nabonassar's reign. Even the
names of several kings who ruled during this interval are unknown (cf.
Oates, 1979, p. 201). However, beginning with Nabonassar, Babylonian
chronology is securely established.
It seems very likely that Ptolemy did not compile the list of Babylonian
96 4 Babylonian and Assyrian records of eclipses

Table 4.1 Egyptian month-names as found in

the Almagest.
Number Name Number Name
I Thoth VII Phamenoth
II Phaophi VIII Pharmouthi
III Athyr IX Pachon
IV Choiak X Payni
V Tybi XI Epiphi
VI Mechir XII Mesore

observations himself. The evidence points instead to Ptolemy's great

predecessor Hipparchus of Rhodes (c. 150 BC). As noted above, Ptolemy
specifically mentions a series of eclipse observations which had been
'brought over from Babylon' and investigated by Hipparchus. Toomer
(1988) is of the opinion that the entire Babylonian record available to
Ptolemy was compiled by Hipparchus, further suggesting that Hipparchus
'arranged them in a form suitable for use by Greek astronomers'. It is
clear from reading the accounts of individual eclipses quoted by Ptolemy
(see section 4.3 below) that he did not receive them at first hand. Sadly,
apart from the material preserved in the Almagest, all trace of Hipparchus'
compilation has long been lost.
No ancient manuscripts of the Almagest now exist; the earliest copies
date from the ninth century AD (Toomer, 1984, pp. 2-4). However, as
Toomer points out, there is in general very close accord between the text
of individual manuscripts.
For the very earliest eclipse observations (721 and 720 BC) which he
cites, Ptolemy numbers the year from the accession of Mardokempad
(Marduk-apla-iddin), who was ruler of Babylon at the time. However, all
later years are counted from the era of Nabonassar. Although the eclipse
dates would originally be expressed in terms of the Babylonian luni-solar
calendar (see section 4.7), Ptolemy invariably specifies the month in terms
of the Egyptian calendar. Numbering days from a fixed epoch was con-
siderably simplified on this latter system. Each Egyptian year contained
12 equal months of 30 days followed by 5 epagomenal ('additional') days.
This fixed year of 365 days was not adjusted to the solar year by interca-
lation so that the first day of the year gradually retrograded through the
seasons, making a complete circuit in 1460 years (the Sothic cycle). Greek
transliterations of the Egyptian month-names - as used by Ptolemy - are
listed in table 4.1.
In quoting days of the month for lunar eclipses, Ptolemy was in the
 4.3 Investigation of the Babylonian lunar eclipses cited by Ptolemy 97

habit of using double dates - e.g. Thoth 18/19. This is because although
the civil day began at sunset in Babylon, according to Egyptian convention
it commenced at the following sunrise (Toomer, 1984, p. 12). As noted in
chapter 3, I have found it convenient for similar reasons to use double
dates on the Julian calendar for all eclipses of the Moon.
Comparison with the many lunar records found on the astronomical
tablets recovered from Babylon (see chapter 6) makes it clear that the
eclipse times quoted by Ptolemy are not in original form (i.e. using
time-degrees) but have been modified to correspond to the Greek method
(equinoctial or seasonal hours). Presumably Hipparchus was responsible
for these reductions. It is a pity that the original measurements are
not preserved; it is likely that some loss of accuracy would occur when
the times were reduced to the Greek system. However, no attempt at
restoration is possible. Only in a single case (523 BC) is there a parallel
inscription on an extant cuneiform tablet (see chapter 6), and even this is
problematical.
Despite the assertion by Diodorus Siculus (Library of History, II, 9)
that the astronomers of Babylon observed from the ziggurat, there is no
confirmatory evidence in other ancient sources. In any case, the tower
is reported by Strabo (Geography, XVI, 1.5) to have been damaged by
Xerxes I after a rebellion in 482 BC, while around 330 BC Alexander the
Great is said to have demolished it (Strabo, loc. cit). Alexander's plans to
rebuild the ziggurat were never carried out. A recent aerial photograph
of the site is shown in figure 4.2.
The geographic co-ordinates of Babylon are: lat. = +32.55 deg, long. =
—44.42 deg. Since the plain in which the site of the city is located is so flat,
there is no need to make any allowance for horizon profile in deducing
the time of sunrise or sunset, etc. Photographs of the horizon taken from
the ruins of the city show that it is remarkably level in all directions (e.g.
Ravn, 1942, plates II, IXb and Xa).

4.3 Investigation of the Babylonian lunar eclipses cited by Ptolemy

The ten Babylonian observations of lunar eclipses which Ptolemy dis-
cusses in his Almagest are all cited as textbook examples to illustrate the
derivation of certain lunar parameters. For instance, several observations
(from 721, 720, 383 and 382 BC) were used by Ptolemy to deduce the
principal lunar anomaly, including correcting errors of calculation made
by Hipparchus. Further data (from 720, 502 and 491 BC) were applied to
the investigation of the Moon's mean motion in latitude, while observa-
tions in 621 and 524 BC yielded a good result (31^ arcmin) for the Moon's
apparent diameter. These various techniques are discussed by Neugebauer
(1975, pp. 81flf.,104 ff., etc.).
98 4 Babylonian and Assyrian records of eclipses

Fig. 4.2 A recent aerial photograph of the site of the Babylonian ziggurat.
 4.3 Investigation of the Babylonian lunar eclipses cited by Ptolemy 99

Each of the Babylonian lunar eclipse observations cited by Ptolemy is

investigated below. In every case, I have given the computed Julian date
and eclipse magnitude, followed by a translation of the text quoted from
Toomer (1984). In some cases it is not clear whether times are expressed
in equinoctial hours (horai isemerinai) or seasonal hours (horai kairikai).
Where necessary I have appended further comments by Ptolemy if these
help clarify the main text. The Julian dates reduced by Toomer from the
historical dates are given in square brackets using negative integers for
years. It will be noted that when compared with the computed Julian
dates, these are invariably accurate. As mentioned in chapter 1, negative
years differ by unity from their BC equivalents.
In each case, Ptolemy only quotes the time of one phase of an eclipse.
This is usually first contact, but on four occasions (in 720, 523, 502 and
490 BC) he implies mid-eclipse (i.e. maximal phase) instead. Fotheringham
(1915), following Nevill (1906b), inferred that the times of all but the first
of these latter events related to first contact. However, since Ptolemy
is our most direct source, there seems little alternative to accepting his
interpretation. For a valuable historical discussion, see Britton (1985).

(1) BC 721 Mar 19/20 (mag. = 1.53)

First the three ancient eclipses which are selected from those observed in
Babylon. The first is recorded as occurring in the first year of
Mardokempad, Thoth [month I] 29/30 in the Egyptian calendar [—720
Mar 19/20]. The eclipse began, it [the report] says, well over an hour after
moonrise and was total. Now since the Sun was near the end of Pisces, and
[therefore] the night was about 12 equinoctial hours long, the beginning of
the eclipse occurred, clearly, A\ equinoctial hours before midnight.
[Almagest, IV, 6; trans. Toomer (1984, p. 191).]
Ptolemy evidently interpreted the phrase 'well over an hour after moon-
rise' as implying an interval of 1^ hours; this will be assumed here. On the
supposition that sunset and moonrise occurred simultaneously, Ptolemy
inferred that the eclipse began l | hours after sunset. However, computa-
tion shows that the Moon would actually rise a significant time (about 0.2
hours) before sunset so that it seems preferable to specifically use moon-
rise as the reference moment. Whether the interval between moonrise and
first contact was expressed in equinoctial or seasonal hours is unimportant
since the Sun was so close to the (vernal) equinox; 1 seasonal hour would
be equal to 1.01 h.

RESULTS
First contact on Mar 19 at 1^ hours after moonrise. LT of moonrise =
17.73 h, hence LT of first contact = 19.23 h, UT = 16.45 h. Computed
TT = 22.46 h, thus AT = 21650 sec.
100 4 Babylonian and Assyrian records of eclipses

(2) BC 720 Mar 8/9 (mag. = 0.12)

... The second eclipse is recorded as occurring in the second year of
Mardokempad, Thoth [month I] 18/19 in the Egyptian calendar [—719
Mar 8/9]. The [maximum] obscuration was 3 digits from the south exactly
at midnight. So since mid-eclipse was exactly at midnight at Babylon ...
[Almagest, IV, 6; trans. Toomer (1984, p. 191).]
Essentially the same record is to be found later in the Almagest (IV, 9)
- Toomer, p. 208.

RESULTS
Maximal phase on Mar 8 at midnight. LT of maximum = 24.00 h on
Mar 8, hence UT = 21.28 h. Computed TT = 2.56 h on Mar 9, thus AT
= 19000 sec.

(3) BC 720 Sep 1/2 (mag. = 0.50)

... The third eclipse is recorded as occurring in the (same) second year
of Mardokempad, Phamenoth [month VII] 15/16 in the Egyptian calendar
[—719 Sep 1/2]. The eclipse began, it says, after moonrise, and the
[maximum] obscuration was more than half [the disk] from the north
... The beginning of the eclipse was about 5 equinoctial hours before
midnight (since it began after moonrise) ...
[Almagest, IV, 6; trans. Toomer (1984, p. 192).]
Ptolemy's rough inference that the eclipse began 'about 5 equinoctial
hours before midnight' will be ignored since this is not based on measure-
ment. However, the observation that commencement did not occur until
after moonrise sets a limit to the value for AT.

RESULTS
First contact on Sep 1 after moonrise. LT of moonrise = 18.52 h, hence
LT of first contact > 18.52 h, UT > 15.58 h. Computed TT = 21.52 h,
thus AT < 21400 sec.
(NB for values of AT > 21400 sec, the Moon would rise already
eclipsed.)

(4) BC 621 Apr 21/22 (mag. = 0.15)

In the fifth year of Nabopolassar, which is the 127th year from
Nabonassar, Athyr [month III] 27/28 in the Egyptian calendar [—620 Apr
21/22], at the end of the 11th hour in Babylon, the Moon began to be
eclipsed; the maximum obscuration was | of the diameter from the south.
Now since the beginning of the eclipse occurred 5 seasonal hours after
midnight...
[Almagest, V, 14; trans. Toomer (1984, p. 253).]
 4.3 Investigation of the Babylonian lunar eclipses cited by Ptolemy 101

A time 5 seasonal hours after midnight corresponds to the end of the

11th (seasonal) hour. Hence the measured time will be taken as exactly
11 seasonal hours after sunset.

RESULTS
First contact on Apr 22 at 5 seasonal hours after midnight. LT of sunset
= 18.48 h, hence length of night = 11.04 h. 1 seasonal hour = 0.92 h, thus
LT of first contact = 4.60 h, UT = 1.60 h. Computed TT = 6.54 h, thus
AT = 17 800 sec.

(5) BC 523 Jul 16/17 (mag. = 0.53)

... Again in the seventh year of Kambyses, which is the 225th year from
Nabonassar, Phamenoth [month VII] 17/18 in the Egyptian calendar
[—522 Jul 16/17], 1 [equinoctial] hour before midnight at Babylon, the
Moon was eclipsed half its diameter from the north. Thus the eclipse
occurred about 1 | equinoctial hours before midnight at Alexandria.
[Almagest, V, 14; trans. Toomer (1984, p. 253).]
It seems evident from the text that the time of maximal phase is
recorded. Since Ptolemy adopted a longitude difference between Alexan-
dria and Babylon of | of an equinoctial hour [IV, 6], it is apparent that
the time of mid-eclipse in Babylon was measured in equinoctial hours.
The extant cuneiform record of this same eclipse seems to contain
predicted rather than observed details (see chapter 6).

RESULTS
Maximal phase on Jul 16 at 1 equinoctial hour before midnight. LT of
mid-eclipse = 23.00 h, UT = 20.05 h. Computed TT = 1.40 h on Jul 17,
thus AT = 19250 sec.

(6) BC 502 Nov 19/20 (mag. = 0.19)

... The second, which Hipparchus too used, occurred in the twentieth
year of that Darius who succeeded Kambyses, Epiphi [month XI] 28/29 in
the Egyptian calendar [—501 Nov 19/20], when 6| equinoctial hours of
the night had passed; at this eclipse the Moon was, again, obscured from
the south ^ of its diameter. The middle of the eclipse was | of an
equinoctial hour before midnight in Babylon (for the length of half the
night was about 6 | equinoctial hours on that date) ...
[Almagest, IV, 9; trans. Toomer (1984, p. 208).]
It is evident from the times given by Ptolemy in the last sentence of
the above quotation that he assumed that the phase which was observed
'when 6^ equinoctial hours of the night had passed' was maximal eclipse.
102 4 Babylonian and Assyrian records of eclipses

RESULTS
Maximal phase on Nov 19 at 6^ equinoctial hours after sunset. LT of
sunset = 17.26 h, hence mid-eclipse = 23.59 h, UT = 20.46 h. Computed
TT = 1.72 h on Nov 20, thus AT = 18 950 sec.

(7) BC 491 Apr 25/26 (mag. = 0.09)

The first eclipse we used is the one in Babylon in the thirty-first year of
Darius I, Tybi [month V] 3/4 in the Egyptian calendar [-490 Apr 25/26],
at the middle of the sixth hour [of night]. It is reported that at this eclipse
the Moon was obscured 2 digits from the south ... For the time of
mid-eclipse was ^-hour before midnight at Babylon, and [hence] \\
equinoctial hours before midnight at Alexandria ...
[Almagest, IV, 9; trans. Toomer (1984, pp. 206-7).]
Use of an ordinal numeral to fix the time (i.e. 'at the middle of the
sixth hour') suggests seasonal hours; this was Ptolemy's normal practice.
However, since the Sun was close to the equinox, the choice of units is
unimportant: half a seasonal hour before midnight would be equal to 0.45
equinoctial hours. A local time of 23.55 h will be assumed.

RESULTS
Maximal phase on Apr 25 at 1/2 seasonal hour before midnight. LT of
mid-eclipse = 23.55 h, hence UT = 20.53 h. Computed TT = 0.37 h on
Apr 26, thus AT = 13 800 sec.

(8) BC 383 Dec 22/23 (mag. = 0.21)

He (Hipparchus) says that these three eclipses which he adduces are
from the series brought over from Babylon, and were observed there; that
the first occurred in the archonship of Phanostratos at Athens in the
month Poseidon, a small section of the Moon's disk was eclipsed from the
summer rising-point [i.e. the north-east] when half an hour of night was
remaining. He adds that it was still eclipsed when it set. Now this moment
is in the 366th year from Nabonassar, in the Egyptian calendar (as
Hipparchus himself says) Thoth 26/27 [-382 Dec 22/23], 5± seasonal
hours after midnight (since half an hour of night was remaining).
[Almagest, IV, 11; trans. Toomer (1984, pp. 211-2).]
Further remarks by Ptolemy indicate that he understood the moment
when 'a small section of the Moon's disk was eclipsed' as meaning first
contact. His statement that this occurred '5^ seasonal hours of night after
midnight since half an hour of night was remaining' implies that the time
before the end of night (i.e. sunrise) was also expressed in seasonal hours.
The fact that the Moon set eclipsed provides independent limits to the
value for AT. In order to determine these, it seems best to ignore the
 4.3 Investigation of the Babylonian lunar eclipses cited by Ptolemy 103

recorded times (in case of possible scribal error) and simply assume that
the Moon set at some time between first and last contact.
NB Phanostratos was archon (chief magistrate) at Athens in the year
383/2 BC. (For a list of the Athenian archons, each of whom held office
for one year, see Bickerman, 1980, pp. 138-9.)

RESULTS
(i) First contact on Dec 23 at \ seasonal hour before sunrise. LT of sunrise
= 7.00 h, hence length of night = 14.00 h. 1 seasonal hour = 1.17 h, thus
LT of first contact = 6.41 h, UT = 3.50 h. Computed TT = 8.43 h, thus
AT = 17 750 sec.

(ii) Moon set eclipsed, (a) First contact: LT of moonset = 6.98 h on

Dec 23, hence for Moon to set eclipsed after first contact, LT < 6.98 h,
UT < 4.07 h. Computed TT = 8.43 h, thus AT > 15 700 sec. (b) Last
contact: LT of moonset = 7.05 h, hence for Moon to set eclipsed before
last contact, LT > 7.05 h, UT > 4.14 h. Computed TT = 10.10 h, thus
AT < 21450 sec. Combining these limits, one obtains 15 700 < AT <
21450 sec. (NB the LT of moonset has been computed iteratively for each
contact, as explained in chapter 3.)

(9) BC 382 Jun 18/19 (mag. = 0.49)

... He (Hipparchus) says that the next eclipse occurred in the archonship
of Phanostratos at Athens in the month Skirophorion, Phamenoth 24/25
in the Egyptian calendar, and that the Moon was eclipsed from the
summer rising-point [i.e. the north-east] when the first hour [of night] was
well advanced. This moment is in the 366th year from Nabonassar,
Phamenoth [month VII] 24/25 [-381 Jun 18/19], about S\ seasonal hours
before midnight.
[Almagest, IV, 11; trans. Toomer (1984, pp. 212).]
Since Ptolemy inferred that the eclipse began 'about 5^ seasonal hours
before midnight', he presumably understood that 'the first hour [of night]'
was also expressed in the same units. As mentioned above, use of ordinal
numerals for times quoted in seasonal hours was his customary style.
Ptolemy thus interprets the clause 'when the first hour [of night] was well
advanced' to mean half an hour after sunset, and this will be assumed.
In addition to the timed observation, which yields an estimate of the
value of AT, the fact the eclipse began soon after sunset may set a useful
limit to AT.

RESULTS
(i) First contact on Jun 18 at \ seasonal hour after sunset. LT of sunset
= 19.15 h, hence duration of night = 9.70 h, 1 seasonal hour = 0.81 h.
104 4 Babylonian and Assyrian records of eclipses

LT of first contact = 19.56 h, UT = 16.51 h. Computed TT = 21.06 h,

thus AT = 16400 sec.

(ii) First contact after sunset. LT of contact > 19.15 h, UT > 16.10 h.
Computed TT = 21.06 h, thus AT < 17 850 sec.
NB for values of AT > 17 850 sec the eclipse would have already begun
before sunset (the Moon would rise several minutes before sunset).

(10) BC 382 Dec 12/13 (mag. = 1.48)

... He (Hipparchus) says that the third eclipse occurred in the
archonship of Euandros at Athens, in the month Poseidon I, Thoth 16/17
in the Egyptian calendar, and that (the Moon) was totally eclipsed,
beginning from the summer rising-point [i.e. the north-east], after four
hours [of night] had passed. This moment is in the 367th year from
Nabonassar, Thoth [month I] 16/17 [-381 Dec 12/13], about l\ hours
before midnight. Now when the Sun is about two-thirds through
Sagittarius, one hour of night at Babylon is about 18 time-degrees. So l\
seasonal hours produce 3 equinoctial hours. Therefore the beginning of the
eclipse was 9 equinoctial hours after noon on the 16th ...
[Almagest, IV, 11; trans. Toomer (1984, pp. 213).]
As Toomer (p. 213n) remarks, the statement that the eclipse began
'after four hours of night' is incompatible with Ptolemy's interpretation
that it commenced 'about l\ hours before midnight'. Ptolemy consistently
assumes l\ hours before midnight in his subsequent argument and it
seems only reasonable to follow him. From Toomer, inspection of the
various extant manuscripts of the Almagest does not clarify this issue
since all give identical readings here.
NB Euandros was archon at Athens in the year 382/1 BC.

RESULTS
First contact on Dec 12 at 2i seasonal hours before midnight. LT of
sunset = 17.03 h, hence length of night = 13.94 h, 1 seasonal hour = 1.16
h. LT of first contact = 21.10 h, UT = 18.10 h. Computed TT = 22.55 h,
thus AT = 16000 sec.

In table 4.2 are summarised the AT results obtained from each individ-
ual observation discussed above. Years are given in terms of the Julian
calendar - using a negative integer rather than BC (see chapter 1). It
should be noted that —719a refers to the first eclipse in the year —719 (i.e.
Mar 8/9), etc. In this table, Ct stands for contact and M for mid-eclipse.
The various values for AT listed in the above table are plotted in figure
4.3. It can be seen from this diagram that with only a single exception
(—490), the results are fairly self-consistent, bearing in mind that - as
 4.4 The astronomical cuneiform texts 105

Table 4.2 AT results from Babylonian lunar eclipse

observations recorded in the Almagest.

Year Ct AT (s)

-720 1 21650
-719a M 19000
-719b 1 <21400
-620 1 17 800
-522 M 19250

-501 M 18950
-490 M 13 800
-382 1 17 750
-382 1 >15 700
-382 4 <21450

-381a 1 16400
-381a 1 <17 850
-381b 1 16000

reported in the Almagest - the measured times are not expressed in their
original form and are further only quoted to the nearest half hour or so.
In particular, the observations in —382 and —381 set rather narrow limits
to AT (15 700 < AT < 17 850 sec) at this specific epoch.

4.4 Historical background to the astronomical cuneiform texts

The history of Babylon from the time of Nabonassar onwards encompasses
several distinct historical periods (for a detailed account, see Oates, 1979).
Soon after the death of this king (733 BC) there followed a century
of Assyrian domination. However, during the rule of Nabopolassar
(Nabu-apla-usur) from 625 to 605 BC, Assyria was eliminated, its capital
of Nineveh being destroyed in 612 BC. Babylon reached its zenith in
the reign of Nabopolassar's son Nebuchadrezzar II (Nabu-kudurra-usur),
when it was the political centre of an extensive empire. Over the following
centuries the city suffered a gradual decline. It was captured by Cyrus
the Great in 538 BC, becoming then part of the large Persian empire.
After more than two centuries of Persian rule, Babylon was annexed by
Alexander the Great in 330 BC. Alexander planned to restore some of
the glory of Babylon by making the city his eastern capital. However,
his untimely death there (in the palace of Nebuchadrezzar II) in 323
BC frustrated these plans. Around 275 BC, Antiochus I ordered much
of the population of Babylon to move to the newly founded metropolis
106 4 Babylonian and Assyrian records of eclipses

30000

V
20000 -
0 ° o
0 sy
Q
Xj
/\
o

10000 --
Key
Individual AT results O
Upper AT Limit V
Lower AT Limit
1 1 I
-800 -700 -600 -500 -400 -300
Year
Fig. 4.3 AT values and limits obtained from the Babylonian eclipse observations
in Ptolemy's Almagest.

of Seleucia, some 100 km to the north. This was a blow from which
Babylon never recovered, although both Antiochus I and his immediate
successors continued to attach special importance to the city. Hellenistic
rule at Babylon continued until 122 BC, when Mithridates II established
Parthian domination there.
Both Diodorus Siculus (II, 9) and Strabo (XVI, 1.5) remark that in
their time (late first century BC) the greater part of Babylon was deserted
and given over to agriculture. About AD 24, merchants from Palmyra
established a trading colony there, but half a century later this was moved
to a site near Seleucia (Oates, 1979, p. 96). The last we hear of Babylon
in the ancient world was in AD 116 when the Roman emperor Trajan
wintered there. He found 'nothing but mounds and stones and ruins' (Dio,
LXVII). Not until relatively recent times was the city rediscovered.
Unlike other great ancient cities such as Nineveh, Babylon was never
finally destroyed by invaders; it gradually died out. Because Ptolemy
wrote several decades after Babylon had ceased to be inhabited, he could
have had no direct contact with the astronomers of that city. This would
explain why he had to rely on Hipparchus for the Babylonian observations
which he used.
After Trajan's visit in AD 116, there is no historical reference to Babylon
for more than a thousand years. Budge (1925, pp. 58 ff.) has given
a fascinating account of the rediscovery of the city, commencing with
Benjamin de Tudela - a twelfth century Jewish rabbi who was a native of
 4.4 The astronomical cuneiform texts 107

Spain. De Tudela journeyed over much of the known world of the time,
visiting the ruins at Babil (as the site was then called) around AD 1173. We
have no definite knowledge of any other traveller reaching Babylon until
1620, when Pietro della Valle collected a number of inscribed bricks from
the site and took them back to his native Italy. Little further was heard
about Babylon for more than a hundred years. However, during the latter
half of the eighteenth century a number of Western explorers reached the
city. These men expressed concern at the extensive pillaging of the site
by workmen from nearby towns and villages for bricks to re-use in new
constructions. Around 1795, the French zoologist Antoine Olivier found
the ruins in utter confusion, further remarking that many neighbouring
towns such as Hillah had been built with bricks from Babylon. About
the same time, Abbe Beauchamp was told by a builder from Hillah that
sometimes the workmen uncovered inscribed cylinders, etc. These were
left among the rubble; only the bricks were salvaged.
Digging for bricks was still continuing a century later when the first
archaeological excavations at Babil began under Sir Austen Layard (in
1851). However, soon afterwards the inhabitants of the area became
aware of the pecuniary value of inscriptions, even when the items were
badly damaged. During the 1870s and early 1880s, numerous clay tablets
found their way to antique dealers in Baghdad. Between 1876 and 1882,
the British Museum purchased virtually all of the available texts from
these merchants via the London dealers Spartali and Shemtob. It was
among this vast collection that the Late Babylonian astronomical texts,
now numbering some 2000 tablets, came to light. At the time, no other
academic institution was concerned to purchase such material. A very
few texts of a similar nature were acquired by other museums such as
the Staatliche Museen, Berlin, and the Morgan Library Collection, New
Haven. However, the British Museum holds at least 98 per cent of the
extant Babylonian tablets devoted to astronomy.
Obviously there is no archaeological context whatever for the Late
Babylonian astronomical texts which were purchased from Baghdad. This
situation was partly remedied as the result of planned excavations at
the site of Babylon during the 1880s. These operations were undertaken
by Hormuzd Rassam on behalf of the British Museum (Rassam, 1897;
Reade, 1986). Some fifty astronomical texts (as well as many inscriptions
of a non-astronomical nature) came to light as the result of his work.
Unfortunately, Rassam did not keep a detailed record of his excavations
so that the precise location in the vast ruins of Babylon where he made
his discoveries is unknown. It appears that no other astronomical tablets
have been unearthed from Babylon since the end of last century, although
in recent years Iraqi archaeologists have undertaken extensive excavations
there.
108 4 Babylonian and Assyrian records of eclipses

Fig. 4.4 (a) For caption see facing page.

4.5 Decipherment of the cuneiform texts

Today, approximately 2000 astronomical texts from Babylon have been

carefully catalogued and many of these are currently undergoing extensive
investigation. These texts are in the form of clay tablets inscribed with
cuneiform writing. Individual signs (which represent syllables) were made
up of triangular wedges, each formed by dipping a pointed reed into moist
clay. The clay was afterwards baked or sun-dried. The name cuneiform
is derived from the Latin for wedge - i.e. cuneus. This name owes its
 4.5 Decipherment of the cuneiform texts 109

Obv.

30

Fig. 4.4 Photograph (a) of an astronomical diary dating from 164 BC together
with a drawing (b) of the same text by T. G. Pinches. (Courtesy: British Museum.)

origin to the Dutch explorer Engelbert Kampfer, who in 1686 copied a

Babylonian inscription. He called the written characters cunatae.
Cuneiform was written from left to right. After one side of a tablet
was fully inscribed, it was turned over in a vertical direction and the
inscription was continued - starting at the top left corner of the reverse
side.
When the astronomical texts were first dug up at the site of Babylon,
many were badly damaged - either accidentally or deliberately - and
110 4 Babylonian and Assyrian records of eclipses

most now consist of mere fragments. As the pecuniary value of the

tablets began to be appreciated, the local inhabitants found that dealers
in nearby Baghdad offered more money for a number of broken pieces
than for a single large text, with the obvious results. By painstaking work
at the British Museum, largely by Theophilus G. Pinches (an assistant
keeper) between 1895 and 1900, many pieces acquired by the museum
were joined together again; an example is illustrated in figure 3.9a. Most
of the inscriptions were hand copied by Pinches, who was an accomplished
scribe. Figures 4.4a and 4.4b show a photograph of a tablet and a Pinches
copy side by side. The original text dates from 164 BC and contains one
of the earliest known records of Halley's comet.
When Pinches left the service of the British Museum in 1900 his numer-
ous sketches were stored in the archives, where they remained untouched
for more than 50 years.
The first decipherment of an astronomical cuneiform text was made by
Father Joseph Epping, SJ in 1881 in collaboration with Father Johann N.
Strassmaier, SJ (Neugebauer, 1952, p. 98). This work was based partly on
existing translations of non-astronomical texts made in the mid-nineteenth
century (for details, see for example Budge, 1925, pp. 39 ff.) and partly
as the result of numerous lunar and planetary calculations by Epping
himself. He was able to investigate Babylonian methods of predicting
lunar phenomena and he correctly identified the names of the planets
and zodiacal constellations as well as the meaning of various Babylonian
astronomical terms.
During the 1880s and 1890s, Strassmaier systematically hand-copied
thousands of Late Babylonian tablets in the British Museum. These texts
were of various kinds, including some on astronomy. Strassmaier sent
annotated copies of his drawings of astronomical texts to Epping, who
was based in Quito, and this led to further research. After Epping's
death in 1894, his work was continued by Father Franz X. Kugler, SJ. By
extensive studies based on sketches supplied by Strassmaier, Kugler (1907,
1909-24, 1913-14) built up a detailed picture of Babylonian mathematical
astronomy. This work was continued by Otto Neugebauer (e.g. 1955) and
Father Johann Schaumberger, SJ.
Until 1948, relatively little interest had been shown in Babylonian
observational astronomy because few relevant texts were known. However,
in that year Abraham Sachs (1948) attempted the first classification of
observational and other non-mathematical cuneiform texts based on the
then known sample of only about 20 tablets of this kind. In 1953, when
Sachs spent a year at the British Museum, he became aware of the full
extent both of the texts themselves and also of the sketches made by
Pinches. Not long afterwards, Sachs and Schaumberger (1955) published
many of the drawings made by Pinches (and also some by Strassmaier) but
 4.6 Origin and classification of the astronomical texts 111

without translations. In their book entitled Late Babylonian Astronomical

and Related Texts, Sachs and Schaumberger cited both the British Museum
(BM) number and their own reference number for each tablet, the latter
usually known as LBAT. Where appropriate, both of these designations
will be used below.
When possible Sachs and Schaumberger assigned a date for each in-
scription. Some dates were still preserved on the tablets, but they derived
many further dates by retrospective calculation of the various planetary
and lunar observations which they contained. Today, some 1500 texts
devoted to observation and other aspects of non-mathematical astronomy
have been catalogued. However, less than half of these have been dated;
numerous small fragments - some no more than 3 cm across - remain
undatable.
Judging from the content of the extant astronomical texts, it may be
conservatively estimated that less than ten per cent of the original material
has so far been found. Whether the remainder was destroyed over the
centuries when the site of Babylon was being pillaged - or whether
further astronomical tablets still lie buried, awaiting future discovery - is
an unanswered question. Nevertheless, cataloguing, dating, transliteration
and translation of the extant material has proved an enormous task,
which is still unfinished. Sachs spent some 30 years on this project and
since his death in 1983 Hermann Hunger of the University of Vienna
has continued his work. Photographs, transliterations and translations of
many texts have already been published (Sachs and Hunger, 1988, 1989,
1996). However, it will be several years before completion is in sight.

4.6 Origin and classification of the Late Babylonian astronomical texts

In principle, the astronomical texts which the British Museum acquired
from antique dealers in Baghdad could either have originated from Baby-
lon or one of the other ancient sites in Southern Iraq such as Borsippa
or Sippar. That they in fact came from Babylon can be established from
the following features: (i) the very obvious continuity of style and content
between these and the tablets excavated at Babylon by Rassam; (ii) the
frequent mention of events occurring in the city of Babylon in the inscrip-
tions (other cities are seldom alluded to); (iii) the specifically Babylonian
deities (e.g. Bel and Beltija) to which the introductory invocations are
addressed; and (iv) the character of the few personal names which occur
in the colophons. For details, see for example Sachs (1948, 1974), and also
the many translations published by Sachs and Hunger (1988, 1989, 1996).
There is, in fact, little evidence that any of the Late Babylonian texts on
non-mathematical astronomy in the British Museum originated anywhere
but Babylon itself.
112 4 Babylonian and Assyrian records of eclipses

The period covered by the tablets is towards the end of the active life
of Babylon, and hence they have become known as the Late Babylonian
astronomical texts. Most dated inscriptions originate from between 400
and 50 BC, but several texts contain older material - one eclipse prediction
dating from 731 BC (first year of Ukin-zer). Huber (1973, p. 1) estimates
that the information on one damaged text must have originally extended
as far back as some time between 750 and 740 BC It thus seems quite
likely that the records could have extended as far back as the reign of
Nabonassar, as Ptolemy asserted.
The most recent surviving astronomical texts, dating from AD 61/62
and 74/75, are almanacs prepared from then contemporary observations
(Sachs, 1976). The latter tablet, now in Dropsie College, Philadelphia, is
also the latest datable text which is written in cuneiform.
The astronomical texts reveal that through centuries of both greatness
and decline, celestial observations continued at Babylon on a regular
basis with little change of pattern. Just when astronomical activity ceased
cannot be established with certainty, but it must have been at some time
between AD 75 and 116 (when Trajan visited the deserted ruins).
In all, the Late Babylonian astronomical texts may be divided into
five main groups: diaries, tables of specific phenomena, 'goal-year texts',
almanacs and texts devoted to mathematical astronomy. Observational re-
ports can be found on tablets in the first three categories. The main motive
for stargazing in Babylon was astrological. In essence, the astronomers
utilised their observations to produce almanacs which in turn formed the
basis of astrological prognostications. It is noteworthy that the casting of
horoscopes originated in Babylon around 400 BC (Sachs, 1967). Celestial
phenomena which were regularly noted include conjunctions of the Moon
with planets and selected stars, conjunctions of planets with one another
with these same stars, the time of moonrise and moonset (measured rela-
tive to sunrise or sunset) at both new and full Moon, eclipses, the heliacal
rising or setting of planets and Sirius, and equinoxes and solstices.
The Babylonian name for the Sun was samas and for the Moon was sin.
An eclipse was identified by the term AN-KUio- (Many different signs
were pronounced as KU.)
The following detailed account of the partial lunar eclipse of Apr 10/11
in 80 BC - recorded on both the obverse and reverse of a British Museum
tablet devoted to this event (BM 33562A; LBAT 1445) - gives some
indication of the variety of information which the Babylonian astronomers
noted:

Year 168 (Arsacid), that is year 232 (SE), Arsaces, king of kings, which
is in the time of King Orodes (I). Month I, night of the 13th, moonrise to
sunset 5,50 (= 5 | ) deg, cloudy (?), measured (?). 5 deg before \i Her
 4.6 Origin and classification of the astronomical texts 113

culminated, lunar eclipse, beginning on the south-east side. In 20 deg of

night it made 6 fingers. 7 deg duration of maximal phase, until it began to
become bright. In 13 deg from south-east to north-west, 4 fingers lacking
to brightness, it set [...]. During this eclipse, north wind (?) went. During
this eclipse, Jupiter, Saturn and Mars stood there. When becoming bright,
Saturn set. The other planets did not stand there. (The Moon was) \ cubit
behind a Lib, 6 fingers toward the south beneath. (Began) at 40 deg before
sunrise. On the 13th, moonset to sunrise 1 deg, mist, measured.
[Trans. Huber (1973 pp. 75-76) - with minor emendations
by the present author.]
In the above text, the year is given in terms of both the Arsacid era
(247 BC) and the Seleucid era (SE: 311 BC). The eclipse was only partial;
when the whole of the Moon (or, rarely, the Sun) was obscured the term
TIL ('total') was used. The various units (deg, cubits, fingers) used will be
discussed below (section 4.10).
The names of the five planets were normally cited in the following
order: MUL-BABBAR (Jupiter), dele-bat (Venus), GU 4 -UD (Mercury),
GENNA (Saturn) and AN (Mars). During the Seleucid period, this
sequence was strictly followed.
Although there was considerable interest in cyclical events - e.g. lunar
and planetary movements and eclipses - little concern was shown for ce-
lestial bodies appearing at irregular intervals such as comets and meteors.
Brief records of Halley's comet have been found on tablets dating from
164 and 87 BC (Stephenson et al.9 1985; Stephenson and Walker, 1985).
Of the various types of astronomical text, the diaries represent the most
original material. They contain a day to day (and night to night) account
of celestial and meteorological observations, each text typically covering a
period of six or seven months. Historical events are also frequently noted
- e.g. the accessions and deaths of kings and major events in Babylon - as
well as changes in the level of the adjacent River Euphrates and prices of
various commodities such as dates and wool. Virtually all of the datable
diaries which were written prior to 164 BC have been published by Sachs
and Hunger (1988, 1989, 1996).
The earliest surviving diaries originate from 652 and 568 BC, the latter
being fairly well preserved. Sachs (1974) gives an interesting bar diagram
showing the years from which fragments of diaries are preserved; this
is reproduced in figure 4.5. Despite recent research by Hunger on addi-
tional diaries, this diagram still closely represents the present temporal
distribution of texts.
During the Hellenistic period (fourth century BC onwards) extensive use
was made of the material in past diaries by the Babylonian astronomers. It
would thus appear that copies of diaries extending over several centuries
were kept on file. Using these primary sources, the astronomers were
114 4 Babylonian and Assyrian records of eclipses

700 -600 -500 -400 -300 -200 -100

690 -590 -490 -390 -290 -190 -90

680 -580 -480 -380 -280 -180 -80

670 -570 -470 -370 -270 -170 -70

660 -560 -460 -360 -260 -160 -60

650 -550 -450 -350 -250 -150 -50

640 -540 -440 -340 -240 -140 -40

630 -530 -430 -330 -230 -130 -30

620 -520 -420 -320 -220 -120 -20

610 -510 -410 -310 -210 -110 -10

Fig. 4.5 Bar diagram showing the years from which fragments of diaries are
preserved (after Sachs, 1974). Each bar represents a fraction of a year covered.

able to compile: (i) tables listing occurrences of a particular phenomenon

(e.g. eclipses or planetary data) and (ii) 'goal-year texts'. The tables often
extended over several centuries, eclipses usually being listed at 18-year
intervals (i.e. the saros period). For example, Huber (1973, p. 1) gives
the following description of one such series of texts - of which the tablets
numbered BM 32238, 45640, 35115, 35789 and 32234 (LBAT 1414, 1415,
1416, 1417 and 1419) are surviving fragments:
A big, tightly organised compilation, originally covering 24 saros cycles,
that is 24 x 38 = 912 eclipse possibilities (i.e. including both observations
and predictions) or 24 x 18 years. Extending over several tablets, it must
have started not earlier than —749 Sep 1 and not later than —739 Mar 20,
and must have ended between —316 Dec 13 and —307 Jul 9.

Some texts cite every eclipse observation and prediction over a selected
short interval - e.g. BM 38462 (LBAT 1420) covers the period from 604
BC to 576 BC. However, as in the case of the eclipse of 80 BC cited above,
other tablets are entirely devoted to a single eclipse.
Goal-year texts represented an intermediate stage in the preparation of
almanacs. On these tablets were assembled observations from past years
which were to form the basis of predictions for a selected future year
 4.7 The Babylonian calendar 115

(the goal-year). Observations would include eclipses 18 years before the

required year, Venus data from 8 years previously and so on. Very few
translations of goal-year texts have so far been published. For sample
transliterations and translations, dealing only with planetary observations,
see Stephenson and Walker (1985, pp. 34-36).
In the study of Babylonian lunar eclipse records, the tables of specific
phenomena - namely those devoted to lunar obscurations - provide a
valuable supplement to the material in the diaries since so many diaries
are missing. In particular, several tables of this kind contain observations
made between about 700 and 400 BC, a period from which very few
diaries are preserved. By chance, no tables of solar eclipses are extant
and there are no known records of these events from before 370 BC. In
later centuries, goal-year texts provide a further source of both lunar and
solar eclipse observations. However, at all periods, surviving lunar eclipse
reports are much more frequent than their solar counterparts.
The compilations of Sachs and Hunger (1988, 1989, 1996) and also Hu-
ber (an unpublished manuscript, produced in 1973 and freely circulated)
have provided transliterations and translations of most of the available
eclipse records. Sachs and Hunger's publications are so far exclusively
devoted to the diaries. Huber's manuscript is specifically concerned with
eclipses, which he extracted from diaries, eclipse tables and goal-year texts.
His work contains many more eclipse observations than the publications
of Sachs and Hunger. However, he relied mainly on the texts for which
drawings were published by Sachs and Schaumberger (1955) so that his
list, although invaluable, contains several important omissions.

4.7 The Babylonian calendar

In common with many ancient calendars, the Babylonian calendar was
luni-solar. The year began with the month Nisannu around the time of
the spring equinox. Until the late fourth century BC, years were counted
from the accession of each ruler, but from 311 BC, the start of the reign
of Seleucus I (known as the Seleucid era - customarily abbreviated to SE),
a continuous system of numbering was adopted. This remained in use at
Babylon for the rest of the period covered by the astronomical texts. The
precise date of the Seleucid era adopted in Babylon was 1 Nisannu (i.e.
Apr 3) in 311 BC However, on the Macedonian calendar it was some 6
months earlier (Dios 1, equivalent to Oct 8 in 312 BC). Even today some
Christian churches in Egypt and Iraq still number years in terms of the
Seleucid era; for medieval examples see chapters 11 and 12. In Babylonian
texts the Arsacid era is occasionally used. Of Iranian origin, the Arsacid
dynasty was established in 247 BC
Most years contained 12 lunar months, each of either 29 or 30 days,
116 4 Babylonian and Assyrian records of eclipses

Table 4.3 Babylonian month-names and their abbreviations as

found in the astronomical cuneiform texts.
Number Name Abbrev Number Name Abbrev
I Nisannu bar VII Tashritu du
II Ayaru VIII Arahsamnu apin
III Simanu sig IX Kislimu gan
IV Du'uzu su X Tebetu ab
V Abu izi XI Shabatu ziz
VI Ululu kin XII Addaru se

making a total of about 354 days. Names of the Babylonian months and
their one-syllable abbreviations as found in the astronomical texts are
listed in table 4.3. In translations, it is customary to give Roman numerals
rather than full month-names.
In order to keep the calendar roughly in step with the seasons, every
two or three years an extra month was added. This intercalary month
was placed immediately after either the 6th or 12th month. (There is
an interesting parallel here with modern leap-seconds, which are always
inserted on Jun 30 or Dec 31.) At first, intercalation in Babylon was
somewhat haphazard, but around 400 BC a regular scheme based on the
Metonic cycle was adopted. Intercalary months are denoted by VI2 or
XII 2 .
At Babylon (as for several other early civilisations), each day began at
sunset - normally a well-defined moment. Towards the close of the 29th
day of a month, a watch was kept for the young crescent Moon low in the
western sky just after sunset. If the crescent was sighted, the new month
was regarded as having just begun; otherwise an extra day was added
to the month. When cloud prevailed, calculation superseded observation
to determine whether or not the crescent should have been visible. Such
calculations seem to have been fairly successful, although the rules are
not known in detail. It has not been ascertained whether, in view of the
more advanced astronomy of later centuries, calculation ever completely
replaced observation for fixing the start of each month.
From an extensive study of the Babylonian calendar, involving detailed
investigation of intercalary months and also the visibility of the crescent
Moon, Parker and Dubberstein (1956) were able to produce comprehensive
tables for the conversion of Babylonian dates to the Julian calendar over
the entire period from 626 BC to AD 75. These tables are highly reliable,
errors seldom exceeding a single day; they have proved very valuable in
the present investigation.
 4.8 Techniques used in dating observational tablets 111

4.8 Techniques used in dating observational tablets

On a well-preserved astronomical diary, the date is typically given in the
first line of the obverse and sometimes in the last line of the reverse. For
example a diary from 247 BC, which now consists of four joined pieces
(BM 32889 + 32967 + 41614 + 41618; LBAT 271, 273 and 272) begins
and ends as follows (Sachs and Hunger (1989, pp. 59-65)):
[First line] Diary for the year 65 [Seleucid], king Antiochus. Month I...
[Last line] Diary from month I to month VI, year 65, [king] Antiochus.
This covers the period from Apr 15 to Oct 8 in 247 BC. For most diaries,
the date is broken away and here it is necessary to resort to computations
based on the planetary and lunar information which the diary contains.
As noted above, much pioneering work in dating damaged texts was done
using this method by Sachs and Schaumberger (1955). In recent years,
I have employed similar techniques to date a number of further diaries
based on transliterations and translations supplied to me by Hermann
Hunger. The following brief comments are based on my experiences.
Often it is necessary to begin with a date range of several hundred
years, as derived from historical and linguistic considerations. Calculation
of the planetary phenomena is best undertaken first and this may indicate
only a very few potentially viable dates, often widely spaced. Afterwards
using the lunar observations it may prove possible to fix a unique date.
Generally speaking, if a single date cannot be fairly quickly derived in
this way no amount of calculation will probably help; the text evidently
contains too little information to allow a unique solution. For examples
of the techniques used, see Stephenson and Walker (1985).
Goal-year texts usually present few dating problems since in addition
to the goal-year each individual eclipse entry carries a separate date which
is 18 years prior to the reference year. Tablets containing lists of lunar
eclipses at 18-year intervals can be fairly easily dated - even if a text
is badly damaged; one or two fully preserved entries or the mention
of the accession or death of a ruler can provide the key to the whole
sequence. If a tablet devoted to a single eclipse is only partially preserved,
the problems of dating are more serious. However, if the position of the
eclipsed Moon relative to a reference star is reported - as in a text from
80 BC (see section 4.7) - it may well prove possible to derive a unique
date by comparing observation with computation. For this purpose a
preliminary value for AT may be utilised since the motion of the Moon
relative to the background stars is fairly slow.
It is important to state here that none of the Babylonian eclipse obser-
vations investigated below have been dated by comparing the observed
time-interval between sunset or sunrise and first contact with the inter-
val computed using a provisional value for AT. Such a technique would
118 4 Babylonian and Assyrian records of eclipses

merely be an example of the 'identification game' (cf. Newton, 1970, p. xiv,

etc.). If an independent date cannot be derived, the observation will not
be considered.

4.9 Units of time

The principal unit of time adopted by the Babylonian astronomers was
the US. It is usually asserted that one US corresponded to four minutes
of time, so that there were precisely 15 of these units in an hour. A
larger measure, the beru, was equal to 30 US and was thus equivalent
to two hours. (For these definitions, see for example Neugebauer (1955,
vol. I, p. 39). The US was subdivided into 60 GAR (sometimes read as
NINDA). Since the US was the time-interval for the celestial sphere to
turn through 1 deg, an appropriate translation of the term is 'time-degree',
although it is customary to abbreviate this to 'degree'. Use of these same
units also spread to Greece, where they were known as chronoi isemerinoi
('equatorial times'), also frequently rendered as 'time-degrees'.
It should be noted that the Babylonians employed separate units for
measuring the angular separation between two celestial bodies; these were
the cubit (KUS), equivalent to about 2 or l\ deg and the finger (SI),
one cubit being divided into 24 fingers (Neugebauer, 1955, vol. I, p. 39).
As discussed in chapter 3, fingers were also used in expressing eclipse
magnitude, but here the definition is quite separate, corresponding to
one-twelfth of the diameter of the luminary obscured.
Eclipse contacts were systematically timed in terms of US relative to
sunrise or sunset, depending on which was nearer. Probably a water clock
was used for this purpose (Neugebauer, 1947) although little is known
regarding the practicalities of time-keeping in Babylon. Eclipses were never
timed to better than the nearest US, but the time of moonrise or moonset
relative to sunrise or sunset was often expressed to the nearest 10 GAR.
In response to a query by Dr C. B. F. Walker of the British Museum,
a recent investigation by Stephenson and Fatoohi (1994b) based on the
recorded durations of lunar eclipses by Babylonian astronomers demon-
strated that the US was indeed of fixed duration, and equal to ^ hour.
It was concluded that down the centuries covered by the Late Babylonian
astronomical texts there is no evidence of changes in the definition of the
unit or of any seasonal variation. Lunar eclipses were made the subject
of this investigation since (unlike in the case of solar obscurations) the
duration of any phase is independent of AT. Hence it is a relatively sim-
ple matter to compare a measured interval with its computed equivalent.
Stephenson and Fatoohi investigated a series of 60 lunar eclipse durations,
timed to the nearest degree. These covered the period from about 560 to
50 BC Although a few rather earlier measurements are preserved on the
 4.10 'Lunar sixes' 119

-600 -500

Fig. 4.6 Constancy of the US during the Late Babylonian period. (Stephenson
and Fatoohi, 1994b.)

Babylonian texts, they are only rounded to the nearest 10 degrees. From
the selected set of data, Stephenson and Fatoohi derived a mean result of
1 hour = 15.6 + 0.4 US, which effectively includes the accepted equivalent
of 15.0 US. It is noteworthy that the more recent observations (after 300
BC) yielded a result of 15.3 + 0.3 US.
A graph showing the individual results obtained from each observation
for the number of US in an hour shows negligible variation down the
centuries - see figure 4.6. In particular, there is no evidence that the
definition of the unit was ever revised during the period covered by the
diagram. (NB in this diagram, M/C stands for measured interval divided
by computed interval.)

4.10 'Lunar sixes'

Each lunar month the Babylonian astronomers determined six time-
intervals, two around new Moon and four around full Moon. These
intervals, appropriately termed by Sachs 'lunar sixes', were quoted to the
nearest degree. They are as follows:

(i) na: Sunset to moonset when the young crescent Moon was first
sighted; this began the month.
120 4 Babylonian and Assyrian records of eclipses

(ii) ME: Moonrise to sunset on the last evening that the almost full
Moon rose before sunset.
(iii) GE6: Sunset to moonrise on the following evening.
(iv) SU: Moonset to sunrise on the last morning that the almost full
Moon set before sunrise.
(v) na: Sunrise to moonset on the following morning.
(vi) KUR: Moonrise to sunrise on the last morning before conjunction
that the crescent Moon was visible.
Where possible, it seems that these intervals were measured, but when
cloud prevented observation the Babylonians calculated them with fair
accuracy. Lunar eclipse reports often specify the ME (moonrise to sunset)
and na (sunrise to moonset) values - respectively at the beginning and
end of the entry. These results may be helpful in fixing the exact date of
an eclipse since the intervals change by some 10 deg from one day to the
next.

4.11 Observation and prediction of eclipses

The various Babylonian non-mathematical texts (diaries, goal-year texts
and tables) contain both observations and (unsuccessful) predictions of
eclipses. If an eclipse was successfully predicted - meaning that it was
observable - this fact is never stated; the texts contain scarcely any com-
parisons between theory and observation. Although the main emphasis in
this chapter will be on observed eclipses, it is necessary to comment also
on the basic form in which predictions are recorded and to discuss how
such accounts differ from observational reports.

4.11.1 Observations
Descriptions of observed eclipses usually contain the more detailed in-
formation. In addition to the time of first contact (expressed relative to
sunrise or sunset) the following details are often given: approximate posi-
tion angle of first and last contact; time interval between first contact and
greatest phase; maximum degree of obscuration of the Sun; time interval
between greatest phase and last contact; whether the Sun or Moon rose
or set whilst eclipsed; wind direction; and which planets were above the
horizon during the eclipse. Ordinarily, the term AN-KU samas or AN-
KU sin indicates a predicted eclipse, whereas the opposite order samas
AN-KU or sin AN-KU denotes an observed event. However, there are
occasional exceptions to this rule, perhaps resulting from scribal errors.
 4.11 Observation and prediction of eclipses 121

The report of the solar eclipse of Mar 14 in 190 BC is particularly

detailed and contains most of the above features. The main description
is found on a tablet specifically devoted to the event (BM 33812, LBAT
1438) which was excavated by Rassam. A badly damaged diary (BM
32951, no LBAT number) duplicates some of this information (Sachs and
Hunger, 1989, pp. 314-15). A translation of the first of these inscriptions
is as follows:

Year 121 (SE), King An(tiochus), month XII, 29 solar eclipse beginning
on the north-west side. In 15 deg day [...] over a third of the disk was
eclipsed. When it began to become bright, in 15 deg day from north-west
to east it became bright. 30 deg total duration. [During this eclipse], east
(wind) went. During this eclipse [...], Venus, Mercury and Saturn [stood
there]. Towards the end of becoming bright, Mars rose (?) The other
planets did not stand there. (Began) at 30 deg (= 1) beru after sunrise.
[Trans. Sachs and Hunger 1989, p. 315).]
The reference to the year on the Seleucid era establishes that this event
occurred during the reign of King Antiochus III. There is no suggestion in
the above text that any of the planets mentioned were actually seen during
the eclipse. Since the phase was only partial, the loss of daylight would
be small. The Babylonian astronomers were in the habit of calculating
which planets, and also whether the bright star Sirius, would be above the
horizon during an eclipse. All the remaining data - including, of course
the wind direction - are apparently based on observation.
When converted to the Julian calendar, the above date (BC 190 Mar
14) proves to be in exact accord with that of a tabular solar eclipse (e.g.
as listed by von Oppolzer, 1887). Such accuracy is characteristic of all
Late Babylonian observations of both solar and lunar eclipses for which
a date is well preserved. On this occasion, sunrise seems to have been
clearly defined. However, when heavy cloud prevailed around sunrise or
sunset, a substantial error in fixing the time of this moment (and hence
the local time of the start of the eclipse) might have been incurred.
Comparison between measured lunar eclipse durations expressed to the
nearest degree and their computed intervals shows that timing errors may
be significant. Employing a mean fit to the AT results obtained from these
data, Stephenson and Fatoohi (1993) showed that such discrepancies aver-
aged 8 deg for intervals up to about 6 hours. Evidently the timing devices
(e.g. water clocks) of the period were crudely designed and presumably
did not incorporate a constant head device. Although these deficiencies
set a limit to the accuracy with which AT can be obtained from individual
Babylonian timed observations, the substantial number of data available
is a major compensating factor.
122 4 Babylonian and Assyrian records of eclipses

4.11.2 Predictions
Reports of eclipse predictions give little more than the date and the
expected time of first contact together with a comment whether the eclipse
was 'watched for but not seen' (ki PAP NU IGI) or merely 'passed by'
(sa DIB). As a general rule, if the former expression was used, the eclipse
was anticipated to occur when the luminary was above the horizon; in the
case of an eclipse which 'passed by' the converse was expected. Examples
are as follows:

(1) BC 731 Apr 9

1,50 Year 1 Ukin-zer, month I, [lunar eclipse] which passes (sa DIB).
(Began) at 1,0 (i.e. 60) deg after sunrise.
[BM 35789 (= LBAT 1414): Obv. col. I; trans. Huber (1973, p. 2).]
Ukin-zer (Nabu-mukin-zeri) reigned from 731 to 729 BC This record
is found on a list of lunar eclipses at 18-year intervals, so that the term
for lunar eclipse is not included in the entry. The number 1,50 in the first
line is unexplained. This eclipse was correctly predicted for the hours of
daylight at Babylon and thus would be expected to be invisible there. (NB
I have given only a single Julian date in this instance - i.e. Apr 9 - since
the eclipse was anticipated during daylight.)

(2) BC 247 Sep 7

[D]iary for the year 65 (SE), king Antiochus... [month V]. The 28th, 74
deg after sunrise, solar eclipse (at) 5 months' distance; when I watched I
did not see it (ki PAP NU IGI).
[BM 32889 + 32967 + 41614 + 41618 (= LBAT 271 + 272 + 273, etc.),
Rev. line 4; trans. Sachs and Hunger (1989, p. 63).]
This event took place during the reign of Antiochus II. Calculations
show that the eclipse, although correctedly expected in the daytime, would
only be visible far to the north of Babylon.
With only a single exception (when cloud prevented observation in
125 BC), no explanation is given in the preserved inscriptions for the
failure of an anticipated eclipse to become visible. However, for such an
early epoch, the Babylonian astronomers were remarkably successful in
their predictions. Investigation of the extant records of solar obscurations
reveals that virtually all of their expected eclipses would be visible from
some part of the terrestrial surface - often remote from Babylon. For
lunar eclipses, the success rate was similar, although about one-third of
their predicted eclipses were actually only penumbral and thus would
very likely pass unnoticed, even if the Moon was above the horizon at the
time. Of course the astronomers themselves could only measure success by
events that were observable at Babylon, representing only a small fraction
 4.11 Observation and prediction of eclipses 123

of the quota for the Earth as a whole. Nevertheless, they would know
on which dates in a year to keep watch for possible eclipses, even though
they could never be sure that their predictions would prove reliable.
Predictions were based on the angular distance of the Moon from either
node at conjunction with the Sun (Aaboe et a/., 1991). The Babylonian
astronomers were well aware that eclipses of the same type - whether
observable or not - tended to recur at intervals of five or six lunar months
on account of the motion of the Sun from one node to the other dur-
ing this time. However, it is not certain just when the Babylonians first
understood the cause of eclipses. Surviving inscriptions do not mention
how eclipses were produced, but the empirical rules of prediction did
not require a physical explanation. Knowledge of the cause of eclipses
may well have passed to the Babylonians during the Hellenistic period
when there was much interaction between Greek and Babylonian astron-
omy.
It will be noted that for the observed solar eclipse of 190 BC the
time of first contact is cited at the end of the account. This practice
was very common for both solar and lunar eclipses. That the quoted
time in such cases is indeed based on observation (rather than a mere
repetition of the originally predicted time) would seem evident from the
fact that the reported durations of the various phases were undoubtedly
measured. Durations are never mentioned in entries which are concerned
only with (unsuccessful) predictions. However, further discussion seems
desirable.
In order to resolve this issue, I have compared (Stephenson, 1997) the
mean error in the recorded times of first contact for both observed and
predicted eclipses listed by Sachs and Hunger (1988, 1989, 1996) and
Huber (1973). As mentioned above, predictions were identified by the use
of either of the terms 'passed by' or 'watched for but not seen' in the
records. In this investigation, I used the preliminary AT formula derived
from a mixture of early observations by Stephenson and Morrison (1984).
This may be expressed as follows:
AT = +1360 + 320t + 44.3t2, (4.1)
where t is measured in Julian centuries from 1800.0. This equation
is adequate for most historical purposes, although refinements will be
made in chapter 14 (see also Stephenson and Morrison, 1995). When a
predicted lunar eclipse was only penumbral, I have made computations
for the moment when the umbral shadow made its closest approach to
the Earth, since there was no true first contact. In the case of a solar
eclipse where the lunar shadow passed fully to the north or the south of
Babylon, I computed for the moment when the shadow made its closest
approach to this site.
124 4 Babylonian and Assyrian records of eclipses

In the above survey, about 50 observations of lunar and solar eclipses

were analysed, giving a result for the mean error in the time of first
contact of 8 deg (see also Stephenson and Fatoohi, 1993). By comparison,
investigation of records of solar eclipse predictions (26 in number) yielded
a mean error as large as 60 deg, while for predicted lunar obscurations
(28 in number) the corresponding result was nearly 40 deg. In both sets
of predictions, the scatter proved to be very large. The implications are
clear: whenever possible, the Babylonian astronomers replaced predicted
times of first contact by observation. This conclusion will be assumed in
all subsequent AT investigations in chapters 5 and 6.
Since observed eclipses were timed relative to sunrise or sunset, it
would seem that when a solar obscuration was expected, the Babylonian
astronomers would start timing as soon as the Sun rose and continue either
until an indentation in the solar limb was observed or until sunset if no
eclipse materialised. In the case of a lunar eclipse, timing would commence
at sunset. When an observable solar eclipse occurred significantly after
noon, the time of onset was measured relative to sunset rather than sunrise,
the converse being the case for a lunar eclipse beginning after midnight.

4.12 Assyrian records of solar and lunar eclipses

Many allusions to both solar and lunar eclipses are found among the div-
ination reports to Assyrian kings around 700 BC. These reports, inscribed
in cuneiform on clay tablets, originate from the royal archives at Nineveh
(lat. = 36.40 deg, long. = —43.13 deg). This city, the last capital of Assyria,
was destroyed by a coalition of Babylonians, Medes and Scythians in 612
BC. Most of the divination texts recovered from the ruins are now in the
British Museum. The Assyrian rulers employed specialists in divination
to record and interpret unusual phenomena of various kinds. Celestial
omens included eclipses, conjunctions of the Moon with planets and stars,
planetary movements, haloes, meteors and comets.
The astrological reports, which cover the period from 709 to 649 BC,
were first investigated by Thompson (1900). A full transliteration, trans-
lation and commentary has recently been published by Hunger (1992).
Many of the allusions to eclipses (identified by the term AN.KI) relate
to attempts at prediction. As most eclipse omens were unfavourable, a
successful prediction might give an opportunity to take action against
whatever danger was expected. An interesting example of a full Moon
when a lunar eclipse was not anticipated is as follows. This is recorded by
the Babylonian astrologer Tab-silli-Marduk.

'[In Iy]yar (month II)... the night of the 14th day, is the [da]y of the
watch (to be held), and there will be no eclipse. I guarantee it seven times,
 4.12 Assyrian records of solar and lunar eclipses 125

an eclipse will not take place. I am writing a definitive word to the king.'
From Tab-silli-Marduk, nephew of Bel-nasir.
[Trans. Hunger (1992, p. 251).]
Few actual observations are noted in the astrological texts, although
prognostications such as the following give such detailed astronomical
circumstances as to imply real observations:
'If the Sun at its rising is like a crescent and wears a crown like the
Moon: the king will capture his enemy's land; evil will leave the land, and
(the land) will experience good...' [From Ras]il the older, servant of the
kmg
' [Trans. Hunger (1992, p. 220).]
The above eclipse has been identified by Parpola (1983) as that of
BC 669 May 27. Rasil was a Babylonian scribe so that the place of
observation may be presumed as somewhere in Babylonia.
Among clearly expressed observations among the Assyrian astrological
reports, the most detailed relates to the solar eclipse of BC 657 Apr
15 (Parpola, 1983). This event is recorded on a badly damaged British
Museum tablet:
'On the 28th day, at 2\ double hou[rs of the day... ] in the west [... ] it
also cover[ed...] 2 fingers towards [...] it made [an eclipse], the east wind
[...] the north wind ble[w. This is its interpretation]...' From [Akkullanu].
[Trans. Hunger (1992, p. 63).]
Akkullanu, the writer of a number of surviving texts, was an Assyrian
scribe. Since the time is only quoted to the nearest hour and the place of
observation is somewhat uncertain, the record is only of minimal value
for the determination of AT. However, this and observations such as the
following are of more general interest:
'The eclipse of the Moon which took place in Marchesvan (month VIII)
began [in the east]. That is bad for Subartu. What [is wrong]? After it,
Jupiter ent[ered] the Moon three times. What is being done to (make) its
evil pass? ...' From the lamentation priest Bel-suma-iskun.
[Trans. Hunger (1992, p. 263).]
Bel-suma-iskun was a Babylonian scribe. According to Parpola (1983),
the text dates from 675 BC.
By far the most well known observation of a solar eclipse from Assyria
occurred in 763 BC. This is recorded not on an astrological text but in
the Assyrian Chronicle:
(Eponym of) Bur-Saggile of Guzana. Revolt in the citadel; in (the
month) Siwan, the Sun had an eclipse (samas attain).
[Assyrian Chronicle; trans. Millard (1994, p. 58).]
126 4 Babylonian and Assyrian records of eclipses

Fig. 4.7 A copy of the Assyrian Chronicle containing a record of the solar
eclipse of 763 BC. (Courtesy: British Museum.)

Several damaged copies of the Assyrian Chronicle are preserved in the

British Museum. A photograph of one of these tablets, recording the
eclipse, is shown in figure 4.7.
The Assyrian Chronicle records very few natural events. It is mainly a
list of the annual limmu, senior officials after whom the year was named.
The practice of appointing Hmmu began as early as the eighteenth or
nineteenth centuries BC In translations, the Greek equivalent of the term
Hmmu - i.e. eponym - is usually preferred. This in turn has found its
way into English and means 'a person who gives his name to something'.
During later centuries in Assyria, the eponym was a provincial governor;
however, several kings also held this office. The Assyrian system of annual
eponyms was probably the origin of the Greek practice of appointing
archons (see section 4.3) and the Roman system of consuls, each of whom
held office for one year.
A complete list of Assyrian eponyms is preserved from 910 to 646 BC
(Millard, 1994, pp. 55-62). Since several kings of Assyria mentioned in the
 4.13 Conclusion 127

list were also rulers of Babylon, it is possible to deduce the year when Bur-
Saggile was eponym from Ptolemy's Canon of Kings (see Toomer, 1984,
p. 11). This year corresponds approximately to BC 763. The month Siwan,
the third month of the year, was equivalent to May-June. Reference to
the charts of Oppolzer (1887) shows that between 777 and 745 BC only
two eclipses could have been large in Assyria: BC 765 Feb 10 and 763 Jun
15. Of these, Feb 10 is far too early for Siwan and hence the date of the
eclipse must be BC 763 Jun 15. This date was accepted by Fotheringham
(1920b).
As no other eclipse is mentioned in the Assyrian Chronicle, Fother-
ingham supposed that it must have been total somewhere in Assyria.
However, this suggestion is unfounded; the record gives no information
regarding magnitude, although the eclipse was presumably very striking.
It may well have been seen at the Assyrian capital of Ashur (lat. = 35.48
deg, long. = —43.23 deg), but the report could have come from some
provincial location instead.

4.13 Conclusion
Although the Assyrian eclipse records are of undoubted historical interest,
they are of negligible significance in the study of the Earth's past rotation.
Fortunately, the roughly contemporaneous Babylonian observations are
of great value in the determination of changes in the length of the
day. Indeed, without them little would have been known concerning the
terrestrial rotation in the BC period. This material forms the subject of
the next three chapters.
 Investigation of Babylonian
observations of solar eclipses

5.1 Introduction
In this chapter, all accessible reports of solar eclipses preserved on the Late
Babylonian astronomical texts which are of value in studying the Earth's
past rotation will be investigated. Compared with the substantial number
of observations which must have originally been made, very few Babylo-
nian records of solar eclipses are now extant. Reasons for this are briefly
discussed in chapter 4. Reliable observations only range in date from 369
to 136 BC - less than half the period covered by lunar eclipse sightings.
The solar eclipses which form the subject of the present chapter will
be divided into four categories: (i) the only known example of a total
obscuration of the Sun observed at Babylon (dating from 136 BC); (ii)
measurements of the local times of a variety of eclipse contacts; (iii)
estimates of solar eclipse magnitude at maximum phase; and (iv) instances
when the Sun rose or set whilst obscured. Several of these observations
have already been discussed in chapter 3, where they were used to illustrate
various analytical techniques. However, they will now be considered in
greater detail. At the close of this chapter, a possible allusion to a
total solar eclipse recorded in the Religious Chronicle of Babylon will be
considered. This event probably took place around 1000 BC - much
earlier than the period covered by the Late Babylonian texts.
All eclipse records on the Late Babylonian texts where the date is
doubtful will be rejected, as will questionable readings of damaged signs.
In some instances, the sign representing the time of first contact is missing
or damaged. Hence even if the durations of other phases are preserved
intact, the observations are of no value for determining AT. As a rule,
only those measurements which seem capable of yielding useful results
for AT are discussed below. In each case, I have cited the Julian date of
the eclipse as derived from modern astronomical computations. When the
recorded date is fully preserved, the equivalent on the Julian calendar is
almost invariably identical to the computed date. In the case of damaged

128
 5.2 The total solar eclipse of BC 136 Apr 15 129

texts, I have given brief comments on the restoration of the original date.
For further details, see Sachs and Schaumberger (1955), Sachs and Hunger
(1988, 1989, 1996) or Huber (1973). Because of the frequent references to
the three volumes published by Sachs and Hunger, these works will be
referred to as SH I, SH II and SH III throughout this chapter and also
in chapters 6 and 7.
Most translations are taken from either SH I, SH II and SH III or
Huber (1973). When these two separate works translate the same texts,
there are often stylistic differences between the individual renderings.
However, these are slight. I have avoided any attempts to standardise the
various translations to a set pattern, preferring when possible to quote
the published versions in SH I, SH II and SH III. This is by no means a
criticism of Huber's excellent manuscript.

5.2 The total solar eclipse of BC 136 Apr 15

Although this is the only total solar eclipse recorded in the Late Baby-
lonian astronomical texts, fortunately the preserved descriptions are quite
detailed. The phenomenon is described in two separate British Museum
texts: (i) BM 34034 - an unusually well-preserved goal-year text - and (ii)
BM 45745 - a small remnant of a diary. The date of the eclipse recorded
in the goal year text is fully specified as the 29th day of the intercalary
12th month of SE 175; this corresponds exactly to BC 136 Apr 15, so that
there can be no doubt about the identification. The year of the diary (SE
175) is also preserved, although the month is broken away. Photographs
of the appropriate sides of the two tablets are shown in figures 3.8a and
3.8b. Translations of the relevant entries are as follows:

(i) SE 175, month XII2. The 29th, solar eclipse. When it began on the
south-west side, in 18 deg daytime in the morning it became
entirely total (TIL ma TIL ti gar AN). (It began) at 24 deg after
sunrise.
[BM 34034 (= LBAT 1285), Rev. 24-28; trans Hunger (1995).]
A similar translation is given by Huber (1973), pp. 93-94. The term
TIL (rendered in the above account as 'total') is commonly used in the
Late Babylonian astronomical texts to identify complete eclipses of the
Moon (see chapter 3); the repetition of TIL here can perhaps best be
rendered 'entirely total'. This same goal-year text also contains a record
of the partial lunar eclipse of BC 136 Apr 1 (see chapter 6).

(ii) SE 175, [king] Arsaces, [month XII2]. The 29th, at 24 deg after
sunrise, solar eclipse; when it began on the south and west side,
[...] [Ven]us, Mercury and the Normal stars were visible; Jupiter
130 5 Investigation of Babylonian observations of solar eclipses

and Mars, which were in their period of invisibility, were visible in

its eclipse [... ] it threw off (the shadow) from west and south to
north and east; 35 deg onset, maximal phase and clearing; in its
eclipse, the north wind which was set [to the west? side blew ...].
[BM 45745 (= LBAT 429), Rev. 13'-15'; trans. SH III, p. 185.]
A comparable translation by Abraham Sachs was published by Clark
and Stephenson (1977, p. 15). Since so much of the original tablet is
missing, there are several lacunae in the text; these are indicated here
(as elsewhere) by [...]. When the top line (or lines) of a text is missing
(as in the above example), it is customary to number lines from the first
preserved line using a prime (e.g. 13').
In the central zone, the computed magnitude of this eclipse was 1.05.
Even without the direct evidence of totality from the goal-year text, it
would be clear from the diary entry that the eclipse was fully complete at
Babylon. As well as stars, as many as four planets were sighted (including
Jupiter and Mars, which were too close to the Sun to be seen under
normal circumstances). The 'Normal Stars' (MUL-SID) mentioned in the
text were a series of 31 zodiacal stars which were often used to indicate
the position of the Moon and planets (see chapter 6). The term as used in
this context presumably means no more than the stars in general, implying
that several stars had become visible.
The computed elongations of the five bright planets from the Sun and
also their stellar magnitudes on this occasion were as follows: Mercury
(20 deg E, mag. = +1.3), Venus (31 deg E, mag. = -4.1), Mars (29 deg
W, mag. = +1.5), Jupiter (1.5 deg E, mag. = —1.5) and Saturn (166 deg
W, mag = +0.4). This last planet would probably be below the western
horizon during the eclipse, which would explain why it was not mentioned.
Both Mercury and Mars were rather faint and hence would be somewhat
difficult to detect. Jupiter, was extremely close to the Sun - only 1.2 deg
from the solar limb. Near the central line, totality would last for about 3
min 25 sec. Although no estimate of the duration of totality is given, it
probably must have lasted at least a minute to enable the various planets
to be distinguished and identified.
The altitude and azimuth for maximal phase of the Sun, four observed
planets and brighter stars in the vicinity of the Sun, are shown in figure
5.1. These positions are computed for the measured time of totality at
Babylon (i.e. 42 deg after sunset or a local time of 8.76 h - see section
5.3).

RESULTS

For totality at Babylon, 11210 < AT < 12140 sec.

 5.3 Timed observations of solar eclipses 131

i i i i i
'• Mars

40° - -

30° -
-
* Mercury OSun
CD
Jupiter

* Venus
20° -

* aTau
10° - -
* y Ori

• a Ori
n° i i i i i i
70° 80° 90° 100° 110° 120° 130° 140°
Azimuth
Fig. 5.1 Visibility of the planets Mercury, Venus, Mars and Jupiter and also
bright stars during the total solar eclipse of 136 BC at Babylon. NB Saturn would
be below the horizon.

If a duration of at least one minute is assumed, the above AT interval

reduces by only 40 sec (to 11230 < AT < 12120 sec), a negligible difference
at this early period.
The measured local times of the various phases lead to independent
results for AT; these will be discussed in section 5.3.

5.3 Timed observations of solar eclipses

At the head of each entry below (after the Julian date), I have given -
for reference only - the computed degree of obscuration of the Sun at
maximum phase; this is based on a provisional AT formula (equation
(4.1)).

(1) BC 322 Sep 26 (computed mag. at Babylon = 0.17)

Year 2 of (king) Philip... (month VI)... The 28th, around 3 deg before
sunset, solar eclipse (ana SU samas AN-KUio) [... ] gusty west and north
wind. It set eclipsed.
[BM 34093 (= LBAT 212), Rev. 23'; trans. SH I, p. 227;
see also Huber, p. 83.]
132 5 Investigation of Babylonian observations of solar eclipses

.,.,..-..-,'

Fig. 5.2 A photograph of the reverse of British Museum Tablet 34093 (an
astronomical diary) containing a record of the sunset eclipse of 322 BC as seen
from Babylon.

A photograph of the reverse of this tablet - containing the above eclipse

record - is shown in figure 5.2.
Although the sign for month VI is missing in this diary, the symbol for
the following month VII is clearly preserved a few lines later (Rev. 26')-
Philip Arrhidaeus was the successor to Alexander the Great. The term
ana SU samas means sunset, so that an extra samas would be expected
before AN-KUio in order to identify the type of eclipse. Perhaps this is
merely an example of haplography, as suggested by Huber.
This observation was used by Fotheringham (1935) in an investigation
of the lunar and solar accelerations on UT However, he used an incorrect
value (4 deg) for the interval between the start of the eclipse and sunset.
The observation that the Sun set eclipsed will be discussed below (section
5.5).

RESULTS
First contact on Sep 26 at 0.20 h before sunset. LT of sunset = 18.08 h,
hence measured LT of first contact = 17.88 h, UT = 14.81 h. Computed
TT = 18.75 h, thus AT = 14150 sec.
 5.3 Timed observations of solar eclipses 133

(2) BC 281 Jan 30 (computed mag. at Babylon = 0.21)

Year 30 (SE), kings Seleucus and Antiochus... (month X)... The 29th,
[solar] eclipse; when the Sun came out (i.e. rose), 2fingerswere obscured
on the south side; at 6 deg daytime, [...] 20 deg daytime onset and
clearing; during its eclipse, the north wind which was slanted to the [...]
blew; during its eclipse, Mars, Venus...
[BM 41660 (= LBAT 232), Rev. 9-10; trans SH I, p. 313.]
In the late summer of the 30th year of his reign (equivalent to 282/1
BC), Seleucus I died and his son Antiochus I ascended the throne; this
is why two kings are mentioned in the above diary entry. Although
the month is missing, it can readily be restored from the frequent lunar
and planetary data. As noted in chapter 4, the Babylonians frequently
calculated which planets would be above the horizon during both solar
and lunar obscurations. The planets would not, of course, be actually
visible at most solar eclipses.
The observation that the Sun rose eclipsed will be discussed below
(section 5.5). It should be noted that the term E-a, which SH translate
as 'came out', is the normal term for sunrise. Presumably the text implies
that maximum phase occurred 6 deg after this moment, following which
the Sun began to clear. However, the relevant part of the entry is only
partially preserved. Hence only the measurement of last contact (20 deg
after sunrise) will be considered here.

RESULTS
Last contact on Jan 30 at 1.33 h after sunrise. LT of sunrise = 6.76 h,
hence measured LT of last contact = 8.09 h, UT = 5.42 h. Computed TT
= 9.02 h, thus AT = 12950 sec.

(3) BC 254 Jan 31 (computed mag. at Babylon = 0.26)

[SE 57]...month X... The 28th, 56 deg before sunset, solar eclipse; when
it began, in 12 deg daytime [...] when it began [to cl]ear, it cleared from
south to north in 11 deg daytime; 23 deg onset and clearing. During its
eclipse blew the west wind which was slanted to the north.
[BM 34278 (= LBAT 596) + BM 35418 (= LBAT 258), Rev. 11, 12;
trans. SH II, p. 29.]
Only the month and day of the eclipse are preserved on this diary, which
consists of two joined fragments, but the year can readily be established
from the many planetary observations which are preserved (SH II, p. 30).
The text indicates that first contact occurred 56 deg before sunset, maximal
phase 44 deg before sunset and end 33 deg before sunset. Although the
reference to maximum is damaged, it is clear from the various recorded
durations that this phase is referred to.
134 5 Investigation of Babylonian observations of solar eclipses

RESULTS
(i) First contact on Jan 31 at 3.73 h before sunset. LT of sunset = 17.26 h,
hence measured LT of first contact = 13.53 h, UT = 10.86 h. Computed
TT = 14.04, thus AT = 11450 sec.
(ii) Maximal phase at 2.93 h before sunset, hence measured LT = 14.33
h, UT = 11.66 h. Computed TT = 14.86 h, thus AT = 11 500 sec.
(iii) Last contact at 2.20 h before sunset hence measured LT = 15.06 h,
UT = 12.39 h. Computed TT = 15.63 h, thus AT = 11650 sec.
These three results for AT are in remarkably close accord with one
another.

(4) BC 249 May 4 (computed mag. at Babylon = 0.80)

Observations of this eclipse are recorded on two separate tablets.

(i) (SE 63, month I)... The 28th, clouds were in the sky; 90 deg
daytime [... solar eclipse ... ] [... ] onset and clearing; during its
eclipse [...].
[BM 45723 (= LBAT 267), Obv 8'; trans. SH II, p. 51;
see also Huber, p. 84.]
BM 45723 is an astronomical diary. Only a small piece remains,
without reference to either year or month, but the few planetary and lunar
observations are sufficient to enable a date to be identified (SH II, p. 53).

(ii) (SE 63, month I), 28 solar eclipse; over two-thirds of the disk it
made. (It began) at 90 deg after sunrise.
[BM 32154 (= LBAT 1216) + BM 32408 (= LBAT 1217),
Obv. 3; trans. Huber, p. 84.]
The joined fragments BM 32154 + BM 32408 are part of a goal-year
text dating from SE 81 (231-230 BC). The eclipse occurred 18 years before
the goal-year. Unfortunately only the local time of first contact - i.e. 90
deg after sunset - is preserved.

RESULTS
First contact on May 4 at 6.00 h after sunrise. LT of sunrise = 5.28 h,
hence measured LT = 11.28 h, UT = 8.22 h. Computed TT = 12.07 h,
thus AT = 13 850 sec.

(5) BC 195 Jun 6 (computed mag. at Babylon = 0.19)

Year 117 (SE), month II 28, solar eclipse, beginning on the south side. 4
fingers it made. (Began) at 2 beru (= 60 deg) after sunrise.

[BM 34048 (= LBAT 1249), Rev. 3'-8'; trans. Huber, p. 86.]

 5.3 Timed observations of solar eclipses 135

This tablet is a goal-year text dating from SE 135 (177-176 BC). The
eclipse, whose date is fully specified on the tablet, occurred 18 years before
the goal-year.

RESULTS
First contact on Jun 6 at 4.00 h after sunrise. LT of sunrise = 4.91 h,
hence measured LT = 8.91 h, UT = 5.83 h. Computed TT = ?? AT = ??
NB no value of AT can satisfy this observation; the computed UT
of first contact (i.e. TT — AT) cannot be earlier than 6.03 h, otherwise
the calculated eclipse magnitude reduces to zero at Babylon. Evidently
either the original measurement was faulty or a scribe made an error in
reporting the time. An upper limit to the value of AT is set by the fact
that the eclipse was visible at all (mag. > 0) at Babylon; this leads to the
result AT < 15 350 sec. The actual magnitude estimate (4 fingers) will be
considered in section 5.4.
(6) BC 190 Mar 14 (computed mag. at Babylon = 0.77)
Year 121 (SE), king An[tiochus] month XII 29 solar eclipse beginning
on the north-west side. In 15 deg day [...] over a third of the disk was
eclipsed. When it began to become bright, in 15 deg of day from
north-west to east it became bright. 30 deg total duration. [During this
eclipse] east (wind) went. During this eclipse [... ] Venus, Mercury and
Saturn [stood there]. Toward the end of becoming bright, Mars rose (?)
The other planets did not stand there. (Began) at 30 deg (= 1) beru after
sunrise.
[BM 33812 (= LBAT 1438), Obv. and Rev.; trans. Huber, p. 87,
with amendment by C. B. F. Walker.]
This clearly dated tablet from the reign of Antiochus III is devoted
specifically to this eclipse. The text is remarkably well preserved. Huber
questioned the reading of 30 deg but direct consultation of the original
tablet by Dr Christopher Walker of the British Museum (personal com-
munication) shows that this figure is clear and is confirmed by further
expressing the interval as 1 beru.
It will be noted that the computed magnitude at Babylon (0.77) is much
greater than the recorded figure. For an investigation of the estimate of
magnitude, see section 5.4. As discussed in chapter 10, this same eclipse
may have been observed to be total in the Hellespont strait (Dardanelles)
and partial at Alexandria. However, in neither case is the time of day
reported.

RESULTS
(i) First contact on Mar 14 at 2.00 h after sunrise. LT of sunrise = 6.10 h,
hence measured LT of first contact = 8.10 h, UT = 5.34 h. Computed
TT = 8.92 h, thus AT = 12900 sec.
136 5 Investigation of Babylonian observations of solar eclipses

(ii) Maximum phase at 3.00 h after sunrise, hence measured LT of

maximum = 9.10 h, UT = 6.34 h. Computed TT = 10.07 h, thus AT =
13400 sec.
(iii) Last contact at 4.00 h after sunrise, hence measured LT of maximum
= 10.10 h, UT = 7.34 h. Computed TT = 11.27 h, thus AT = 14150 sec.

(7) BC 170 Jul 28 (computed mag. at Babylon = 0.44)

Year 142 (SE) of the kings An(tiochus) and An(tiochus) his son, month
IV 28 solar eclipse, beginning on the north-west side. In 12 deg day it
made one third of the disk. (Began) at 20 deg day before sunset.
[BM 35387 (= LBAT 1263), Rev. 3'-7; trans. Huber, p. 90.]
At the time, Antiochus IV reigned jointly with his son (Sachs and
Wiseman, 1954). This tablet is a goal-year text dating from SE 160 (152-
151 BC). The eclipse, whose date is fully preserved, occurred 18 years
before the goal-year.

RESULTS
(i) First contact on Jul 28 at 1.33 h before sunset. LT of sunset = 18.98 h,
hence measured LT of first contact = 17.65 h, UT = 14.74 h. Computed
TT = 18.16 h, thus AT = 12 300 sec.
(ii) Maximal phase at 0.53 h before sunset, hence measured LT of
maximum = 18.45 h, UT = 15.54 h. Computed TT = 18.93 h, thus AT =
12200 sec.

(8) BC 136 Apr 15 (computed mag. at Babylon = 1.05)

SE 175 month XII2. The 29th, at 24 deg after sunrise, solar eclipse;
when it began on the south-west side, in 18 deg day towards noon it
became entirely total. (Time interval of) 35 deg for onset, maximal phase
and clearing.
[Composite description based on translations by Hunger (1995)
and SH III, p. 185 of BM 34034 (= LBAT 429), Rev. 24-28
and BM 45745 (= LBAT 1285), Rev. 13'-15'J
The observation of totality has already been discussed above (section
5.2).

RESULTS
(i) First contact on Apr 15 at 1.60 h after sunrise. LT of sunrise = 5.56 h,
hence measured LT of first contact = 7.16 h, UT = 4.19 h. Computed TT
= 7.69 h, thus AT = 12 600 sec.
(ii) Total at 2.80 h after sunrise, hence measured LT of maximum =
8.36 h, UT = 5.39 h. Computed TT = 8.76 h, thus AT = 12 100 sec.
 5.4 Estimates of solar eclipse magnitude 137

Table 5.1 AT results from Late Babylonian timings of

solar eclipses.
Year* Ct* SR/SS C Int (deg)^ AT (sec)e
-321 1 SS -3 14150
-280 4 SR +20 12950
-253 1 SS -56 11450
-253 M ss -44 11500
-253 4 ss -33 11650

-248 1 SR +90 13 850

-189 1 SR +30 12900
-189 M SR +45 13400
-189 4 SR +50 14150
-169 1 SS -20 12 300

-169 M ss -8 12200
-135 1 SR +24 12 600
-135 2 SR +42 12100
-135 4 SR +59 12250
a
The year of each eclipse on the Julian calendar, using negative
integers.
b
The contact, etc., whose time was measured (M = maximal
phase).
c
Whether the eclipse was timed relative to sunrise (SR) or sunset
(SS).
d
The measured interval in degress (positive if after sunrise, nega-
tive if before sunset).
e
The calculated value of AT (in sec) which satisfies the
observation.

(iii) Last contact at 3.93 h after sunrise, hence measured LT of end =

9.49 h, UT = 6.52 h. Computed TT = 9.93 h, thus AT = 12250 sec.

Here is another set of very self-consistent values for AT. These results
closely correspond to the limits set by the observation of totality at the
same eclipse - i.e. 11210 < AT < 12140 sec (see section 5.2)
The results obtained in this section from timed observations are listed
in table 5.1; the AT results in this table are in very good accord with one
another.

5.4 Estimates of solar eclipse magnitude

Six preserved Babylonian records of solar eclipses contain estimates of the
maximum degree of obscuration of the Sun: BC 369, 249, 195, 190, 170
and 166. Timings of all but the first and last of these eclipses have been
138 5 Investigation of Babylonian observations of solar eclipses

discussed earlier (section 5.3) so that for these only the essential details
will be repeated below. However, full accounts of the eclipses of 369 and
166 BC will be given.

(1) BC 369 Apr 11

[Artaxerxes II, year 35, month XII, end... solar eclipse... ]. In 6 deg
daytime | of the disk was covered ...[...] At that time Jupiter was in Leo
[... to] the 7th day of month I of year 36, the river level [... ] 2 cubits 8
fingers.
[BM 37097 + 37211 (no LBAT number), Rev. 4'-7; trans. SH I, p. 125.]
The date of this small diary fragment may be derived from the reference
to year 36 on line T'. This mentions a measurement of the river level at
Babylon from some day of the current month to the 7th day of month I
in year 36. Only the 36th year of either Artaxerxes I (i.e. 430-429 BC) or
Artaxerxes II (370-369 BC) - both of whom had unusually long reigns -
or 36 SE (277-276 BC) can be intended. In their commentary (SH I, p.
125), Sachs and Hunger state that:

... it appears that the reverse (of the tablet) deals with the last month of
a year 35, because a measurement of the river level from some day of the
current month 'to the 7th of month I of year 36' is mentioned.

There were solar eclipses in the last month of both the 35th year of
Artaxerxes II (i.e. BC 369 Apr 11) and SE 35 (i.e. BC 276 Apr 3).
However, no similar event took place after the 9th month in the 35th year
of Artaxerxes I (BC 429 Jan 17). Hence only the eclipses of BC 369 and
276 would seem viable (see also SH II, p. 125). The location of Jupiter in
Leo (a sign of the zodiac covering the longitude range 120-150 deg) fits
only the earlier year. At the eclipse of BC 369 the longitude of Jupiter
was 136 deg, whereas at the eclipse of 276 BC it was 82 deg. It should be
pointed out that when the eclipse of BC 429 occurred, the longitude of
the planet was 120 deg, but as noted above, this eclipse occurred at the
wrong time of year.

RESULTS
To attain the observed magnitude of 0.33 in 369 BC, AT = ?? sec.
NB no value of AT can produce a maximum visible obscuration as
small as the observed figure of 0.33 unless greatest phase occurred a little
before sunrise. If this were the case, the eclipse would already be declining
when the Sun rose. However, the record seems clear that the eclipse was
already visible for 6 deg before maximal phase was reached. For a value
of AT (16 500 sec) which produces greatest phase 6 deg (i.e. 0.4 h) after
sunrise, the magnitude would be 0.39. Although the Babylonian estimate
 5.4 Estimates of solar eclipse magnitude 139

of '^ of the disk' would thus be quite reasonable, a result for AT cannot
be derived.

(2) BC 249 May 4

Two observations of this eclipse are recorded on separate tablets - a diary
and a goal-year text (see section 5.3) - but only the latter inscription
contains an estimate of magnitude.

(SE 63, month I), 28 solar eclipse; over two-thirds of the disk it made.
(Began) at 90 deg after sunrise.
[BM 32154 (= LBAT 1216) + BM 32408 (= LBAT 1217), Obv. 3;
trans. Huber, p. 84.]
It will be assumed that the magnitude estimate of 'over two-thirds'
means somewhere between \I (8
(8 fingers)
fingers) and
and | \ (9 fingers) - i.e. approxi-
mately %\ twelfths or 0.71.

RESULTS
To attain the observed magnitude of 0.71, AT = 14650 sec (only one
possible solution).

(3) BC 195 Jun 6

Year 117 (SE), month II 28 solar eclipse, beginning on the south side. 4
fingers it made. (Began) at 2 beru (= 60 deg) after sunrise.
[BM 34048 (= LBAT 1249), Rev. 3/ -8/ ; trans. Huber, p. 86.]

RESULTS
To attain the observed magnitude of 0.33, with the lower limb of the Sun
covered, AT = 10250 sec.

(4) BC 190 Mar 14

Year 121 (SE)...month XII 29 solar eclipse beginning on the north-west
side. In 15 deg day [... ] over a third of the disk was eclipsed. When it
began to become bright... (Began) at 30 deg (= 1) beru after sunrise.
[BM 33812 (= LBAT 1438), Obv. and Rev.; trans. Huber, pp. 87-88.]
It will be assumed that the magnitude estimate of 'over a third' means
somewhere between \ (4 fingers) and 5 fingers - i.e. some 4^ twelfths or
0.38.

RESULTS
To attain the observed magnitude of 0.38 with the upper limb of the Sun
covered, AT = 6850 sec.
140 5 Investigation of Babylonian observations of solar eclipses

(5) BC 170 Jul 28

Year 142 (SE)... month IV 28 solar eclipse beginning on the north-west
side. In 12 deg day it made one third of the disk. (Began) at 20 deg day
before sunset.
[BM 35387 (= LBAT 1263), Rev. 3'-7; trans. Huber, p. 90.]

RESULTS

To attain the observed magnitude of 0.33 with the upper limb of the Sun
covered, AT = 35 850 sec.
NB this result is highly discordant, as comparison with the fairly self-
consistent AT values in table 5.1 shows. Furthermore with the derived
figure for AT of 35 850 sec, the computed LT would be around 10.00
h instead of the recorded time of about 8 deg (i.e. 20-12 deg) before
sunset (18.45 h). Even if it is supposed that the lower limb of the Sun was
obscured, no value of AT can produce a maximum visible obscuration
quite so small as the observed figure unless the eclipse reached its peak
some time after sunset. This is not in accord with observation. It would
appear that the magnitude estimate is unreliable - presumably representing
a scribal error.

(6) BC 166 May 17

[Y]ear 146, [king] Antiochus [month I, end...solar eclipse]. In 13 deg of
day more than one-third of the disk...
[BM 32844 (= LBAT 376), Obv. A13; trans. SH II, p. 486.]
This entry is towards the end of the first month covered by a diary
for the year 146 SE (166/165 BC). At the time, Antiochus IV was sole
ruler. Although the month is missing, assumption of month I is readily
confirmed from the various lunar observations noted in the text.
As above, it will be assumed that the magnitude estimate of 'over a
third' means somewhere between | (4 fingers) and 5 fingers - i.e. 4j
twelfths or 0.38.

RESULTS

To attain the observed magnitude of 0.38, either AT = 15 250 sec or

—6850 sec. Only the first value for AT need be considered; the alternative
is quite discordant.

The rather mediocre set of results obtained in this section is summarised

in table 5.2. This table gives the year (expressed as a negative integer),
observed magnitude and AT result for each observation.
 5.5 Solar eclipses occurring near sunrise or sunset 141

Table 5.2 AT results from

Babylonian estimates of solar
eclipse magnitude.

Year Est. mag. AT (sec)

-368 1/3 ??
-248 >2/3 14650
-194 4/12 10250
-189 >l/3 6850
-169 1/3 ??
-165 >l/3 15 250

5.5 Solar eclipses occurring near sunrise or sunset

There are only two known Babylonian records in this category (322 and
281 BC). In the latter instance, the degree of obscuration of the Sun
is specified at sunrise; this estimate will also be considered. A further
observation (241 BC) is also worth investigating. Although computation
based on a preliminary value of AT (equation (4.1)) indicates that the
eclipse occurred near sunset, the text itself mentions the clearance of
the Sun. Pre-selection of this kind in no way introduces bias; it merely
pinpoints potentially viable observations.
Timings of both the eclipses of 322 and 281 BC have already been
discussed in section 5.3. Hence only partial quotations, omitting irrelevant
details, will be given here. However, the record from 241 BC will be
translated in full. Following earlier practice, computed magnitudes are
given purely for reference.

(1) BC 322 Sep 26 (computed mag. at Babylon = 0.17)

Year 2 of (king) Philip, (month VI). The 28th, around 3 deg before
sunset, solar eclipse [...]... It set eclipsed.

[BM 34093 (= LBAT 212), Rev. 23'; trans. SH I, pp. 226-227;

see also Huber, p. 83.]

As discussed in some detail in chapter 3, for the present purpose the

most viable inference is that the Sun set at some time between first and
last contact. Although the '3 deg' of the text suggests that sunset occurred
long before maximal phase, there is always a slight possibility that this
could represent a scribal error for a much larger number.
142 5 Investigation of Babylonian observations of solar eclipses

RESULTS

(1) First contact on Sep 26 before sunset. LT of sunset = 18.08 h, hence

LT of first contact < 18.08 h, U T < 15.01 h. Computed TT = 18.74 h,
thus A T > 13 400 sec.
(ii) Last contact after sunset. LT of contact > 18.08 h, U T > 15.01 h.
Computed TT = 20.02 h, thus A T < 18 000 sec.
Combining the above limits, one obtains 13400 < A T < 18 000 sec.

(2) BC 281 Jan 30 (computed mag. at Babylon = 0.21)

Year 30 (SE)... (month X). The 29th [solar] eclipse; when the Sun came
out (i.e. rose), 2fingerswere obscured on the south side; at 6 deg daytime,
[...]... 20 deg day onset and clearing ...
[BM 41660 (= LBAT 232), Rev. 9-10;
trans SH I, p. 313.]
One-sixth of the solar diameter was estimated to be obscured at sunrise,
after which the phase apparently increased and then decreased. Hence it
may reasonably be concluded that the Sun rose between first contact and
maximum phase.

RESULTS

(i) First contact on Jan 30 before sunrise. LT of sunrise = 6.78 h, hence

LT of first contact < 6.78 h, UT < 4.11 h. Computed TT = 7.53 h, thus
AT > 12 350 sec.
(ii) Maximum phase after sunrise. LT of maximum > 6.78 h, UT >
4.11 h. Computed TT = 8.13 h, thus AT < 14450 sec.
Combining the above limits, one obtains 12 350 < AT < 14450 sec.
(iii) Fraction of Sun covered at rising. For any value of AT in the above
interval, the computed eclipse magnitude ranges from 0.17 to 0.22. These
figures are probably too close to the estimated degree of obscuration at
sunrise (2 fingers or 0.17) for a useful independent value of AT to be
deduced from this observation.

(3) BC 241 Nov 28 (computed mag. at Babylon = 0.38)

[SE 70, month VIII, end...Solar eclipse ...] cleared; 30 deg onset and
clearing; during the eclipse [...].
[Rm 720 + 732 + BM 41522 (= LBAT 278, 277, 883), Obv. 3';
trans. SH II, p. 79.]
This is a fairly well preserved diary consisting of several joining frag-
ments, two of which were uncovered by Rassam at the site of Babylon.
Although both the year and month have been broken off, the planetary
and lunar observations enable the date to be unambiguously derived (SH
 5.6 A possible ancient allusion to a total solar eclipse 143

Table 5.3 AT limits from Babylonian obser-

vations of solar eclipses.

AT range (sec)
Year Observation LL (sec) UL (sec)
-321 sunset 13 400 18 000
-280 sunrise 12 350 14450
-240 sunset 12 800 —
-194 mag.>0 — 15 350
-135 total 11210 12140

II, p. 83). The time of first contact for the solar eclipse relative to sunrise
or sunset is, unfortunately, wanting.
The value of AT computed according to equation (4.1) indicates that
the eclipse ended close to sunset. However, since clearing is mentioned
(twice), it must have been seen to end before sunset. In this instance, it is
only possible to determine one useful (lower) limit to AT.

RESULTS
Last contact on Nov 28 before sunset. LT of sunset = 17.11 h, hence LT
of contact < 17.11 h, UT < 14.01 h. Computed TT = 17.57 h, thus AT
> 12 800 sec.

In table 5.3 are assembled the results obtained from the three observa-
tions analysed in this section together with the AT limits derived from the
total eclipse of 136 BC (section 5.2) and the very small eclipse of 195 BC
(section 5.3). Individual columns of this table contain the year (negative),
type of observation, lower limit (LL) to AT, and upper limit (UL) to AT.
In the year —194 (i.e. 195 BC), the fact that the eclipse was visible from
Babylon at all provides an upper limit only to AT.

5.6 A possible allusion to a total solar eclipse in the Babylonian

Religious Chronicle
The Religious Chronicle of Babylon, a fragmentary copy of which is
preserved in the British Museum (tablet BM 35968), makes reference to
darkness by day accompanied by 'fire in the sky'. This has been interpreted
as a large eclipse of the Sun. Consecutive extracts from the chronicle may
be translated as follows:
On the sixteenth day of the month Ab, in the seventh year, two deer
entered Babylon and were killed.
144 5 Investigation of Babylonian observations of solar eclipses

On the twenty-sixth day of the month Sivan, in the seventh year, day
turned to night and (there was) a fire (isatu) in the sky.
In the month Elul, in the eleventh year, water flowed within the wall of
the lower forecourt
[Religious Chronicle, ii 12-15; trans. Grayson (1975, p. 135).]
NB as pointed out by Grayson, the scribe put the events in Ab (the 5th
month) and Sivan (the 3rd month) in the wrong order.
The unknown king in whose reign the above events occurred was on
the throne for at least 17 years. The chronicle is mainly concerned with
portents observed at Babylon, although there are occasional references to
other places. Fotheringham (1920b) noted that as a rule no more than
two or three portents are recorded in any one year and he thus inferred
that the celestial phenomenon described 'was a rare one, was observed at
Babylon, and may reasonably have been a total eclipse of the Sun'. He
identified the 'fire in the sky' (rendered in his source as 'fire in the midst of
heaven') as the solar corona. On the presumption that the event occurred
during the eleventh century BC, he calculated that the eclipse of BC 1063
Jul 31, was 'the only one which can have been total at Babylon within
that century'.
However, this identification is far from secure. For instance:
(i) The name of the king is not known.
(ii) July is too late for the month Sivan, the third month of the Babylo-
nian year (which began in March-April).
(iii) Since the start of each month was fixed by visibility of the crescent
Moon, a solar eclipse could not occur as early as the 26th day. Thus
in the Late Babylonian texts, solar eclipses are invariably recorded
on the 28th or 29th days (see the various examples quoted earlier in
this chapter).

(iv) Many of the portents recorded in the chronicle are quite mundane,
as can be seen in the above extract. They closely resemble the omens
found in the Late Babylonian texts - translations of which are given
by SH I, SH II and SH III. Thus there is no need to assume that
the phenomenon causing darkness was unusually rare.
(v) It is by no means certain that the occurrence described was indeed
an eclipse. It could have been a severe storm, in which case isatu
would mean lightning (Grayson, 1975, 135n; see also Brinkman,
1968, p. 68n). The Babylonians at this period should have been very
familiar with eclipses but in the above description there is nothing
which could be identified as a technical term for an event of this
 5.6 A possible ancient allusion to a total solar eclipse 145

i • • • I ' ' I . . . .

16000 V
X
v X
14000 o o
O o
/x
o o
/\
12000

X
10000 - -

8000-- -

6000 - Key -

I Total solar eclipse

4000 o Tinned measurement -

V Upper limit to AT
2000 /\ Lower limit to AT _

x Magnitude estimate
1 1 1 1 1 1 1 1 1 I , , , , 1 , . . , 1 , , , ,

-350 -300 -250 -200 -150 -100

Year
Fig. 5.3 Diagram showing the AT values and limits obtained from Late
Babylonian observations of solar eclipses.

kind - such as AN-KUio found in the Late Babylonian texts. Even

if the event was an unusually spectacular obscuration of the Sun, it
would begin and end like any other eclipse.
(vi) Explanation of the 'fire in the sky' as the corona seems very unlikely.
The corona is, in fact, scarcely ever alluded to in ancient and medieval
literature, even when the effects of a total eclipse are vividly described
(see especially chapter 11).
Rowton (1946) favoured a solar eclipse at sunset and proposed BC 1012
May 9. However, according to Wiseman (1965), this was in the 15th year
of a king (namely Simbar-sipak), rather than the 7th. It should be stressed
that the chronology of this period is still poorly established. Although
calculations using a value for AT derived from equation (4.1) (i.e. 28000
sec) indicate that the eclipse of 1012 BC would be fairly large at sunset
in Babylon, no figure for AT can lead to a visible magnitude greater than
0.96 there. A significantly greater phase would be needed to effectively
turn the day into night.
Similar calculations based on equation (4.1) (AT = 29000 sec) indicate
that the eclipse of 1063 BC would reach a magnitude of 0.85 at Babylon
in the early morning (6.0 h). Decreasing AT to about 25000 sec would
146 5 Investigation of Babylonian observations of solar eclipses

produce totality at Babylon. However, this result is of little significance;

Fotheringham derived the date by playing the 'identification game' with
the minimum of constraints. In view of the uncertain chronology of the
period and the doubtful interpretation of the phenomenon described in
the Religious Chronicle, further discussion seems profitless.

5.7 Conclusion
The miscellaneous results obtained in this chapter, most of which are
listed in tables 5.1 through 5.3, are plotted in figure 5.3. Apart from the
magnitude estimates (represented by crosses), this diagram - which also
depicts the mean for all the timed data - reveals that there is close accord
between both the individual values for AT and the various limits. The
magnitude estimates will not be considered further.
 6
Timed Babylonian lunar eclipses

6.1 Introduction
Because so many observations of lunar eclipses are preserved on the Late
Babylonian astronomical texts, it is necessary to devote two chapters of
this book to their discussion. This is a reflection of the great importance of
the lunar eclipse records from Babylon in the study of long-term changes
in the length of the day. The present chapter is restricted to timed contacts,
while in chapter 7 a variety of untimed observations of eclipses of the
Moon will be considered.
Most Babylonian lunar eclipse timings are expressed relative to sunrise
or sunset, although very occasionally they are referred to moonrise or
moonset instead. I have divided these measurements into four categories:
(i) those for which only a single contact timing is preserved; (ii) timings of
two separate contacts (mainly partial eclipses); (iii) three or four contact
timings (total eclipses only); and (iv) timings of eclipse maxima (partial
eclipses only). Observations in these categories will be discussed in turn in
sections 6.2 through 6.5. In addition, a number of eclipses were also timed
in relation to the culmination of certain reference stars (see section 6.6).
Times of the very earliest eclipses (whether observed or predicted)
were nearly always quoted with low precision. Between 731 BC (the
earliest surviving record, representing a prediction) and just before 600
BC times were almost exclusively expressed to the nearest 10 deg. The
only known exception to this rule is an observation dating from 666 BC;
here it is said that the Moon commenced to be obscured only 3 deg after
sunset. Evidently 10 deg was the maximum precision that the Babylonian
astronomers felt able to achieve under normal circumstances at this archaic
period. However, two significant improvements in accuracy can be traced.
Between about 600 and 560 BC (a period which roughly coincides with
the reign of Nebuchadrezzar II), rounding of times to the nearest 5 deg
was the norm, but after 560 BC virtually all measurements were estimated
to the nearest degree. It may be presumed that solar eclipse timings

147
148 6 Timed Babylonian lunar eclipses

40°—,

30° -

CD
"D

C
CD

2
20° -

10° -

0° 10° 20° 30° 40° 50° 60°

Computed time interval for all phases of eclipse (deg)
Fig. 6.1 Errors in measured lunar eclipse durations (Stephenson and Fatoohi,
1993).

originally followed much the same pattern but - as noted in chapter 5 -

no observations of this kind have survived from before 370 BC.
Despite the marked decrease in rounding errors around 560 BC, during
later centuries progress in measuring time-intervals appears to have been
relatively slow. This is apparent from figure 6.1, which plots the errors
in measuring lunar eclipse durations made by the Late Babylonian as-
tronomers from 560 to 50 BC (Stephenson and Fatoohi, 1993). It thus
seems a pity that observers at earlier periods were not more optimistic
when rounding their measurements.
As is true of several solar eclipse records on the astronomical texts,
the date of a lunar obscuration is often broken away. Techniques of
 6.1 Introduction 149

restoration have already been discussed (chapter 4). When considering an

observation which is reported on either a diary, goal-year text, fragmentary
lunar eclipse table, or single tablet devoted to the event, I have given
specific comments with regard to the date of that record. However, a
substantial number of observations (as well as predictions) are reported
on a series of extensive lunar eclipse tables; these typically cite events over
many years. Some preliminary remarks on these tables seem necessary
here in order to avoid needless repetition when discussing the observations
themselves. The following comments are largely based on the scholarship
of Huber (1973).

(i) All of the surviving observations (and predictions) of lunar eclipses

from earliest times (731 BC) to 609 BC - as well as many later
observations down to 317 BC - are recorded on a series of five
British Museum tablets. Their reference numbers are: BM 32238
(= LBAT 1414), BM 45640 + 35115 + 35789 (= LBAT 1415 +
1416 + 1417: three joining pieces) and BM 32234 (= LBAT 1419).
This major compilation, which lists eclipses at 18-year intervals,
originally covered 24 saros cycles or 432 years and extended from
some time between 749 and 740 to between 317 and 308 BC. The
extant remnant listed as BM 32238 cites eclipses from 731 to 659
BC (obverse) and from 389 to 317 BC (reverse). Tablets BM 45640
+ 35115 + 35789 contain data from 703 to 632 BC (obverse) and
from 415 to 360 BC (reverse), while BM 32234 extends from 609
to 537 BC (obverse) and from 519 to 447 BC (reverse). Many
names of rulers are preserved on these tablets: e.g. Nabu mukin-zeri
(who reigned from 731 to 726 BC), Bel-ibni (702-699 BC), Samas-
sum-ukin (667-647 BC), Kandalanu (647-625 BC), Nebuchadrezzar
II (604-562 BC), Xerxes I (485-465 BC) and Philip (323-316 BC).
From the well-defined chronological sequence on this series of texts,
virtually all eclipse dates can be confidently restored.

(ii) BM 38462 (= LBAT 1420) reports lunar eclipses for almost every
year from the beginning of the reign of Nebuchadrezzar II (604/3
BC) to his 29th year (576/5 BC). The damaged (but still recognisable)
name of Nebuchadrezzar is given on the first line of the tablet.

(hi) Between them, BM 34787 (= LBAT 1426) and BM 34684 (= LBAT

1427) cover the period from the 22nd year of Artaxerxes I (443/2
BC) to the 18th year of his successor Darius II (406/5 BC). Eclipses
are mentioned in most years. Darius II became king during months
IX/X of the 41st year of Artaxerxes I. His accession ('Ochos, whose
name is Darius') is noted in column II of tablet BM 34787.
150 6 Timed Babylonian lunar eclipses

* pAri*

a +5
*tiTau" °
#
BSco * *CnC
P
» $Ll* P^gpLeo aLeo^Cnc 60°
Long 360° 330° 300° 270° 240°»6S £° 180°» 150° 120° * 90° * r *L Gem 30°
9Cnc n G
8Cap**yCap ^ph ^ aVir »™ Tau

a Sco # a T a u -5°
#
yGem
J
-10°

Fig. 6.2 Co-ordinates of the Babylonian 'Normal Stars' at 164 BC.

Further justification for the dates of lunar eclipses mentioned on any

of the above tablets will not usually be given below.
In later texts (from the fifth century BC onwards), the location of
the eclipsed Moon relative to the nearest 'Normal Star' (MUL-SID) is
frequently given. The Normal Stars were a series of 32 stars in the
zodiacal belt. Conjunctions between the Moon and planets with these
stars are mentioned very frequently on the Late Babylonian astronomical
texts and calculations of the circumstances of these events have readily
revealed their identity. For a full list of names and identifications (in
terms of Bayer Greek letters) see Stephenson and Walker (1985), p. 16
and Sachs and Hunger (1988), pp. 17-19. The celestial co-ordinates of
the Normal Stars for the epoch 164 BC (the date of the earliest recorded
sighting of Halley's comet in Babylonian history) are shown in figure 6.2.
Spacing of these stars was quite irregular and in particular there was a
significant gap between about 290 and 350 deg longitude; the reason for
this gap is unknown.
Many examples of eclipse observations containing allusions to Normal
Stars will be found in this chapter. For instance, at the eclipse of BC 317
Dec 13/14 the Moon was said to be '\\ cubits in front of (3 Gem', while
at the eclipse of BC 80 Apr 10/11 the Moon was ' | cubit behind a Lib'.
As mentioned in chapter 4, 1 cubit was roughly equivalent to 2 or 2^ deg.
Estimates of the position of the eclipsed Moon in relation to a Normal Star
can sometimes provide valuable aids to verifying dates (see section 6.3).
The terms 'in front of (ina IGI) and 'behind' (ar) - as used in the
above examples - respectively imply locations 'to the west of and 'to the
east of a celestial body - following the diurnal motion of the celestial
sphere. Similarly the expressions 'above' (e) and 'below' (SIG) are also
frequently used in the Late Babylonian astronomical texts. It has yet
to be established whether an ecliptic frame of reference or some other
system was used in these cases. In making calculations, I have somewhat
arbitrarily assumed an ecliptic framework.
For each individual eclipse investigated below, the computed magnitude
 6.2 Lunar eclipses with a single contact measurement 151

is given in parentheses for reference, following the calculated Julian date.

Where only one contact measurement is extant (as in section 6.2) I have
rejected the observation if the magnitude is very small (less than about
0.10). Individual contacts for small partial eclipses are probably difficult to
define with the unaided eye since the Moon then enters the Earth's umbral
shadow obliquely. Under these circumstances, the boundary between the
umbra and the deep penumbral shadow is likely to be poorly defined.
However, where both first and last contact are preserved for an eclipse of
small magnitude, I have utilised the pair of observations to derive a single
mean result for AT (section 6.3).

6.2 Lunar eclipses for which only a single contact measurement relative
to sunrise or sunset is preserved
In this section those measurements which are rounded to the nearest degree
will be considered first (section 6.2.1) followed by the older observations of
lower precision (section 6.2.2). A continuous sequence of numbering entries
will nevertheless be adopted. In some instances, dates have been derived
by Sachs and Schaumberger (1955). This reference will be abbreviated
to SS in the remainder of this chapter. As mentioned in chapter 5, the
publications by Sachs and Hunger (1988, 1989 and 1996) are cited as
SH I, SH II and SH III.
For partial eclipses, a duration of maximal phase is often recorded -
e.g. 5 deg in 143 BC. This is not a real effect and merely arises from
the limited resolution of the human eye. A detailed discussion is given in
section 6.5.

6.2.7 Measurements expressed to the nearest degree

(1) BC 666 Apr 10/11 (mag. = 0.48)
[Samas-sum-ukin], year 2, month I, beginning(?) on [south-east(?)].
(Began) at 3 deg after sunset.
[BM 45640 (= LBAT 1415), Obv. Ill, bottom; trans. Huber, p. 8.]
NB direct consultation of the original tablet by Dr Christopher Walker
(personal communication) shows that the 3 deg (questioned by Huber) is
clear.

RESULTS

First contact on Apr 10 at 0.20 h after sunset. LT of sunset = 18.29 h,

hence LT of first contact = 18.49 h, UT = 15.57 h. Computed TT = 21.42
h, thus AT = 21050 sec.
152 6 Timed Babylonian lunar eclipses

(2) BC 562 Mar 2/3 (mag. = 1.77)

[Nebuchadrezzar] year 42, month XII 15, 1,30 (= 90) deg after sunset
[...] 25 deg duration of maximal phase. In 18 deg it became bright. West
(wind) went. 2 cubits below y Vir eclipsed.
[BM 41536 (= LBAT 1421), col. II, 5'-8'; trans. Huber, pp. 22-23.]
This small tablet is part of a lunar eclipse table. It now records only two
eclipses: in the 6th and 12th months of the 42nd year of an unspecified
king. Such a long reign could only refer to either Nebuchadrezzar II (42nd
year = BC 563/2) or Artaxerxes II (BC 363/2). However, the terminology
is early, and in any case there were only eclipses in the 3rd and 9th months
of the appropriate year of Artaxerxes II. Hence the former year must be
intended. Since the interval between first and second contact in the above
text is not preserved, only the timing of the start of the eclipse can
be utilised. Calculation shows that when the eclipse occurred, the Moon
would be about 3 deg to the south of y Vir, in close accord with the record.

RESULTS
First contact on Mar 2 at 6.00 h after sunset. LT of sunset = 17.66 h,
hence LT of first contact = 23.66 h, UT = 20.97 h. Computed TT = 1.41
h (Mar 3), thus AT = 15 950 sec.

(3) BC 537 Oct 16/17 (mag. = 1.50)

[Cyrus, year 2], month VII [...it made] \ of the disk towards totality.
Not (yet) total, it set eclipsed... 5 deg in front of the Pleiades eclipsed.
(Began) at 14 deg before sunrise.
[BM 32234 (= LBAT 1419), Obv. V, middle; trans. Huber, p. 25.]
In this text, the term US is used both as an angular unit and as a
measure of time. Before dawn on the morning of Oct 17, the Moon would
be about 8 deg to the west of r| Tau, rather more than - but still in fair
agreement with - the estimated figure.

RESULTS
First contact on Oct 17 at 0.93 h before sunrise. LT of sunrise = 6.24 h,
hence LT of first contact = 5.30 h, UT = 2.17 h. Computed TT = 7.39 h,
thus AT = 18 800 sec.

(4) BC 483 Nov 18/19 (mag. = 1.47)

[Xerxes, year 3], month VIII 13. Beginning on the south side, maximal
phase not observed, it set eclipsed. During the eclipse [Venus] stood (there),
the other planets did not stand there. (Began) at 10 deg before sunrise.
[BM 32234 (= LBAT 1419), Rev. Ill; trans. Huber, pp. 27-28.]
 6.2 Lunar eclipses with a single contact measurement 153

The death of Xerxes 18 years later (465 BC) is recorded in the next
entry of this tablet, following a note of a lunar eclipse in that same year
(BC 465 Jun 5/6). Unfortunately, the account of this latter eclipse is too
badly damaged to provide usable information on AT.

RESULTS
First contact on Nov 19 at 0.67 h before sunrise. LT of sunrise = 6.74 h,
hence LT of first contact = 6.07 h, UT = 2.93 h. Computed TT = 7.74 h,
thus AT = 17 300 sec.
(5) BC 421 Feb 2/3 (mag. = 1.64)
[Darius II], year 2, month XI, 15. (At) 19 deg after sunset, beginning on
the east. In 14 [+x deg... ] [... cubits] in front of e Leo eclipsed...
[BM 34787 (= LBAT 1426), col. II, ll'-12'; trans. Huber, p. 34.]
The accession of Darius II (423 BC) is noted in line 4' of this same
column. Since the interval between first and second contact is damaged,
only the timing of the start of the eclipse can be utilised. In the early
evening of Feb 2, the Moon would be more than 20 deg to the east of
e Leo, but only about 1.3 deg to the west of 0 Leo (one of the adjacent
Normal Stars). Possibly the star has been misidentified in the text.
RESULTS
First contact on Feb 2 at 1.27 h after sunset. LT of sunset = 17.26 h,
hence LT of first contact = 18.53 h, UT = 15.86 h. Computed TT = 20.17
h, thus AT = 15500 sec.
(6) BC 406 Oct 9/10 (mag. = 0.97)
[Darius II, year 18, month VII...] (At) 14 deg before sunrise, beginning
on the east [... ] Mercury and [...].
[BM 34684 (= LBAT 1427), Rev. 9-10; trans. Huber, p. 41.]

RESULTS
First contact on Oct 10 at 0.93 h before sunrise. LT of sunrise = 6.12 h,
hence LT of first contact = 5.19 h, UT = 2.07 h. Computed TT = 6.66 h,
thus AT = 16 500 sec.
(7) BC 371 May 17/18 (mag. = 0.78)
[Artaxerxes II, year 34, month II...eclipsed]. During the eclipse, Saturn
stood above a Sco. (Began) at 1,6 (= 66) deg after sunset.
[BM 32238 (= LBAT 1414), Rev. II, top; trans. Huber, pp. 45-46.]
On the above date, Saturn would be almost in conjunction with a Sco
and about 7 deg to the north of it - in satisfactory agreement with the
record.
154 6 Timed Babylonian lunar eclipses

RESULTS
First contact on May 17 at 4.40 h after sunset. LT of sunset = 18.88 h,
hence LT of first contact = 23.28 h, UT = 20.19 h. Computed TT = 23.73
h, thus AT = 12750 sec.
(8) BC 367 Aug 29/30 (mag. = 1.35)
[Artaxerxes II, year 38] month V, night of the 13th, moonrise to sunset:
9 deg. When 56 deg night were left to sunrise, lunar eclipse; on the
south-east side [... ] cubits behind the 'rear container' (qup-pu ar) of
Aquarius it was eclipsed; the north wind blew. During its eclipse Mars (?)

[BM 35184 (= LBAT 186), Rev. 5-6; trans. SH I, p. 135;

see also Huber, p. 48.]
On this astronomical diary, the date is given at the lower edge of the
tablet as: '[...] to the end of month VI, year 38 of king Arses, who is
ca[lled] king Artaxerxes'. The star referred to as the 'rear container' of
Aquarius has not been identified.

RESULTS
First contact on Aug 30 at 3.73 h before sunrise. LT of sunrise = 5.44 h,
hence LT of first contact = 1.71 h, UT = 22.76 h (Aug 29). Computed
TT = 4.19 h (Aug 30), thus AT = 19 550 sec.

(9) BC 317 Jun 18/19 (mag. = 0.37)

[Philip, year 7, month III...] ...4(?) [cubits] in front of P Cap eclipsed.
(Began) at 10 deg after sunset.
[BM 32238 (= LBAT 1414), Rev. IV + V, top; trans. Huber, p. 50.]
The date (year 7 of Philip) is given at the end of the next entry, which
records the eclipse of BC 317 Dec 13/14 (see section 6.4). Soon after
sunset on Jun 18, the Moon would be about 10 deg to the west of P Cap,
in close accord with observation.

RESULTS
First contact on Jun 18 at 0.67 h after sunset. LT of sunset = 19.15 h,
hence LT of first contact = 19.82 h, UT = 16.78 h. Computed TT = 21.11
h, thus AT = 15 550 sec.

(10) BC 308 Jul 8/9 (mag. = 0.99)

[SE 4, month IV, night of the 13th...] 10 deg before sunrise, lunar
eclipse [...] behind the 'rear horn of the goat-fish' (P Cap). The 13th,
sunrise to moonset: 50' ( | deg). Gusty wind.
[BM 40122 (= LBAT 218), 5'-6'; trans. SH I, p. 243.]
 6.2 Lunar eclipses with a single contact measurement 155

Although this tablet is only a small fragment of an astronomical diary,

it contains a variety of planetary and lunar observations. From this
information, the date may be confidently derived by calculation (see SH I,
p. 243). With regard to the eclipse report, before dawn on the morning of
Jul 9 the Moon would be about 8 deg to the east of (or 'behind') P Cap.
The computed interval between sunrise and moonset on this occasion is
0.3 deg, in adequate accord with the recorded | deg.

RESULTS
First contact on Jul 10 at 0.67 h before sunrise. LT of sunrise = 4.86 h,
hence LT of first contact = 4.19 h, UT = 1.22 h. Computed TT = 5.14 h,
thus AT = 14100 sec.

(11) BC 240 Nov 2/3 (mag. = L40)

[SE 72], month VIII 14, at 3 deg before sunrise, beginning on the east
side. It set eclipsed. Month XII2 (? or: cloudy).
[BM 32236) (= LBAT 1218), Rev. 3-5; trans. Huber, p. 55.]
This tablet is a goal-year text for SE 90 (222-221 BC) - (SS, p. xxv). As
is typical of such texts, the eclipse observation was made 18 years before
this goal-year. The last two signs in the entry (SE? DIR) most probably
imply that there was an intercalary 12th month in the year of the eclipse;
this is confirmed for SE 72 (BC 240/239) by the tables of Parker and
Dubberstein (1956). However, the sign DIR can also mean 'cloudy'.

RESULTS
First contact on Nov 3 at 0.20 h before sunrise. LT of sunrise = 6.54 h,
hence LT of first contact = 6.34 h, UT = 3.17 h. Computed TT = 7.12 h,
thus AT = 14200 sec.

(12) BC 212 Apr 29/30 (mag. = 0.62)

Year 100 (SE), king Antiochus (III). Month I 13 moonrise to sunset: 9
deg, measured. Lunar eclipse, beginning on the south side. Around(?)
maximal phase cloudy, not observed. It set eclipsed. (Began) at 20 deg
before sunrise.
[BM 32222 (= LBAT 1237), Rev. 48-53; trans. Huber, p. 56.]
This accurately dated observation is reported on a goal-year text for SE
118 (194-193 BC) - (SS, p. xxv). The computed interval between moonrise
and sunset on Apr 29 is 8.1 deg.

RESULTS
First contact on Apr 30 at 1.33 h before sunrise. LT of sunrise = 5.34 h,
hence LT of first contact = 4.01 h, UT = 0.97 h. Computed TT = 4.25 h,
thus AT = 11800 sec.
156 6 Timed Babylonian lunar eclipses

(13) BC 212 Oct 24/25 (mag. = 0.94)

Year 100 (SE), king Antiochus (III). Month VII 15 moonrise to sunset:
3 deg, measured. Lunar eclipse, beginning on the north-east side. 2 fingers
were lacking to totality. (Began) at 28 (or 27?) deg after sunset.
[BM 32222 (= LBAT 1237), Rev. 56-59; trans. Huber, p. 57.]
This record immediately follows the entry just cited (i.e. BC 212 Apr
30). Inspection of the original tablet by Walker (personal communication)
indicates that despite slight damage, the reading 'after sunset' (i.e. GE6
gin) is quite certain. However, Walker notes that the text is not clear
whether the interval after sunset is 28 or 27 deg, although he prefers the
former. In calculations, I have thus assumed a mean of 27.5 deg. The
computed interval between moonrise and sunset on Oct 24 is 2.3 deg.

RESULTS
First contact on Oct 24 at 1.83 h after sunset. LT of sunset = 17.60 h,
hence LT of first contact = 19.43 h, UT = 16.27 h. Computed TT = 22.20
h, thus AT = 21 350 sec.

(14) BC 194 Nov 4/5 (mag. = 0.92)

The observations of this eclipse are preserved in more than one text:

(i) [SE 118, month VIII.] Night of the 14th moonrise to sunset: 6 deg,
measured (despite) clouds [...lunar eclipse... when ...culminated;
when it began on the east side, in 12 deg night [it made...] it
entered a cloud. In its eclipse, Jupiter, Venus, Mars and Sirius stood
there; the remainder of the planets did not stand there... ] [... ] it
was eclipsed 1 cubit [...] oc Tau; (began) at 12 deg before sunrise.
[BM 35331 (= LBAT 324), Obv. 17'-19'; trans. SH II, p. 279;
see also Huber, p. 58.]
Although the date of this astronomical diary is broken away, it has been
restored using the numerous planetary and lunar observations which are
cited (SH II, 285). When the eclipse occurred, the Moon would be rather
less than 2 deg to the west of a Tau, in close accord with the record. The
computed interval between moonrise and sunset on Nov 4 is 5.0 deg.

(ii) SE 118, month VIII 14, when 'the 4 of (the Lion's) breast' (a, y, r\,
£ Leo) culminated, (lunar eclipse), beginning on the east side. In 12
deg of night it made two-thirds of the disc [...] 1,52 (?)...
[BM 34236 (= LBAT 1436), Obv. 4-6; trans. Huber, p. 58.]
This lunar eclipse text contains only two records: a prediction for the
13th day of month II and the above observation.
 6.2 Lunar eclipses with a single contact measurement 157

RESULTS
First contact on Nov 5 at 0.80 h before sunrise. LT of sunrise = 6.57 h,
hence LT of first contact = 5.77 h, UT = 2.59 h. Computed TT = 6.38 h,
thus AT = 13 650 sec.

(15) BC 160 Jan 26/27 (mag. = 1.45)

[SE 151...] king Demetrius... [month IX] when Gemini culminated,
[lunar eclipse...] the north? wind blew; in its eclipse clouds were in the
sky [...] a Leo it was eclipsed; (began) at 48 deg after sunset.
[BM 46003 (= LBAT 385), Obv. 3'-5'; trans. SH III, p. 41.]
The date of this astronomical diary has been established by calculation
using the numerous planetary and lunar observations which it contains
(SH III, p. 41). The eclipse actually took place in the 2nd year of king
Demetrius I.

RESULTS
First contact on Jan 26 at 3.20 h after sunset. LT of sunset = 17.21 h,
hence LT of first contact = 20.41 h, UT = 17.73 h. Computed TT = 21.64
h, thus AT = 14050 sec.

(16) BC 143 Feb 17/18 (mag. = 0.88)

[SE 168], month XI 13 moonrise to sunset: 3 deg, measured. 5 deg after
(3 Aur culminated, lunar eclipse, beginning on the south-east side. In 20
deg night it made 9 fingers. 5 deg duration of maximal phase. (Began) at 7
deg after sunset.
[BM 35787 (= LBAT 1278), Rev. 3'-8'; trans. Huber, p. 68.]
Although the year in which this eclipse occurred is no longer preserved,
the event is reported in a goal-year text for SE 186 (126-125 BC). The
computed interval between moonrise and sunset on Feb 17 is 2.3 deg. A
photograph of BM 35787 and the joining fragment BM 34658 is shown
in figures 6.3 (obverse) and 6.4 (reverse).

RESULTS
First contact on Feb 18 at 0.47 h after sunset. LT of sunset = 17.51 h,
hence LT of first contact = 17.98 h, UT = 15.31 h. Computed TT = 18.80
h, thus AT = 12 550 sec.

(17) BC 134 Mar 9/10 (mag. = 0.26)

SE 177, month XII [...] when a Lyr culminated, lunar eclipse; when it
began on the east side, it covered 2 fingers in 9 deg of night [... ] [behind] y
Vir it became eclipsed; (began) at 9 deg before sunrise. The 13th, sunrise to
moonset: 1 deg, measured (despite) clouds and mist; the north wind blew.
[BM 34669 (= LBAT 433), Rev. 16'-17; trans. SH III, p. 197.]
158 6 Timed Babylonian lunar eclipses

Fig. 6.3 Photograph of BM 35787 and the joining fragment BM 34658 - obverse.
(Courtesy: British Museum.)

This tablet is an astronomical diary. Its year (177 SE), during the
reign of King Arsaces, is preserved in two places. The computed interval
between sunrise and moonset on Mar 10 is 0.3 deg.

RESULTS

First contact on Mar 10 at 0.60 h before sunrise. LT of sunrise = 6.16 h,

hence LT of first contact = 5.56 h, UT = 2.81 h. Computed TT = 5.87 h,
thus AT = 10950 sec.

(18) BC 109 May 1/2 (mag. = 0.52)

SE 203, Arsaces, king of kings [month I]. Night of the 15th, moonrise to
sunset: 3 deg, measured (despite) mist. 8 deg after sunset, lunar eclipse on
the south side [... ] when it began, it made 6 ringers in 8 deg of night [... ]
maximal phase and clearing; its eclipse was red, its redness was red brown,
the east wind blew; in its eclipse, Jupiter, Venus, Saturn [and Mars stood
there . . . ] .
[BM 40622 (= LBAT 469), Obv. 7 - 8 ' + BM 45646 (+ LBAT 466),
Obv2'; trans SH III, p. 355.]
 6.2 Lunar eclipses with a single contact measurement 159

w?

Fig. 6.4 Photograph of BM 35787 and the joining fragment BM 34658 - reverse.
(Courtesy: British Museum.)

The computed interval between moonrise and sunset on May 1 is 2.5

deg.
The red colour of the Moon was presumably due to a local meteoro-
logical phenomenon (e.g. a dust storm). Unless an eclipse is almost total,
reddening of the Moon caused by dispersion of sunlight in the Earth's
atmosphere before it reaches the lunar surface is relatively insignificant.

RESULTS
First contact on May 1 at 0.53 h after sunset. LT of sunset = 18.69 h,
hence LT of first contact = 19.22 h, UT = 16.17 h. Computed TT = 19.53
h, thus AT = 12050 sec.

(19) BC 96 Aug 3/4 (mag. = 0.71)

[SE 216, month V...] when it began to clear, in 20 deg night [it cleared]
from south to north and we[st ... ] [... the remainder of the planet] s did
not stand there; in Aquarius it became eclipsed; (began) at 57 deg after
sunset. The 13th [...]
[BM 45847 (= LBAT 492), 7 - 8 ' ; trans. SH III, p. 425; see also Huber, p. 73.]
160 6 Timed Babylonian lunar eclipses

This astronomical diary has been dated by planetary and lunar calcu-
lations (SH III, p. 425).

RESULTS
First contact on Aug 3 at 3.80 h after sunset. LT of sunset = 18.91 h,
hence LT of first contact = 22.71 h, UT = 19.81 h. Computed TT = 23.47
h, thus AT = 13 150 sec.

(20) BC 80 Apr 10/11 (mag. = 0.60)

Year 168 (Arsacid), that is year 232 (SE), Arsaces, king of kings, which
is in the time of king Orodes (I). Month I 13 moonrise to sunset: 5,50 (=
5|) deg, cloudy(?), measured(?). 5 deg before \i Her culminated, lunar
eclipse, beginning on the south-east side. In 20 deg night it made 6 fingers.
7 deg duration of maximal phase, until it began to become bright. In 13
deg from south-east to north-west, 4 fingers lacking to brightness, it set
[...] During this eclipse, north wind (?) went. During this eclipse, Jupiter,
Saturn and Mars stood there. When becoming bright, Saturn set. The
other planets did not stand there. | cubits behind a Lib, 6 fingers towards
the south beneath. (Began) at 40 deg before sunrise.
[BM 33562A (= LBAT 1445), Obv. and Rev.; trans. Huber, pp. 75-76.]
This tablet is devoted to this eclipse. Years are reckoned both from the
Arsacid era (247 BC) and the Seleucid era. Orodes I, a Parthian king,
began to reign in 80 BC. When the eclipse occurred, the Moon would be
about 0.6 deg to the east of oc Lib and about 2 deg to the south of it
- in fair (but not good) accord with the record. The computed interval
between moonrise and sunset on Apr 10 is 7.4 deg.

RESULTS
First contact on Apr 11 at 2.67 h before sunrise. LT of sunrise = 5.62 h,
hence LT of first contact = 2.95 h, UT = 0.00 h. Computed TT = 3.19 h,
thus AT = 11450 sec.

(21) BC 80 Oct 5/6 (mag. = 0.39)

[...] year 232 (SE) [king Arsa]ces [...] which is in the time of king
Orodes (I) and Is [... his wife] [... month VII... ] (Began) at 30 deg after
sunset.
[BM 42073 (= LBAT 1446), Obv. and Rev.; trans. Huber, p. 76.]
This badly damaged tablet is concerned specifically with this particular
eclipse. There were only two lunar eclipses in the year 232 SE. The earlier
of these two events, which occurred in month I, has already been discussed
in the previous entry. In view of the differing details, the present record
must relate to the later eclipse.
 6.2 Lunar eclipses with a single contact measurement 161

Fig. 6.5 Photograph of BM 41565 - obverse. (Courtesy: British Museum.)

RESULTS
First contact on Oct 5 at 2.00 h after sunset. LT of sunset = 17.90 h,
hence LT of first contact = 19.90 h, UT = 16.78 h. Computed TT = 20.13
h, thus AT = 12050 sec.

(22) BC 67 Jan 19/20 (mag. = 0.81)

Year 180 (Arsacid), that is year 244 (SE), king Arsaces (IV) and
Pir'ustana, his wife, the lady, month X 15. Moonrise to sunset: 1 deg, mist,
measured. As the Moon rose, two-thirds of the disk on the north-east side
were eclipsed. 6 deg night duration of maximal phase, until it began to
become bright. In 16 deg night from south-east to north-west it became
bright. 23 deg total duration. During this eclipse the sky was overcast.
During this eclipse, north (wind) went. During this eclipse, Venus, Saturn
and Sirius stood there, the other planets did not stand there. 11/2 cubits
in front of oc Leo. (Began) at 16 deg before sunset.
[BM 45628 (= LBAT 1448) + Dupl. BM 41565 (= LBAT 1447), Obv.
and Rev; trans. Huber, pp. 77-78.]
These two tablets are duplicates of one another. Each records only this
particular eclipse. A photograph of the surviving fragment of BM 41565
(obverse) is shown in figure 6.5. It is evident that the recorded time of
first contact (i.e. 16 deg before sunset) represents a prediction. When the
162 6 Timed Babylonian lunar eclipses

Moon rose it was said to be already largely obscured. The eclipse appears
to have ended 22 or 23 deg after moonrise. The discrepancy of 1 deg in
the arithmetic (i.e. 6 + 16 = 22 rather than 23) is presumably due to a
scribal error. A measured time-interval of 22.5 deg (1.50 h) after moonrise
will be assumed for last contact. When the eclipse occurred, the Moon
was 2 deg to the west of oc Leo, which accords well with the text. The
computed interval between moonrise and sunset on Jan 19 is 0.7 deg.

RESULTS
Last contact on Jan 19 at 1.50 h after moonrise. LT of moonrise = 17.10
h, hence LT of last contact = 18.60 h, UT = 15.89 h. Computed TT =
18.71 h, thus AT = 10150 sec.
The various results obtained above are summarised in table 6.1.

6.2.2 Other less accurate measurements (to the nearest 5 or 10 degrees)

(23) BC 695 Apr 30/May 1 (mag. = 0.74)
[Assur-nadin-sum], year 5, month II l[0+x...not] complete, it set
eclipsed. (Began) at 30 deg before sunrise.
[BM 32238 (=LBAT 1414), Obv. Ill; trans. Huber, p. 4.]

RESULTS
First contact on May 1 at 2.00 h before sunrise. LT of sunrise = 5.40 h,
hence LT of first contact = 3.40 h, UT = 0.36 h. Computed TT = 5.01 h,
thus AT = 16 750 sec.

(24) BC 686 Apr 21/22 (mag. = 0.55)

[Se]nnacherib [year 3], month II(?)...5] months [...] beginning on the
[north-] east side. \ of the disk [...] west (wind) went. (Began at) 1,40 (=
100) deg after sunset. Month VI2.
[BM 35789 (= LBAT 1417), Obv. I + BM 35115 (= LBAT 1416),
Obv. I, top; trans. Huber, p. 5.]
Sennacherib, who became ruler of Assyria in 704 BC, assumed the
ancient title of 'King of Sumer and Akkad' in 688 BC following his
capture of the city of Babylon (Oates, 1986, pp. 119-20). The VI 2 at the
end of the above entry identifies the intercalary month in that year; this
occurred several months after the eclipse.

RESULTS
First contact on Apr 22 at 6.67 h after sunset. LT of sunset = 18.46 h,
hence LT of first contact = 1.13 h, UT = 22.14 h (Apr 21). Computed TT
= 4.40 h (Apr 22), thus AT = 22 500 sec.
 6.2 Lunar eclipses with a single contact measurement 163

Table 6.1 AT results from Babylonian timings of lunar eclipses

estimated to the nearest degree (single contacts only).

Yeara Ct SR/SS* Int (deg)c AT (sec)*

-665 1 SS +3 21050
-561 1 SS +90 15 950
-536 1 SR -14 18 800
-482 1 SR -10 17 300
-420 1 SS +19 15 500

-405 1 SR -14 16 500

-370 1 SS +66 12 750
-366 1 SR -56 19 550
-316 1 SS +10 15 550
-307 1 SR -10 14100

-239 1 SR -3 14200
-211a 1 SR -20 11800
-211b 1 SS +27.5 21350
-193 1 SR -12 13 650
-159 1 SS +48 14050

-142 1 SS +7 12 550
-133 1 SR -9 10950
-108 1 SS +8 12050
-95 1 SS +57 13150
-79a 1 SR -40 11450

-79b 1 SS +30 12050

-66 4 MR +22.5 10150
a
The year of the eclipse on the Julian calendar, using negative integers.
b
Whether the eclipse was timed relative to sunrise (SR), sunset (SS), or moonrise
(MR) - the last instance only in the year —66.
c
The measured interval in degrees (positive if after sunset or moonrise, negative
if before sunrise).
d
The value of AT (in seconds) which satisfies the observation.
NB -211a refers to the first eclipse in the year -211 (Apr 29/30), etc.

(25) BC 685 Oct 3/4 (mag. = 1.83)

[Sennacherib, year 4, month VI(?)... ] 6(?) deg [... ] middle, Aries [... ]
south. (Began) at 20 deg after sunset.

[BM 45640 (= LBAT 1415), Obv. II, top; trans. Huber, p. 6.]
164 6 Timed Babylonian lunar eclipses

RESULTS
First contact on Oct 3 at 1.33 h after sunset. LT of sunset = 18.01 h,
hence LT of first contact = 19.34 h, UT = 16.27 h. Computed TT = 21.55
h, thus AT = 19000 sec.

(26) BC 667 Oct 14/15 (mag. = 1.86)

[Samas-sum-ukin, year 1, month VII(?)...] It set eclipsed. (Began) at
20 [+x] deg before sunrise.
[BM 45640 (= LBAT 1415), Obv. Ill, top; trans. Huber, pp. 7-8.]
The immediately following entry on the same tablet (Obv. Ill, bottom)
records an eclipse in month I of the second year of the same ruler - see
entry (1) above. Hence, although the month is not specified in the text
under discussion, it should be either 5 or 6 months previously and thus
in month VII or VIII. Only the eclipse of BC 667 Oct 14/15 can fit this
date.
According to a personal communication from Walker, who has exam-
ined the original tablet on my behalf, the 20 in 20 [+x] is clear and almost
certainly not a damaged 30. He is in doubt whether the sign between the
20 and GE6 ana ZALAG ('before sunrise') represents a numeral or not. I
have taken 25 deg as a fair average.

RESULTS
First contact on Oct 15 at 1.67 h before sunrise. LT of sunrise = 6.18 h,
hence LT of first contact = 4.51 h, UT = 1.39 h. Computed TT = 6.19 h,
thus AT = 17 250 sec.
NB if the text had actually read 20 deg before sunrise, the corresponding
value of AT would have been 16050 sec.

(27) BC 601 Apr 10/11 (mag. = 0.84)

[Nebuchadrezzar II, year 4] month I, 13, middle watch, 3 beru, 5 US (=
95 deg) after sunset [... ] north-east side, 3 (fingers) lacked to totality [... ]
on north it became bright. North (wind) went.
[BM 38462 (= LBAT 1420), Obv. I, 8-10; trans. Huber, p. 13.]
The three watches into which the night was divided - USAN ('first part
of the night'), MURUB 4 ('middle part') and ZALAG (last part') - are
only mentioned in fairly early texts.

RESULTS
First contact on Apr 11 at 6.33 h after sunset. LT of sunset = 18.31 h,
hence LT of first contact = 0.64 h, UT = 21.71 h (Apr 10). Computed TT
= 2.72 h (Apr 11), thus AT = 18000 sec.
 6.2 Lunar eclipses with a single contact measurement 165

(28) BC 599 Feb 19/20 (mag. = 0.75)

[Nebuchadrezzar II, year 5], month VI2 passed. [Month XI] (after) 5
months, 3^ beru (= 105 deg) after sunset [...] | covered, on the south
covered [...during] this eclipse [...].
[BM 38462 (= LBAT 1420), Obv. I, 12-15; trans. Huber, p. 13.]
The true lunar month (XI) can be readily restored since it was said to
be 5 months after a predicted eclipse in the intercalary 6th month.

RESULTS
First contact on Feb 20 at 7.00 h after sunset. LT of sunset = 17.49 h,
hence LT of first contact = 0.48 h, UT = 21.83 h (Feb 19). Computed TT
= 1.83 h (Feb 20), thus AT = 14400 sec.

(29) BC 587 Jan 7/8 (mag. = 1.80)

[Nebuchadrezzar II, year] 17, month IV (eclipse) passed. [Month X] 13,
morning watch(?) 1 beru 5 U[S] (= 35 deg) [before sunrise...] it set totally
[eclipsed].
[BM 38462 (= LBAT 1420), Obv. II, 16-18; trans. Huber, p. 17.]
Although the interval to which the measurement of 35 deg refers is
broken away, it can confidently be restored as 'before sunrise'. The
characteristic style of BM 38462 is to cite the time interval measured after
sunset or before sunrise near the start of a record. Since the Moon set
eclipsed, it is clear that the 35 deg cannot have been measured after sunset.

RESULTS
First contact on Jan 8 at 2.33 h before sunrise. LT of sunrise = 6.97 h,
hence LT of first contact = 4.64 h, UT = 1.86 h. Computed TT = 7.10 h,
thus AT = 18 850 sec.

(30) BC 580 Aug 14/15 (mag. = 1.82)

[Nebuchadrezzar II] year 25, month V, 1 \ beru (= 45 deg) after sunset.
[BM 38462 (= LBAT 1420), Rev. I, 3'; trans. Huber, p. 18.]
This very brief record is intact. Evidently the scribe who compiled the
eclipse table could obtain no information on the times of the later phases.
There is no reference to the eclipse 'passing by', which suggests that it was
observed.

RESULTS
First contact on Aug 14 at 3.00 h after sunset. LT of sunset = 18.81 h,
hence LT of first contact = 21.81 h, UT = 18.89 h. Computed TT = 0.19
h (Aug 15), thus AT = 19050 sec.
166 6 Timed Babylonian lunar eclipses

(31) BC 573 Apr 1/2 (mag. = 1.73)

[Nebuchadrezzar II, year 31, month XII...] South (wind) went (?) 1 (?)
cubit in front of Libra eclipsed. Saturn in Capricorn... Mars 2 cubits in
front of a Sco. (Began) at 1,30 (= 90) deg after sunset.
[BM 32234 (= LBAT 1419), Obv. Ill, middle; trans. Huber, p. 21.]
The immediately following entry in the same column of this text specif-
ically mentions Nebuchadrezzar in recording an eclipse which 'passed by'
in month VI of his 32nd year (i.e. BC 573 Sep 26). On the night of Apr
1/2 in BC 573, Mars was about 5 deg (or roughly 2 cubits) distant from
a Sco but it was located to the north of this star rather than west of it (as
implied in the text).

RESULTS

First contact on Apr 2 at 6.00 h after sunset. LT of sunset = 18.16 h,

hence LT of first contact = 0.16 h, UT = 21.28 h (Apr 1). Computed TT
= 2.30 h (Apr 2), thus AT = 18050 sec.

(32) BC 523 Jul 16/17 (mag. = 0.54)

Year 7 (Kambyses), month IV, night 14, l | beru (= 50 deg) after sunset,
the Moon makes a total eclipse, (but) a little is left over; north (wind) went.
[BM 33066 (= LBAT 1477), Rev. 19-20; trans. Huber, p. 25.]
Both this and the following entry are recorded on the same tablet from
the reign of king Kambyses II (529-522 BC). The terminology of the
entry ('the Moon makes a total eclipse, (but) a little is left over') is most
unusual. The eclipse described above is also reported in the Almagest (see
chapter 4), but with differing details. There it is stated that first contact
took place 1 hour before midnight and that the Moon was eclipsed by
half its diameter. As sunset occurred at 19.11 h, the local time of first
contact as recorded on the cuneiform tablet would actually correspond
to 1.56 h before midnight, while an almost total eclipse is (incorrectly)
implied. Regardless of which measurement of time is the more accurate,
the Almagest provides a sound estimate of magnitude.
As the recorded magnitude on the cuneiform text is considerably greater
than the true value, Huber was of the opinion that this record represents
a prediction, although the text mentions wind and the phrase sin AN
KU normally relates to an observation. He further remarked that the
obverse of the same tablet certainly contains predictions. Because of these
uncertainties, the record will not be analysed for AT. Similar remarks
apply to the immediately following report on BM 33066 (entry 33 below).
 6.3 Eclipses with two contact measurements 167

Table 6.2 AT results from Babylonian timings of lunar

eclipses estimated to the nearest 5 or 10 deg (all single con-
tacts).

Year Ct SR/SS Int. (deg) AT (sec)

-694 1 SR -30 16 750
-685 1 ss +100 22 500
-684 1 ss +20 19000
-666 1 SR -25 17 250
-600 1 SS +95 18000

-598 1 ss +105 14400

-586 1 SR -35 18 850
-579 1 SS +45 19050
-572 1 ss +90 18 050

(33) BC 522 Jan 9/10 (mag. = 1.85)

Year 7 (Kambyses), month X, night 14, 2^ beru (= 75 deg) to sunrise are
left over, the Moon makes a total eclipse. South and north, clouded, went.
[BM 33066 (= LBAT 1477), Rev. 21-22; trans. Huber, pp. 25-26.]
Totality is correctly described in this brief account, but for the reasons
given in the previous entry the record will not be considered further.

Table 6.2 gives a summary of the results obtained in this section. The
form of this table is similar to that of table 6.1.
The AT values listed in tables 6.1 and 6.2 are plotted in figure 6.6. This
diagram provides clear evidence of a gradual decline in AT down the
centuries, and with a single exception (the high value in —211), the results
form a remarkably self-consistent set.

6.3 Eclipses for which two contact measurements relative to sunrise or

sunset are extant
Most of the observations discussed in this section are of partial lunar
eclipses, the times of both first and last contact being preserved. However,
also considered here are two total eclipses in 189 and 120 BC. On the
former occasion the Moon set whilst eclipsed and only the first two
contacts were timed. Although the eclipse of 120 BC is computed to have
been just total (magnitude 1.02), it was actually described as incomplete
at maximal phase. Two observations, dated by calculation as BC 239 Apr
28/29 (SH II, p. 85) and BC 154 Mar 21/22 (Huber, pp. 66-67) have
been rejected. In each case, the eclipse is reported on a tablet specifically
168 6 Timed Babylonian lunar eclipses

20000
o cP •
o o

16000
° *

3 12000

8000 -

4000 Key
• Measurement expressed to nearest 1°
: o Measure rounded to 5° or 10°
n • i i i

-700 -600 -500 -400 -300 -200 -100 0

Year

Fig. 6.6 AT values obtained from timed lunar eclipses - single contact only.

devoted to that event but the date is broken away. Although the derived
dates are probably sound, there is very little independent information
in either text to assist dating; without pre-supposing a value for AT,
alternative identifications cannot be ruled out.
I have included in this investigation two eclipses of very small magnitude
(close to 0.10). These date from 397 and 163 BC. Timings of both first
and last contact are available for each event. As noted above (section 6.1),
definition of the individual contacts seems likely to have been poor. This
may explain why the duration of the eclipse of 397 BC - estimated by the
Babylonian astronomers as 27 deg - contrasts with the computed figure
of only 16 deg. Yet in 163 BC there was remarkably close agreement
between observation (20 deg) and theory (21 deg) for this parameter. It is
perhaps best to use only the mean result for both contacts in each case.
This will be identified by the letter m in the summary table (6.3) at the
end of this section.

(1) BC 424 Sep 28/29 (mag. = 0.93)

[Artaxerxes I, year 41], month VI 14, 50 deg after sunset, beginning on
the north-east. After 22 deg, 2 fingers lacked to totality. 5 deg duration of
maximal phase. In 23 deg towards [... ] 50 deg total duration. The sky was
overcast... 3 cubits below oc + (3 Ari.
[BM 34787 (= LBAT 1426), col. II, 2'-3'; trans. Huber, pp. 32-33.]
 6.3 Eclipses with two contact measurements 169

Direct consultation of the original tablet by Walker (personal commu-

nication) shows that the 50 deg (questioned by Huber) is very clear. When
the eclipse occurred, the Moon was about 10 deg to the south of a and (3
Ari, in fair agreement with the record.

RESULTS

(1) First contact on Sep 28 at 3.33 h after sunset. LT of sunset = 18.07 h,

hence LT of contact = 21.40 h, U T = 18.33 h. Computed T T = 22.98 h,
thus A T = 16700 sec.
(ii) Last contact on Sep 29 at 6.67 h after sunset. LT of contact = 0.74
h, U T = 21.66 h (Sep 28). Computed T T = 2.40 h (Sep 29), thus A T =
17050 sec.

(2) BC 408 Oct 31/Nov 1 (mag. = 0.18)

[Darius II, year 16], month VIII 15, 15 deg after sunset, beginning on
the south-east. A quarter of the disk was eclipsed [... ] 27 deg total
duration... 1 cubit in front of Taurus eclipsed [... ] and Mars stood there

[BM 34684 (= LBAT 1427), Obv. lff-13'; trans. Huber, pp. 38-39.]
When the eclipse occurred, the Moon was about 2.5 deg to the west
of a Tau (the principal star of Taurus). This is in good accord with
observation.

RESULTS

(i) First contact on Oct 31 at 1.00 h after sunset. LT of sunset = 17.51 h,

hence LT of contact = 18.51 h, UT = 15.34 h. Computed TT = 19.57 h,
thus AT = 15 250 sec.
(ii) Last contact at 2.80 h after sunset. LT of contact = 20.31 h, UT =
17.14 h. Computed TT = 21.11 h, thus AT = 14 300 sec.

(3) BC 397 Apr 5/6 (mag. = 0.09)

[Artaxerxes II, year 7], month XII2, 14. Beginning on the south side, a
quarter of the disk covered. It became bright towards the west. 27 deg
total duration. The sky was overcast, south (wind) went. (Began) at 48 deg
after sunset.
[BM 35115 (= LBAT 1416), Rev. II, middle; trans. Huber, p. 41.]

RESULTS

(i) First contact on A p r 5 at 3.20 h after sunset. LT of sunset = 18.25 h,

hence LT of contact = 21.45 h, U T = 18.54 h. Computed T T = 22.84 h,
thus A T = 15 450 sec.
170 6 Timed Babylonian lunar eclipses

(ii) Last contact at 5.00 h after sunset. LT of contact = 23.25 h, UT =

20.34 h. Computed TT = 23.92 h, thus AT = 12 850 sec.
Since the magnitude was so small, only the mean AT value of 14150
sec will be retained in further analysis.

(4) BC 189 Feb 16/17 (mag. = 1.28)

[Year 122 (SE)] month XI 14. Moonrise to sunset 9 deg, mist, measured.
Lunar eclipse beginning on the east side. After 16 deg of night totally
covered. It set eclipsed. (Began) at 34 deg before sunrise.
[BM 34519 (= LBAT 1251), Rev. 23-26; trans. Huber, p. 61.]
This tablet is a goal-year text for SE 140 (172-171 BC). Following the
usual practice on this type of tablet, the eclipse observation was made 18
years prior to this goal-year. The computed interval between moonrise
and sunset on Feb 16 (8.2 deg) is close to the recorded figure.

RESULTS
(i) First contact on Feb 17 at 2.27 h before sunrise. LT of sunrise = 6.51
h, hence LT of contact = 4.24 h, UT = 1.58 h. Computed TT = 4.57 h,
thus AT = 10750 sec.
(ii) Second contact at 1.20 h before sunrise. LT of contact = 5.31 h, UT
= 2.65 h. Computed TT = 5.78 h, thus AT = 11250 sec.

(5) BC 163 Mar 30/31 (mag. = 0.12)

The observations of this eclipse are preserved in three separate texts. Each
tablet is damaged but the surviving accounts supplement one another.
(i) [Year 148 (SE) king Antiochus IV/V, month XII2 15...], cloudy,
measured. During the whole night cirrus clouds (?). 3 deg after u
Boo culminated, [... ] From south to west it became bright. 20 deg
total duration. During this eclipse the sky was overcast.... During
this eclipse north wind... [...] before sunrise. On the 15th sunrise
to moonset 5,30 (=5^) deg.
[Composite translation of BM 41628 (= LBAT 378), Rev. 12'-13'
and BM 41462 (= LBAT 380), Rev. 20/-21/; trans. Huber, p. 64.]
Both BM 41628 and BM 41462 are astronomical diaries. Their dates,
which prove to be identical, have been independently derived by calcula-
tion using the planetary and lunar observations which they contain (SS, p.
xvi). They thus represent separate copies of the same diary. The computed
interval between sunrise and moonset on Mar 31 is 4.1 deg.
(ii) Year 14[8 (SE) king Antiochus IV/V], month XII2 [15]. 3 deg after
i) Boo culminated, lunar eclipse, beginning on the south side. In
10 deg of night it made 3fingers.(Began) at 85 deg before sunrise.
[BM 34037 (= LBAT 1264), Rev. 3'-9'; trans. Huber, pp. 63-64.]
 6.3 Eclipses with two contact measurements 111

This tablet is a goal-year text for SE 167 (145-144 BC). The eclipse
observation was made 18 years before this goal-year.
NB Antiochus IV died a few months before the eclipse, in the same
year (month IX of SE 148). This is why Huber names both Antiochus IV
and his son Antiochus V in restoring the text.

RESULTS
(i) First contact on Mar 31 at 5.67 h before sunrise. LT of sunrise = 5.81
h, hence LT of contact = 0.14 h, UT = 21.27 h (Mar 30). Computed TT
= 23.94 h (Mar 30), thus AT = 9600 sec.
(ii) Last contact at 4.33 h before sunrise. LT of contact = 1.48 h, UT =
22.61 h (Mar 30). Computed TT = 1.32 h (Mar 31), thus AT = 9750 sec.
Since the magnitude was so small, only the mean AT value of 9700 sec
will be retained in subsequent analysis.

(6) BC 129 Nov 4/5 (mag. = 0.63)

Year 120 (Arsacid), that is year [183] (SE), king Arsaces, month VIII
13. Moonrise to sunset 4,30 (= 4^) deg, measured [...] after (a, (3) Gem
culminated, lunar eclipse beginning on the [north-east] side. [In 18 deg]
night [it made ...fingers]6 deg duration of maximal phase [... ] In 16 deg
from north(?) [...]. 40 deg total duration [...] distant; in the beginning of
a halo [...]. During this eclipse, Sirius [stood] there. The other planets did
not [stand there]. 2 cubits above a Tau [...]. (Began) at 55 deg before
sunrise. On the 13th, sunrise to moonset 3,30 (=3^) deg, cloudy, not
observed.
[BM 33982 (= LBAT 1441), Obv. and Rev.; trans. Huber, pp. 70-71.]
This tablet is devoted to the record of a single eclipse. Huber notes an
error of one year in the date relative to the Arsacid era. He proposes year
119 rather than 120. There was no eclipse in month VIII of the latter
year, and although one did occur in month VII - i.e. BC 128 Oct 25 - the
calculated circumstances do not accord with observation. For instance, the
computed duration of maximal phase (totality) would be fully 25 deg; this
is much greater than the 6 deg of the text. Further, using a preliminary
value for AT of 11900 sec (from equation (4.1)) - a figure supported by
other roughly contemporary observations discussed in this chapter - this
eclipse would occur around noon at Babylon. It would thus be entirely
invisible from this site. At greatest phase, the Moon would be about 11
deg to the west of oc Tau (rather than 2 cubits above this star), while the
reported intervals between moonrise and sunset and sunrise to moonset
are considerably at variance with theory.
On the contrary, the eclipse of BC 129 Nov 4/5 occurred in the correct
lunar month. At the time, the Moon would be about 1 deg to the west
of oc Tau and some 6 deg to the north of this star - in good accord with
172 6 Timed Babylonian lunar eclipses

observation. Further, at the measured time of 55 deg before sunrise (LT

= 2.92 h), all of the planets (but not the star Sirius) would indeed be
below the horizon. The computed interval between moonrise and sunset
on Nov 4 was 6.3 deg, while that between sunrise and moonset on Nov 5
was 2.9 deg. Both of these are in fairly good agreement with the record.
For these various reasons, the date proposed by Huber would seem to be
satisfactory.

RESULTS

(i) First contact on Nov 5 at 3.67 h before sunrise: LT of sunrise = 6.58

h, hence LT of contact = 2.92 h, UT = 23.74 h (Nov 4). Computed TT =
3.04 h (Nov 5), thus AT = 11 850 sec.
(ii) Last contact at 1.00 h before sunrise. LT of contact = 5.58 h, UT =
2.41 h. Computed TT = 5.91 h, thus AT = 12600 sec.

(7) BC 120 Jun 1/2 (mag. = 1.02)

Year 128 (Arsacid), that is year 192 (SE), king Arsaces. Month II 14.
Moonrise to sunset 6,30 (=6^) deg, measured. 5 deg after a Her
culminated, lunar eclipse, beginning on the north-east side. After 24 deg, 1
UD lacked to totality. 6 deg duration of maximal phase, until it began to
become bright. In 24 deg, it became bright from north-east to south-west.
54(?) deg total duration... (Began) at 66 deg after sunset.
[BM 45845 (= LBAT 1442), Obv. and Rev.; trans. Huber, pp. 71-72.]
This tablet records no more than the above eclipse. Although the eclipse
is calculated to have been just total, the record implies that a portion of
the Moon ('1 UD': probably a small fraction) remained unobscured. The
interval from start to finish is damaged but this can readily be restored
from the sum of the durations of the individual phases (i.e. 54 deg). The
computed interval between moonrise and sunset on Jun 1 is 5.5 deg.

RESULTS

(i) First contact on Jun 1 at 4.40 h after sunset. LT of sunset = 19.06 h,

hence LT of contact = 23.46 h, UT = 20.38 h. Computed TT = 23.86 h,
thus AT = 12 500 sec.
(ii) Last contact on Jun 2 at 8.00 h after sunset. LT of contact = 3.06 h,
UT = 23.98 h (Jun 1). Computed TT = 3.01 h (Jun 2), thus AT = 10900
sec.

The results derived in this section are summarised in table 6.3.

 6.4 Sets of three or four timings 173

Table 6.3 AT results from Babylonian timings of lunar eclipses

for which two contacts were observed.
Yearfl Ct* SR/SS C Int (deg)^ AT (sec)"
-423 1 SS +50 16700
-423 4 ss +100 17050
-407 1 ss +15 15 250
-407 4 ss +42 14 300
-396 m ss +61.5 14150

-188 1 SR -34 10750

-188 2 SR -18 11250
-162 m SR -75 9 700
-128 1 SR -55 11850
-128 4 SR -15 12600

-119 1 SS +66 12 500

-119 4 ss +120 10900
a
The year of the eclipse on the Julian calendar, using negative integers.
b
The contact; for the very small eclipses of —396 and —162, mid-eclipse (m) is
used.
c
Whether the eclipse was timed relative to sunrise (SR) or sunset (SS).
d
The measured interval in degrees (positive if after sunset or moonrise, negative
if before sunrise).
e
The value of AT (in seconds) which satisfies the observation.

6.4 Three or four timings relative to sunrise or sunset preserved

(total eclipses only)
(1) BC 555 Oct 6/7 (mag. = 1.53)
[Nabunaid, year 1], month VII 13. In 17 deg (from the) east side total.
28 deg duration of the maximal phase. In 20 deg from east to west it
became bright. This eclipse was clouded (?). Behind oc and (3 Ari eclipsed.
Towards becoming bright west (wind) went. (Began) at 55 deg before
sunrise.
[BM 32234 (= LBAT 1419), Obv. IV, middle; trans. Huber, p. 23.]
Although the sum of the durations of the individual phases is 65 deg,
this eclipse is said to have begun at 55 deg before sunrise. There is clearly
a textual error somewhere; otherwise last contact would occur long after
moonset. Walker has kindly verified the reading of the various numbers
in the text. It will be seen that the measured interval between first and
second contact (17 deg) is in fair accord with that between third and
fourth contact (20 deg); these figures also agree well with the computed
interval of 17 deg. In addition, the recorded duration of totality (28 deg)
174 6 Timed Babylonian lunar eclipses

is close to the calculated figure of 25 deg. Hence it seems likely that

the timing of first contact is incorrect. Huber (p. 23) preferred to read
'after sunset' (GE6 gin) for 'before sunrise' (GE6 ana ZALAG). This is
not unreasonable, but it is also plausible that the scribe made a numerical
error in recording the interval between first contact and sunrise. It thus
seems best to reject this record.
(2) BC 501 Nov 7/8 (mag. = 1.47)
[Darius I, year 21], month VIII. In 15 deg (from) the east total. 25 deg
duration of the maximal phase. In 25 deg from east to west it became
bright. Clouded (?) Towards becoming bright, south (wind) went. (Began)
at 1,17 (= 77) deg after sunset.
[BM 32234 (= LBAT 1419), Rev. II; trans. Huber, p. 27.]
Direct consultation of the original tablet by Walker (personal commu-
nication) shows that the first '1' in '1,17 deg' (questioned by Huber) is
fairly clear.
RESULTS
(i) First contact on Nov 7 at 5.13 h after sunset. LT of sunset = 17.41 h,
hence LT of contact = 22.55 h, UT = 19.39 h. Computed TT = 23.54 h,
thus AT = 14950 sec.
(ii) Second contact at 6.13 h after sunset. LT of contact = 23.55 h, UT
= 20.39 h. Computed TT = 0.65 h (Nov 8), thus AT = 15 300 sec.
(iii) Third contact on Nov 8 at 7.80 h after sunset. LT of contact = 1.22
h, UT = 22.06 h (Nov 7). Computed TT = 2.22 h (Nov 8), thus AT =
14 950 sec.
(iv) Fourth contact at 9.47 h after sunset. LT of contact = 2.88 h, UT
= 23.72 h (Nov 7). Computed TT = 3.33 h (Nov 8), thus AT = 13 000
sec.
(3) BC 407 Oct 20/21 (mag. = 1.38)
[Darius II, year 17], month VII 14, 48 deg before sunrise, beginning in
the east. In '21' deg total, 12(?) deg [duration of maximal phase]. In 15 deg
not [complete it set(?)...].
[BM 34684 (= LBAT 1427), Rev. 4-6; trans. Huber, pp. 39-40.]
Examination of the original tablet by Walker (personal communication)
indicates that the '21' (questioned by Huber) is clear, but the subsequent
interval - '12'(?) - is indeed doubtful. However, Huber's interpretation
that the Moon set at 15 deg after the end of totality is probably correct
so that the '12' may be restored (i.e. 21 + 12 + 15 = 48). This interval
almost certainly relates to the duration of maximal phase, as inferred by
Huber.
 6.4 Sets of three or four timings 175

RESULTS

(i) First contact on Oct 21 at 3.20 h before sunrise. LT of sunrise = 6.31

h, hence LT of contact = 3.11 h, UT = 23.97 h (Oct 20). Computed TT
= 4.12 h (Oct 21), thus AT = 14950 sec.
(ii) Second contact at 1.80 h before sunrise. LT of contact = 4.51 h, UT
= 1.37 h. Computed TT = 5.23 h, thus AT = 13 850 sec.
(iii) Third contact at 1.00 h before sunrise. LT of contact = 5.31 h, UT
= 2.17 h. Computed TT = 6.68 h, thus AT = 16200 sec.

(4) BC 378 Apr 6/7 (mag. = 1.32)

[Artaxerxes II, year 26], month XII2 15, beginning on the south side.
[After] 15 deg total, 21 deg duration of maximal phase. When it began to
become bright from the east, [in] 19 deg toward the west it became bright.
45 deg (error for 55 deg ?) total duration. 14fingersin front of a Lib
eclipsed... (Began) at 37 deg [after sunset or before sunrise].
[BM 45640 (= LBAT 1415), Rev. Ill, middle; trans. Huber, p. 45.]
Huber was of the opinion that the sign for 15 deg (i.e the onset of
totality) was indistinct. However, inspection of the tablet by Walker
(personal communication) reveals that the sign is clear. The sum of the
durations of the individual phases is 55 deg rather than the stated 45; the
latter thus probably represents a scribal error. Whether the eclipse began
37 deg before sunrise or after sunset is not preserved. However, if this
interval was measured relative to sunrise, the end would occur long after
moonset and thus be unobservable. Hence it may reasonably be inferred
that the time of first contact was measured after sunset. This conclusion
is confirmed by calculation of the position of the Moon in relation to oc
Lib.

RESULTS

(i) First contact on Apr 6 at 2,47 h after sunset. LT of sunset = 18.26 h,

hence LT of contact = 20.73 h, UT = 17.82 h. Computed TT = 22.27 h,
thus AT = 16000 sec.
(ii) Second contact at 3.47 h after sunset. LT of contact = 21.73 h, UT
= 18.82 h. Computed TT = 23.32 h, thus AT = 16200 sec.
(iii) Third contact at 4.87 h after sunset. LT of contact = 23.13 h, UT
= 20.22 h. Computed TT = 0.60 h (Apr 7), thus AT = 15 750 sec.
(iv) Last contact on Apr 7 at 6.13 h after sunset, hence LT of contact
= 0.40 h, UT = 21.49 h (Apr 6). Computed TT = 1.65 h (Apr 7), thus
AT = 14950 sec.
176 6 Timed Babylonian lunar eclipses

(5) BC 353 Nov 21/22 (mag. = 1.35)

[Artaxerxes III, year 6], month VIII 14, beginning on the south-east
side. After 23 deg total. 18 deg duration of maximal phase. After 6 deg of
night, a quarter of the disk had become bright (?) and it set eclipsed. The
eclipse was clouded (?). l\ cubits behind (3 Tau eclipsed. During the
eclipse, Saturn stood there; the other planets did not stand there... (Began)
at 47 deg before sunrise.
[BM 32238 (= LBAT 1414), Rev. Ill, bottom, 1-9;
trans. Huber, pp. 49-50.]
When the eclipse occurred, the Moon would be 4.5 deg to the east of (3
Tau, which agrees well with the 1 \ cubits of the text. Since the Moon set
eclipsed, only the times of the first three contacts could be measured by
the Babylonian astronomers.

RESULTS
(i) First contact on Nov 22 at 3.13 h before sunrise. LT of of sunrise =
6.79 h, hence LT of contact = 3.66 h, UT = 0.53 h. Computed TT = 4.96
h, thus AT = 15 950 sec.
(ii) Second contact at 1.60 h before sunrise. LT of contact = 5.19 h, UT
= 2.06 h. Computed TT = 6.10 h, thus AT = 14 550 sec.
(iii) Third contact at 0.40 h before sunrise. LT of contact = 6.39 h, UT
= 3.26 h. Computed TT = 7.50 h, thus AT = 15 250 sec.

(6) BC 317 Dec 13/14 (mag. = 1.34)

[Philip Arrhidaeus, year 7], month IX 15, beginning on the south-east
side. After 19(?) deg total. 5 deg duration of maximal phase. In 16 deg on
the north-east side it became bright. 40 deg total duration... This eclipse
was clouded (?). l\ cubits in front of (3 Gem eclipsed. (Began) at 44 deg
after sunset. Month IX, year(?) 7 of Philip(?), (the following year is) year 2
of Antigonus, son of [...].
[BM 32238 (= LBAT 1414), Rev. V, bottom, 1-10; trans. Huber, p. 51.]
When the eclipse occurred, the Moon was 3 deg to the west of P
Gem, in fair agreement with the record. Although the reading of the
interval between first and second contact as 19 deg is doubtful, it may be
restored from the recorded duration of 40 deg. It should be noted that
the recorded duration of totality is extremely short; for comparison, the
calculated figure is about 21 deg. However, the text is quite legible at this
point. The results given below are derived on the assumption that the
individual measurements are correct.
NB an independent cuneiform text confirms that Antigonus began to
rule in 316 BC. However, he is never given the title of king in contemporary
documents (Parker and Dubberstein, 1956, p. 20).
 6.4 Sets of three or four timings 111

RESULTS
(i) First contact on Dec 13 at 2.93 h after sunset. LT of sunset = 17.02 h,
hence LT of contact = 19.96 h, UT = 16.98 h. Computed TT = 21.58 h,
thus AT = 16 550 sec.
(ii) Second contact at 4.20 h after sunset. LT of contact = 21.22 h, UT
= 18.25 h. Computed TT = 22.72 h, thus AT = 16100 sec.
(iii) Third contact at 4.53 h after sunset. LT of contact = 21.55 h, UT
= 18.58 h. Computed TT = 0.12 h (Dec 14), thus AT = 19950 sec.
(iv) Last contact at 5.60 h after sunset. LT of contact = 22.62 h, UT =
19.65 h. Computed TT = 1.26 h (Dec 14), thus AT = 20200 sec.
Because of the serious error in the duration of maximum phase, the last
two values for AT are in poor accord with the first two of the same set.
However, there do not seem to be sufficient grounds for rejecting these
results.

(7) BC 226 Aug 1/2 (mag. = 1.21)

[SE 86], month IV, night of the 14th, moonrise to sunset: 4 deg,
measured (despite) mist; at 52 deg after sunset, when a Cyg culminated,
lunar eclipse; when it began on the east side, In 17 deg night time it
covered it completely; 10 deg night time maximal phase; when it began to
clear, it cleared in 15 deg night time from south to north...42 deg onset,
maximal phase and clearing; its eclipse was red(?); (in) its eclipse, a gusty
north wind blew; (in) its eclipse all of the planets did not stand there; 5
cubits behind 5 Cap it became eclipsed.
[BM 33655 (no LBAT number), Rev. 3-8; trans. SH II, p. 141.]
This tablet, which was discovered at the site of Babylon by Rassam, is
an astronomical diary. Its date has been derived by calculation using the
numerous planetary and lunar observations which are preserved (SH II,
p. 138). When the eclipse occurred, the Moon would be 11 deg to the east
of 5 Cap, which agrees well with the 5 cubits of the record. The computed
interval between moonrise and sunset on Aug 1 is 3.5 deg.

RESULTS
(i) First contact on Aug 1 at 3.47 h after sunset. LT of sunset = 18.94 h,
hence LT of contact = 22.41 h, UT = 19.51 h. Computed TT = 0.38 h
(Aug 2), thus AT = 17 550 sec.
(ii) Second contact at 4.60 h after sunset. LT of contact = 23.54 h, UT
= 20.63 h. Computed TT = 1.50 h (Aug 2), thus AT = 17 550 sec.
(iii) Third contact on Aug 2 at 5.27 h after sunset. LT of contact = 0.21
h, UT = 21.29 h (Aug 1). Computed TT = 2.57 h, thus AT = 18 950 sec.
(iv) Last contact at 6.27 h after sunset. LT of contact = 1.21 h, UT =
22.30 h (Aug 1). Computed TT = 3.68 h, thus AT = 19 350 sec.
178 6 Timed Babylonian lunar eclipses

(8) BC 215 Dec 25/26 (mag. = 1.37)

Year 97 (SE), month IX, night of the l[3th (?)...] measured; the bright
star of the Old Man (oc Per) stood in culmination, lunar eclipse; on the east
side when it began, in 21 deg of night, all of it became covered; 16 deg of
night totality; when it began to clear, it cleared in 19 deg of night from
east and north to the west(?); 56 deg onset, totality [and clear]ing; (began)
at one-half beru (i.e. 15 deg) after sunset. [...] eclipse; in its eclipse, Sirius
[stood there; Sa]turn set; Mars came out (i.e. rose); the remainder of the
planets did not stand there. In its eclipse, the north wind which was set to
the west side blew; it was cold. 2 cubits in front of the front stars of
Cancer (0 and y Cnc) it became eclipsed. The 13th, sunrise to moonset: 9
deg 10' (=9^ deg), measured; the north wind blew, the cold became severe.
[BM 36402 (= LBAT 294) + BM 36865 (no LBAT number), Obv. 1-9
and Rev. 1-5; trans. SH II, p. 157.]
This remarkably detailed description is recorded on an astronomical
diary, consisting of two joining fragments. When the eclipse occurred, the
Moon would be about 5 deg to the west of 0 and y Cnc, in close accord
with the text. The computed interval between sunrise and moonset on
Dec 26 is 8.2 deg.
RESULTS
(i) First contact on Dec 25 at 1.00 h after sunset. LT of sunset = 16.99 h,
hence LT of contact = 17.99 h, UT = 15.10 h. Computed TT = 20.04 h,
thus AT = 17 750 sec.
(ii) Second contact at 2.40 h after sunset. LT of contact = 19.39 h, UT
= 16.50 h. Computed TT = 21.09 h, thus AT = 16 500 sec.
(iii) Third contact at 3.47 h after sunset. LT of contact = 20.46 h, UT
= 17.57 h. Computed TT = 22.47 h, thus AT = 17 650 sec.
(iv) Last contact at 4.73 h after sunset. LT of contact = 21.72 h, UT =
18.83 h. Computed TT = 23.52 h, thus AT = 16 850 sec.

In table 6.4 are summarised the various results obtained in this section.
The form of this table is similar to that of table 6.3
The results from tables 6.3 and 6.4 are plotted in figure 6.7. In this
diagram, the general trend towards decreasing AT values in more recent
centuries is obscured by the anomalously high results derived from both
the eclipses of - 2 2 5 and -214.

6.5 Eclipse maxima timed relative to sunrise or sunset

(partial eclipses only)
As a rule, the Babylonian astronomers do not seem to have determined
the moment when a partial lunar eclipse was at its height. Instead they
distinguished a discrete period around mid-eclipse during which no change
 6.5 Timings of eclipse maxima 179

Table 6.4 AT results from Babylonian timings of lunar eclipses

for which three or four contacts were observed.
Year Ct SR/SS Int. (deg) AT (sec)
-500 1 ss +77 14950
-500 2 ss +92 15 300
-500 3 ss +117 14950
-500 4 ss +142 13 000

-406 1 SR -48 14950

-406 2 SR -27 13 850
-406 3 SR -12 16200

-377 1 SS +37 16000

-377 2 ss +52 16200
-377 3 ss +73 15 750
-377 4 ss +92 14950

-352 1 SR -47 15 950

-352 2 SR -24 14 550
-352 3 SR -6 15 250

-316 1 SS +47 16 550

-316 2 ss +63 16100
-316 3 ss +68 19950
-316 4 ss +84 20200

-225 1 ss +52 17 550

-225 2 ss +69 17 550
-225 3 ss +79 18 950
-225 4 ss +94 19 350

-214 1 ss +15 17 750

-214 2 ss +36 16 500
-214 3 ss +52 17 650
-214 4 ss +71 16 850

in the degree of obscuration of the Moon was noticeable. This typically

lasted around 6 deg, after which a gradual decline began to be observed.
In sections 6.2 and 6.3 above, several examples have already been cited in
passing - e.g. BC 143 Feb 17/18, when it was recorded that:

... In 20 deg night it made 9fingers.5 deg duration of maximal phase.

180 6 Timed Babylonian lunar eclipses

t
18000 -
* o *
16000
8 8# : •
14000
• • o
12000 o
-
o •
Ji 10000
I

8000
1

6000
I

4000
I

Key
2000 o First contact measurement
• Other measurement
0 1 1 1 1 1
-500 -400 -300 -200 -100 0
Year
Fig. 6.7 AT values obtained from timed lunar eclipses - two, three and four
contact measurements.

No other ancient or medieval civilisation seems to have assigned an

actual duration to the greatest phase of a partial eclipse. Although purely
artificial, and arising from the limited acuity of the unaided eye, this
implies considerable care in observation. Computation of the change in
the degree of obscuration of the Moon over an interval of 3 deg (half
of the typical duration) yields a result close to two per cent of the lunar
diameter for all but the smallest eclipses. This is illustrated in figure
6.8, which depicts an eclipse of central magnitude 0.75 at 3 deg (12 min)
before maximum, at greatest phase itself, and at 3 deg after maximum.
In angular measure, the change corresponds to only about 0.5 arcmin.
Interestingly, the Babylonian astronomers never seem to have reported a
duration of maximal phase for a solar eclipse (as is clear from the records
quoted in chapter 5). Probably the great brilliance of the Sun prevented
the necessary close scrutiny. For further discussion, see Stephenson and
Fatoohi (1993).
Assuming that greatest phase occurred near the mid-point of the interval
during which no change in the lunar obscuration was noticed, such records
may enable a reasonable estimate of the moment of mid-eclipse to be
inferred directly, as in the examples discussed below. For all but the last
of these observations, timings of first and/or last contact have already
been considered in sections 6.2 and 6.3. Hence in these cases only those
 6.5 Timings of eclipse maxima 181

3° before maximum Maximum phase 3° after maximum

Fig. 6.8 Change in magnitude of a partial lunar eclipse around maximal phase.

portions of the records which need to be repeated will be cited here.

However, the remaining account - from 66 BC - will be quoted in full.
(1) BC 424 Sep 28/29 (mag. = 0.93)
[Artaxerxes I, year 41], month VI 14, 50 deg after sunset, beginning on
the north-east. After 22 deg, 2 fingers lacked to totality. 5 deg duration of
maximal phase. In 23 deg towards [... ] 50 deg total duration...
[BM 34787 (= LBAT 1426), col. II, 2'-3'; trans. Huber, pp. 32-33.]
Halving the estimated duration of greatest phase (i.e. 5 deg), mid-eclipse
may be assumed to have occurred 50 + 22 + 2.5 = 74.5 deg after sunset.

RESULTS
Maximum on Sep 28 at 4.97 h after sunset. LT of sunset = 18.07 h, hence
LT of maximum = 23.04 h, UT = 19.96 h. Computed TT = 0.69 h (Sep
29), thus AT = 17000 sec.

(2) BC 239 Apr 28/29 (mag. = 0.41)

[SE 73], month I, night of the 13th... At 80 deg after sunset, lunar
eclipse; it began on the south and east side, in 15 deg night it made a little
over |(?) of the disk; 10 deg of night maximal phase. When it began to
clear, in 15 deg of night it cleared from the east to the west; 40 deg onset,
maximal phase and clearing...
[BM 55511 (no LBAT number); Obv. and Rev.; trans. SH II, pp. 84^85.]
On this occasion, an unusually long duration of maximal phase is
recorded. As in the previous example, it will be assumed that mid-eclipse
occurred 80 + 15 + 5 = 100 deg after sunset.

RESULTS
Maximum on Apr 29 at 6.67 h after sunset. LT of sunset = 18.63 h, hence
LT of maximum = 1.30 h, UT = 22.26 h (Apr 28). Computed TT = 0.56
h (Apr 29), thus AT = 8250 sec.
182 6 Timed Babylonian lunar eclipses

(3) BC 154 Mar 21/22 (mag. = 0.85)

[Year 157 (SE), king Demetrius, month XII] 15...Lunar eclipse,
beginning on the south-east side. In 20 deg of night it made 10fingers;6
deg duration of maximal phase. In 18 deg. ..it became bright. 44 deg total
duration... (Began) at 4 deg after sunset.
[BM 41129 (= LBAT 1440), Obv. and Rev.; trans. Huber, pp. 66-67.]
Mid-eclipse may be estimated to have occurred 4 + 20 + 3 = 27 deg
after sunset.

RESULTS

Maximum on Mar 21 at 1.80 h after sunset. LT of sunset = 18.02 h, hence

LT of maximum = 19.82 h, UT = 17.01 h. Computed TT = 20.67 h, thus
AT = 13 150 sec.

(4) BC 143 Feb 17/18 (mag. = 0.88)

[SE 168], month XI 13...lunar eclipse.. .In 20 deg of night it made 9
fingers. 5 deg duration of maximal phase. (Began) at 7 deg after sunset.
[BM 35787 (= LBAT 1278), Rev. 3/-8/; trans. Huber, p. 68.]
A time for mid-eclipse of 7 + 20 + 2.5 = 29.5 deg after sunset will be
adopted.

RESULTS

Maximum on Feb 17 at 1.97 h after sunset. LT of sunset = 17.51 h, hence

LT of maximum = 19.48 h, UT = 16.81 h. Computed TT = 20.40 h, thus
AT = 12900 sec.

(5) BC 129 Nov 4/5 (mag. = 0.63)

Year 120 (Arsacid)... month VIII, night of the 13th... lunar eclipse... [In
18 deg] of night [it made ...fingers]6 deg duration of maximal phase [... ]
In 16 deg ...40 deg total duration...(Began) at 55 deg before sunrise.
[BM 33982 (= LBAT 1441), Obv. and Rev.; trans. Huber, pp. 70-71.]
Mid-eclipse will be assumed to have occurred 55 — 18 — 3 = 34 deg
before sunrise.

RESULTS

Maximum on Nov 5 at 2.27 h before sunrise. LT of sunrise = 6.58 h,

hence LT of maximum = 4.31 h, UT = 1.13 h. Computed TT = 4.48 h,
thus AT = 12050 sec.
 6.5 Timings of eclipse maxima 183

(6) BC 120 Jun 1/2 (mag. = 1.02)

Year 128 (Arsacid), that is year 192 (SE)...month II 14...lunar
eclipse... After 24 deg, 1 UD lacked to totality. 6 deg duration of maximal
phase, until it began to become bright. In 24 deg, it became bright... 54(?)
deg total duration... (Began) at 66 deg after sunset.
[BM 45845 (= LBAT 1442), Obv. and Rev.; trans. Huber, pp. 71-72.]
It will be inferred that mid-eclipse occurred 66 + 24 + 3 = 93 deg after
sunset.

RESULTS

Maximum on Jun 2 at 6.20 h after sunset. LT of sunset = 19.06 h, hence

LT of maximum = 1.26 h, UT = 22.18 h (Jun 1). Computed TT = 1.44
h, thus AT = 11700 sec.

(7) BC 80 Apr 10/11 (mag. = 0.60)

Year 168 (Arsacid), that is year 232 (SE)...month I 13...lunar
eclipse... In 20 deg night it made 6 fingers. 7 deg of night duration of
maximal phase, until it began to become bright. In 13 deg...4 fingers
lacking to brightness, it set... (Began) at 40 deg before sunrise.
[BM 33562A (= LBAT 1445), Obv. and Rev.;
trans. Huber, pp. 75-76.]
Mid-eclipse will be assumed to have occurred 40 — 20 — 3.5 = 16.5 deg
before sunrise.

RESULTS

Maximum on Apr 11 at 1.10 h after sunset. LT of sunrise = 5.62 h, hence

LT of maximum = 4.52 h, UT = 1.57 h. Computed TT = 4.58 h, thus
AT = 10 800 sec.

(8) BC 67 Jan 19/20 (mag. = 0.81)

Year 180 (Arsacid), that is year 244 (SE)...month X, 15...As the Moon
rose, two-thirds of the disk... were eclipsed. 6 deg night duration of
maximal phase, until it began to become bright. In 16 deg night ... it
became bright. 23 deg total duration... (Began) at 16 deg before sunset.
[BM 45628 (= LBAT 1448) + Dupl. BM 41565 (= LBAT 1447 0, Obv.
and Rev.; trans. Huber, pp. 77-78.]
Since the Moon rose so close to mid-eclipse, it would probably be
difficult for the observers to define this phase. Hence the record is best
rejected.
184 6 Timed Babylonian lunar eclipses

(9) BC 66 Dec 28/29 (mag. = 0.34)

Year 182 (Arsacid), that is year 246 (SE), Arsaces, king of kings, and
[Pir'ustana], Istar, his wife(?), the lady. Month IX 14 [...] When the Moon
rose, 2 fingers on the south side [were eclipsed]. In 9 deg night, over a third
of the disk [was eclipsed]; 8 deg duration of maximal phase, until it began
to become bright, in 11 [+x deg from... ] to south-west it became bright.
22(?) [total duration]. During this eclipse the sky was overcast... During
this eclipse, north (wind) [...] During this eclipse, a halo enci[rcled(?)...]
Jupiter and Saturn stood there [...] rose. The other p[lanets did not stand
there]. 4 | cubits behind (3 Gem [...]. (Began) at 6 deg before [sunset].
[BM 32845 (= LBAT 1450), Obv. and Rev.; trans. Huber, pp. 79-80.]
This tablet is devoted specifically to this eclipse. Its date is fully
preserved. The text has not been considered previously since first contact
was unobservable (estimated to occur 6 deg before sunset) while the time
of last contact is only partially preserved. Although the magnitude was
small, it is clear that the Moon rose well before maximal phase. Halving
the estimated duration of greatest phase (i.e. 8 deg), mid-eclipse may be
assumed to have occurred 9 + 4 = 13 deg after moonrise.

RESULTS
Maximum on Dec 28 at 0.87 h after moonrise. LT of moonrise = 16.85 h,
hence LT of maximum = 17.72 h, UT = 14.85 h. Computed TT = 17.75 h,
thus AT = 10450 sec.

In table 6.5 are summarised the results obtained in this section. This
table follows much the same form as the previous tables in this chapter.
Measurements were all made relative to sunrise (SR) or sunset (SS) with
the exception of the eclipse of —65, for which the time was determined
relative to moonrise (MR).

6.6 First contacts timed relative to culmination of stars

By the third century BC, in addition to the traditional sunrise and sunset
measurements, the Babylonian astronomers began to time the beginning
of lunar eclipses relative to the culmination of certain stars or small
star groups. (Several examples have already been cited above in other
contexts.) There were 26 of these ziqpu ('culminating') stars, occupying
a rather broad zone which lay significantly to the north of the celestial
equator, but was roughly parallel to it. According to the detailed study by
Schaumberger (1952), declinations of individual ziqpu stars ranged from
about +20 to +40 deg. Spacings in RA between adjacent reference objects
varied from about 5 to 30 deg.
Just when this method of timing eclipses was introduced is not known,
but the earliest surviving example dates from 226 BC. Over the next 150
 6.6 First contacts timed relative to culmination of stars 185

Table 6.5 AT results from Babylonian timings of

lunar eclipse maxima.

Year SR/SS Int. (deg) AT (sec)

-423 SS +74.5 17000
-238 ss +100 8250
-153 SS +27 13150
-142 ss +29.5 12900

-128 SR -34 12050

-119 SS +93 11700
-79a SR -16.5 10800
-65 MR +13 10450

years or so (down to 80 BC), some 15 observations are extant. Probably

this represents only a small fraction of the original material. In some cases
it is simply asserted that a particular eclipse began 'when' a certain ziqpu
star was on the meridian. However, often a time interval 'before' (ina IGI)
or after (ar) culmination is specified. Such intervals are typically of the
order of 5 to 10 deg. I have rejected those records for which the interval
of time between the culmination of the selected ziqpu star and the start of
the eclipse is missing or damaged.
I have disregarded three lunar eclipses of very small magnitude since
first contact would probably be poorly defined. The eclipse of BC 188 Aug
1/2 was said to begin when P Cyg culminated (BM 42053, Obv. 1; Huber,
p. 61). This was in fact a penumbral eclipse, although the computed
magnitude came very close to zero. Interestingly, this represents the only
known ancient or medieval report of a penumbral eclipse. The eclipse of
BC 185 Nov 24/5, which reached a calculated magnitude of only 0.03, was
stated to have begun when the star group consisting of n9 o, E, and v Cas
culminated (BM 35330, Obv. 7 ; SH II, p. 355; Huber, p. 62). Finally, the
eclipse of BC 163 Mar 30/31, which began 3 deg after u Boo culminated,
attained a magnitude of only 0.12.
The calculated magnitudes of all other eclipses which were timed in
relation to star culminations exceeded 0.25. Each of the observations for
which a reliable date can be deduced will be considered below.
It is unfortunate that no Late Babylonian lists of ziqpu stars exists. The
most recent list probably dates from the eighth century BC and is found
on tablet AO 6478 (Antiquite orientale) in the Louvre. This was published
with a detailed commentary by Schaumberger (1952). An earlier list is
contained on MUL.APIN ('Plough Star'), several partial copies of which
are preserved in the British Museum (Hunger and Pingree, 1989).
186 6 Timed Babylonian lunar eclipses

AO 6478 gives an estimate of the difference in RA (i.e. difference

in the local time of culmination) between each of the 26 ziqpu stars
- usually to the nearest 5 US. However, no values for declination are
preserved. Identities of the brighter stars (such as a Boo, oc Cor, a Her,
oc Lyr, a Cyg, oc Per, oc Aur, and a and P Gem) appear reasonably well
established, but some of the fainter ones seem decidely questionable. In
the discussion of individual eclipse records below, I have accepted the
results of Schaumberger (1952), but with some hesitation in certain cases.
In general, Schaumberger's identifications have been accepted by both
Huber (1973) and SH.
For most of these observations, contact timings relative to sunrise
or sunset have already been considered in sections 6.2 to 6.4. In such
instances, only those parts of the record which need to be repeated will
be cited below; otherwise the full text will be quoted.

(1) BC 226 Aug 1/2 (mag. = 1.21)

[SE 86], month IV, night of the 14th... At 52 deg after sunset, when oc
Cyg culminated, lunar eclipse; when it began on the east side....
[BM 33655 (no LBAT number), Rev. 3-8; trans. SH II, p. 141.]

RESULTS
RA of star = 19.43 h, RA of Sun = 8.44 h; hence LT of first contact =
12.00 + 11.00 h = 23.00 h, UT = 20.09 h. Computed TT = 0.38 h, thus
AT = 15450 sec.

(2) BC 215 Dec 25/26 (mag. = 1.37)

Year 97 (SE), month IX, night of the 1 [3th(?)...] measured; the bright
star of the Old Man (a Per) stood in culmination, lunar eclipse; when it
began on the east side...
[BM 36402 + BM 36865 (= LBAT 294), Obv. 1-9 and Rev. 1-5;
trans. SH II, p. 157.]

RESULTS
RA of star = 1.07 h, RA of Sun = 18.07 h; hence LT of first contact =
12.00 + 7.00 h = 19.00 h, UT = 16.11 h. Computed TT = 20.04 h, thus
AT = 14150 sec.

(3) BC 194 Nov 4/5 (mag. = 0.92)

[SE 118], month VIII 14. When 'the 4 of (the Lion's) breast' (oc, y, r|, £
Leo) culminated, (lunar eclipse) beginning on the [north-]east side....
[BM 34236 (= LBAT 1436), Obv. 4-6; trans. Huber, p. 58.]
 6.6 First contacts timed relative to culmination of stars 187

RESULTS

Mean RA of star group = 8.12 h, RA of Sun = 14.45 h; hence LT of first

contact = 12.00 + 17.67 h = 5.67 h, UT = 2.49 h. Computed TT = 6.38
h, thus AT = 14000 sec.

(4) BC 150 Jul 2/3 (mag. = 1.11)

[SE 162] king Alexander... king Demetrius... [month III... ] moonrise to
sunset: [x]+6? deg, measured (despite mist); the north wind blew. When
(the point) 4 deg in front of 15 Lac culminated, lunar ec[lipse ...] when it
began, in 20 deg night it was completely covered; 12 deg of night maximal
phase; it set eclipsed. 32 deg onset and [maximal phase ...] in its eclipse,
Jupiter and Saturn stood there [... ] the remainder of the planets did not
stand there; l\ cubits below p Cap, [having passed a little to the east, [it
became eclipsed ... ]
[BM 34632 (= LBAT 400), Obv. 5'-8'; trans. SH III, p. 85]
This tablet is an astronomical diary. Its date has been established from
the numerous planetary and lunar observations which it contains and the
references to both Alexander and Demetrius (SH III, p. 87). No previous
reference has been made to this eclipse since the measurement of first
contact relative to sunrise or sunset is not preserved. Accordingly, the
full record is translated above. It is most unusual for a text to record the
sum of the individual phases for an eclipse whose visibility was eventually
interrupted by moonset. When the eclipse occurred, the Moon was 6 deg
to the south of P Cap, i.e. roughly l\ cubits.
An alternative rendering of the statement referring to the ziqpu star
would be: '4 deg before 15 Lac culminated, lunar eclipse...'. The star 15
Lac is very faint (magnitude 4.9) and it may well be doubted whether its
identification by Schaumberger is correct. Hence this observation will not
be considered further.

(5) BC 143 Feb 17/18 (mag. = 0.88)

[SE 168], month XI 13, moonrise to sunset 3 deg measured. 5 deg after
p Aur culminated, lunar eclipse beginning... In 20 deg of night it made 9
fingers
[BM 35787 (= LBAT 1278), Rev. 3'-8'; trans. Huber, p. 68.]

RESULTS

RA of star = 3.47 h, RA of Sun = 21.89 h; hence LT of first contact =

12.00 + 5.59 + 0.33 = 17.92 h, UT = 15.25 h. Computed TT = 18.80 h,
thus AT = 12 750 sec.
188 6 Timed Babylonian lunar eclipses

Table 6.6 AT results from Babylonian timings of first

contacts for lunar eclipses measured relative to meridian
transits of ziqpu stars.

Year Int. (deg) AT (sec)

-225 0 15450
-214 0 14150
-193 0 14000
-142 +5 12750

-135 0 11800
-119 +5 11950
-79a -5 11300

(8) BC 136 Mar 31/Apr 1 (mag. = 0.73)

Year 175 (SE), month XII/2 15 moonrise to sunset 17,40 (= 17§) deg,
cloudy, not observed. When oc Cor culminated, lunar eclipse, beginning on
the south-east side. In 18(?) deg of night it made 7 fingers. At 1(?) beru
before sunrise.

[BM 34034 (= LBAT 1285), Rev. 17-23; trans. Huber, p. 69.]

This tablet is a goal-year text for the year SE 194 (118-117 BC). A total
solar eclipse recorded on the same tablet (date equivalent to BC 136 Apr
15) was considered in chapter 5. Both events occurred 18 years before
the goal-year. The lunar eclipse was not considered earlier in this chapter
since the measurement of first contact relative to sunrise is damaged;
although not questioned by Huber, Walker (personal communication)
doubts the reading, which the scribe has written over an erasure. The full
record is translated above. Fortunately, the reference to the culminating
star is intact; this implies that the eclipse began when oc Cor was on the
meridian.
The recorded interval between moonrise and sunset (17§ deg) is much
too large for the evening before a lunar eclipse; the computed equivalent
is 6.2 deg. Perhaps the text should have read 7,40 rather than 17,40.

RESULTS

RA of star = 14.08 h, RA of Sun = 0.47 h; hence LT of first contact =

12.00 + 13.61 = 1.61 h, UT = 22.73 h. Computed TT = 2.00 h, thus AT
= 11800 sec.
 6.6 First contacts timed relative to culmination of stars 189

IOUUU
o
16000
X

x
14000 - x

12000 -
**%

& 10000 -
o
8000 -

6000 -

4000 Key
o Lunar eclipse maxima
2000 -
x Ziqpu star timings
0 i i 1 1
-500 -400 -300 -200 -100 0
Year
Fig. 6.9 AT values obtained from timed lunar eclipses - maximal phase and
ziqpu star observations.

(7) BC 120 Jun 1/2 (mag. = 1.02)

Year 128 (Arsacid), that is year 192 (SE)... Month II 14 moonrise to
sunset 6,30 {= 6\ deg) measured. 5 deg after a Her culminated, lunar
eclipse, beginning on the north-east side...
[BM 45845 (= LBAT 1442), Obv. and Rev.; trans. Huber, pp. 71-72.]

RESULTS
RA of star = 15.65 h, RA of Sun = 4.35 h; hence LT of first contact =
12.00 + 11.30 + 0.33 = 23.63 h, UT = 20.53 h. Computed TT = 23.86 h,
thus AT = 11950 sec.

(8) BC 80 Apr 10/11 (mag. = 0.60)

Year 168 (Arsacid), that is year 232 (SE)... Month I 13, moonrise to
sunset 5,50 (= 5§ deg), cloudy(?), measured(?)... 5 deg before \i Her
culminated, lunar eclipse, beginning on the south-east side...
[BM 33562A (== LBAT 1445), Obv. and Rev.; Huber, pp. 75-76.]
Although the '5' of '5 deg before \i Her culminated' is slightly damaged,
it is fairly certainly '5' (the only alternative is '4'; Walker, personal
communication).
190 6 Timed Babylonian lunar eclipses

22000,

20000 m 1
x
c • •• # x t
18000 x • • x it
S
16000 . ' ' . ' * . !
i « •: •
14000 °* * C *•
o * •
12000 • • ° 8
-
10000
•• i.
- •
8000 •
o Solar eclipses
6000
• Lunar eclipses (cuneiform texts)
4000 x Lunar eclipses (Almagest)
2000
n
-700 -600 -500 -400 -300 -200 -100
Year

Fig. 6.10 AT values obtained from timed Babylonian observations of both lunar
and solar eclipses.

RESULTS
RA of star = 16.43 h, RA of Sun = 1.10 h; hence LT of first contact =
12.00 + 15.33-0.33 = 3.00 h, UT = 0.05 h. Computed TT = 3.19 h, thus
AT = 11 300 sec.

The various results obtained in this section are summarised in table 6.6.
In figure 6.9 are plotted the AT results listed in both tables 6.5 and 6.6.
With the obvious exception of —238, the results follow much the same
pattern as in figure 6.6.

6.7 Conclusion
Figure 6.10 displays all of the AT results obtained in this chapter along
with values obtained from eclipse timings shown in figure 4.3 (Babylonian
lunar eclipse observations recorded in the Almagest) and figure 5.3 (solar
eclipses cited on the Late Babylonian astronomical texts). These various
data represent over 95 per cent of the carefully timed observations of
eclipses preserved from the ancient world (up to about AD 400). In
general, there is close accord between the AT values obtained from such
a diverse assortment of measurements. However, since many of the time-
intervals measured by the Babylonian astronomers amounted to several
hours, there is a distinct possibility of significant clock drift; this may well
 6.7 Conclusion 191

Table 6.7 AT results from Babylonian lunar and solar eclipse

timings estimated to the nearest degree (interval measured < 25
deg).

Year Type SR/SS Int. (deg) AT (sec)

-665 Moon SS +3 21050
11
-536 SR -14 18 800
11
-482 SR -10 17 300
11
-420 SS +19 15 500
11
-407 SS +15 15 250

-406 Moon SR -12 16 200

-405 SR -14 16 500
ii
-352 SR -24 14 550
ii
-352 SR -6 15 250
-321 Sun SS -3 14100

-316 Moon SS +10 15 550

-307 SR -10 14100
-280 Sun SR +20 12950
-239 Moon SR -3 14200
ii
-214 SS +15 17 750

-211 Moon SR -20 11800

-193 SR -12 13 650
-188 SR -18 11250
-169 Sun SS -20 12 300
-142 Moon SS +7 12 550

-135 Sun SR +24 12 600

-133 Moon SR -9 10950
ii
-128 SR -15 12 600
-108 "
SS +8 12050
-66 Moon MR +22.5 10150

account for much of the scatter in figure 6.10. On the other hand, some
intervals were very short (in some cases as small as 3 deg), and in such
cases clock drift should have been minimal. I have somewhat arbitrarily
selected 25 deg (1.67 hours) as a dividing line between 'short' and long'
intervals.
Table 6.7 lists the AT values obtained from measurements of both lunar
and solar eclipse times where the recorded interval is less than 25 deg.
(NB results from times rounded to the nearest 5 or 10 deg, observations
192 6 Timed Babylonian lunar eclipses

' Key
• . Lunar eclipse times
20000 o Solar eclipse times

15000
0 • •

#
•
8 °* .

10000 *t

5000
-700 -600 -500 -400 -300 -200 -100 +100
Year

Fig. 6.11 AT values obtained from timed Babylonian observations of lunar and
solar eclipses for which the measured interval was less than 25 deg.

of mid-eclipse and all ziqpu star measurements are not included.) These
various values are plotted in figure 6.11. With only a very few exceptions,
the AT results derived from these data are remarkably self-consistent, and
in the final analysis (chapter 14) they will be assigned double weight. All
other AT values as displayed in figure 6.10 will be given only unit weight.
 Untimed Babylonian observations of
lunar eclipses: horizon phenomena

7.1 Introduction
In the previous chapter, a variety of Babylonian timings of lunar eclipse
contacts were analysed. Several of these records also noted that the Moon
rose (ki E-a) or set (SU) whilst eclipsed. Additionally, some damaged texts
do not contain useful measurements of time but nevertheless affirm that
the Moon was eclipsed at its rising or setting. Such fairly straightforward
observations (which enable limits on the value of AT to be deduced),
would require no instrumental aid. In the following pages these various
observations will be investigated together with a few rather more careful
reports which give an estimate of the fraction of the Moon covered at
moonrise or moonset.
Since the eclipsed Moon is in direct opposition to the Sun, it invariably
rises close to sunset or sets near sunrise. It thus usually reaches the
horizon when the sky is quite bright - often when the Sun is above the
opposite horizon. However, as noted in chapter 4, despite these seemingly
unfavourable conditions the Babylonians systematically measured the time
of moonrise (relative to sunset) and moonset (in relation to sunrise) around
full Moon with considerable care. Many examples of this practice are
found in one of the earliest surviving astronomical diaries - dating from
568 BC, the 37th year of Nebuchadrezzar II (see SH I, pp. 47 ff.) - and it
may extend back much further in time.
Because of this regular observing routine, if the sky was clear around
the middle of a lunar month the Moon would be systematically scrutinised
at its rising and setting. Hence any small indentation on the lunar limb
would probably be noticed at these moments, whether or not an eclipse
was expected at this time. The fact that Babylonian and also Chinese
astronomers (see chapter 9) were in the habit of estimating the proportion
of the Moon which was eclipsed when it rose or set suggests that they
were able to discern the Moon fairly clearly when on the horizon.
Three Babylonian records - from 587, 171 and 134 BC - assert that the

193
194 7 Untimed Babylonian lunar eclipses

Moon rose or set totally eclipsed. For example, the report of the event
whose date corresponds to BC 171 Aug 23/24 reads as follows:
[SE 141, Month V...] When the Moon came out (i.e rose), it was
completely covered; when it began to clear, [... the remainder of the
pla]nets did not stand there; it was eclipsed in the end of Aquarius.
(Began) at 42 deg before sunset. The 15th [...].
[BM 45654 (= LBAT 365), Rev. 12'-13'; trans. SH II, p. 451.]
As noted in chapter 3, it seems unlikely that the lunar disk would be
visible under these conditions. However, because of the monthly routine
practised by the astronomers of Babylon, if the Moon were to rise totally
obscured, it should have been detected soon after emersion. Equally,
if the Moon set totally eclipsed, its failure to reappear after immersion
should have been noticed. Nevertheless, observations of this kind, based
on indirect evidence, seem likely to be of reduced reliability, especially if
cloud or mist prevailed at the time. Accordingly, reports that the Moon
rose or set totally eclipsed will not be considered here.
The observations discussed in this chapter are divided into two cate-
gories :
(i) Moon rising or setting eclipsed (section 7.2);
(ii) Estimates of the proportion of the Moon covered at its rising or
setting (section 7.3).
If the timing aspects of a certain text have already been discussed
in chapter 6, only the necessary details regarding rising or setting will
usually be reproduced below. Such entries will be marked with an asterisk
[*]. However, if a record has not yet been considered, a full translation
together with comments on dating (where appropriate) will be given. As
in the previous chapter, all translations are either by Huber or SH.

7.2 Moon rising or setting eclipsed

Since many texts are badly damaged, often it can only be inferred that the
Moon rose or set visibly eclipsed - i.e. between first and last contact. Unless
an observation indicates the contrary, these circumstances will normally
be assumed without comment. Since the local time of moonrise and
moonset is a function of AT, these times have been computed iteratively,
as discussed in chapters 3 and 4.
(1) BC 702 Mar 19/20 (mag. = 0.22)
Bel-ibni, year 1 [month I ... beginning] on the south side [... it] set
eclipsed [...before sunri]se.
[BM 45640 (= LBAT 1415), Obv. I, bottom; trans. Huber, p. 4.]
 7.2 Moon rising or setting eclipsed 195

The eclipse of BC 702 M a r 19/20 was the first to occur in the short
reign of Bel-ibni (702-699 BC). Its record represents the oldest extant ob-
servation on the cuneiform texts which has been reliably dated. However,
this is fully twenty years after the earliest Babylonian eclipse reported by
Ptolemy (see chapter 4).

RESULTS

(1) First contact on M a r 20 before moonset. LT of moonset = 6.17 h.

Hence LT of contact < 6.17 h, U T < 3.39 h. Computed T T = 8.49 h, thus
A T > 18 350 sec.
(ii) Last contact after moonset. LT of moonset = 6.21 h. Hence LT of
contact > 6.21 h, U T > 3.43 h. Computed TT = 10.37 h, thus AT <
25 000 sec.
Combining these limits yields 18 350 < AT < 25 000 sec.

(2) BC 695 Apr 30/May 1 (mag. = 0.74) [*]

(Assur-nadin-sum), year 5, month II, 1 [0+x, not] complete, it set
eclipsed. (Began) at 30 deg before sunrise.
[BM 32238 (= LBAT 1414), Obv. Ill; trans. Huber, p. 4.)
This eclipse was only partial. It is not clear whether the statement that
it was '(not) complete' means that the Moon set before the eclipse had
attained its maximum or before the disk had fully cleared. Hence it will
only be assumed the Moon set between first and last contact.

RESULTS

(i) First contact on May 1 before moonset. LT of moonset = 5.36 h.

Hence LT of contact < 5.36 h, UT < 2.32 h. Computed TT = 5.01 h, thus
AT > 9700 sec.
(ii) Last contact after moonset. LT of moonset = 5.43 h. Hence LT
of contact > 5.43 h, UT > 2.38 h. Computed TT = 7.90 h, thus AT <
19 850 sec.
Combining these limits yields 9700 < AT < 19 850 sec.

(3) BC 667 Oct 14/15 (mag. = 1.86) [*]

[Samas-sum-ukin, year l]...It set eclipsed. (Began) at 20[+x] deg before
sunrise.
[BM 45640 (= LBAT 1415), Obv. Ill, top; trans. Huber, p. 7.)
Although this eclipse was total, so little information is preserved that it
can only be inferred that the Moon set between first and last contact.
196 7 Untimed Babylonian lunar eclipses

RESULTS

(i) First contact on Oct 15 before moonset. LT of moonset = 6.15 h.

Hence LT of contact < 6.15 h, UT < 3.02 h. Computed TT = 6.19 h, thus
AT > 11400 sec.
(ii) Last contact after moonset. LT of moonset = 6.31 h. Hence LT of
contact > 6.31 h, UT > 3.20 h. Computed TT = 9.73 h, thus AT < 23500
sec.
Combining these limits yields 11400 < AT < 23 500 sec.

(4) BC 591 Mar 22/23 (mag. = 1.84)

[Nebuchadrezzar] year 13, month VI [eclipse passed]. [Month XII
1]4()?, it rose eclipsed... 1 beru 10 deg [duration?] ...
[BM 38462 (= LBAT 1420), Obv. II, 13'-15'; trans. Huber, p. 15.]
As pointed out in chapter 4, the Babylonian astronomers knew that
lunar eclipses occurred at intervals of five or six months. Hence although
the month in which the observed eclipse reported in the above text is
missing, it may be concluded that it would take place five or six months
after the predicted event in month VI. The date can thus be restored with
fair confidence. The time-interval of '1 beru 10 deg' may well indicate the
measured duration but this is only conjecture.

RESULTS

(i) First contact on Mar 22 before moonrise. LT of moonrise = 17.87 h.

Hence LT of contact < 17.87 h, UT < 15.07 h. Computed TT = 18.56 h,
thus AT > 12 550 sec.
(ii) Last contact after moonrise. LT of moonrise = 18.02 h. Hence LT
of contact > 18.02 h, UT > 15.22 h. Computed TT = 22.09 h, thus AT
< 24 700 sec.
Combining these limits yields 12550 < AT < 24 700 sec.

(5) BC 588 Jan 18/19 (mag. = 0.54)

[Nebuchadrezzar] year 16, month IV [eclipse passed]. [Month X], 14,
morning watch(?)> | (?) beru night [before sunrise]. By its f (??) covered,
set eclipsed.
[BM 38462 (= LBAT 1420), Obv. II, 13ff.; trans. Huber, p. 17.]
Both the time interval and degree of obscuration of the Moon are
damaged so that it may be only reliably inferred that the Moon set at
some time between first and last contact.
 12 Moon rising or setting eclipsed 197

RESULTS

(i) First contact on Jan 19 before moonset. LT of moonset = 6.88 h.

Hence LT of contact < 6.88 h, UT < 4.16 h. Computed TT = 7.22 h, thus
AT > 11000 sec.
(ii) Last contact after moonset. LT of moonset = 6.95 h. Hence LT of
contact > 6.95 h, UT > 4.24 h. Computed TT = 10.03 h, thus AT <
20 850 sec.
Combining these limits yields 11000 < AT < 20850 sec.

(6) BC 577 Jun 13/14 (mag. = 0.99)

[Nebuchadrezzar] year 27, month III 14? [...north] east side [...] it set
eclipsed.
[BM 38462 (= LBAT 1420), Rev. I, 7'-9'; trans. Huber, p. 19.]

RESULTS

(i) First contact on Jun 14 before moonset. LT of moonset = 4.85 h.

Hence LT of contact < 4.85 h, UT < 1.77 h. Computed TT = 4.06 h, thus
AT > 8250 sec.
(ii) Last contact after moonset. LT of moonset = 4.96 h. Hence LT
of contact > 4.96 h, UT > 1.89 h. Computed TT = 7.53 h, thus AT <
20 300 sec.
Combining these limits yields 8250 < AT < 20300 sec.

(7) BC 563 Sep 5/6 (mag. = 0.35)

[Nebuchadrezzar] year 42, month VI 14. It rose eclipsed [...] and
became bright. 6 US to become bright. (Began) at 35 deg [before sunset].
[BM 41536 (= LBAT 1421), II, 2'ff.; trans. Huber, p. 22.]
The time of beginning is based on prediction rather than observation.
Quite possibly the eclipse ended 6 deg after moonrise but since the text is
damaged this inference may be incorrect. It will only be assumed that the
Moon rose between first and last contact.

RESULTS
(i) First contact on Sep 5 before moonrise. LT of moonrise = 18.46 h.
Hence LT of contact < 18.46 h, UT < 15.49 h. Computed TT = 19.12 h,
thus AT > 13 050 sec.
(ii) Last contact after moonrise. LT of moonrise = 18.50 h. Hence LT
of contact > 18.50 h, UT > 15.54 h. Computed TT = 21.26 h, thus AT
< 20600 sec.
Combining these limits yields 13050 < AT < 20600 sec.
198 7 Untimed Babylonian lunar eclipses

(8) BC 537 Oct 16/17 (mag. = 1.50) [*]

[Cyrus, year 2] month VII 10[+x...it made] | of the disk towards
totality. Not (yet) total, it set eclipsed ... Began at 14 deg before sunrise.
[BM 32234 (= LBAT 1419), Obv. V, middle; trans. Huber, p. 25.]
Although the observers did not witness totality, apparently they were
able to judge from the almost central entry of the Moon into the shadow
that the eclipse would ultimately be complete. The Moon evidently set
between first and second contact.

RESULTS
(i) First contact on Oct 17 before moonset. LT of moonset = 6.24 h.
Hence LT of contact < 6.24 h, UT < 3.10 h. Computed TT = 7.39 h, thus
AT > 15 450 sec.
(ii) Second contact after moonset. LT of moonset = 6.27 h. Hence LT
of contact > 6.27 h, UT > 3.13 h. Computed TT = 8.50 h, thus AT <
19 300 sec.
Combining these limits yields 15450 < AT < 19 300 sec.

(9) BC 483 Nov 18/19 (mag. = 1.47) [*]

[Xerxes, year 3] month VIII 13. Beginning on the south side. Maximal
phase not observed; it set eclipsed... (Began) at 10 deg before sunrise.
[BM 32234 (= LBAT 1419), Rev. Ill; trans. Huber, p. 27.]
As in the previous example, the observers were aware that the eclipse
had not reached its height by the time the Moon set. It is thus clear that
moonset took place between first and second contact - an interpretation
in accord with the very short time-interval between the start of the eclipse
and sunrise.

RESULTS
(i) First contact on Nov 19 before moonset. LT of moonset = 6.73 h.
Hence LT of contact < 6.73 h, UT < 3.59 h. Computed TT = 7.74 h, thus
AT > 14900 sec.
(ii) Second contact after moonset. LT of moonset = 6.78 h. Hence LT
of contact > 6.78 h, UT > 3.64 h. Computed TT = 8.84 h, thus AT <
18 700 sec.
Combining these limits yields 14900 < AT < 18 700 sec.

(10) BC 442 Mar 24/25 (mag. = 1.04)

[Artaxerxes (I), year 22...] it set eclipsed [...] \ (?) it made and then
the Sun rose (?) 5 us(?) behind oc Vir (?).
[BM 34787 (= LBAT 1426), col. I, top; trans. Huber, p. 30.]
 7.2 Moon rising or setting eclipsed 199

Since the text is damaged, it seems best to assume no more than that
the Moon set at some time between first and last contact. When the eclipse
began, the Moon was about 7 deg to the east of oc Vir, in fair accord with
observation.

RESULTS

(i) First contact on Mar 25 before moonset. LT of moonset = 5.98 h.

Hence LT of contact < 5.98 h, UT < 3.16 h. Computed TT = 6.87 h, thus
AT > 13 350 sec.
(ii) Last contact after moonset. LT of moonset = 6.07 h. Hence LT of
contact > 6.07 h, UT > 3.25 h. Computed TT = 10.10 h, thus AT <
24 650 sec.
Combining these limits yields 13 350 < AT < 24650 sec.

(11) BC 353 Nov 21/22 (mag. = 1.35) [*]

[Artaxerxes III, year 6] Month VIII 14. Beginning on the south-east
side. After 23 deg total. 18 deg duration of maximal phase. After 6 deg
night, one quarter of the disc had become bright (?) (and) it set eclipsed...
[BM 32238 (= LBAT 1414), Rev. Ill, bottom; trans. Huber, p. 49.]
It is clear that the Moon set between third and last contact.

RESULTS

(i) Third contact on Nov 22 before moonset. LT of moonset = 6.89 h.

Hence LT of contact < 6.89 h, UT < 3.76 h. Computed TT = 7.50 h, thus
AT > 13 450 sec.
(ii) Last contact after moonset. LT of moonset = 6.94 h. Hence LT
of contact > 6.94 h, UT > 3.81 h. Computed TT = 8.64 h, thus AT <
17 350 sec.
Combining these limits yields 13450 < AT < 17 350 sec.

(12) BC 248 Oct 2/3 (mag. = 1.01)

[SE 64, month VII... ] Night of the 14th, [... ] a little cloudburst,
overcast; it set eclipsed; (during) its eclipse, Venus, Mercury and Ma[rs?
stood there...].
[BM 45949 (= LBAT 268), A3-A4; trans. SH II, p. 55;
see also Huber, p. 55.]
This tablet is an astronomical diary. Although its date has been broken
away, this has been derived by calculation using the various planetary and
lunar observations which are preserved (Sachs and Schaumberger, 1955,
p. xiv; SH II, p. 58).
200 7 Untimed Babylonian lunar eclipses

RESULTS

(i) First contact on Oct 3 before moonset. LT of moonset = 6.01 h. Hence

LT of contact < 6.01 h, UT < 2.90 h. Computed TT = 6.09 h, thus AT
> 11450 sec.
(ii) Last contact after moonset. LT of moonset = 6.15 h. Hence LT
of contact > 6.15 h, UT > 3.06 h. Computed TT = 9.48 h, thus AT <
23 100 sec.
Combining these limits yields 11450 < AT < 23 100 sec.

(13) BC 240 Nov 2/3 (mag. = 1.40) [*]

[SE 72], Month VIII 14, at 3 deg before sunrise, beginning on the east
side. It set eclipsed. Month XII2 (? or cloudy).
[BM 32286 (= LBAT 1218), Rev. 3-5; trans. Huber, pp. 55-56.]
Taken by itself, the extremely short duration of visibility is insufficient
to prove that the Moon set before totality (because of the slight possibility
of a scribal error in reporting this time). However, when combined with
the lack of further description in the text (which is preserved intact) it
may be reasonably inferred that the Moon reached the western horizon
between first and second contact.

RESULTS

(i) First contact on Nov 3 before moonset. LT of moonset = 6.50 h. Hence

LT of contact < 6.60 h, UT < 3.33 h. Computed TT = 7.12 h, thus AT
> 13 650 sec.
(ii) Second contact after moonset. LT of moonset = 6.52 h. Hence LT
of contact > 6.52 h, UT > 3.35 h. Computed TT = 8.16 h, thus AT <
17 300 sec.
Combining these limits, 13 650 < AT < 17 300 sec.

(14) BC 212 Apr 29/30 (mag. = 0.62) [*]

Year 100 (SE)...month I 13...Lunar eclipse, beginning on the south
side. Around(?) maximal phase cloudy, not observed. It set eclipsed.
(Began) at 20 deg before sunrise...
[BM 32222 (= LBAT 1237), Rev. 48-53; trans. Huber, p. 56.]
Apparently the sky was clear when the eclipse began and also when the
Moon set. On account of the clouds around maximal phase, the observers
may not have been able to judge whether the eclipse was on the decline
when the Moon reached the horizon. It thus seems best to assume no
more than that moonset occurred between first and last contact.
 7.2 Moon rising or setting eclipsed 201

RESULTS
(i) First contact on Apr 30 before moonset. LT of moonset = 5.40 h.
Hence LT of contact < 5.40 h, UT < 2.36 h. Computed TT = 4.25 h, thus
AT > 6800 sec.
(ii) Last contact after moonset. LT of moonset = 5.47 h. Hence LT
of contact > 5.47 h, UT > 2.43 h. Computed TT = 7.03 h, thus AT <
16 550 sec.
Combining these limits yields 6800 < AT < 16 550 sec.

(15) BC 190 Feb 27/28 (mag. = 1.05)

SE 121 month XII 15 [...] lunar eclipse, beginning on the south-east(?)
side. After 20(?) deg total [...] duration of maximal phase. When it began
to get bright, from the east side to the west. Toward [...] thunder(?) [...]
becoming bright not observed. Mars and Mercury in Capricorn. During
the eclipse [... ] eclipsed [... Began at ... ] night before sunrise. It set
eclipsed. On the 15th, sunrise to moonset 3,30 (i.e. 3^) deg, cloudy(?)
measured(?).
[BM 33643 (= LBAT 1437); Obv. and Rev.; trans. Huber, pp. 59-60.]
This tablet, part of the surface of which is badly damaged, is devoted
specifically to this eclipse. Although the date is fully preserved, the time of
onset is missing. It is clear from the text - which mentions the brightening
of the Moon after maximal phase - that moonset took place between
emersion and last contact.

RESULTS
(i) Third contact on Feb 28 before moonset. LT of moonset = 6.42 h.
Hence LT of contact < 6.42 h, UT < 3.72 h. Computed TT = 6.70 h, thus
AT > 10700 sec.
(ii) Last contact after moonset. LT of moonset = 6.44 h. Hence LT
of contact > 6.44 h, UT > 3.75 h. Computed TT = 8.15 h, thus AT <
15 850 sec.
Combining these limits, 10700 < AT < 15 850 sec.

(16) BC 190 Aug 23/24 (mag. = 1.02)

SE 122, King An[tiochus III], month V 14. Sunset to moonrise 1 deg,
cloudy, measured. When the Moon rose from a cloud(?), 2 fingers on the
[west] side lacked to brightness. (Began) at 1 beru before sunset.
[BM 34579 (= LBAT 1251), Rev. 14-18: trans. Huber, p. 60.]
This tablet is a goal-year text for SE 140 (172-171 BC). The eclipse
occurred 18 years before the goal-year. Although it was apparently cloudy
when the Moon rose, it was possible to measure the interval between sunset
and moonrise. Hence the Moon must have been discernible at the time.
202 7 Untimed Babylonian lunar eclipses

Normally only thin cloud prevails in Iraq in August. In describing the

eclipse, the standard term for moonrise (sin ki E-a) is used. The statement
that the eclipse began at 1 beru before sunset clearly represents a prediction
of an unobservable event.
The lack of reference to maximum phase in the above text and the
statement that 2 fingers lacked to brightness' (ana ZALAG kat) both
suggest that the eclipse was already nearly over by moonrise. It will thus
be assumed that the Moon rose at some time between emersion (although
the eclipse was only marginally total) and last contact.

RESULTS

(i) Third contact on Aug 23 before moonrise. LT of moonrise = 18.63 h.

Hence LT of contact < 18.63 h, UT < 15.70 h. Computed TT = 17.99 h,
thus AT > 8250 sec.
(ii) Last contact after moonrise. LT of moonrise = 18.68 h. Hence LT
of contact > 18.68 h, UT > 15.75 h. Computed TT = 19.39 h, thus AT
< 13 100 sec.
Combining these limits, 8250 < AT < 13 100 sec.

(17) BC 189 Feb 16/17 (mag. = 1.28) [*]

[SE 122] month XI 14...Lunar eclipse, beginning on the east side. After
16 deg night totally covered. It set eclipsed. (Began) at 34 deg before
sunrise.
[BM 34579 (= LBAT 1251), Rev. 23-26: trans. Huber, p. 61.]
This entry is recorded on the same goal-year text as the previous
record, the eclipse also occurring 18 years before the goal-year. Recovery
of the Moon after totality is not mentioned, suggesting that the Moon set
totally eclipsed. However, this is not definitely implied and it seems best
to conclude no more than that the Moon set between second and last
contact.

RESULTS

(i) Second contact on Feb 17 before moonset. Local time of moonset =

6.53 h. Hence LT of contact < 6.53 h, UT < 3.86 h. Computed TT = 5.78
h, thus AT > 6900 sec.
(ii) Last contact after moonset. LT of moonset = 6.60 h. Hence LT
of contact > 6.60 h, UT > 3.93 h. Computed TT = 8.33 h, thus AT <
15 800 sec.
Combining these limits, 6900 < AT < 15 800 sec.
 7.2 Moon rising or setting eclipsed 203

(18) BC 150 Jul 2/3 (mag. = 1.11) f*]

[SE 162 month III... ] lunar ecflipse... ] when it began, in 20 deg night it
was completely covered; 12 deg of night maximal phase; it set eclipsed. 32
deg onset and [maximal phase ... ]
[BM 34632 (= LBAT 400), 5'-8'; trans. SH III, p. 85.]
Although recovery of the Moon after totality is not mentioned, the
text does not definitely imply that it set totally eclipsed; we can only
conjecture what the lacuna at the end of the preserved text may have
originally contained. As in the previous example, it seems best to assume
no more than that the Moon set between second and last contact.

RESULTS
(i) Second contact on Jul 3 before moonset. Local time of moonset = 4.87
h. Hence LT of contact < 4.87 h, UT < 1.88 h. Computed TT = 4.21 h,
thus AT > 8350 sec.
(ii) Third contact after moonset. LT of moonset = 4.95 h. Hence LT
of contact > 4.95 h, UT > 1.97 h. Computed TT = 6.43 h, thus AT <
16050 sec.
Combining these limits yields 8350 < AT < 16050 sec.

(19) BC 99 Apr 11/12 (mag. = 1.82)

Year 213 (SE), month I ... Night of the 14th, when the Moon came out
(i.e. rose), it was eclipsed 3 fingers on the east side in the end of Libra, 2
cubits behind oc Lib; in its eclipse, Jupiter in Gemini and Sirius stood there.
[BM 140677 (no LBAT number), Obv. 4-6; trans. SH III, p. 407.]
This tablet is an astronomical diary. When the eclipse occurred, the
Moon was in longitude 200 deg, and thus approaching the end of Libra
(210 deg); it would be about 4 deg to the east of oc Lib (in close accord
with observation.) The statement that when the Moon rose it was partially
obscured on the east side implies that the eclipse had just begun. (A lunar
eclipse starts on the eastern edge of the disk and ends on the western limb,
the opposite to a solar obscuration.) However, it is surprising that the
text (which is intact) makes no reference to totality. Rather than attach
too much weight to the recorded position angle of first contact - in case
of possible scribal error - it would seem better to assume either that the
Moon rose between first and second contact or between third and last
contact.

RESULTS
(i) First contact on Apr 11 before moonrise. LT of moonrise = 18.26 h.
Hence LT of contact < 18.26 h, UT < 15.30 h. Computed TT = 15.36 h,
thus AT > 200 sec.
204 7 Untimed Babylonian lunar eclipses

(ii) Second contact after moonrise. LT of moonrise = 18.29 h. Hence

LT of contact > 18.29 h, UT > 15.33 h. Computed TT = 16.43 h, thus
AT < 3950 sec.
(iii) Third contact before moonrise. LT of moonrise = 18.35 h. Hence
LT of contact < 18.35 h, UT < 15.40 h. Computed TT = 18.21 h, thus
AT > 10100 sec.
(iv) Last contact after moonrise. LT of moonrise = 18.40 h. Hence LT
of contact > 18.40 h, UT > 15.45 h. Computed TT = 19.29 h, thus AT
< 13 800 sec.
Combining these limits yields either 200 < AT < 3950 sec or 10100 <
A T < 13 800 sec.
NB the second set of values is in much better accord with results
obtained from other roughly contemporaneous observations and only this
set will be retained.

(20) BC 80 Apr 10/11 (mag. = 0.60) [*]

Year 168 (Arsacid), that is year 232 (SE)...month I 13 ...Lunar eclipse
beginning on the south-east side. In 20 deg night it made 6fingers.7 deg
duration of maximal phase until it began to become bright. In 13 deg from
south-east to north-west, 4 fingers lacking to brightness, it set...
[BM 33562A (= LBAT 1445), Obv. and Rev.; trans. Huber, pp. 75-76.]
It is evident that the Moon set between maximum and last contact.

RESULTS
(i) Maximum phase on Apr 11 before moonset. Local time of moonset =
5.63 h. Hence LT of contact < 5.63 h, UT < 2.68 h. Computed TT = 4.58
h, thus AT > 6850 sec.
(ii) Last contact after moonset. LT of moonset = 5.66 h. Hence LT of
maximum > 5.66 h, UT > 2.71 h. Computed TT = 5.97 h, thus AT <
11 700 sec.
Combining these limits yields 6850 < AT < 11 700 sec.

(21) BC 67 Jan 19/20 (mag. = 0.81) [*]

Year 180 (Arsacid), that is year 244 (SE)...month X 15. Moonrise to
sunset 1 deg, mist, measured. As the Moon rose, two thirds of the disk on
the north-east side were eclipsed. 6 deg night duration of maximal phase,
until it began to become bright...
[BM 45628 (= LBAT 1448) + Duplicate BM 41565 (= LBAT 1447),
Obv. and Rev.; trans. Huber, pp. 77-78.]
The text implies that the Moon rose close to greatest phase. However,
under these circumstances it would probably be difficult to judge whether
 7.2 Moon rising or setting eclipsed 205

the eclipse had already passed maximum by this time or whether this
phase was still to come. Hence it seems best to conclude only that the
Moon rose at some time between first and last contact.

RESULTS

(i) First contact on Jan 19 before moonrise. LT of moonrise = 17.05 h.

Hence LT of contact < 17.05 h, UT < 14.33 h. Computed TT = 15.49 h,
thus AT > 4150 sec.
(ii) Last contact after moonrise. LT of moonrise = 17.15 h. Hence LT
of contact > 17.15 h, UT > 14.44 h. Computed TT = 18.71 h, thus AT
< 15 350 sec.
Combining these limits yields 4150 < AT < 15350 sec.

(22) BC 66 Dec 28/29 (mag. = 0.34) [*]

Year 182 (Arsacid), that is year 246 (SE)...Month IX 14 [...] When the
Moon rose, 2 fingers on the south side [were eclipsed]. In 9 deg night, over
a third of/the disk [was eclipsed]; 8 deg duration of maximal phase, until it
began to become bright...
[BM 32845 (= LBAT 1450), Obv. and Rev.; trans. Huber, pp. 79-80.]
It is apparent that the Moon rose between first contact and maximum.

RESULTS

(i) First contact on Dec 28 before moonrise. LT of moonrise = 16.87 h.

Hence LT of contact < 16.87 h, UT < 14.00 h. Computed TT = 16.64 h,
thus AT > 9500 sec.
(ii) Maximum phase after moonrise. LT of moonrise = 16.90 h. Hence
LT of contact > 16.90 h, UT > 14.14 h. Computed TT = 17.75 h, thus
AT < 13 350 sec.
Combining these limits yields 9500 < AT < 13 350 sec.

In addition to the above observations, Huber (pp. 17 and 22) concluded

that the Moon rose or set eclipsed on BC 588 Jan 18/19 and BC 563
Sep 5/6. However, both texts are damaged at the appropriate point
and Huber's inferences, although quite reasonable, cannot be reliably
confirmed.
In the following example, the recorded time after sunset when the
eclipse began is unusually small (only 3 deg). The mere fact that the
eclipse was seen to commence after sunset provides a limit on AT. Only a
single meaningful limit may be derived in this case (corresponding to first
contact after sunset).
206 Untimed Babylonian lunar eclipses

v
sy
24000
\y

22000 V

20000 sy
sy
18000 A
\
sy
/ sy
V
_ 16000 V V

T
g 14000 A
/ \ y\ v
"5 12000
/s
^ /\ 1
10000 /s

8000
Key
6000 v Upper limit to AT
/N Lower limit to AT
4000 • Critical upper limit /\

A Critical lower limit

2000 1 Total solar eclipse

0
-700 -600 -500 -400 -300 -200 -100
Year
Fig. 7.1 AT limits obtained from rising and setting observations in chapters 4,
5 and 7, together with the total solar eclipse of 136 BC.

(23) BC 666 Apr 10/11 (mag. = 0.48) [*]

[Samas-sum-ukin] year 2, month I, beginning(?) on [...]. (Began) at 3
deg after sunset.
[BM 45640 (= LBAT 1415), Obv. Ill, bottom; trans. Huber, p. 8.]

RESULTS

First contact on Apr 10 after sunset. LT of sunset = 18.29 h. Hence LT

of contact > 18.29 h, UT > 15.37 h. Computed TT = 21.42 h, thus AT
< 21 750 sec (only one limit).
NB on this occasion, the Moon would rise some 2 deg before sunset, so
that it would be well clear of the horizon by the time the Sun set.

In table 7.1 are summarised the various results obtained in this section.
The results listed in table 7.1, along with the few AT ranges obtained
from rising and setting observations in chapters 4 and 5 are shown in
figure 7.1. (Also shown, for comparison, is the range derived from the
total solar eclipse of 136 BC.) Although it is clear that many of the
wider limits are redundant, several observations provide valuable bounds
on AT at specific epochs. It should be noted that the limits set by a
few observations between about —400 and —200 seem unusually critical.
This question will be considered in chapter 14 when the full suite of data
obtained throughout this book is analysed.
 7.2 Moon rising or setting eclipsed 207

Table 7.1 AT limits from Babylonian observations where the

Moon was seen to rise or set whilst eclipsed.

AT Range (sec)
Yearfl MR/MS* Ctc Ctc LI/ UL'
-701 MS 1 4 18 350 25 000
-694 MS 1 4 9 700 19 850
-666 MS 1 4 11400 23 500
-665 MS 1 — — 21750
-590 MR 1 4 12 550 24700

-587 MS 1 4 11000 20 850

-576 MS 1 4 8 250 20300
-562 MR 1 4 13050 20600
-536 MS 1 2 15 450 19 300
-482 MS 1 2 14900 18 700

-441 MS 1 4 13 350 24650

-352 MS 3 4 13 450 17 350
-247 MS 1 4 11450 23100
-239 MS 1 2 13 650 17 300
-211 MS 1 4 6 800 16 550

-189a MS 3 4 10700 15 850

-189b MR 1 4 8 250 13100
-188 MS 2 4 6900 15 800
-149 MS 2 4 8 350 16050
-98 MR 3 4 10100 13 800

- 79 MS M 4 6 850 11700
- 66 MR 1 4 4150 15 350
- 65 MR 1 M 9 500 13 350
a
Year (using negative integers rather than BC).
b
Whether the eclipse was observed near moonrise (MR) or moonset (MS).
c
The contacts between which the Moon rose or set.
d
The range of AT indicated, from lower limit (LL) to upper limit (UL).
NB —189a and —189b refer to the first and second lunar eclipses in that year.
The combined limits for that year are 10700 < AT < 13 100 sec.
208 7 Untimed Babylonian lunar eclipses

7.3 Estimates of the proportion of the Moon obscured at its

rising or setting
In this section, I have discussed all reliably dated Babylonian observations
for which the proportion of the Moon covered at moonrise or moonset
is clearly stated. Texts in which the sign for the appropriate number of
digits (or fraction) is damaged will be ignored. In each case a correction
to the observed degree of obscuration derived from figure 3.23 (although
scarcely significant) will be applied. Since all of the observational records
have been translated in section 7.2, only summaries containing the relevant
details will normally be cited here.

(1) BC 537 Oct 16/17 (mag. = 1.50)

[Cyrus, year 2], month VII, 10[+x. ..it made] | of the disk towards
totality. Not (yet) total, it set eclipsed (Began) at 14 deg before sunrise.
[BM 32234 (= LBAT 1419), Obv. V; trans. Huber, p. 25.]
It is clear that the estimate of phase was made before immersion. The
proportion of the lunar diameter covered at moonset, adjusted according
to the line of best fit in figure 3.23, will be taken as 0.70.

RESULTS

LT of moonset on Oct 17 = 6.22 h, UT = 3.08 h. When 0.70 of

lunar diameter obscured (phase increasing), TT = 8.14 h. Hence AT =
18 200 sec.

(2) BC 353 Nov 21/22 (mag. = 1.35)

[Artaxerxes III, year 6] month VIII 14. Beginning on the south east side.
After 23 deg total. 18 deg duration of maximal phase. After 6 deg night,
one quarter of the disk had become bright (?) (and) it set eclipsed...
[BM 32238 (= LBAT 1414), Rev. Ill, bottom; Huber, p. 49.]
NB from Huber, the translation of line 4 - 'one quarter of the disk had
become bright' - is pure guesswork. However, this interpretation seems
reasonable since the interval between emersion and moonset is roughly
one-quarter of the interval from first contact to immersion. The proportion
of the lunar diameter covered at moonset, adjusted as above, is 0.80.

RESULTS

LT of moonset on Nov 22 = 6.87 h, UT = 3.74 h. When 0.80 of lunar di-

ameter obscured (phase decreasing), TT = 7.75 h. Hence AT = 14450 sec.
7.3 Estimates of the proportion of the Moon obscured at its rising or setting 209

(3) BC 190 Aug 23/24 (mag. = 1.02)

SE 122... month V 14. Moonrise to sunset 1 deg, cloudy, measured.
When the Moon rose from a cloud(?), 2 fingers on the [west] side lacked to
brightness. (Began) at 1 beru before sunset.
[BM 34579 (= LBAT 1251), Rev. 14-18: trans. Huber, p. 60.]
As discussed in section 7.2, it seems most likely from the lack of reference
to further phases and the terminology that the phase was decreasing at
moonrise. The (adjusted) proportion of the lunar diameter covered at
moonrise will be taken as 0.08.

RESULTS
LT of moonrise on Aug 23 = 18.70 h, UT = 15.77 h. When 0.08 of lunar di-
ameter obscured (phase decreasing), TT = 19.31 h. Hence AT = 12 700 sec.

(4) BC 99 Apr 11/12 (mag. = 1.82)

Year 213 (SE), month I .. .Night of the 14th, when the Moon came out
(i.e. rose), it was eclipsed 3 fingers on the east side ...
[BM 140677 (no LBAT number), Obv. 4-6; trans. SH III, p. 407.]
As discussed in section 7.2, the statement that when the Moon rose it
was partially obscured on the east side is at variance with the lack of
reference to further stages in the text. Hence it will be assumed that the
estimate of 3 fingers could refer to either an increasing or decreasing phase
at moonrise. The (adjusted) proportion of the lunar diameter covered at
moonrise will be taken as 0.18.

RESULTS
(i) LT of moonrise on Apr 11 = 18.29 h, UT = 15.33 h. When 0.18 of
lunar diameter obscured at moonrise (phase increasing), TT = 15.54 h.
Hence AT = 750 sec.
(ii) LT of moonrise = 18.42 h. When 0.18 of lunar diameter obscured
(phase decreasing), TT = 19.09 h. Hence AT = 13 050 sec.
The former result is highly discordant. However, the latter figure is in
close agreement with AT values derived from roughly contemporaneous
observations and will thus be retained.

(5) BC 80 Apr 10/11 (mag. = 0.60)

Year 168 (Arsacid), that is year 232 (SE)...month I 13 ...Lunar eclipse
beginning on the south-east side. In 20 deg night it made 6fingers.7 deg
duration of maximal phase until it began to become bright. In 13 deg from
south-east to north-west, 4fingerslacking to brightness, it set...
[BM 33562A (= LBAT 1445), Obv. and Rev.; trans. Huber, pp. 75-76.]
The (adjusted) proportion of the lunar diameter covered at moonset
will be taken as 0.28.
210 7 Untimed Babylonian lunar eclipses

RESULTS
LT of moonset on Apr 11 = 5.62 h, UT = 2.67 h. When 0.28 of lunar
diameter obscured (phase decreasing), TT = 5.55 h. Hence AT = 10350
sec.

(6) BC 67 Jan 19/20 (mag. = 0.81)

Year 180 (Arsacid), that is year 244 (SE)... month X 15... As the Moon
rose, two thirds of the disk on the north-east side were eclipsed. 6 deg
night duration of maximal phase, until it began to become bright...
[BM 45628 (= LBAT 1448) + Duplicate BM 41565 (= LBAT 1447),
Obv. and Rev.; trans. Huber, pp. 77-78.]
The (adjusted) proportion of the lunar diameter covered at moonrise
will be taken as 0.70.

RESULTS
LT of moonrise on Jan 19 = 17.15 h, UT = 14.44 h. When 0.70 of lunar
diameter obscured (phase increasing), TT = 17.61 h. Hence AT = 11400
sec.

(7) BC 66 Dec 28/29 (mag. = 0.34)

Year 182 (Arsacid), that is year 246 (SE)...Month IX 14 [...] When the
Moon rose, 2fingerson the south side [were eclipsed]. In 9 deg night, over
a third of the disk [was eclipsed] 8 deg duration of maximal phase, until it
began to become bright
[BM 32845 (= LBAT 1450), Obv. and Rev.; trans. Huber, pp. 79-80.]
The (adjusted) proportion of the lunar diameter covered at moonrise
will be taken as 0.08.

RESULTS
LT of moonrise on Dec 28 = 16.90 h, UT = 14.14 h. When 0.08 of lunar
diameter obscured (phase increasing), TT = 16.89 h. Hence AT = 9900
sec.

In table 7.2 are summarised the results obtained in this section. The
following details are listed for each observation: year (negative); whether
the eclipse was seen at moonrise (MR) or moonset (MS); the estimated
fraction of the Moon obscured; and the resulting value for AT.
The few results in the above table are plotted in figure 7.2. It can be
seen that these are in good accord with both one another and also the
much larger quantity of data plotted in figure 6.10. Since they lead to
discrete values of AT (rather than limits), they will be grouped with timed
data in subsequent analysis.
 7.4 Conclusion 211

Table 7.2 AT values from Babylonian observations

for lunar eclipses in which the phase was estimated at
moonrise or moonset.
Year MR/MS Est. mag. AT (sec)

-536 MS 3 18 200
3
-352 MS 4 14450
1
-189 MR 6 12 700
3
-98 MR 12 13 050

4
- 79 MR 12 10 350
2
-66 MR 3 11400
2
-65 MR 9900

-600 -500 -400 -300 -200 -100

Year
Fig. 7.2 AT values obtained from lunar eclipses in which the fraction of the
Moon obscured at its rising or setting was estimated.

7.4 Conclusion
In the four chapters (4-7) of this book devoted to Babylonian data, a
large number of valuable results and limits for AT in the period from
700 to 50 BC were deduced. The agreement between the AT values from
timed measurements (figures 6.10 and 6.11) and untimed observations
212 7 Untimed Babylonian lunar eclipses

(figure 7.1 is remarkably good throughout the whole period covered by

the data. Further results from ancient observations will be derived in
chapters 8 (Chinese) and 10 (European), but there can be no doubt
that the Babylonian data provide by far the most important source for
information on AT before the medieval period.
 8
Chinese and other East Asian
observations of large solar eclipses

8.1 Introduction
More solar eclipses are recorded in the history of China than in the annals
of any other civilisation. Not only were these events regarded as important
astrological omens by the Chinese from an early period, but they also
played a major role in the maintenance of the calendar.
Eclipses of the Sun have been systematically observed in China from
at least the eighth century BC. The many hundreds of reports which are
preserved since then are part of a huge corpus of accounts of celestial
phenomena of various kinds (including eclipses of the Moon, lunar and
planetary conjunctions, comets, novae and supernovae, meteors, sunspots
and the aurora borealis). Most of these events were noted by official
astronomers, who were employed by the ruler to keep a regular watch
of the sky for ominous happenings. Nearly all of the original reports
have long since been lost. Existing records - as found in the standard
dynastic histories and other historical compendia - are usually no more
than summaries of what may well have been detailed descriptions. These
secondary sources are readily accessible in major libraries throughout the
world, having been printed and reprinted many times. Block printing was
discovered quite early in China (eighth century AD) and as a result older
manuscripts have been phased out and are now relatively rare.
During the early centuries of its history, Chinese writing gradually
evolved from simple pictograms to an advanced form of ideographic
script. This script has remained essentially unchanged since the first
century AD. Around 1950, the more complex characters were simplified
in the People's Republic, but this revision has not found favour in Taiwan
and Hong Kong, for example.
In my search for records of eclipses, I have made extensive use of the
dynastic histories. A recent compilation of astronomical records in Chinese
history entitled Zhongguo Gudai Tianxiang Jilu Zongji ('A Union Table
of Ancient Chinese Records of Celestial Phenomena') has also proved

213
214 8 East Asian observations of large solar eclipses

extremely valuable. This compendium, published by Beijing Observatory

(1988), is based on a systematic search of both national and local histories
of China by a team of scholars.
Around the middle of the first millennium AD, Korean and Japanese
astronomers independently began to make celestial observations, in each
case following very much the traditional Chinese pattern. Numerous
records of celestial phenomena are still preserved in the history of both
countries, and these are also very largely found in printed works. Ancient
and medieval Korean histories are invariably written in Classical Chinese
and this is also true of most early histories of Japan. Although the Korean
alphabet of 24 letters (known as Harigul) was invented in the fifteenth
century AD, it was not popular with the literati of the country until the
present century. Medieval Japanese texts occasionally employ the kana
syllabic signs in addition to Chinese characters; these were developed in
the ninth century AD.
Among the accounts of solar eclipses from East Asia, more than fifty
relate to instances where the Sun was said to be either totally or largely
obscured. Observations of this type - most of which are untimed -
will be the subject of the present chapter. After Chinese data have first
been considered, Korean and Japanese material will then be discussed.
However, it should be emphasised that as a rule the observations from
China are superior to those from elsewhere in East Asia. Measurements
of solar and lunar eclipse times, estimates of solar eclipse magnitudes and
instances where the Sun or Moon was seen to rise or set eclipsed will be
the subject of chapter 9.
Throughout this chapter (and also chapter 9), the Wade-Giles system
of Chinese romanisation will be generally adopted; the only exceptions
will be the personal names of current Chinese authors and the names of
modern cities (for example Beijing) and provinces, for which the pinyin
equivalent has become fairly standard. For Korean names the McCune-
Reischauer method of romanisation is used, and for Japanese names the
Hepburn system. Translations, except where otherwise indicated, are by
the author. The help of Dr Kevin K. C. Yau (of Jet Propulsion Laboratory,
Pasadena) with difficult texts is gratefully acknowledged.
A list of the principal dynasties in Chinese history - together with
the intervening periods when the country was divided into two or more
independent states - is given in table 8.1. This table also lists the inclusive
dates for each era. The earliest dynasty whose existence can definitely
be established is the Shang; numerous original inscriptions have survived
from this period. According to tradition, the Shang was preceded by the
Hsia dynasty. However, no contemporary Hsia documents are extant, and
the very existence of this dynasty has been called into question (see, for
example, Fitzgerald, 1976, pp. 26 ff.). The dates for the beginning and end
 8.2 Records of solar eclipses from the Shang dynasty 215

Table 8.1 Major Chinese dynasties and periods of partition.

Dynasty etc. Date range
Shang c. 1500-1050 BC
Chou c. 1050-480 BC
Chan-kuo c. 480-221 BC
Ch'in 221-206 BC
Former Han 206 BC-9 AD

Hsin AD 9-23
Later Han AD 23-220
San-kuo 220-265
Chin 265-420
Nan-pei 420-589

Sui 581-618
T'ang 618-907
Wu-tai 907-960
Sung 960-1279
Kin 1115-1234

Yuan 1271-1368
Ming 1368-1644
Ch'ing 1644-1911

of the Shang are uncertain, but most subsequent periods are accurately
dated. As will be seen from table 8.1, there are occasional periods of
overlap between dynasties. For example, although the Sui dynasty was
founded in AD 581, it did not establish rule over the whole of China until
589.

8.2 Records of solar eclipses from the Shang dynasty

Only a very few allusions to eclipses survive from the Shang. Like other
records of the period, these are inscribed on a variety of animal bones, etc.
Since the beginning of the present century, vast numbers (some 150000 in
total) of Shang inscriptions have come to light as the result of excavations
near the city of An-yang in north-east China. This location is generally
regarded as the site of the Shang capital of Yin. The various texts which
have so far been studied probably all range in date from c. 1350 BC to
1050 BC. At some time around the former epoch, Yin became the royal
216 8 East Asian observations of large solar eclipses

residence, while the most likely date for the end of the Shang dynasty is
close to 1050 BC.
Shang inscriptions are largely in the form of oracular texts; it is clear
that divination was practised on a major scale. The diviners of the period
attempted to foretell the future by making cracks with a hot needle on
the prepared surfaces of animal bones etc., such as ox shoulder blades
and turtle plastrons (the relatively flat lower shell), and interpreting the
form of these cracks (Keightley, 1978a; Chou, 1979). The results of
their divination were inscribed on these same bones using a script which,
although far from primitive, is much closer to picture writing than the
present highly evolved system of ideographs (see figure 8.1). After use, the
bones were buried to prevent defilement; this has led to their preservation
in such large numbers.
The following text is fairly typical:
The divination on day chi-mao (the 16th day of a 60-day cycle) was
performed by Kuo. The King, after examining the crack forms,
commented that it would rain on daymen (-wu). On day jen-wu (the 19th
day of the cycle) indeed it did rain.
[Trans. Xu et al (1989).]
Many other inscriptions of the period are concerned with similar mun-
dane matters. Evidently the diviners were astute; they managed to achieve
sufficient success in their predictions to ensure the continuation of this
practice for several centuries. Since the inscriptions are largely devoted to
various aspects of divination, there are few reports of historical events;
astronomical observations are especially rare. In addition, most texts have
suffered extensively with the passage of time and are now mere fragments.
Despite these drawbacks, the names of nearly all of the Shang rulers have
been discovered on the oracle bones, thus confirming the names in the
traditional king list as found in later works.
The Shang calendar was basically lunar, but days were usually specified
according to a 60-day cycle, examples of which are given in the above
quotation. The origin of this cycle, which is independent of any astronom-
ical parameters, is obscure. Nevertheless, the scheme continued in use -
probably without any interruption - in the succeeding Chou dynasty and
throughout later Chinese history; it still remains in operation today at the
popular level. Ten 'celestial stems' (t'ien-kari) are combined consecutively
with twelve 'earthly branches' (ti-chih) making 60 combinations in all. The
ten kan are as follows: chia, i, ping, ting, wu, chi, keng, hsin, jen, kuei; the
twelve chih are tzu, ch'ou, yin, mao, ch'en, szu, wu, wei, shen, yu, hsu, hai. In
any particular cycle, the first day was chia-tzu, the second i-cKou, and so
on until the 60th day kuei-hai was reached, after which a new cycle began.
The full kan-chih cycle (as it is usually termed) is listed in table 8.2.
 8.2 Records of solar eclipses from the Shang dynasty 217

3 3694

Fig. 8.1 Example of oracle bone inscription.

An eclipse was identified by the Shang people using the term shih
('to eat'); they evidently believed that some monster was devouring the
luminary. This became the standard term to describe eclipses at all
subsequent periods in Chinese history, long after the true explanation was
understood (approximately first century AD).
218 8 East Asian observations of large solar eclipses

Table 8.2 The Chinese sexagenary cycle.

tzu ch'ou yin mao ch'en szu wu wei shen yu hsu hai
chia 1 51 41 31 21 11
i 2 52 42 32 22 12
ping 13 3 53 43 33 23
ting 14 4 54 44 34 24
wu 25 15 5 55 45 35
chi 26 16 6 56 46 36
keng 37 27 17 7 57 47
hsin 38 28 18 8 58 48
jen 49 39 29 19 9 59
kuei 50 40 30 20 10 60

Several references to solar and lunar eclipses have been identified on

the Shang oracle bones (e.g. Pang et a/., 1988b; Xu et a/., 1989; Xu et a/.,
1995). However, the records are often incomplete, while the meanings of
many Shang ideographs are still poorly understood. In addition, eclipse
dates are only partly preserved. Although the cyclical day is usually
specified, in no case is the year cited - while only a single report gives the
lunar month. Hence attempts to date a text which is believed to refer to
an eclipse involves 'playing the identification game' (Newton, 1970, p. xiv).
Shang observations of lunar eclipses are briefly discussed in chapter 9.
Among possible references to solar obscurations, only a single Shang text
has ever been seriously proposed as referring to a total eclipse, but the
interpretation involves a strong element of speculation. The phenomenon,
recorded on a turtle plastron is stated to have occurred on the day i-mao
(the 52nd day of the sexagenary cycle).
An eclipse identification was first proposed by Liu Chao-yang (1945).
Recently, Pang et al. (1988b) revived this hypothesis. They gave a partial
translation of the text as follows:
(i) Diviner Ko: ... day i-mao [cyclic day 52] to [next] dawn, fog. Three
flames ate the Sun. Big stars [seen].
However, Prof. David N. Keightley, of the University of California,
Berkeley - who has made an extensive study of Shang inscriptions - has
suggested the following alternative translation of the full inscription in a
personal communication to me:
(ii) On the next day i-mao it may not be sunny. The King read the
cracks and said, 'There will be disaster but it will not rain'. On day
i-mao at dawn it was foggy; when it came to the time of the
... meal, the day greatly cleared.
 8.3 The Chou dynasty and Warring States period 219

Although the term 'eclipse' is not mentioned in the text, Pang and
his colleagues assumed that the phenomenon described was a total solar
eclipse with a vivid corona. They accordingly computed the local circum-
stances of all solar obscurations which could have been visible at An-yang
(lat. = 36.07 deg, long. =—108.88 deg) for several decades on each side of
1300 BC. No large eclipse proved to be visible at this location on any date
whose cyclical day number was 52, but they identified a total obscuration
of the Sun on a certain occasion when the day number was 53. The
calculated Julian date of this event was BC 1302 Jun 5. Values of AT
which produce totality at An-yang lead to a local time around 11 h.
Pang et a\. did not attempt to justify their selection of day 53 (instead
of the previous cyclical day). If the phenomenon had indeed occurred on
day 53, a reference to this date would have been expected in the text since
the oracle bones are careful to specify individual dates. As it happens,
references to the corona - even in a clear sky - are extremely rare in the
history of all ancient and medieval cultures (see also chapters 10 and 11).
If the translation by Keightley is correct, the text resembles the weather
prediction quoted near the beginning of this section. However, it is evident
that scholars are in dispute over the rendering of several key characters
in the inscription. Until these difficulties are resolved, the record would
appear to be of negligible value in the study of long-term changes in the
rate of rotation of the Earth.

8.3 The Chou dynasty and Warring States period (c. 1050-221 BC)
Around 1050 BC, the inhabitants of the Chou region in western China
overthrew the Shang kingdom and established a new dynasty. Early Chou
history is very fragmentary and there are no direct references to eclipses
until the eighth century BC. Various attempts have been made to explain
a supposed 'double-dawn' occurring around 900 BC as produced by a
major eclipse at sunrise. This phenomenon is recorded in the 'Bamboo
Annals' (Chu-shu Chi-nien), a chronicle recovered in AD 281 from the
tomb of a prince who had died in 296 BC. Although the chronicle has
long since been lost again, its content was partially reconstructed during
the last (Ch'ing) dynasty from preserved quotations in a variety of writings
(Keightley, 1978b). It might be mentioned here that before the invention
of paper in the first century BC, bamboo was a common writing medium,
a book being formed by tying together a number of strips, each of typical
length about 50 cm. The record of interest, which like other entries in the
Bamboo Annals is very brief, may be translated as follows:

During the first year of King I (of western Chou), the day dawned twice
in Cheng.
220 8 East Asian observations of large solar eclipses

King I, the seventh Chou ruler, is believed to have reigned at some time
between about 966 and 895 BC (Pang et a/., 1988a).
Cheng was located in central China, not far from present-day Lo-
yang. (The approximate geographical co-ordinates of the site are: lat. =
34.50 deg, long. = —109.80 deg.) In principle, it would perhaps not
be unreasonable to attribute the phenomenon described in the Bamboo
Annals to a total eclipse occurring at sunrise. Rather similar phenomena
were recorded during medieval times in England (AD 1230) and in Russia
(1476) - see chapter 11. However, it should be stressed that in the Chou
texts there is no direct reference to the cause of the phenomenon. If the
Sun rose fully obscured, the closing stages of the event would resemble any
other eclipse so that its true nature would be evident. It is clear from the
descriptions given by the medieval European observers mentioned above
that they were well aware that an eclipse was responsible for the darkness
which they experienced. In contrast, the absence of the term shih in the
Chinese account materially weakens an eclipse interpretation.
Several investigations on the eclipse theme have been published since
the pioneering work of Liu Chao-yang (1944), but none can be regarded
as successful. The most recent attempt is that of Pang et al. (1988a).
From a computer check of all eclipses during the selected interval from
966 to 895 BC, these authors chose the sunrise eclipse of Apr 21 in 899
BC as the 'only possible match'. However, this eclipse was merely annular
and even at those places where the central phase was witnessed fully ten
per cent (by area) of the solar disk would remain unobscured. Hence the
reduction in the level of illumination would be scarcely noticeable - less
than on a typical overcast day.
As is apparent from the historical eclipse maps for East Asia produced
by Stephenson and Houlden (1986), no eclipse in the preferred date range
could possibly have been total at Cheng, even if large uncertainties in
the value of AT at this epoch were supposed. It seems doubtful whether
any alternative natural explanation (e.g. meteorological) can be found.
Possibly the original account in the Bamboo Annals was merely drawing
on legend rather than having a real, factual basis. For a recent critical
re-appraisal of the evidence, see Stephenson (1992).
Commencing in the eighth century BC, we encounter the first reliable
observations of solar eclipses in Chinese history. These are recorded in
a Chou dynasty chronicle known as the Ch'un-ch'iu (Spring and Autumn
Annals). This intriguing compilation, which is reputed to have been
edited by Confucius, extends from 722 to 481 BC. As one of the accepted
Confucian Classics, the text has been studied and commented on by
generations of scholars. Between AD 837 and 841 the entire set of these
Classics was engraved on series of stone tablets at the T'ang dynasty
capital of Ch'ang-an (now Xi'an) - (Carter, 1925, p. 212). These stelae are
 8.3 The Chou dynasty and Warring States period 221

Fig. 8.2 Section of a ninth century stele in Xi'an Museum containing a record
of the total solar eclipse of 709 BC. (Courtesy: Dr Liu Ciyuan.)

now preserved in the Beilin Museum (Museum of the Forest of Stelae)

in Xi'an.
A photograph of a section of one of these stelae, containing the record
of the total solar eclipse of 709 BC, is shown in figure 8.2.
The Ch'un-ch'iu has been translated into English by Legge (1872). The
chronicle is largely concerned with the State of Lu (the birthplace of
Confucius) and its interaction with the other semi-independent states into
which China was divided at this period. Only a formal allegiance was
paid to the Chou king by the rulers of individual states.
The only state chronicle of ancient China which is still intact - possibly
as a result of the 'Burning of the Books' at the command of Emperor
Ch'in Shih Huang-ti in 213 BC - the Spring and Autumn Annals records
222 8 East Asian observations of large solar eclipses

as many as 36 eclipses of the Sun. This series of observations, which

commences with the event of Feb 22 in 720 BC, is the earliest from
any part of the world. Only isolated eclipse sightings of more ancient
date are preserved from elsewhere, notably an observation in Assyria in
763 BC (see chapter 5). Lately, a detailed discussion of the eclipses and
other astronomical records in the Ch'un-ch'iu (including a few sightings of
comets and meteors) has been published by Stephenson and Yau (1992a).
Despite the interest in solar eclipses at this early period, there are no
allusions to lunar obscurations in the Ch'un-ch'iu. By at least the eighth
century BC, eclipses of the Sun were regarded as major omens of ill,
whereas their lunar counterparts were considered to be of minor signifi-
cance. This is probably because lunar eclipses were already recognised to
be periodic phenomena, and were therefore not considered worth report-
ing. Thus in one of the early poems in the Shih-ching (Book of Odes) we
find the following remark concerning two successive eclipses that occurred
during the eighth century BC:
That this Moon is eclipsed is but an ordinary matter; but that this Sun
is eclipsed - wherein lies the evil?
[Trans. Karlgren (1950, p. 100).]
In the Spring and Autumn Annals, years are numbered from the ac-
cession of the current 'Duke' (Kung) of Lu. Within any particular year,
entries are grouped in four conventionalised seasons (spring, summer,
autumn and winter), each of duration three lunar months. The calendar
was so regulated that the year began around - or rather earlier than -
the time of the winter solstice. Hence the four divisions of the year do
not correspond well with the true seasons. Most years contained twelve
lunar months, but an occasional extra month was inserted to keep the
calendar in step with the seasons. This intercalary month could be added
at any time of year. There is no evidence that months ever began with
the first sighting of the lunar crescent. Since solar eclipses recorded in
the Ch'un-ch'iu invariably began on the first day of the lunar month, it
may be supposed that they were used to regulate the calendar. How the
start of a month was fixed during the long intervals between observable
eclipses is not known. Use of a luni-solar calendar continued at all later
stages in Chinese history.
Many dates in the chronicle are also expressed in terms of the sex-
agenary cycle and this is especially true of solar eclipses. Although
intercalation may have been somewhat irregular at this early period, the
use of the sexagenary cycle enables accurate conversion of dates to the
Julian calendar with the minimum of assumptions. Nearly all dates, when
so reduced, prove to be in exact accord with those of eclipses listed in
modern tables. This is a remarkable result for such an ancient work and
 8.3 The Chou dynasty and Warring States period 223

it gives strong support to the authenticity of the purely historical data

recorded in the Classic.
The main focus of the Ch'un-ch'iu is on Lu affairs. Eclipse ceremonies
held at the state capital of Ch'u-fu are described in the Ch'un-ch'iu on
several occasions (BC 669 May 27, 664 Aug 28 and 612 Apr 28). It seems
likely that all or nearly all of the eclipse observations in the chronicle
were made at Ch'u-fu. This was situated on the site of the present city
of the same name (pinyin equivalent Qifu) in north-east China. Already
by the Chou dynasty, individual states are known to have had official
astronomers, and the names of some of these are still preserved. These
are recorded in the Shih-chi ('Historical Record': chaps. 6 and 27), a work
compiled by the Grand Historian Szu-ma Ch'ien between 104 and 87 BC.
In each case the record of the eclipse ceremonies at Ch'u-fu reads as
follows:

The Sun was eclipsed; drums were beaten and oxen were sacrificed at
the temple.

Presumably, similar rites took place at other eclipses. Even in relatively

recent centuries, it was the official custom to beat drums and gongs during
an eclipse (in an attempt to rescue the Sun from the dragon which was
imagined to be devouring it). Yet, already by the Ch'un-ch'iu period,
eclipses seem to have been recognised as natural phenomena, at least
among certain individuals. For example, the Tso-chuan (an enlargement
of the Ch'un-ch'iu mainly composed around 300 BC) records the following
extract from a speech made in 597 BC in support of a general who had
recently suffered defeat and was in danger of execution:

His defeat is like an eclipse of the Sun or Moon; does it harm the
brightness (of these bodies)?

In view of these well-chosen words, the general was reinstated!

The style of the Ch'un-ch'iu is laconic and this applies to the eclipse
records; in most cases it is merely reported that on a certain day the Sun
was eclipsed. Although on three occasions (BC 709 Jul 17, 601 Sep 20 and
549 Jun 19) the Sun was said to be totally obscured, further details - e.g.
darkness or the appearance of stars - are lacking. In each case the term
chi is used to indicate totality. This character originally was a pictogram
of a man turning his head away from a dish of food, indicating that he
was replete; its astronomical meaning is thus fairly obvious. Throughout
subsequent Chinese history this same expression continued in usage to
describe major solar obscurations.
There are no references to large partial obscurations of the Sun in the
Ch'un-ch'iu - e.g. using terms such as 'almost complete' or like a hook',
224 8 East Asian observations of large solar eclipses

Table 8.3 Geographical co-ordinates of Chinese cities where the eclipse

observations analysed in this book were made.

Place name Modern name Lat. (deg) Long, (deg)

Ch'ang-an (Han) Xi'an 34.35 -108.88
Ch'ang-an (T'ang) Xi'an 34.27 -108.90
Chien-k'ang Nanjing 32.03 -118.78
Ch'u-fu Qifu 35.53 -117.02

Lin-an Hangzhou 30.25 -120.17

Lo-yang Luoyang 34.75 -112.47
Pien Kaifeng 34.78 -114.33
P'ing-ch'eng Datong 40.20 -113.20

Sung-chiang Songjiang 31.00 -121.22

Ta-hsing Ch'eng Xi'an 34.27 -108.90
Ta-tu Beijing 39.92 -116.42

as found in later works from the Han dynasty onwards. Hence at this
early period it is possible that the expression chi might have been loosely
used to describe both total and nearly-total eclipses. However, even if
the original reports had been fairly specific, loss of detail would almost
certainly occur when they were included in the Ch'un-ch'iu in order to
conform to its style. For these reasons, it might well be doubted whether
the observers in 709, 601 and 549 BC actually witnessed the central phase
or merely saw an eclipse in which the Sun was very largely covered.
At subsequent periods in Chinese history, records of totality are some-
times qualified by allusions to darkness or the visibility of stars by day. If
so, an observation will be assigned a high rating (class 'A'). A brief account
of a total eclipse which is supported by extra details in other sources will
usually be included in category 'B', whereas one which merely notes the
occurrence of totality without providing any further information will be
assigned to the lowest category (class *C). In the case of the eclipse of 709
BC, a later text provides additional particulars. Hence the observation will
be given a 'B' classification. However, the two records from 601 and 549
BC are not supported by any other sources and hence will be allocated
to the lowest grade. The utility of this grading scheme will be reviewed at
the close of this chapter.
The location of Ch'u-fu and those cities where later eclipse observations
analysed in this and the following chapter were made are shown in figure
8.3; geographical co-ordinates (in degrees and decimals) are listed in
table 8.3. On the map, the modern place names are used.
 83 The Chou dynasty and Warring States period 225

/iOO°E 110°E I 120° E

40° N
A
I Beijing ,—f
r' I
Xi'an •
Luoyang
•
#Qifu

•Kaifeng

I\
f
Nanjing
• \Songjiang
Hangzh
—

I 20° N
4—_ y \5
/
-I
—
Fig. 8.3 Chart showing the locations of Chinese cities where eclipses were
observed.

For each of the eclipses of 709, 601 and 549 BC, the following summary
information is given below: calculated Julian date; type of eclipse (includ-
ing computed central magnitude); presumed place of observation (namely
Ch'u-fu); and classification. These details are followed by a full translation
of the record - including the original date - together with brief additional
remarks and also the computed range of AT (in seconds) necessary to
render the eclipse total at Ch'u-fu. The calculated approximate local time
and solar altitude at maximum phase, based on equation (4.1), are also
given for reference; these latter details will not be used to interpret the
records in any way. This same formula will also be used to calculate
local times and altitudes in later sections of this chapter for all dates prior
to AD 950. However, for all later eclipses, these peripheral details will
be computed using a further AT equation derived by Stephenson and
Morrison (1984). This is as follows:
AT = 25.5f2, (8.1)
where t is in Julian centuries from the epoch AD 1800.
226 8 East Asian observations of large solar eclipses

(1) BC 709 Jul 17 (total mag. = 1.06): Ch'u-fu [class B]

Duke Huan, 3rd year, 7th month, day jen-ch 'en [cyclical day number =
29], the first day (of the month). The Sun was eclipsed and it was total.
[Ch'un-ch'iu, book I.]
This is the earliest direct allusion to a complete obscuration of the
Sun in any civilisation. The recorded date, when reduced to the Julian
calendar, agrees exactly with that of a computed solar eclipse.
The Han-shu ('History of the Former Han Dynasty'), compiled in the
first century AD, gives additional information about this event in its
'Five-Element Treatise' (chap. 27c). Here it is stated that:
...the eclipse threaded centrally through the Sun; above and below it
was yellow.
Just possibly, there might be an allusion to the corona here. Unfortu-
nately, the source of this information is obscure.
RESULTS
For totality at Ch'u-fu, 20230 < AT < 21170 sec.
(NB computed local time of maximum = 15.3 h; solar altitude = 45
deg.)
(2) BC 601 Sep 20 (total mag. = 1.04): Ch'u-fu [class C]
Duke Hsuan, 8th year, 7th month, day chia-tzu [1]. The Sun was
eclipsed and it was total.
[Ch'un-ch'iu, book VIL]
In order for the recorded and computed dates to agree, the 7th month
of the text should be replaced by the 10th month. However, the written
characters for these numerals are very similar. If this alteration is made,
the reduced date is in precise accord with computation.
RESULTS
For totality at Ch'u-fu: 21140 < AT < 21960 sec.
(NB computed local time of maximum = 15.6 h; solar altitude = 31
deg.)
(3) BC 549 Jun 19 (total, mag. = 1.08): Ch'u-fu [class C]
Duke Hsiang, 24th year, 7th month, day chia-tzu [1], the first day of the
Moon. The Sun was eclipsed and it was total.
[Ch'un-ch'iu, book IX.]
Here the recorded and computed dates are in exact agreement. Since the
track of totality was unusually broad and ran nearly parallel to the equator
as it crossed China, the range of AT which satisfies the observation is
rather wide.
 8.3 The Chou dynasty and Warring States period 227

RESULTS
For totality at Ch'u-fu: 16 150 < AT < 21 710 sec.
(NB computed local time of maximum = 13.9 h; solar altitude = 63
deg.)

The above AT results, along with those derived from later Chinese (and
other East Asian) solar eclipse observations, will be summarised at the
end of this chapter (tables 8.8 and 8.9).
The Chan-kuo or Warring States period which followed the Ch'un-ch'iu
era lasted until 221 BC. Only nine solar eclipses are recorded during this
interval of 260 years. These are cited briefly in the Piao or Chronological
tables of the Shih-chi (chap. 15). They are presumably taken from a
lost chronicle of the State of Ch'in in western China, whose capital was
Hsien-yang (close to the site of modern Xi'an). Years are numbered from
the accession of each Ch'in ruler. The earliest eclipse date corresponds to
444/443 BC, and the latest to 249/248 BC.
On three occasions (444/443, 383/382 and 302/301 BC) the Shih-chi
records that the daylight was darkened by an eclipse, while on the first of
these dates it is stated that stars were also seen. However, in each case, no
more than the year of occurrence is specified and there is no indication
as to whether the phase was central or only partial. The earliest of these
records may be translated as follows:
Duke Li (of the Ch'in dynasty), 34th year. The Sun was eclipsed. It
became dark in the daytime and stars were seen.
[Shih-chi, chap. 15.]
On the Ch'in calendar, the year began with the 10th lunar month,
and thus commenced either in late October or November on the Julian
calendar. The eclipse may be identified as that of BC 444 Oct 24. However,
as this was annular, only 93 per cent of the Sun being obscured in the
central zone, the allusion to darkness must be exaggerated. Both Venus
(magnitude —3.3; 20 deg east of the Sun) and Mercury (—0.5; 14 deg east)
would be well placed for visibility. It might be mentioned that the eclipse
is the earliest in any civilisation for which the visibility of stars is reliably
reported. (A similar occurrence, dating from 431 BC, is described by the
ancient Greek historian Thucydides - see chapter 10.)
The other two eclipses were reported in the third year of Duke Hsi-an
(302/1 BC) and the sixth year of Duke Chao Hsiang. The corresponding
Julian dates can be readily established as BC 382 Jul 3 and 300 Jul 26,
respectively. Both of these eclipses were generally total. Since all three
Chan-kuo observations make no direct reference to magnitude, they are
probably valueless for determining AT and hence will not be considered
further.
228 8 East Asian observations of large solar eclipses

8.4 The Ch'in and Han dynasties (221 BC to AD 220)

Towards the end of the Chan-kuo, Ch'in became the dominant state and
eventually, in 256 BC, the last Chou king abdicated. Not long afterwards
(221 BC) the Ch'in ruler unified the whole country. He then adopted the
title Shih Huang-ti ('First Sovereign Emperor'). One of the most important
archaeological finds in recent decades - the so-called Terracotta Army' -
was made in an annexe of his tomb near Xi'an. No eclipses are recorded
during the brief Ch'in dynasty, which lasted until 206 BC. Many historical
documents were probably destroyed when, following the collapse of Ch'in
rule, the capital of Hsien-yang was stormed by rebel troops led by Hsiang
Yu (Dubs, 1938, vol. I, p. 7). Fortunately, the two official histories of the
subsequent Han dynasty, covering the period from 206 BC to AD 220,
contain many accounts of eclipses. From the Han onwards, it became the
practice for scholars to be commissioned to write a comprehensive history
after the end of each dynasty, the compilers usually having free access to
official records including those of the imperial astronomers.
The Han-shu, compiled by Pan Ku between AD 58 and 76, covers the
Former Han dynasty together with the brief Hsin interregnum (AD 9
to 23). During the Former Han the capital was Ch'ang-an ('Everlasting
Peace'), whose site lies a little to the north-west of the present city of
Xi'an. Wang Mang usurped the throne in AD 9 but his reign proved
extremely unpopular and when he was killed by Han loyalist forces in
AD 23, Ch'ang-an was so ruined that the capital had to be transferred to
Lo-yang - far to the east. Events of the second phase of the Han dynasty
are reported in the Hou-han-shu ('History of the Later Han Dynasty'),
which was composed by Fan Yeh towards the middle of the fifth century.
In common with most later dynastic histories, the Han-shu and Hou-
han-shu consists of four main divisions. These are: Pen-chi ('Basic Annals'),
Piao ('Chronological tables'), Chih ('Treatises') and Lieh-chuan ('Biogra-
phies'). In both Han official histories, solar eclipse records are assembled
(along with certain other astronomical observations - as well as meteoro-
logical data) in special monographs known as Wu-hsing Chih or 'Treatises
on the Five Elements'. (NB the five elements of Chinese philosophy were
earth, metal, water, fire and wood.) In each treatise, much stress is placed
on the omen-value of the various natural phenomena which are reported.
There are many other treatises on a wide variety of topics: astrology, the
calendar, music, rites, food and commodities, and offices, to name just a
few (see Han, 1955, pp. 200-201).
The lengthy Five-Element monograph in the Han-shu forms chapter 27
of this work and eclipse observations are found towards the end of this
treatise (section 27c). In the corresponding treatise of the Hou-han-shu,
which covers several chapters, eclipse records are confined to chapter 28.
 8.4 The Ch'in and Han dynasties 229

Table 8.4 Treatises of Chinese dynastic histories which contain

series of solar eclipse records.

Name of history Chapter Title

Han-shu 27c Wu-hsing Chih
Hou-han-shu 28 Wu-hsing Chih
Chin-shu 12 T'ien-wen Chih
Sung-shu 34 Wu-hsing Chih
Nan-ch'i-shu 12 T'ien-wen Chih

Wei-shu 105 T'ien-hsiang Chih

Sui-shu 21 T'ien-wen Chih
Chiu-t'ang-shu 36 T'ien-wen Chih
Hsin-t'ang-shu 32 T'ien-wen Chih
Chiu-wu-tai-shih 139 T'ien-wen Chih

Hsin-wu-tai-shih 59 Szu-t'ien K'ao

Sung-shih 52 T'ien-wen Chih
Kin-shih 20 T'ien-wen Chih
Yuan-shih 48 T'ien-wen Chih

References to lunar obscurations are still extremely rare at this period.

Brief notices of solar eclipses also frequently occur in the Pen-chi ('Basic
Annals'), the opening chapters of these same histories; these compilations
describe the main events of each emperor's reign. Very occasionally, an
eclipse may be alluded to in another part of an official history - for instance
in the Lieh-chuan or biographical section. Incidentally, observations of
celestial phenomena other than eclipses are mainly reported in the Tien-
wen Chih (Treatises on Astrology') of the two Han histories. These form
chapter 26 of the Han-shu and chapters 20-22 of the Hou-han-shu.
In most later histories, eclipse observations are cited in the Astrological
Treatise rather than in the Five-Element section. This usually goes under
the title Tien-wen Chih but other variants occasionally occur. The chapter
numbers in the appropriate histories of the various chih which contain
series of solar eclipse observations are listed in table 8.4.
During the early decades of the Former Han, years were numbered from
the accession of each ruler. However, before long the use of subdivisions
of a reign - or reign periods - was established. This practice continued
down to the end of the last dynasty (it also spread to Japan, but reign
periods were never used in Korea). Dates within a given year were usually
expressed in terms of the lunar month, cyclical day and day of the month
- as in the Ch'un-ch'iu - but the year now began in the early spring,
230 8 East Asian observations of large solar eclipses

rather than near the winter solstice. Only minor changes were made to
the calendar in succeeding dynasties. During the Former Han and the
first few decades of the Later Han, most solar eclipses occurred on the
last day of a lunar month. Afterwards the calendar was so adjusted that
the first day became much more common.
In expressing dates of Former Han eclipses in terms of the Julian
calendar, I have followed Dubs (1938-1955). However, in each case I
have verified his reduction of cyclical day numbers using a computer
program. I have effected date conversion in the Later Han - and all
subsequent dynasties - with the aid of a specially designed computer
program based on the extensive chronological tables of Hsueh and Ou-
yang (1956). It should be emphasised that throughout Chinese history
from the Han onwards, recorded dates of solar eclipses, when converted
to the Julian calendar, usually agree precisely with the calculated dates of
these phenomena.
In common with the numerous other celestial events (such as occul-
tations of stars and planets by the Moon, planetary conjunctions, etc.)
reported in the astrological treatises, the observations of eclipses are prob-
ably largely based on the records of the court astronomers. Because of
the emphasis placed on astrology by the Chinese court, these officials were
employed to maintain a systematic watch of the sky - both day and night
- at the imperial observatory for any unusual phenomenon which might
be regarded as a celestial omen. They were required to keep their business
secret so that the detailed astronomical knowledge that they possessed
could not be passed to the general public (see, for example, Needham,
1959, p. 193). Soon after the first Han emperor Kao Tzu ascended the
throne in 202 BC, an observatory was built at his capital of Ch'ang-an
(present-day Xi'an). This observatory remained active for more than two
centuries - until Ch'ang-an was sacked when the reign of the usurper
Wang Mang was forcibly brought to an end in AD 23. Following the
re-establishment of Han rule, a new imperial observatory was founded at
Lo-yang; the site of this can still be identified today.
Usually, it seems that the official histories take for granted that a reader
would assume that an astronomical observation was made at the capital
itself. Only very rarely does a text directly mention the capital - for
example, to comment on the visibility of an eclipse at this city and its
invisibility in the provinces, or vice versa. Thus, on a date corresponding
to BC 16 Nov 1 it is stated in the Five-Element Treatise of the Han-shu
that:
Heaven caused the capital alone to know of it; the states in the four
(quarters) did not see it.
In the whole of the Han-shu, only a single solar eclipse (BC 15 Mar 29)
 8.4 The Ch'in and Han dynasties 231

is stated to have been observed in the provinces and on this occasion the
weather was cloudy in Ch'ang-an. Here, the Wu-hsing Chih notes that:
The four quarters all saw it, but in the capital it was overcast.
Provincial observers seem to have been unusually active in the Later
Han and as many as 17 accounts (nearly one quarter of the total) in the
Five-Element Treatise of this work are said to have been reported from a
provincial location. At no other period in Chinese history do we find a
comparable situation.
A typical record of this kind in the Hou-han-shu treatise takes the
following form:
The Sun was eclipsed; it was not seen by the official astronomers but
was communicated from ... province.
In each case the precise date is given and the name of the province
is usually specified. However, there is nothing in these brief records to
suggest that any of the eclipses were unusually large. In the absence of a
reference to a provincial location, it seems reasonable to assume that an
observation was made at the capital itself.
Many solar eclipse reports in both the Han-shu and Hou-han-shu are
accompanied by astrological comments. Since at least the time of the
philosopher Mencius (fourth century BC), the ruler was regarded as the
Son of Heaven and it was thus believed that celestial portents could either
(a) warn him of coming danger, or (b) level criticism at him for mis-rule.
Thus following an obscuration of the Sun on a date corresponding to Nov
14 in AD 30, the Bearer of the Gilded Mace memorialised to the throne
that this event was a warning against too frequent transfer of officials.
After a further eclipse six months later (AD 31 May 10), the Grand
Palace Grandee memorialised that this portent was a warning against the
Emperor's unwillingness to accept candidates recommended by the high
officials. (Both of these accounts are given in chapter 36 of the Hou-han-
shu; see also Bielenstein, 1979.) Shrewd courtiers must often have taken
advantage of the opportunities for criticising the ruler presented by the
occurrence of eclipses and other celestial omens.
In both Han Five-Element Treatises the right ascension (RA) of the
Sun at an eclipse is frequently recorded to the nearest degree (tu) in one
of the 28 lunar lodges. (NB there were 365.25 Chinese degrees to a circle.)
Often an astrological prognostication is based on the lunar lodge in which
the Sun happened to be. These constellations were associated both with
different aspects of Chinese administration and the regions occupied by
the ancient states into which China had once been divided. Thus the total
eclipse of AD 65 Dec 16 was said to have occurred while the Sun was 11
deg in Nan-tou. The Hou-han-shu (chap. 28) text continues:
232 8 East Asian observations of large solar eclipses

Nan-tou represents (the ancient state of) Wu. Kuang-ling as far as the
constellations are concerned belongs to Wu. Two years later, Ching, King
of Kuang-ling, was accused of plotting rebellion and committed suicide.

Kuang-ling lay much to the east of Lo-yang, not far from the site of
present-day Nanjing. A delay of two years between the occurrence of a
celestial portent and its acknowledged fulfilment is not unusual.
On two occasions (in 181 BC and AD 120 - see below), a major
eclipse was regarded as presaging the death of a member of the Han
imperial family. A passage in a later work, the Chin-shu ('History of the
Chin Dynasty'), shows that the omen-value of an eclipse was believed to
depend very much on its magnitude. Commenting on the solar eclipse
of AD 360 Aug 28, the Astrological Treatise (chap. 12) of this history
records the following remarks:

Whenever an eclipse covers a small portion of the Sun the calamity it

brings will be relatively small, but when it covers a large portion of the
Sun the consequences will be much more serious.
[Trans. Ho 1966, p. 159.]
Han and later records indiscriminately apply the same expression chi
(i.e. 'total') to both total and annular eclipses. Thus both the eclipses
of 198 and 181 BC are described in this way in the Han-shu treatise,
but whereas the former event was, in fact, only annular (no more than
95 per cent of the Sun being covered by the Moon), the latter was fully
total. Similar ambiguity occurs in several later dynasties, possibly because
during a central annular eclipse the Sun is reduced to a complete ring.
Not until as late as AD 1292 do we find a separate expression to describe
an annular obscuration: 'the Sun was like a golden ring' (see section 8.7
below). However, even long after that date there was still no separate
term in standard usage. It should be noted that accounts of large partial
obscurations of the Sun in the Five-Element Treatises normally avoid the
use of chi. Instead, the expression chin ('complete') - for example in the
sense pu-chin (not complete) - is usually preferred.
Although the accounts of eclipses in the Han treatises are often paral-
leled by records in the annals of the same dynastic histories, these latter
works normally omit almost all technical details. For instance, whereas the
Wu-hsing Chih frequently specify the right ascension of the Sun during an
eclipse (to the nearest degree in a particular lunar lodge), the Pen-chi are
silent on such information. It is also noteworthy that the annals of both
Han histories never describe partial eclipses; unless totality is alleged, no
mention is made of magnitude. Thus, while the Five-Element Treatise in
the Hou-han-shu records that in AD 120 the eclipse was 'almost complete;
on the Earth it became like evening', the Annals (chap. 5) merely state
 8.4 The Ch'in and Han dynasties 233

that the Sun was totally obscured. Again, the eclipses of both 80 and 28
BC were said to be 'not complete, and like a hook' in the Five-Element
Treatise of the Han-shu whereas the Annals simply record the total phase.
The lower reliability of accounts of celestial phenomena in the Pen-chi
of both the Han-shu and Hou-han-shu is not confined to eclipses. It is
also evident in the case of the various cometary records. For example,
the report of the 12 BC apparition of Halley's comet in the Astrological
Treatise of the Han-shu cites precise dates and carefully describes the
motion of the comet through several constellations (see, for example,
Stephenson and Yau, 1985). Such detailed information is doubtless based
on the observations of the court astronomers. However, the corresponding
entry in the Basic Annals (chap. 10) merely notes the month of visibility
and only mentions one constellation through which the comet passed.
Many similar illustrations could be cited and for these reasons notices of
eclipses in the Han annals will not be considered further.
In the Five-Element Treatise of the Han-shu, three solar eclipses (198
BC, 181 BC and AD 2) were reported as total while as many as seven (188,
147, 89, 80, 35, 28 and 2 BC) were stated to be not complete. Only one
eclipse of each kind (total in AD 65 and incomplete in AD 120) is noted
in the corresponding treatise of the Hou-han-shu. It is unfortunate that all
four records of totality in both the Han-shu and Hou-han-shu treatises are
devoid of any supporting description; in each case it is merely asserted
that an eclipse was 'total'. With the exception of the event of 181 BC,
which is described in more detail in the Shih-chi, I have included each of
these observations in the lowest category ' C
Where careful descriptions of partial eclipses are available, such as jo-
k'ou ('like a hook'), I have ranked each observation in class 'A'. However,
it is sometimes merely stated that an eclipse was 'almost complete' or 'not
complete'. In the case of an eclipse which was total on the Earth's surface,
the meaning of these terms is fairly obvious. Nevertheless, ambiguity might
well occur in the case of an obscuration of the Sun which was generally
annular. Since it is not known whether the term chi was consistently
applied to central annular eclipses, in principle a description such as
'almost complete' could imply either the ring phase or a large partial
eclipse. For these reasons, I have adopted a 'B' rating for generally total
eclipses which - without any further details - were said to be either 'almost
complete' or 'not complete' However, I have rejected all observations of
annular eclipses which were described in this way.
Full translations of all 'A' and 'B' ranked accounts of large solar eclipses
from both phases of the Han dynasty - as reported in the respective Five-
Element Treatises - are given below, along with comments and results of
calculation. The form of each entry follows that in section 8.2 above. For
brevity, all records in category ' C are merely summarised, along with the
234 8 East Asian observations of large solar eclipses

corresponding AT limits, in table 8.5. Throughout the Former Han, the

place of observation will be taken as Ch'ang-an; in the Later Han, Lo-
yang will be assumed. It should be noted that whereas an observation of
totality yields an inclusive range of AT, a partial eclipse is best interpreted
as excluding a set of AT values.
(1) BC 188 Jul 17 (total mag. = 1.02): Ch'ang-an [class B]
Emperor Hui, 7th year, 5th month, day ting-mao [4], the last day of the
month. The Sun was eclipsed; it was almost complete. It was in the
beginning of (the lunar lodge) Ch'i-hsing.
[Han-shu, chap. 27c]
The recorded RA of the Sun corresponds to 115 deg. This is in good
agreement with the calculated figure of 112 deg.
RESULTS
For a partial eclipse at Ch'ang-an, either AT < 13 830 or > 14140 sec.
(NB computed local time of maximum = 15.2 h; solar altitude = 47
deg.)
(2) BC 181 Mar 4 (total, mag. = 1.04): Ch'ang-an [class B]
Empress of Kao-tzu, 7th year,firstmonth, day chi-ch'ou [26], the last
day of the month. The Sun was eclipsed; it was total; it was 9 deg in (the
lunar lodge) Ying-shih, which represents the interior of the Palace
chambers. At that time the (Dowager) Empress of Kao-[tzu] was upset by
it and said, This is on my account'. The next year it was fulfilled.
[Han-shu, chap. 27c]
The Empress Dowager died nearly 18 months afterwards, on 18 August
in 180 BC The eclipse and the Empress' reaction are described in more
detail in the Annals of the Shih-chi, a work composed some 150 years
before the Han-shu. This is clearly based on an eyewitness report:
On the day chi-ch'ou, the Sun was eclipsed and it became dark in the
daytime. The Empress Dowager was upset by it and her heart was ill at
ease. Turning to those around her she said, This is on my account'.
[Shih-chi, chap. 9.]
Although in his Shih-chi Szu-ma Ch'ien notes the occurrence of many
solar eclipses during the first century or so of the Former Han dynasty,
only in this single instance does he give any descriptive details. His account
suggests that the eclipse was extremely large at Ch'ang-an, but unfortu-
nately there is no direct reference to the total phase. Conversely, although
the Han-shu account states that the eclipse was total, it omits to give any
qualifying information. Hence category 'B' seems most appropriate. The
recorded RA of the Sun corresponds to 329 deg. This is in poor accord
with the calculated figure of 342 deg.
 8.4 The Ch'in and Han dynasties 235

RESULTS
For totality at Ch'ang-an, 11 800 sec < AT < 12 720 sec.
(NB computed local time of maximum = 14.7 h; solar altitude = 33
deg.)
(3) BC 147 Nov 10 (total mag. = 1.02): Ch'ang-an [class B]
Emperor Ching, 3rd year, 9th month, day wu-hsi [35], the last day of the
month. The Sun was eclipsed; it was almost complete. It was 9 deg in (the
lunar lodge) Wei.
[Han-shu, chap. 27c]
The recorded RA of the Sun corresponds to 228 deg. This is in excellent
agreement with the calculated figure of 227 deg.
On this occasion, the track of totality did not extend further south than
36.2 deg N latitude at any point on the Earth's surface. Hence no value of
AT will render the eclipse central at Ch'ang-an. In consequence, no limits
to AT can be derived from the observation that the phase was partial there.
(NB computed local time of maximum = 11.1 h; solar altitude = 38
deg.)

(4) BC 89 Sep 29 (barely annular, mag. = 0.997): Ch'ang-an [class A]

Cheng-ho reign period, 4th year, 8th month, day hsin-yu [58], the last
day of the month. The Sun was eclipsed; it was not complete and like a
hook. It was 2 deg in K'ang. At the hour of/w (= 15-17 h) the eclipse
began from the north-west. Towards the hour of sunset it recovered.
[Han-shu, chap. 27c]
As this eclipse was only said to be partial and the zone of annularity
was of negligible width, virtually any value of AT would be acceptable.
The time of day when the eclipse began is too rough to provide any useful
information regarding AT. However, the fact that the eclipse is said to
have ended before sunset enables a limiting value for AT to be deduced.
This observation will be considered in chapter 9 below. The recorded RA
of the Sun is equivalent to 188 deg. This is in fair agreement with the
calculated value of 183 deg.
(NB computed local time of maximum = 15.7 h; solar altitude = 27
deg.)

(5) BC 80 Sep 20 (total mag. = 1.02): Ch'ang-an [class B]

Yuan-feng reign period, first year, 7th month, day chi-hai [36], the last
day of the month. The Sun was eclipsed; it was almost complete. It was 20
deg in (the lunar lodge) Chang.
[Han-shu, chap. 27c]
The recorded RA of the Sun corresponds to 143 deg. This is in very
poor accord with the calculated figure of 175 deg.
236 8 East Asian observations of large solar eclipses

RESULTS
For a partial eclipse at Ch'ang-an, either AT < 8200 or > 8520 sec.
(NB computed local time of maximum = 13.2 h; solar altitude = 54
deg.)
(6) BC 35 Nov 1 (?) (annular, mag. = 0.985): Ch'ang-an [class B]
Chien-chao reign period, 5th year, 6th month, day jen-shen [9], the last
day of the month. The Sun was eclipsed; it was not complete and like a
hook; then (yin) it set.
[Han-shu, chap. 27c]
The date in the above text, which is repeated in the Basic Annals
(chap. 9), cannot be correct. It corresponds to Aug 23 in 34 BC, but
no eclipse was visible on the Earth's surface on that date. Unfortunately,
the RA of the Sun is not specified; this would have materially helped
in establishing the correct date. From the eclipse maps of Stephenson
and Houlden (1986), it can be seen that between 39 and 29 BC only a
single eclipse could have been large in China: that of Nov 1 in 35 BC,
which would indeed occur near sunset. This event, first identified by Dubs
(1944, vol. II, p. 355), took place on the day ting-ch'ou, the last day of the
ninth month in the fourth year of the Chien-chao reign period. Since the
recorded and calculated dates are only some ten months apart, this is a
viable alternative, but it is difficult to explain how such a rare combination
of major calendrical errors occurred. Despite the careful description of
the appearance of the Sun, I have therefore included the observation in
category 6B\
As the zone of annularity in 35 BC would not reach further north than
about latitude 27 deg N anywhere on the Earth's surface, the eclipse can
only have been partial at Ch'ang-an for any value of AT. Although this
result is in keeping with observation, it is not possible to use the record to
set limits to AT. The reference to the Sun setting largely eclipsed is more
useful and this will be discussed in chapter 9.
(NB computed local time of maximum = 17.6 h; solar altitude = —3
deg.)
(7) BC 28 Jun 19 (total mag. = 1.07): Ch'ang-an [class A]
Ho-p'ing reign-period, first year, 4th month, day chi-hai [36], the last
day of the month. The Sun was eclipsed; it was not complete and like a
hook; it was 6 deg in Tung-ching... When the eclipse first began, it started
from the south-west.
[Han-shu, chap. 27c]
The above entry is followed by a lengthy astrological commentary. This
eclipse is also noted in the Biographical Section of the Han-shu in a
 8.4 The Ch'in and Han dynasties 237

chapter dealing with the maternal relatives of the emperors. The following
account from this source implies that the magnitude was particularly large
at Ch'ang-an:
The Sun was eclipsed in Tung-ching. A moment later it was almost
exhausted; it was not much different from a total eclipse.
[Han-shu, chap. 97b.]
The solar RA reported in the Five-Element Treatise corresponds to 72
deg. This is in poor accord with the calculated figure of 84 deg.

RESULTS
For a partial eclipse at Ch'ang-an, either AT < 8090 sec or > 9530 sec.
(NB computed local time of maximum = 9.6 h; solar altitude = 57
deg.)

(8) BC 2 Feb 5 (barely total, mag. = 1.006): Ch'ang-an [class A]

Yuan-shou reign period, first year, first month, day hsin-ch'ou [38], the
first day of the month. The Sun was eclipsed; it was not complete and like
a hook. It was 10 deg in Ying-shih.
[Han-shu, chap. 27.]
Since this eclipse was said to be partial and the central zone was
extremely narrow, only a very restricted range of AT (between about
2050 and 2150 sec) is excluded. The observation is thus only of minimal
value. The solar RA corresponds to 332 deg - considerably more than
the calculated value of 317 deg.
(NB computed local time of maximum phase = 7.8 h; solar altitude =
11 deg.)

(9) AD 120 Jan 18 (total mag. = 1.03): Lo-yang [class A]

Yuan-ch'i reign-period, 6th year, 12th month, day wu-wu [55], the first
day of the month. The Sun was eclipsed. It was almost complete; on Earth
it was like evening. It was 11 deg in Hsu-nu. The Female Ruler was upset
by it; two years and three months later, Teng, the Empress Dowager, died.
[Hou-han-shu, chap. 28.]
The Empress Dowager's death actually occurred 15 months later, on
April 17 in AD 121 (inclusive counting of years is adopted in the text).
An almost contemporary text, the Fu-hou Ku-chin-chu ('Lord Fu's Com-
mentary on Things Old and New'), written c. AD 140, notes that 'the
stars were all visible' but no other details are preserved. This is included
along with the Hou-han-shu report. Other commentaries speak indirectly
of the astrological consequences of a total eclipse but this may be merely
a generalisation.
238 8 East Asian observations of large solar eclipses

A relatively recent block-printed version of the text, together with the

various commentaries (shown in smaller letters) is illustrated in figure 8.4.
In the Hou-han-shu account we have a good description of an eclipse
which fell only marginally short of totality. Further, there is a particularly
strong suggestion that the eclipse was indeed seen at the capital. Around
this date the Five-Element Treatise is careful to specify when an eclipse was
not seen at Lo-yang and in such instances the true place of observation (in
the provinces) is usually given. This is the case for both the immediately
preceding eclipse of AD 118 Sep 3 and the following one of AD 120 Aug
12. However, no such statement accompanies the AD 120 Jan 18 record.
The RA of the Sun is equivalent to 297 deg, very close to the calculated
value of 299 deg.
RESULTS
For a partial eclipse at Lo-yang, either AT < 8150 sec or > 8970 sec.
(NB computed local time of maximum = 13.7 h; solar altitude = 29
deg.)

In table 8.5 are given the following details for each of the remaining Han
observations in categories 'B' and ' C : Julian date; type of eclipse (whether
annular or total); computed magnitude in the central zone; summary of
observation; lower and upper limit to AT (in seconds) necessary to satisfy
the record.
Referring to the results in table 8.5, the following specific comments
seem appropriate. In 198 BC, the fairly broad zone of annularity ran
almost parallel to the equator; as a result, the calculated AT range to
satisfy the observation is unusually wide. The belt of totality in AD 2 was
of almost negligible width (some 5 km in the latitude of Ch'ang-an) and
thus only an extremely narrow range in AT would satisfy the observation.
The probability of totality actually being witnessed and reported - whether
at Ch'ang-an or anywhere else in China - is very low. On this occasion,
the most plausible assumption would seem to be that an account of an
almost total eclipse from Ch'ang-an became corrupted into an assertion
of totality by the time it found its way into the official history. The
observation would thus appear to be valueless for determining AT.

8.5 From the San-kuo to the Sui dynasty (AD 220-617)

After the fall of the Han dynasty in AD 220, China was divided into
three kingdoms or San-kuo (namely Shu Han, Wei and Wu). This period
of partition lasted for only 45 years. In AD 265 the country was again
unified, this time under the Chin dynasty, whose capital was initially
Lo-yang - as in most of the Later Han. However, the northern half of
the country was overrun by invaders half a century later and Lo-yang
 8.5 From the San-kuo to the Sui dynasty 239

Fig. 8.4 Nineteenth century block print containing AD 120 solar eclipse record.
240 8 East Asian observations of large solar eclipses

Table 8.5 AT limits derived from 'C'-rated Han records of large

solar eclipses.

AT Range (sec)
Julian date Type Mag. Record LL UL
BC 198 Aug 7 A 0.95 total 6350 13110
AD 2 Nov 23 T 1.001 total 8570 8 590
AD 65 Dec 16 T 1.02 total 8450 8 900

was destroyed. As a result, in AD 316, Chien-k'ang (modern Nanjing)

became the new Chin metropolis. On the demise of the Chin dynasty
in AD 420, a series of relatively short-lived dynasties (Sung, Ch'i, Liang
and Ch'en) followed in south China. In each case the capital was still
Chien-k'ang. In the north, the principal dynasty was the Northern Wei,
centred on the cities of P'ing-ch'eng (near the site of modern Da-t'ung)
and later a rebuilt Lo-yang. Eventually, in AD 589, the country was re-
unified under the Sui dynasty. A vast metropolis named Ta-hsing Ch'eng
('City of Great Revival'), close to the site of the Former Han capital of
Ch'ang-an, was built soon afterwards. However, following the death of the
first Sui Emperor in AD 605, Lo-yang became the new imperial residence.
Although of brief duration (lasting only until AD 617), the Sui gave way
to the long-lived T'ang dynasty.
This turbulent period of almost four centuries is covered by several
dynastic histories, which (as at all periods down to the end of the Yuan)
provide the most direct source of eclipse reports. In most of these histories,
the major source of solar eclipse observations is the astrological treatise;
the only exception is the Sung-shu ('History of the Liu-sung Dynasty'),
which incorporates these data in its Five-Element section instead (see
table 8.4). Records from the Chin, Liu-sung, Southern Ch'i, Northern Wei
and Sui dynasties are contained in the treatises of the official histories
devoted to these same dynasties. However, several official histories do
not contain any treatises. The corresponding observations from the San-
kuo (Three Kingdoms Period) are found in both the Tien-wen Chih of
the Chin-shu and the Wu-hsing Chih of the Sung-shu (this latter work
also cites many Chin observations). Additionally, the Sui-shu astrological
monograph covers the brief Liang, Ch'en and Northern Chou dynasties.
This official history was compiled between AD 629 and 636 - soon after
the end of the Sui dynasty - under Wei Cheng.
Four total eclipses (in AD 306, 454, 562 and 616) and two large partial
obscurations of the Sun (in AD 360 and 429) are reported in the various
treatises of the period. Of these, the eclipses of AD 360, 429, 454 and 616
 8.5 From the San-kuo to the Sui dynasty 241

are all recorded as total in the Imperial Annals of the period, but without
any further details. In particular, the reports of totality in the Chin annals
for AD 360 and in the Sung annals for AD 429 both conflict with the
rather careful descriptions of partial eclipses found in the treatises. Two
additional eclipses which occurred on AD 516 Apr 16 (this was generally
annular on the Earth's surface) and 522 Jun 10 (generally total) were
both described as complete - but without any further information - in the
Basic Annals of the Nan-shih ('History of the Southern Dynasties'). The
second record is also to be found in the Annals of the Liang-shu. These
observations do not find support in any astronomical treatise. As for the
Han dynasty, these observations - as well as the other eclipse reports in
the various annals of the period - will not be considered further.
Of the six observations which either affirm or deny a central eclipse
in the various treatises, three - AD 360, 429 and 454 - are especially
detailed. All are to be found in a section of the Five-Element Treatise
of the Sung-shu (chap. 34). I have included these in category 'A, and
for each a full translation, together with commentary, is given below. A
further eclipse, occurring in AD 494, was said to reach only a magnitude
of one-third between 9 and 11 h. However, if the recorded local time of
maximum is accurate, the computed magnitude would actually lie between
0.65 and 0.9. The fact that totality is denied may be of some consequence
and this observation will be included in category 'B'. The remaining three
rather brief records noted above (AD 306, 562 and 616) are assigned to
class ' C and will be discussed at the end of this section. Each simply
notes the occurrence of a total obscuration of the Sun.
During the eclipse of AD 523 Nov 23 it is reported in the Sui-shu
Astronomical Treatise (chap. 21) that Venus was seen by day, but the
eclipse magnitude is not specified. This observation, although interesting,
is of little significance in the present context since it does not necessarily
imply a very large eclipse; numerous sightings of Venus in full daylight
- sometimes for many days at a time - are recorded throughout Chinese
history from the Han onwards.

(1) AD 360 Aug 28 (annular, mag. = 0.93): Chien-k'ang [class A]

Sheng-p'ing reign period, 4th year, 8th month, day hsin-ch'ou [38], the
first day of the month. The Sun was eclipsed. It was not complete and like
a hook.
[Sung-shu, chap. 34: section on Chin eclipse records.]
In part of the Astrological Treatise (chap. 12) of the Chin-shu, this
eclipse is reported as follows:
It was almost total and was in Chueh. Whenever an eclipse covers a
small portion of the Sun the calamity it brings will be relatively small, but
242 8 East Asian observations of large solar eclipses

when it covers a large portion of the Sun the consequences will be much
more serious. Chueh forms the 'Celestial Entrance', and hence misfortune
would fall upon the Head of State - the next year the Emperor died.
[Trans. Ho (1966, p. 159).]
A portion of the Chin-shu account has already been quoted (section 8.4)
to illustrate the extent to which magnitude was regarded as affecting the
omen-value of an eclipse. The young Emperor Mu died on 10 July in AD
361; he was then aged 19. Although the expression 'almost complete' in
the Chin-shu text is obscure when applied to an eclipse which is generally
annular, the Sung-shu terminology ('not complete and like a hook') is
quite specific.
RESULTS
For a partial eclipse at Chien-k'ang, either AT < 7120 sec or > 9400 sec.
(NB computed local time of maximum = 11.3 h; solar altitude = 65
deg.)
(2) AD 429 Dec 12 (total, mag. = 1.04): Chien-k'ang [class A]
Yuan-chia reign period, 6th year, 11th month, day chi-ch'ou [26], the
first day of the month. The Sun was eclipsed; it was not complete and like
a hook. During the eclipse, stars were seen. At the hour of/w (= 15-17 h),
then it disappeared (i.e. ended). In Ho-pei (province) the Earth was in
darkness.
[Sung-shu, chap. 34.]
This eclipse was evidently total in Ho-pei (= Hebei) province but
unfortunately a precise location is not specified. Situated far to the north
of the Liu-Sung capital of Chien-k'ang, Ho-pei was then part of the
Northern Wei Empire. Since the track of totality did not extend further
south than 35 deg N latitude at any point on the Earth's surface, no
alteration in AT will render this eclipse central at Ch'ien-k'ang. Hence
no limits to AT can be derived from the observation that the phase was
incomplete there.
(NB computed local time of maximum = 12.6 h; solar altitude = 34
deg.)
(3) AD 454 Aug 10 (total mag. = 1.03): Chien-k'ang [class A].
Hsiao-chien reign period, first year, 7th month, day ping-shen [33], the
first day of the month. The Sun was eclipsed; it was total; all the
constellations (i.e. lunar lodges) were brightly lit.
[Sung-shu, chap. 34.]
The previous entry in chap. 34 of the Sung-shu also notes a total eclipse,
this time on the day hsin-ch'ou (38), the first day of the seventh month
 8.5 From the San-kuo to the Sui dynasty 243

in the 30th year of the Yuan-chia reign period. On this occasion it was
said that 'the stars were all seen'. However, no eclipse was visible in
China on this date - which corresponds to AD 453 Aug 20 - and apart
from AD 454 Aug 10 itself, none could have been total in China at any
time between AD 438 and 494. This latter record may well represent a
further reference to the eclipse of AD 454, misplaced one calendar year
by the compilers of the Wu-hsing Chih, and with the cyclical day adjusted
retrospectively.
Whatever the explanation of the AD 453 record, in order for many
stars to have appeared 'brightly lit' at the eclipse of AD 454 Aug 10, the
eclipse must have been fully total.

RESULTS
For a total eclipse at Chien-k'ang in AD 454, either +70 < AT < 1800
sec or 6130 < AT < 7900 sec.
As discussed in chapter 3, two possible ranges of AT are indicated by
this observation. However, comparison with the AT values obtained from
roughly contemporaneous Chinese observations in both this chapter and
chapter 9 indicates that only the latter range (i.e. 6130 < AT < 7900 sec)
is acceptable.
(N.B. computed local time of maximum = 8.9 h; solar altitude = 45
deg.)

(4) AD 494 Jun 19 (total: mag. = 1.04): Chien-k'ang [class B]

Lung-ch'ang reign period, first year, 5th month, day chia-hsu [11], the
first day of the month. At the hour of szu (9-11 h) the Sun was one-third
eclipsed. At the hour of wu (11-13 h) it was restored to fullness.
[Nan-ch'i-shu, chap. 12.]
No possible variation in AT will lead to a visible magnitude as small
as 0.33 at the Southern Ch'i capital of Chien-k'ang. For any value of AT
which places the Sun above the horizon at this site, the actual magnitude
cannot have been less than 0.56. Furthermore, as noted above, if the
recorded local time of maximum is accurate, the computed magnitude
then would be at least 0.65. Hence, although the magnitude estimate must
be erroneous, the fact that there is no mention of the total phase seems
worth investigating. It seems clear from the text that the Sun was visible
when the eclipse was at its height; there is nothing to suggest that cloud
intervened. Computation shows that for values of AT in the range 5980
to 6600 sec the eclipse would be total at Chien-k'ang. Since the eclipse
appears to have been only partial there, a value for AT outside these
limits is required. In view of the textual error in recording the magnitude,
the observation will be included in category 'B'.
244 8 East Asian observations of large solar eclipses

Table 8.6 AT limits derived from 'C'-rated records of solar eclipses

in the period from AD 220 to 616.

AT Range (sec)
Julian date Type Mag. Record LL UL
AD 306 Jul 27 A 0.94 total 6550 7890
AD 562 Oct 14 A 0.99 total
AD 616 May 21 A 0.97 complete 2300 3030

RESULTS
For a partial eclipse at Chien-k'ang, either AT < 5980 or > 6600 sec.
(NB computed local time of maximum = 8.9 h; solar altitude = 48
deg.)

Each of the eclipses of AD 306, 562 and 616 was recorded as total (using
the term chin in AD 616). The first of these records is to be found in chap.
34 of the Sung-shu while the two later observations are cited in different
sections of chap. 21 of the Sui-shu. Each event occurred in a different
dynasty (respectively Chin, Northern Chou and Sui). The appropriate
capitals were Lo-yang (AD 306), Ch'ang-an (AD 562) and again Lo-yang
(AD 616). Although the original Sui metropolis was Ta-hsing Ch'eng, by
AD 616 it had been moved to Lo-yang. As it happens, all three eclipses
were generally annular rather than total.
Assumption of a central eclipse in both AD 306 and 616 leads to
specific limits for AT. However, no value of this parameter can render
the eclipse of 562 central anywhere in China; the track of annularity
never reached any further north than latitude +17 deg in the longitudes
of China. It seems highly unlikely that an observation was communicated
from some remote site outside China. Instead, there is a real possibility
that the record represents a fairly successful prediction cited as though
it were an observation. There is evidence of eclipse prediction in China,
using numerical cycles based on past observation, from the San-kuo
period onwards (Foley and Stephenson, 1997). For further details of Sui
predictions, see chapter 9 below; examples of T'ang and Sung predictions
are given in sections 8.6 and 8.7.
In table 8.6 are listed the following details for the three ' C rated
eclipses discussed above: Julian date; eclipse type; magnitude in central
zone; observation; AT range to satisfy the record.
 8.6 The T'ang dynasty and Wu-tai period 245

8.6 The T'ang dynasty and Wu-tai period (AD 617-960)

There are two official T'ang histories: the 'Old History of the T'ang
Dynasty' (Chiu-t'ang-shu), written between AD 940 and 945 (soon after
the dynasty had ended) under the direction of Liu Hsu, and the more
systematically compiled 'New History of the T'ang Dynasty' (Hsin-t'ang-
shu). The latter work was composed by a team of scholars supervised by
Ou-yang Hsiu and Sung Ch'i between AD 1043 and 1060. The earlier
history often contains details which are omitted in the revision - and this
is true of several eclipse records.
The extensive astrological treatises of the Chiu-t'ang-shu and Hsin-t'ang-
shu both contain numerous eclipse observations. These are to be found in
chapters 36 and 32 of the respective works. Contrary to normal practice,
the Annals of the Chiu-t'ang-shu sometimes give more extensive accounts
of celestial phenomena than the Astrological Treatise of the same his-
tory. However, in recording solar eclipses, the treatises and annals of
both histories tend to support one another; they do not contain mutually
contradictory information as in earlier works. A roughly contempora-
neous work, the Tang-hui-yao ('Collected Essentials of T'ang History'),
completed in AD 961, also cites many eclipse records but these do not
yield any information which is not found in the dynastic histories.
Throughout most of the T'ang, the capital was Ch'ang-an, the former
Sui metropolis of Ta-hsing Ch'eng. This city was re-named after the
Former Han capital, although it lay a few kilometres toward the south-
east of that site. The present-day city of Xi'an is centred on T'ang
Ch'ang-an. Several large solar eclipses are noted in T'ang history (see
below). However, the two official histories of the brief Wu-tai ('Five
Dynasties') era, which commenced on the fall of the T'ang in AD 906,
do not specify the magnitudes of the eclipses which they cite. In the
remainder of this section, we shall concentrate exclusively on the T'ang
observations.
Eight eclipses of the Sun are described as either total or very large in
T'ang history. These were recorded in the years AD 756, 761, 879 and 888
(total) and 702, 729, 754 and 822 (partial). The observations of the eclipses
of 702, 729, 754, 761 and 822 are fairly detailed and I have included all
these in category 'A'. Since the account of the total obscuration in AD 888
is very brief and is only cited in the Annals of the Hsin-t'ang-shu, I have
not considered it further.
Although the remaining two records (AD 756 and 879) are brief, both
present unusual problems and will therefore be discussed first.
On a date equivalent to AD 756 Oct 28, both the Hsin-t'ang-shu
Astrological Treatise (chap. 32) and the Chiu-t'ang-shu Annals (chap.
10) report that 'The Sun was eclipsed and it was total', the former work
246 8 East Asian observations of large solar eclipses

adding that 'the Sun was 10 deg in (the lunar lodge) 77. Although the date
is accurately noted, the place of observation on this occasion cannot be
established. Only a few months previously, the capital of Ch'ang-an had
been captured by rebels led by An Lu-shan; the Emperor and his court
had fled to the provinces. At the time of the eclipse there was turmoil.
Ch'ang-an was not recovered until the following year.
Towards the end of the T'ang dynasty, on the date AD 879 Apr 25,
another eclipse is recorded in the Astrological Treatise of the Hsin-t'ang-
shu, in the following words: The Sun was eclipsed; it was total. It was
8 deg in Wei\ However, although the Moon and Sun prove to be in
conjunction on that day, and the estimated RA of the Sun (32 deg) is
in good accord with computation (36 deg), no eclipse in fact occurred.
Further, none would be central in China for several years around this time.
Rather than make a major adjustment to the date, it would seem most
reasonable to assume that this entry represents an abortive prediction.
The distance of the Moon from the node at conjunction was 18 deg, only
just beyond the limit for a real eclipse to be possible. Direct evidence of
solar eclipse prediction during the T'ang is to be found in the Astrological
Treatise of the Chiu-t'ang-shu; thus eclipses in AD 778 and 790 were said
to have been predicted but not seen. However, by making appropriate
computations, it can be readily shown that the records of eclipses in both
official T'ang histories contain many other false sightings which are cited
as if they were observations. In some instances, the reported events did
not occur; in several others, an eclipse was only visible in areas of the
Earth's surface remote from China (for example in Antarctica). There is
thus ample evidence of prediction on a large scale (Cohen and Newton,
1981-3; Foley and Stephenson, 1997).
The following translations of the five 'A rated T'ang eclipse records
are mainly based on the accounts in the Astrological Treatises of the
Chiu-t'ang-shu and Hsin-t'ang-shu. However, when the Imperial Annals
of the two histories provide additional information, this will also be
considered.

(1) AD 702 Sep 26 (total, mag. = 1.04): Ch'ang-an [class A]

Ch'ang-an reign period, 2nd year, 9th month, day i-ch'ou [2]. The Sun
was eclipsed; it was almost complete. It was in Chueh.
[Hsin-t'ang-shu, chap. 32.]

The Annals of the Chiu-t'ang-shu (chap. 6), although omitting the solar
co-ordinate, state that the eclipse was 'not complete and like a hook; it
was seen at the capital and in the four directions'. Here we have a rare
instance where observation at the capital is expressly stated.
 8.6 The T'ang dynasty and Wu-tai period 247

RESULTS
For a partial eclipse at Ch'ang-an, either AT < 1460 sec or > 2760 sec.
(NB computed local time of maximum = 15.6 h; solar altitude = 28
deg.)
(2) AD 729 Oct 27 (total mag. = 1.03): Ch'ang-an [class A]
K'ai-yuan reign period, 17th year, 10th month, day wu-wu [55], the first
day of the month. The Sun was eclipsed; it was not complete and like a
hook. It was 9 deg in Ti.
[Hsin-t'ang-shu, chap. 32.]
This same account is found in the Annals of the Chiu-t'ang-shu (chap.
8), apart from the reference to the solar co-ordinate. The recorded RA
of the Sun is equivalent to 206 deg, considerably less than the calculated
value of 215 deg.
RESULTS
For a partial eclipse at Ch'ang-an, either AT < 420 sec or > 1190 sec.
(NB computed local time of maximum = 7.3 h; solar altitude = 8 deg.)
(3) AD 754 Jun 25 (total, mag. = 1.04): Ch'ang-an [class A]
T'ien-pao reign period, 13th year, 6th month, day i-ch'ou [2], the first
day of the month. The Sun was eclipsed; it was almost total. It was 19 deg
in Tung-ching.
[Hsin-t'ang-shu, chap. 32.]
The Annals of the Chiu-t'ang-shu (chap. 9), while omitting the position
of the Sun, assert that the eclipse 'was not complete and like a hook'. The
reported RA of the Sun, which corresponds to 97 deg, is identical to the
calculated figure.
As the track of totality did not reach further north than latitude 30
deg, for any value of AT the eclipse would be no more than partial at
Ch'ang-an. Hence calculation is in accord with observation but no limits
on AT can be set.
(NB computed local time of maximum = 11.8 h; solar altitude = 79
deg.)
(4) AD 761 Aug 5 (total mag. = 1.05): Ch'ang-an [class A]
Shang-yuan reign period, 2nd year, 7th month, day kuei-wei [20], the
first day of the month. The Sun was eclipsed; the large stars were all seen.
The Astronomer-Royal Ch'u T'an reported (to the Emperor): 'On day
kuei-wei the Sun diminished. The loss began at 6 marks (k'o) after the hour
of ch'en. At 1 mark after the hour of szu it was total. At 1 mark before the
hour of wu it was restored to fullness. The eclipse was 4 deg in Chang9.
[Chiu-t'ang-shu, chap. 36.]
248 8 East Asian observations of large solar eclipses

A block-print of the Chinese text of this account is shown in figure 8.5.

Both the Annals of the Chiu-t'ang-shu (chap. 10) and the Astrological
Treatise of the Hsin-t'ang-shu affirm that the eclipse was total and that
'all the large stars were seen', as does the Astrological Treatise of the
Tang-hui-yao (chap. 42). The local times of the various phases recorded
in the Chiu-t'ang-shu Treatise are ambiguously expressed - see chapter 9
below. It should be noted that the unit of time known as the k'o (here
translated 'mark') represented 1/100 of a (24-hour) day, or 0.24 h; see
also chapter 9. The RA of the Sun, reported in the Astrological Treatises
of the Hsin-t'ang-shu and Chiu-t'ang-shu as 4 deg in Chang (i.e. 137 deg)
is in close agreement with the calculated value of 138 deg.
For totality at Ch'ang-an, 1720 sec < AT < 3290 sec.
(NB computed local time of maximum = 9.8 h; solar altitude = 55
deg.)

(5) AD 822 Apr 25 (total, mag. = 1.07): Ch'ang-an [class A]

Ch'ang-ch'ing reign period, 2nd year, 4th month, day hsin-yu [58], the
first day of the month. The Sun was eclipsed. It was 12 deg in Wei; \ of it
was not complete. In (the regions) Yen and Chao it was seen to be total.

[Chiu-t'ang-shu, chap. 36.]

Yen and Chao represent the regions covered by the former Warring
States bearing these names. These correspond roughly to the provinces
of Ho-pei (= Hebei) and Shan-hsi (= Shanxi) in northern China. It
is regrettable that more precise locations where totality was witnessed
are not given. Evidently the compilers of the Chiu-t'ang-shu thought it
unnecessary to state where the partial phase was seen and also considered
this observation to be more important than the provincial reports of
totality. There is thus a very strong presumption that the eclipse was partial
(magnitude | ) at the capital itself. The reported solar RA, equivalent to
37 deg, is close to the calculated figure of 36 deg.
Since totality was witnessed in north-east China, the central zone must
have passed to the north of Ch'ang-an. Hence the observation from the
capital sets a single (upper) limit to the value of AT.

RESULTS

For a partial eclipse at Ch'ang-an, AT < 4020 sec.

The magnitude estimate of | will be considered in chapter 9.
(NB computed local time of maximum = 12.2 h; solar altitude = 70
deg.)
 8.6 The T'ang dynasty and Wu-tai period 249

S fe Ar ML S — tTt ^ f T l i^ JY '>V
7C/

!4kH-ff. ^^

Fig. 8.5 Nineteenth century block print containing AD 761 total solar eclipse
text.
250 8 East Asian observations of large solar eclipses

8.7 The Sung, Kin and Yuan dynasties (AD 960-1368)

The brief and turbulent Wu-tai period was succeeded by the long-lasting
Sung dynasty in AD 960. For the first half of the new dynasty, the capital
was situated at Pien. However, in 1127 the capital, along with most of
China lying north of the river Ch'ang-chiang (Yang-tze) fell to invaders -
the Juchen - who had established the Kin dynasty 12 years before. As a
result, the Sung capital was removed south of the Yang-tze to Lin-an and
thus began the period known as the Southern Sung. During the remainder
of this dynasty, the territories occupied by the Juchen were never regained.
In AD 1276, Lin-an was captured by the Mongols, who had overrun the
northern half of China half a century before. Three years later the last
Sung prince was drowned during a naval battle and the dynasty came to
an end.
The original capital of the Kin dynasty - at Shang-ching in Manchuria
- was replaced in 1153 by Yen-ching (a city occupying the site of modern
Beijing). Afterwards, in 1214, the Kin metropolis was further transferred
to the former Sung imperial residence of Pien, whose name was changed
to Pien-ching (ching means capital). Twenty years later, this city fell into
Mongol hands and the Kin dynasty was extinguished.
Astronomical observations commenced at Pien at the very beginning
of the Sung dynasty. However, between AD 1115 and 1234, a variety
of celestial phenomena were recorded independently in southern and
northern China by the Sung and Kin astronomers. After the fall of the
Kin dynasty, Sung observations continued until the Mongol Conquest.
The whole of China now remained under the domination of the Mongols
- who established the Yuan dynasty - until 1368 when the invaders were
finally driven out.
Kubilai became the first Yuan emperor. Although he had become Khan
in AD 1260, he did not found the Yuan dynasty until 1271. However,
eclipse observations are recorded in the official history of the dynasty
(the Yuan-shih) from AD 1261 onwards. In 1267, Kubilai transferred his
capital from Karakorum in Mongolia to the former Kin capital of Yen-
ching, which he renamed Ta-tu ('Great Capital'). As its name suggests,
this city was rebuilt on a vast scale. Its site is near the centre of modern
Beijing.
The principal sources of eclipse observations in all three dynasties are
the astrological treatises of the official histories. Each of these treatises is
very extensive; the chapters containing the eclipse records are as follows:
Sung-shih (chap. 52), Kin-shih (chap. 20), Yuan-shih (chap. 48). Between
1343 and 1345 the Sung-shih and Kin-shih were compiled by a team of
Yuan scholars led by T'o T'o. The Yuan-shih itself was composed soon
after the fall of the dynasty under the direction of Sung Lien. It is on
 8.7 The Sung, Kin and Yuan dynasties 251

record that this task took only 370 days. These various works describe
total eclipses in AD 977, 1221 and 1275, and annularity in 1292. Large
partial obscurations of the Sun were noted in 1022, 1054 and 1135. Two
further eclipses of unspecified magnitude during which either 'the large
stars were all visible' (AD 1214) or darkness occurred (1367) were also
reported. No reference to the eclipse of AD 1022 occurs in the Sung
Annals, but the other eclipses listed above are all mentioned in the annals
of the appropriate history. An unusually vivid account of totality in AD
1275 is also to be found in the Five-Element Treatise of the Sung-shih
(chap. 67).
During this same period, a total solar eclipse was also recorded without
any further description in the Annals of the Sung-shih (chap. 18), but
there is no corresponding entry in the Astrological Treatise. The recorded
date, when converted to the Julian calendar (i.e. AD 1099 Nov 30), is in
exact agreement with that of a total lunar eclipse which proves to have
been visible in China. This entry must have been wrongly filed by the
compilers of the history.
Complete obscurations of the Sun, on dates corresponding to AD 1005
Jan 13 and 1122 Mar 10 are briefly noted in the Annals (chaps. 14
and 29) of the Liao-shih, the official history of the Liao dynasty - also
compiled under the direction of T'o T'o. A semi-nomadic kingdom, the
Liao was established in the early tenth century in the extreme north
of China. It was brought to an end by the Juchen in AD 1125. The
Liao-shih does not contain an astronomical treatise. Even if the reports
of totality were reliable, the place of observation cannot be established
with any confidence; the Liao rulers had five separate residences scattered
throughout Manchuria.
The careful accounts of totality in AD 1221 and 1275, of annularity
in 1292, and of the partial phase in AD 1054 and 1135 will each be
assigned to category 'A. However, the observations in 1214 and 1367,
although noting either the visibility of stars or darkness, do not mention
the complete disappearance of the Sun directly. These and two other
reports (AD 977 and 1022), which although merely noting that the phase
was either 'total' or 'almost complete' require special comment, will be
considered first in chronological order.
On a date corresponding to AD 977 Dec 13, both the Annals (chap.
4) and the Astrological Treatise (chap. 52) of the Sung-shih note the
occurrence of a total eclipse of the Sun. Since the computed track of
totality on this occasion did not reach further north than latitude 19
deg N, no variation in AT will render the eclipse central in any part of
China. An 'almost complete' obscuration of the Sun on a date which is
equivalent to AD 1022 Aug 1 is reported in the Astrological Treatise of the
same history. Calculation shows that this eclipse, which is not mentioned
252 8 East Asian observations of large solar eclipses

in the Annals, would be visible mainly in the southern hemisphere and

could not have been seen in China. In the Sung-shih, as in the official
histories of earlier dynasties, a number of spurious sightings of eclipses
are recorded along with genuine observations. The Sung astronomers are
known to have attained a fair degree of success in predicting eclipses. Thus
according to the Sung-hui-yao chi-kao (History of Sung Adminstration),
the Sung Bureau of Astronomy predicted a total eclipse in AD 1021, but
this was seen to be only partial (magnitude ^ ) at the capital. Hence in
both AD 977 and 1022, rather than assume a chance sighting at some
remote location which somehow was communicated to the Sung capital,
it would seem more reasonable to assume predictions recorded as if they
were observations.
The astrological treatise of the Kin-shih (chap. 20) contains the following
record, whose date corresponds to AD 1214 Oct 5: 'The Sun was eclipsed;
the large stars were all visible'. The Annals of the Kin-shih (chap. 14)
mention no more than the occurrence of this eclipse. If the brighter
stars were indeed 'all visible', as the astrological treatise of the Kin-shih
asserts, no phase less than totality would suffice. However, the fact
that the record fails to mention the complete disappearance of the Sun
considerably weakens its value. There is the additional complication that
in the very year that the eclipse occurred the Kin capital was transferred
from Yen-ching to Pien-ching so that the place of observation is uncertain.
Both the Annals of the Yuan-shih (chap. 47) and the Ming-shih-lu
(chap. 24) note the occurrence of a solar eclipse which caused 'darkness
by day' on a date corresponding to AD 1367 Jun 28. However, this
eclipse was only partial on the Earth's surface and would be invisible in
China. Presumably the record represents a faulty prediction which to the
Ming historians who compiled the Yuan-shih became associated with the
downfall of the Yuan dynasty very soon afterwards.
The four eclipses of AD 977, 1022, 1214 and 1367 will not be considered
further. However, the more reliable records from AD 1054, 1135, 1221,
1275 and 1292 are translated in full below, along with comments. Also
included in this section is a discussion of the solar obscuration of AD
1361. A vivid description of an almost complete eclipse in that year is
quoted in a history of the town of Sung-chiang (near Shanghai). This is
the earliest such account to appear in a surviving fang-chih or local history.

(1) 1054 May 10 (total mag. = 1.03): Pien [class A]

Huang-yu reign period, 6th year, 4th month, day chia-wu [31], the first
day of the month. The Sun was eclipsed. The loss began on the south-west
side. The Sun was more than ^ eclipsed and it reached its maximum at 1
mark in the central half of the hour of shen.
[Yuan-shih, chap. 53.]
 8.7 The Sung, Kin and Yuan dynasties 253

This careful account of a large partial eclipse which fell short of totality
at the Sung capital of Pien is recorded in the calendar treatise of the Yuan-
shih. In this treatise, many past observations of eclipses were compared
with the results derived from contemporary tables (see chapter 9). A
very similar account to the above is to be found in chapter 283 of the
Wen-hsien T'oung-k'ao ('Comprehensive History of Civilisation'), compiled
c. AD 1300.
Since the track of totality did not reach further north than latitude +26
deg at any point on the Earth's surface, totality at Pien (lat. 34.8 deg) is
out of the question; the magnitude cannot have exceeded 0.76. A range
of AT cannot be derived. The timing of maximum phase, corresponding
to 16.36 h, will be considered in chapter 9.
(NB computed local time of maximum = 16.5 h; solar altitude = 29
deg.)

(2) 1135 Jan 16 (annular, mag. = 0.93): Lin-an [class A]

Shao-hsing reign period, 5th year, 1st month, the first day of the month.
A man named Ch'en Te-i predicted that the Sun should be 8^ tenths
eclipsed with the beginning of loss in the initial half of the hour of szu.
(These predictions) were verified by observation.
[Sung-shih, chap. 81.]
This account is to be found in the calendar treatise of the Sung-shih.
Although the observational record is brief, it would seem that a large
partial eclipse was witnessed at Lin-an; there is no mention of annularity.
The local time of beginning corresponds to some time between 9 and 10 h.

RESULTS
For a large partial eclipse at Lin-an, AT < 1840 sec or > 3810 sec.
(NB computed local time of maximum = 11.4 h; solar altitude = 39
deg.)

(3) AD 1221 May 23 (total; mag. = 1.07): Kerulen River [class A]

An extensive record is to be found in one of the books which make up the
Tao-ts'ang (Taoist Patrology). The book is entitled Ch'ang-ch'un Chen-jen
Hsi-yu-chi (The Journey of the Adept Ch'ang-ch'un to the West). It is a
detailed narrative of the travels of the elderly Taoist leader Ch'ang-ch'un
from north-east China to Samarkand between AD 1220 and 1224 to meet
Genghis Khan. The Ch'ang-ch'un Chen-jen Hsi-yu-chi (which is listed as
No. 1410 in L. Wieger's catalogue of the Tao-ts'ang) was composed by Li
Chih-ch'ang, one of the disciples of Ch'ang-ch'un who accompanied him
on his long journey. A full translation of Li Chih-ch'ang's account has
been made by Waley (1934). The observation of totality by the Kerulen
254 8 East Asian observations of large solar eclipses

River in eastern Mongolia has been analysed in some detail by Stephenson

and Yau (1992b) and the following discussion (including translations) is
largely based on their investigation.
The eclipse is described in two separate sections of Li Chih-ch'ang's
book. The year is given as jen-wu, the 19th year of the cycle, equivalent
to AD 1221-2.
(i) On the first day of the fifth month (May 23), at noon, the Sun was
eclipsed and it was total. All the stars were therefore seen. A short
while later the brightness returned. At that time we were on the
southern bank of the river. The eclipse (began) at the south-west
and (the Sun) reappeared from the north-east. At that place it is
cool in the morning and warm in the evening; there are many
yellow flowers among the grass. The river flows to the north-east.
On both banks there are many tall willows. The Mongols use them
to make their tents.
[Trans. Stephenson and Yau (1992b).]
While Ch'ang-ch'un was passing the following winter at Samarkand, he
discussed the eclipse with an astronomer:
(ii) (Ch'ang-ch'un) asked (him) about the solar eclipse on the first day
of the fifth month (May 23). The man replied: 'Here the Sun was
eclipsed up to 6 fen (-j|) at the hour of ch'en (7-9 h)'. The Master
continued, 'When we were by the Lu-chu Ho (Kerulen River),
during the hour wu (11-13 h) the Sun was seen totally eclipsed and
also south-west of the Chin-shan the people there said that the
eclipse occurred at the hour szu (9-11 h) and reached 1 fen. At
each of these three places it was seen differently. According to the
commentary on the Ch'un-ch'iu by K'ung Ying-ta, when the body
(of the Moon) covers the Sun, then there will be a solar eclipse.
Now I presume that we must have been directly beneath it; hence
we observed the eclipse to be total. On the other hand, those
people on the sides (of the shadow) were further away and hence
(their view) gradually became different. This is similar to screening
a lamp with a fan. In the shadow of the fan there is no light or
brightness. Further away from the sides (of the fan) then the light
of the lamp gradually becomes greater.'
[Trans. Stephenson and Yau (1992b).]
In both of the above descriptions, the standard term chi was used to
indicate totality. Both the detail 'all the stars were then seen' and Ch'ang-
ch'un's subsequent explanation of the cause of the loss of daylight provide
emphatic evidence in favour of totality having been witnessed.
When the recorded date of the observation is converted to the Julian
calendar (i.e. AD 1221 May 23), it agrees exactly with that of a calculated
 8.7 The Sung, Kin and Yuan dynasties 255

108° 110' 112° 114° 116°

Longitude

Fig. 8.6 Route of Ch'ang-ch'un up the Kerulen river valley in AD 1221. The
computed track of the total solar eclipse of AD 1221 May 23 is also shown
(based on a preliminary value for AT of 850 sec).

eclipse. The main difficulty lies in the fixing of the geographic location
of Ch'ang-ch'un at the time. We know from the detailed narrative that
on the 17th day of the fourth month (May 10) his party left the camp of
Prince Tamuga, younger brother of Genghis Khan, having been given a
farewell gift of several hundred oxen and horses, as well as ten wagons.
Five days later - on the 22nd day of the month (May 15) - it is recorded
that they reached Lake Hulun Nur at the place where the Kerulen River
outflows. From here they proceeded in a westerly direction along the
southern bank of the river. After the eclipse, they continued along the
Kerulen valley for a further 16 days until they 'came to a point where the
Kerulen makes a loop to the north-west, skirting some mountains'. Here,
on a date corresponding to Jun 7, the party turned south-west.
In order to derive a fairly precise location for the travellers at the time
of the eclipse, Stephenson and Yau (1992b) utilised the following maps of
the area: (i) charts on a scale of 1:1000000 published by the US Army
Map Service in 1944 (these were derived from Russian maps of the area
on the same scale produced in 1936); (ii) a map on a scale of 1:2 500000
published in the USSR in 1967; (iii) a 1:5 000000 map in the Times
Atlas. Using this source material, the course of the Kerulen River can be
accurately delineated (see figure 8.6). It is apparent from the topography
shown on the maps that the river has not changed its course appreciably
on a time-scale of millennia.
In figure 8.6, the points A through G (working upstream) denote major
256 8 East Asian observations of large solar eclipses

bends in the river. There is nothing in the narrative to suggest any major
halt between A and G. Colleagues who have journeyed in Mongolia state
that the normal method of travel in this part of the world would be to
rise at dawn and, apart from occasional short stops, progress all day from
sunrise to sunset when camp would be set up. This same procedure would
be repeated daily. Such a routine is implied by Li Chih-ch'ang himself.
It thus seems reasonable to suppose a roughly uniform rate of advance
between points A and G.
Li Chih-ch'ang makes no mention of any diversion of the party from
the course of the river between May 15 and Jun 7. After they reached
the Kerulen (at A), the travellers are stated to have proceeded along the
southern bank of the river. Again, when the eclipse occurred, we are
reminded that the party was 'on the southern bank of the river'. No
mention is made of a short-cut at point B (by-passing the large bend in
the vicinity of C and continuing directly to D); if so, they would have had
to cross a desert region.
It may be estimated that if Ch'ang-ch'un indeed followed the entire
course of the river from A to G, the total distance travelled would be
about 830 km. This corresponds to an average daily rate of progress (over
23 days) of 36 km. Since at this time of year and latitude there would
be nearly 16 hours of daylight per day, the rate of travel would seem
quite reasonable. Although the members of the delegation travelled on
horseback, their baggage was conveyed in ox-carts. Assuming that this
average speed was maintained, when the eclipse occurred the party would
be some ^ of this distance from A or about 290 km from Lake Hulun.
This site is roughly halfway between C and D. Over the whole of the
section CD, the river flows north-east, as asserted in the narrative.
In addition, it is extremely unlikely that the eclipse could have taken
place while Ch'ang-ch'un was still between A and B in figure 8.6. Point
B is only about 145 km from A and thus only 0.17 of the full distance
between A and G. If the party took 8 days to reach B, the average speed
would only be 18 km per day, after which a mean daily rate of progress
of as much as 46 km would be needed in order to arrive at G by Jun 7.
Further, an observation between B and C is ruled out by the reference to
the north-east flow of the Kerulen where the eclipse occurred. A location
somewhere between C and D will thus be assumed.
In figure 8.7, which shows the eastern half of the Kerulen River on a
larger scale, the computed northern boundary of the belt of totality using
AT = 980 sec is marked. This boundary just touches point D. For any
value of AT greater than 980 sec, the track would lie further south and
the eclipse would not be total anywhere along the Kerulen westward of B.
(As discussed above, a location to the east of B is highly unlikely.) Since
the location of Ch'ang-ch'un at the time of the eclipse cannot be deduced
 8.7 The Sung, Kin and Yuan dynasties 257

49°.00

Longitude
Fig. 8.7 Eastern half of Kerulen river valley showing the northern edge of the
track of totality in AD 1221 computed for AT = 980 sec.

more accurately than some point between C and D, it is not possible to

derive a more refined upper limit for AT than this. The lower limit to AT
for totality in this region is extremely negative (about —8100 sec). Hence
only the upper limit of +980 sec is viable.

RESULTS

For totality to be observed by Ch'ang-ch'un on the Kerulen River, AT <

980 sec.
(NB computed local time of maximum at 48 deg N, 115 deg E = 12.9
h; solar altitude = 63 deg.)
Using the value of AT derived from equation (8.1) (i.e. 850 sec), both the
computed magnitude (0.63) and local time (8.1 h) at Samarkand (lat. =
39.67 deg, long. = —66.95 deg) are in good accord with observation.
Similar remarks apply to the report from near the Chin-shan (T'ien-shan)
mountains in Central Asia (approx. lat. = 49 deg, long. = —88 deg). Here
the computed magnitude was 0.73 and local time 10.1 h.

(4) 1275 Jun 25 (mag. = 1.07): Lin-an [class A]

Te-yu reign period, 1st year, month VI, day keng-tzu [37], the first day
of the month. The Sun was eclipsed; it was total. The sky and Earth were
in darkness. People could not be distinguished within a foot. The chickens
and ducks returned to roost. (It lasted) from the hour szu (9-11 h) to the
hour wu (11-13 h); then it regained its brightness.
[Sung-shih, chap. 67.]

The above account is from the Five-Element Treatise of the Sung-shih.

A less extensive description of this event is to be found in the Astrological
Treatise of the same history and this is also worth citing in full:
258 8 East Asian observations of large solar eclipses

The Sun was eclipsed; it was total. Stars were seen. The chickens and
ducks all returned to roost. In the following year the Sung dynasty was
extinguished.

[Sung-shih, chap. 52.]

The Mongol historians who compiled the Sung-shih evidently regarded

the eclipse as an omen of disaster for the Sung regime. The Annals (chap.
47) confirm that the eclipse was total, adding that 'it became as dark as
night'.

RESULTS

For totality at Lin-an -700 < AT < 1300 sec.

By this period, AT had become so small that the set of values needed
to render an eclipse total might well encompass negative figures, as is the
case here.
(NB computed local time of maximum = 10.1 h; solar altitude = 63
deg.)

(5) AD 1292 Jan 21 (annular, mag. = 0.94): Ta-tu [class A]

Chih-yuan reign-period, 29th year, first month, day chia-wu [31]. The
Sun was eclipsed. A darkness invaded the Sun, which was not totally
covered. It was like a golden ring. There were vapours like golden earrings
on the left and right and a vapour like a halo completely surrounding it.

[Yuan-shih, chap. 48.]

The description of the Sun as like a golden ring' is a very clear

allusion to the ring phase - the only such example in East Asian history.
The aptness of this description is clear from figure 8.8, which shows a
photograph of a recent annular eclipse. Evidently a cloud of ice-crystals
in the atmosphere was responsible for the halo display in AD 1292; the
weather in Beijing (formerly Ta-tu) can be extremely cold at this time of
year. The Annals (chap. 14) of the same history note that because of the
eclipse the normal daily greetings to the Emperor were not offered by the
court. However, in this latter account, the ring phase is not described,
only the accompanying atmospheric phenomenon.

RESULTS

For annularity at Ta-tu - 5 0 < AT < 1830 sec.

(NB computed local time of maximum = 12.9 h; solar altitude = 31
deg.)
 8.7 The Sung, Kin and Yuan dynasties 259

Fig. 8.8 Photograph by Loredano Ceccaroni of annular solar eclipse of 1994

May 10 showing the ring phase. (Courtesy: Vastronomia.)

(6) AD 1361 May 5 (total mag. = 1.05): Sung-chiang [class A]

Chih-cheng reign period, 21st year, 4th month, day hsin-szu [18], the
first day of the month. As the Sun was about to sink (i.e. set), suddenly it
lost its light. It took the shape of a plantain (chiao) leaf. The sky was as
dark as night and the stars were shining brightly. For a short time
(literally: for the duration of a meal), the sky became bright again. Then in
a short time (the light) disappeared once more.
[Sung-chiang Fu-chih, chap. 47.]
The Sung-chiang Fu-chih is a history of the town of Sung-chiang (pinyin:
Songjiang) lying a little to the south-west of Shanghai. This history work
was compiled by local government officials in the late Ming dynasty (some
time between 1628 and 1644) using original documents.
The plantain leaf is oval in shape, with many fronds on its lower edge;
it resembles a banana leaf and has often been used as a fan. In comparing
the eclipsed Sun with a plantain leaf - rather than some other shape -
the observer may have had in mind the very obvious fronds of this leaf.
260 8 East Asian observations of large solar eclipses

Perhaps the most likely inference would seem to be that at maximum

phase the Sun was reduced to a fine crescent edged with Baily's beads.
However, it would probably be unwise to overstate this case, and hence I
shall assume only a very large partial eclipse at Sung-chiang.

RESULTS
For a partial eclipse at Sung-chiang, AT < 500 sec or > 1760 sec.
A result in the higher range would be discordant with many roughly
contemporaneous observations; equation (8.1) leads to a value for AT of
approximately 650 sec. Hence it seems reasonable to select only the limit of
AT < 500 sec. If it is only assumed that the Sun appeared crescent-shaped
at maximum phase, it is not possible to obtain a more precise value for
AT than this. However, since the description indicates an extremely large
eclipse, a result for AT close to 500 sec is probably indicated.
(NB computed local time of maximum = 18.6 h; solar altitude = 2
deg.)

8.8 The Ming dynasty (AD 1368-1644)

The Ming dynasty, which was established in 1368 after the collapse of
Mongol rule in China, had its first capital at Ying-t'ien (modern Nanjing).
In 1420 the capital was removed to Pei-ching ('Northern Capital', i.e.
Beijing). The astronomical treatise of the official history, the Ming-shih,
does not contain any solar eclipse records; these are almost entirely
confined to the Imperial Annals instead but there are no allusions to
large eclipses in the annals. Although systematic observations are to be
found in the Ming-shih-lu ('Veritable Records of the Ming Dynasty') -
an extensive chronicle of events covering each reign of the dynasty - this
compilation is also devoid of any references to either total or near-total
solar obscurations. Hence we must turn to other Ming sources.
From about AD 1500 onwards, Ming provincial histories (fang-chih)
often report large eclipses accompanied by darkness and the appearance
of stars. However, by no means all of these records are reliable; frequently
no eclipse took place on or near the stated date so that the event is
unidentifiable. Possibly the compilers were careless in filing documents
or individual reign periods became confused. Nevertheless there are
many reliable accounts of total eclipses in the local histories, sometimes
providing considerable detail. At least ten separate descriptions for each
of the total solar eclipses of 1514 Aug 20, 1542 Aug 11, 1575 May 10 and
1641 Nov 3 are preserved from different parts of China. Although minor
dating errors occur in certain texts, in most cases the date (when reduced
to the Julian or Gregorian calendar) is in exact agreement with the results
of modern computation.
 8.8 The Ming dynasty 261

The following descriptions of the total eclipse of AD 1514 Aug 20 are

fairly typical. Both are translated from the texts quoted in the compilation
by Beijing Observatory (1988). In each case the dates is correctly given
as the ninth year of the Cheng-te reign period, eighth month, first day
hsin-mao [28].
(i) At the hour of wu (i.e. between 11 and 13 h) the Sun was eclipsed.
The sky and Earth became dark in the daytime. All the birds flew
about in alarm. The domestic animals went into the forest. At the
hour of yu (17-19 h) the light came back.
[Fu-ning Chou-chih: local history of Fu-ning county,
Fu-chien ( = Fujian) province.]

(ii) At the hour of wu suddenly the Sun was eclipsed; it was total.
Stars were seen and it was dark. Objects could not be discerned at
arm's length. The domestic animals were alarmed and people were
terrified. After one (double-) hour it became light.
[Tung-hsiang Hsien-chih: local history of Tung-hsiang county,
Chiang-hsi ( = Jiangsi) province.]
The precise place of observation in each of the above accounts cannot
be determined. It could have been anywhere in an area covering several
thousand square kilometres. This is true for nearly all records in the Ming
provincial histories. By the sixteenth century, AT would be so small (less
than about 300 sec) that only a reliable observation at a precisely known
place would be of any value in refining this parameter. Hence most
observations are of little more than historical interest. The succeeding
Ch'ing dynasty (AD 1644-1911) - during which AT was smaller still -
lay entirely within the telescopic period and is thus beyond the scope of
this book. Although the telescope was introduced into China around AD
1630 (D'Elia, 1960, pp. 41-2), eclipse and other astronomical observations
continued to be made in China with the unaided eye until relatively recent
times. These, of course, could not compete in precision with telescopic
observations made in the West.
One of the very few Ming records of a total eclipse seen at a definite
site comes from the town of Sung-chiang, near Shanghai. The observation
is found in the Sung-chiang Fu-chih, the same source as the detailed Yuan
account of the eclipse of AD 1361 (see section 8.7) and dates from 1575.
The text may be translated as follows.

AD 1575 May 10 (total mag. = 1.07): Sung-chiang [class A]

Fang-li reign period, 3rd year, 4th month, day chi-szu [6], the first day
of the month. The Sun was eclipsed. At about the hour of wu (roughly
noon) the eclipse was total. The bright day became dark.
[Sung-chiang Fu-chih, chap. 47.]
262 8 East Asian observations of large solar eclipses

RESULTS

For totality at Sung-chiang, -2150 < AT < 1390 sec.

The rather wide zone of totality ran nearly parallel to the equator in
the longitudes of eastern China.
(NB computed local time of maximum = 14.8 h; solar altitude = 51
deg.)

8.9 Korea
The earliest history of Korea covers the interval from legendary times
to AD 935. This work, entitled the Samguk Sagi (History of the Three
Kingdoms), was compiled by Kim Pu-sik in AD 1145. It is written in
Classical Chinese - a feature which it shares with virtually all other major
Korean literary works composed before the present century. Very early
in the first millennium AD the country was divided into the kingdoms of
Koguryo, Paekche and Silla. Around AD 670 Silla subjugated its rivals
and the era of Silla ascendancy lasted until AD 935. During both the
Three Kingdoms (Samguk) period and the subsequent Silla domination,
the Samguk Sagi chronicles many solar eclipses, using precisely the same
terminology as found in Chinese records. However, the records are very
brief and there are no references to either total or near-total obscurations.
In AD 935, the kingdom of Koryo, which had been founded some
years before in northern Korea by Wang Kun, overthrew Silla. Koryo
established a unified kingdom throughout the Korean peninsula which
was to endure until AD 1392. During most of the Koryo dynasty, the
capital was Songdo (lat. = 37.97 deg, long. = —126.57 deg), on the site
of modern Kaesong. However, when the Mongols invaded the country
in AD 1229, King Kojong and his court fled to the island of Kangwha
(lat. = 37.73 deg, long. = —126.48 deg) where they remained in exile for
many years. In 1270, Kojong's successor returned to the former capital
of Songdo as a vassal of the Mongols. Not until 1364 were the Mongols
finally driven out of Korea.
Early in the Koryo dynasty, a royal observatory was founded at Songdo
and from this site numerous observations of celestial phenomena were
made. These are reported in the extensive Astrological Treatise (chaps.
47-49) of the Koryo-sa (History of Koryo), the official history of the
period. This work, compiled under the direction of Chong Inji in 1451, is
modelled on a typical Chinese dynastic history. As in China, the motive
for skywatching was largely divination, but no surviving eclipse record is
accompanied by an astrological prognostication. Occasional lunar eclipse
reports in the Koryo-sa describe ceremonies in which the king accompanied
by his ministers attempted to rescue the Moon from its eclipse. Possibly
 8.9 Korea 263

these represented little more than the maintenance of a long-established

tradition. However, no such solar rites are on record.
The Koryo-sa treatise contains a special section (in chap. 47) devoted
to solar eclipse observations. Virtually identical records to these are to be
found in the Basic Annals at the beginning of the same history (chaps.
1-46). In all, five solar eclipses (occurring in the years AD 1245, 1321,
1361, 1366 and 1390) were said to be complete. Each is briefly announced
in both the Astrological Treatise and the Annals in the same form: namely
that on a certain date there was a total obscuration of the Sun. There
is nothing to suggest observation outside the capital. In each instance,
totality is simply identified by the same term chi as in Chinese history (but
read as hi); no further description - such as time of occurrence - is given.
The many references to other solar eclipses in the Koryo-sa are devoid of
any allusion to magnitude or other details
Since the pattern of the five Korean accounts of totality is very repetitive,
a single example will suffice. The following observation was recorded on
a date corresponding to AD 1245 Jul 25:

King Kojong, 32nd year, 7th month, day kuei-szu [30], the first day of
the month. The Sun was eclipsed and it was total.
[Koryo-sa, chap. 47.]
As was customary from earliest times, the date follows the Chinese
style, except that the name of the Korean king is used. (NB I have given
the Chinese name of the cyclical day, rather than its Korean equivalent.)
It was the practice each year for the Chinese emperor to present copies
of the new calendar to the Korean envoy for use in his homeland. At
this particular eclipse, the place of observation was presumably the island
refuge of Kangwha. Although the phase was described as total, the above
eclipse was in fact only annular. Here we have another instance of the
use of chi to denote annularity.
Of the five eclipses said to be total in the Koryo-sa, only that of AD
1361 May 5 was also observed to be large in China. As noted above
(section 8.7), this was described as virtually total just before sunset in
Sung-chiang, with stars shining on account of the darkness. If this eclipse
was indeed total in Songdo, the magnitude at Sung-chiang (nearly 1000
km to the south-west) would have been relatively small - less than about
0.8 - and maximum phase would occur more than two hours before
sunset. Neither of these deductions are in accord with the circumstances
so vividly described at Sung-chiang. Hence considerable doubt must be
cast on the brief Korean allegation of totality. The reverse situation - an
eclipse on the verge of totality at Sung-chiang and only partial at Songdo
- is illustrated in figure 8.9.
264 8 East Asian observations of large solar eclipses

45° N 1 1

(—^ ^1 K
-> MSongdo «/

35° N -

^Sung-chiang

25° N -
Sunset position
-
j
15° N
90° E 105° E 120° E 135° E 150°E

Fig. 8.9 Map showing the computed track of totality for the eclipse of AD 1361
May 5 in relation to Sung-chiang and Songdo (based on AT = 0 sec).

On the remaining three occasions when the Sun was said to be com-
pletely obscured (AD 1321 Jun 26, 1366 Aug 7 and 1390 Oct 9) the belt
of totality was extremely narrow - between about 10 and 20 km wide.
The probability of these three marginal eclipses all being central at the
capital of Songdo is very small. Further, it seems highly unlikely that
chance observations of totality from random sites in the provinces would
reach the capital and eventually be included in the official history. Hence,
in view of the lack of reference to any phase but totality in the Koryo-sa,
it seems plausible that at least some of the five records may merely repre-
sent abbreviated descriptions of eclipses which were originally described
as almost complete. As in the case of the less detailed Chinese accounts
discussed above, each of these observations will be included in category
'C\ Summary information is given in table 8.7.
The Koryo dynasty endured until AD 1392 when King Kongyangwang
was deposed and General Yi Sung-gae established a new dynasty known as
Yi. Soon afterwards, the capital was transferred from Songdo to Hanyang
(modern Seoul, lat. = 37.55 deg, long. = —126.97 deg), where it has since
remained. The Yi dynasty itself lasted until the Japanese annexation in
1910.
There is no official Yi history similar to the Koryo-sa. The most
direct source of astronomical records, including solar eclipses, for the
earlier part of this period is the Choson Wangjo Sillok (Veritable Records
 8.9 Korea 265

Table 8.7 AT limits deduced from Korean observations of solar eclipses

described as total (all ' C rated).

AT Range (sec)
Julian date Type Mag. Record Capital LL UL
1245 Jul 25 A 0.95 total Kangwha 20 1200
1321 Jun 26 T 1.002 total Songdo 110 150
1361 May 05 T 1.05 total Songdo 7000 8790
1366 Aug 07 T 1.004 total Songdo -910 -890

1390 Oct 09 T 1.005 total Songdo 740 830

1397 May 27 T 1.06 total Hanyang 280 1620
1460 Jul 18 T 1.06 total Hanyang -3070 -1750

of the Choson Dynasty), a detailed Korean chronicle resembling the

Ming-shih-lu and extending down to the mid-nineteenth century. An
even more extensive chronicle known as the Sungjongwon Ilgi (Diaries
of the Court Secretariat) formerly covered the dynasty but owing to a
serious of unfortunate events (notably the Japanese invasions of 1592
and 1596) very little of this work now survives from before AD 1625.
Park (1979) gives an interesting comparison between the detail in sev-
enteenth century records of the same celestial phenomena as found in
the Sungjongwon Ilgi and Choson Wangjo Sillok; it is evident that the
latter work omits many observations and condenses other reports. The
loss of so many earlier volumes of the Sungjongwon Ilgi is much to be
regretted.
In the Choson Wangjo Sillok, references to large partial obscurations
of the Sun are lacking, while only two total eclipses are reported prior
to AD 1600 - in the years AD 1397 and 1460. Both of these accounts
are very brief, once again noting no more than the occurrence of totality
using the term ki (i.e chi). Each observation will therefore be assigned
a ' C grading, and details summarised along with the Koryo material in
table 8.7.
Since the paths of the umbral shadow in AD 1321, 1366 and 1390
were so narrow, only a very restricted range of AT can satisfy the record
of totality in each case. Most of the results in table 8.7 are mutually
discordant. In particular, the observation in AD 1361 indicates a very
large positive set of values in which AT should lie whereas in 1460 the
corresponding range is markedly negative. In summary, the contribution
of the Korean records of total solar eclipses to the study of Earth's past
rotation is virtually negligible.
266 8 East Asian observations of large solar eclipses

8.10 Japan
Japanese observations of solar eclipses are also numerous, but unlike
those from China and Korea they are scattered - along with references to
other celestial phenomena - in a large number of diverse writings. These
works include privately compiled histories, diaries and temple records
as well as official histories such as the Dainihonshi (History of Great
Japan). Fortunately, early in the present century Kanda (1934) made
an exhaustive collection of astronomical observations from this varied
material; this extends down to AD 1600. Kanda's extensive publication
contains a major section devoted to solar eclipses. I have used this as my
principal source of Japanese data (see also Kanda, 1935).
Japanese history emerges from legend around AD 600. The ancient
chronicle known as the Nihongi (Annals of Japan, also known as the
Nihon Shoki), which covers the period from earliest times to AD 693,
contains several records of celestial phenomena, including a few solar
obscurations from the seventh century. Among these, the eclipse of AD
628 Apr 10 is reported as total (using the character chi), but further details
are absent. At this early period, the precise place of observation cannot be
established; it was the custom to move the imperial residence to another
site on the death of each ruler. However, it is known that the various
capitals were all situated in the Yamato Plain, east of present-day Osaka
(approximate lat. = 34.6 deg, long. = —135.8 deg).
By AD 710, the first fixed capital was built - at Nara (lat. = 34.68
deg, long. = —135.82 deg). This was modelled on the T'ang metropolis of
Ch'ang-an, but its active life was short. In AD 784 the imperial residence
was transferred to Heian-kyo (lat. = 35.03 deg, long. = —135.75 deg),
later renamed Kyoto (the 'Capital'). Here it remained for more than a
thousand years - until 1868, when it was replaced by Edo (Tokyo). In
1192, a Shogunate or military government was founded at Kamakura
(lat. = 35.22 deg, long. = —139.55 deg). This was to last for 140 years,
although Kyoto appears to have remained the more important cultural
centre throughout this period. Edo (lat. = 35.67 deg, long. = —139.75
deg) did not attain prominence until AD 1590.
As was the practice in contemporary China during the T'ang dynasty,
for several centuries after eclipse observations began in Japan, offices were
closed on the day of a solar eclipse. The Gehpei seiseiki (which narrates
the rise and fall of the Minamoto and Taira clans) relates that on a day
corresponding to 1183 Nov 17 the Minamoto army fled, frightened by a
solar eclipse (Nakayama, 1969, p. 51). This eclipse was only annular, 95
per cent of the solar diameter being covered in the central zone.
The following solar eclipses are reported as either total or very large in
Japanese history before AD 1600: 628, 975, 1069, 1366, 1413, 1415 and
 8.10 Japan 267

1460. Of these, both the eclipses of 1366 and 1460 were also said to be
total in Korea (see section 8.9 above). Some of the Japanese descriptions
are fairly detailed, and frequently several independent reports of the same
event are available. The brief report in AD 628 most likely stems from the
Yamato region while the various accounts of totality in AD 975 almost
certainly originate from Heian-kyo, the capital of the time. However,
because of the varied nature of the sources on the remaining dates there
is considerable doubt regarding the place(s) of observation. I shall discuss
the observations made in AD 628 and 975 in detail, but shall only abstract
from the later texts.

(1) AD 628 Apr 10 (total mag. = 1.05): Yamato Plain [class C]

36th year of Empress Suiko, spring, 2nd month, 27th day. The Empress
took to her sick bed. 3rd month, 2nd day. There was a total eclipse of the
Sun. 6th day. The Empress' illness became very grave and (death) was
unmistakably near ... 7th day. The Empress died at the age of seventy-five.
[Trans. Aston, 1972, p. 155.]
The recorded date proves to be in accurate accord with that of a
computed solar eclipse. Although the obscuration of the Sun was said
to be total, further description is lacking, so that I have included the
observation in the lowest category. It is noteworthy that the event does
not seem to be linked in any way with the Empress' death by the chronicler
(unlike several similar occurrences in China - see above).

RESULTS
For totality in the Yamato Plain, 2270 < AT < 2960 sec.
(NB computed local time of maximum = 9.6 h; solar altitude = 48
deg.)

(2) AD 975 Aug 9 (total, mag. = 1.06): Heian-kyo [class A]

Kanda (1934) uncovered as many as 14 separate descriptions of this event.
Three of these, which give detailed descriptions of totality, are translated
below. In each case the date is correctly given as the 3rd year of the Ten-en
reign period, 7th month, day hsin-wei [8], the first day of the month. Like
the Koreans, the Japanese adopted the Chinese calendar from an early
period in their history.
(i) The Sun was eclipsed.... Some people say that it was entirely total.
During the hours mao and ch 'en (some time between 5 and 9 h) it
was all gone. It was the colour of ink and without light. All the
birdsflewabout in confusion and the various stars were all visible.
There was a general amnesty (on account of the eclipse).
[Nihon Kiryaku]
268 8 East Asian observations of large solar eclipses

(ii) At the hour ch'en (7-9 h), the Sun was eclipsed; it was completely
total. All under heaven became entirely dark and the stars were all
visible.
[Fuso Ryakki]

(iii) The Sun was eclipsed; it was all gone. It was like ink and without
light. The stars were all visible (or: stars were visible in the daytime
- text corrupt).
[Hyaku Rensho]
NB I have given the cyclical day and hour names in Chinese rather
than their Japanese equivalents.
Brief comments on the above historical sources are as follows. The
Nihon Kiryaku is a privately compiled history written about AD 1028.
Of unknown authorship, it is believed to be historically important and
trustworthy from about AD 887, and to be based on contemporary diaries
of court officials at Heian-kyo. The Fuso Ryakki is a history of Buddhist
affairs written by the monk Koen of Heizan (the mountains near Heian-
kyo) in AD 1094. Finally, the Hyaku Rensho is a history of the imperial
court from AD 960 to 1269. It was written at the end of this period.

RESULTS
For totality at Heian-kyo, 1230 < AT < 4480 sec.
This range of AT is very wide; the broad zone of totality ran nearly
parallel to the equator in the longitudes of Japan.
(NB computed local time of maximum = 7.9 h; solar altitude = 31
deg.)

Darkness like that of night is briefly mentioned at the marginal (annular-

total) eclipses of AD 1069 Jul 21 and 1366 Aug 7, but in neither case
is the degree of obscuration of the Sun recorded. Both central zones
would be extremely narrow. Darkness is also described at the eclipse of
AD 1413 Feb 1, which was said to be total. There is a single report
of totality for the eclipse of 1415 Jun 7, but no further details are
given and - as in 1413 - the place of observation cannot be established
from the text. The situation on Jul 18 in AD 1460 is more interesting.
On this occasion, as many as ten reports note either the occurrence of
totality and/or darkness. In the sources compiled by Kanda, individual
descriptions include remarks such as: 'It was like a moonlight night
and the stars appeared'; Tt was like a cloudless night for about half an
hour'; There was a total eclipse and the stars were all visible'. It is
unfortunate that the place where any of these records originated cannot
be established. For example, there is nothing to relate them to the capital
of Kyoto.
 8.11 Conclusion 269

Table 8.8 AT limits derived from 'A' and 'B' rated East
Asian records of solar eclipses.
AT Range (sec)
c
Year* Country** Phase UI/
-708 China central 20230 21170
-187 China partial 14140 13 830
-180 China central 11800 12720
-79 China partial 8 520 8 200
-27 China partial 9 530 8 090

-1 China partial 2150 2050

+120 China partial 8 970 8150
+360 China partial 9400 7120
+454 China central 6130 7900
+494 China partial 6600 5 980

+702 China partial 2 760 1460

+729 China partial 1190 420
+761 China central 1720 3 290
+822 China partial — 4020
+975 Japan central 1230 4480

1135 China partial 3 810 1840

1221 Mongolia central — 980
1275 China central -700 1300
1292 China central -50 1830
1361 China partial 1760 500

1575 China central -2150 1390

a
Year.
b
Country where the observation was made.
c
Whether the eclipse was said to be central or partial.
d
Lower limit (LL) to the value of AT deduced from the observation.
e
Corresponding upper limit (UL) to AT.

8.11 Conclusion

The results of the investigation of 'A' and 'B' rated solar eclipse obser-
vations from China and Japan (there are no reports from Korea in this
category) are listed in table 8.8. In table 8.9 are assembled the correspond-
ing data obtained from the analysis of observations of lower reliability
(class 'C'). Both of these tables use negative and positive integers for years
270 8 East Asian observations of large solar eclipses

Table 8.9 AT limits derived from ' C rated East Asian records
of solar eclipses.

AT Range (sec)
Year Country Phase LL UL
-600 China central 21140 21960
-548 China central 16150 21710
-197 China central 6 350 13110
+2 China central 8 570 8 590
+65 China central 8 450 8 900

+306 China central 6 550 7 890

+616 China central 2 300 3 030
+628 Japan central 2270 2960
1245 Korea central 20 1200
1321 Korea central 110 150

1361 Korea central 7000 8 790

1366 Korea central -910 -890
1390 Korea central 740 830
1397 Korea central 280 1620
1460 Korea central - 3 070 -1750

(rather than BC and AD) since their main purpose is to summarise the
results obtained in earlier sections.
Following the adopted convention for a partial eclipse (see chapter 3),
the lower AT limit in column 4 of table 8.8 is actually greater than the
upper limit in column 5. This is because a range of AT values is excluded
by the observation.
Most of the AT ranges in table 8.8 are plotted in figure 8.10. In this
diagram, full vertical lines (usually short) represent the range of AT values
derived from central solar eclipses. Broken vertical lines denote the range
of AT values obtained from partial solar eclipses. NB in both cases, blank
regions of solution space are void. For the partial eclipses in the years
—187, —79 and —1, the excluded zones are so narrow that they cannot
be satisfactorily displayed on the scale of the diagram. For all practical
purposes, each observation is effectively redundant; virtually any value of
AT would render each eclipse partial at the Chinese capital (Ch'ang-an)
in accord with the record. It is evident that all of the data displayed in
figure 8.10 form a fairly self-consistent set.
All of the AT limits in table 8.9 (category 'C) are depicted in figure
8.11. It is clear from this diagram that the Korean data (between the
 8.11 Conclusion 111

+20000

+16000

+12000

< +8000

+4000

-800 -600 -400 -200 0 +200 +400 +600+800+1000+1200+1400+1600

Year
Fig. 8.10 AT ranges obtained from East Asian observations of central and large
partial solar eclipses (categories 'A' and 'B').

+24000

+20000

+16000 -

+12000 -

< +8000

-600 -400 -200 0 +200 +400 +600 +800 +1000 +1200 +1400
Year
Fig. 8.11 AT ranges obtained from East Asian observations of central and large
partial solar eclipses (category ' C only).

years +1245 and +1460) lack mutual consistency. The results from the
Chinese observations (between the years —600 and +607) and the solitary
Japanese observation (+628) seem fairly self-consistent. However, it should
be emphasised (as discussed in sections 8.5 to 8.7), that further Chinese
records in category ' C (in the years +562, +879, +977 and +1022) have
been rejected because the eclipses proved to be invisible in China. In
addition, several of the AT ranges shown in figure 8.11 (especially for the
years +2 and +65) are extremely narrow and it would seem unrealistic to
272 8 East Asian observations of large solar eclipses

insist that the AT curve must pass between the limits indicated. Clearly
the use of laconic records which assert no more than that a certain eclipse
was complete is hazardous. The whole set of data in this category will
thus be rejected.
In subsequent analysis, only the results obtained from material rated
'A' and 'B' will be considered further. In retrospect, it would appear
that there is little to choose between the quality of the observations in
these two classes and hence a distinction will no longer be made. Both
sets of records will be ranked equally with descriptions of central and
near-central eclipses from other parts of the world.
 Other East Asian observations
of solar and lunar eclipses

9.1 Introduction
In the previous chapter, East Asian records of total and near-total solar
eclipses were analysed. Emphasis was placed on observations which either
affirmed or specifically denied that the Sun was completely obscured.
Other reported details - such as measurements of the times of the various
contacts or rising or setting of the luminary whilst eclipsed - were not
considered. The theme of the present chapter is the investigation of
further observations of both solar and lunar eclipses from the same part
of the world which in principle are of value in the determination of
AT. These may be grouped in three main categories: (i) timed contacts;
(ii) estimates of the proportion of the Sun covered at maximum phase;
and (iii) instances where the Sun or Moon was said to rise or set whilst
visibly obscured.
In the pre-telescopic period, there is a significant number of Chinese
observations in each of the above categories. However, accounts of
eclipses from Korea tend to be extremely brief and no useful records
are preserved in the history of this country - whether in the Koryo-sa
('History of Koryo') or the Choson Wangjo Sillok ('Veritable Records of
the Choson Dynasty'). Although many early Japanese observations of
both solar and lunar eclipses give estimates of magnitude or local time,
these are contained in a miscellany of sources. Not only is the place
of origin often obscure (see also chapter 8), it is frequently difficult to
distinguish observation from prediction. Despite the extensive compila-
tion of Japanese eclipse data by Kanda (1934), a thorough investigation
will be needed before useful AT results can be obtained from this ma-
terial. The author eventually hopes to undertake such a project, but at
present only the Chinese observations seem worthy of attention. The
remainder of this chapter will thus be devoted exclusively to Chinese
data.

273
274 9 Other East Asian observations of solar and lunar eclipses

9.2 General remarks on recorded eclipse timings

More than one hundred separate timings of eclipse contacts (both for
the Sun and Moon) are recorded in Chinese history - the earliest dating
from 134 BC. Yet compared with the total number of eclipses noted in
the same sources, such measurements are still relatively rare. Sufficient
measurements are preserved from certain dynasties (especially the Sung)
to suggest that the timing of eclipses was practised fairly frequently.
However, it would appear that most of the original observations have not
survived. The majority of extant timings recorded in the official histories
are not included in the astrological sections (in contrast to more general
eclipse records) but tend to be found in calendar treatises (li-shih) instead.
In these works, observations by the court astronomers were used to check
the reliability of existing almanacs. Many further measurements of eclipse
times (especially for the Moon) are noted in the Wen-hsien T'ung-k'ao
('Comprehensive History of Civilisation'). This work, compiled by Ma
Tuan-lin (AD 1254-1323), was completed in AD 1307. Ma presumably
obtained many of his detailed astronomical records from the now lost
Sung Shih-lu (Annals of the Sung Dynasty).
Before considering individual measurements, some discussion of Chinese
units of time seems necessary. Solar eclipse times recorded in Chinese
history are almost invariably expressed in terms of the twelve shih or
'double hours'. Although lunar eclipse times are often cited in this same
form, they are also frequently quoted relative to the five keng or night
watches instead - especially before AD 600.

9.2.1 Double hours and their subdivisions

During the Former Han dynasty (c. 100 BC), the interval from one
midnight to the next was divided into twelve equal parts known as shih.
There is no evidence that the origin of the twelve shih is related to
the occidental system of 24 hours in a day. Individual double hours
were named according to the twelve ti-chih (terrestrial branches) of the
sexagenary cycle - see chapter 8. The first double hour (tzu) was centred
on midnight and thus began at 23 h local time. A full list of the shih
and their equivalent local times is given in table 9.1. This table also
includes the twelve symbolic animals with which the double hours were
popularly associated; of course, these names were not used in astronomical
practice.
From at least AD 1000, each double hour was bisected into an 'initial'
and 'central' half. These intervals of time were named for the 'beginnings'
(ch'u) and 'mid-points' (cheng) of the shih. The initial half of the first
double hour (tzu-ch'u) corresponded to between 23 and 24 h, while the
 9.2 General remarks on recorded eclipse timings 275

Table 9.1 The twelve double hours

Name Animal Time (h) Name Animal Time (h)
tzu rat 23-01 wu horse 11-13
ch'ou ox 01-03 wei sheep 13-15
yin tiger 03-05 shen monkey 15-17
mao hare 05-07 yu cock 17-19
ch'en dragon 07-09 hsu dog 19-21
szu snake 09-11 hai Pig 21-23

central half (tzu-cheng) lasted from 24 h to 1 h - and so on throughout

the other eleven shih.
An independent division of the day, known as a k'o, was also in
regular usage. The literal meaning of this term is a 'notch' or 'mark'
on a time-keeping scale (Needham, Lu, Combridge and Major, 1986, p.
9). Although k'o will be translated as 'mark' throughout this chapter, it
should be emphasised that it relates specifically to an interval of time.
During the Former Han, the natural day (midnight to midnight) was first
divided into 100 k'o. Each of these units was thus equivalent to a 0.24 h so
that 1 shih or double hour contained 8 | marks. Despite the inconvenience
of not having an integral number of marks to the double hour, in all but
a few brief periods down to the late Ming dynasty there remained exactly
100 k 'o to a 24-hour day.
In 6 BC Emperor Ai decreed that there should be 120 marks in a
day, one double hour thus containing exactly ten k'o. However, this
revision proved unpopular and only two months later the k'o was restored
to its original length. Wang Mang, who usurped the Han throne in
AD 9, repeated this experiment, but his revision was abolished on his
assassination in AD 23. No further changes were made for nearly five
centuries. In AD 507, the number of marks in a day was first reduced to
96, but afterwards (AD 544) it was increased to 108. When the traditional
convention was reverted to by popular demand in AD 563, it was retained
without any alteration until as late as AD 1628. In the latter year, Emperor
T'ai-tsung ordained that there should be 96 marks in a day - each exactly
one-quarter of an hour in length - and this definition continued to be
adopted until the present century.
No accurate eclipse timings are reported in the early periods during
which anomalous definitions of the k'o were in vogue, while the most
recent period is after the introduction of the telescope and thus outside
the scope of this book. Hence all subsequent discussion will be based on
the basic figure of 100 of these units to a day. Measurements of eclipse
276 9 Other East Asian observations of solar and lunar eclipses

(a)

ch'u (initial) 1st 2nd 3rd 4th 5th 6th 7th hsiao (small)
k-oV k'o k'o k'o k'o k'o k'o k'o y k'o
I
1 shih (double hour)

(b)

ch'u (initial) hsiao (small)

k'o k'o
1st 2nd 3rd /
k'o k'o k'o /
i II ill
1 single hour
Fig. 9.1 Chinese double (a) and single (b) hours and their divisions.

times were made with the aid of water clocks - such devices were standard
equipment at the imperial observatory from the Han dynasty onwards.
Early descriptions of water clocks often mention graduations in both shih
and k'o (Needham, Wang and Price, 1986, chaps. 4 and 6).
In most instances, k'o were counted from the beginning of each double
hour. However, alternatively k'o could be measured relative to the start
of each half-s/n/z. Which of these systems was used in timing a particular
eclipse is normally clear from the context. Individual 'hours' (whether
double or single) commenced with an 'initial' (ch'u) mark, covering the
first 0.24 h. This was followed by the first mark, which embraced the next
0.24 h, then the second mark and so on up to the eighth mark (for a
double hour) or the fourth mark (for a single hour).
Since 1 shih was equal to 8^ k'o, the eighth and last mark in a double
hour was only one-third of the normal length, covering the final 0.08 h
before the start of the next shih. This was often known as the hsiao-k'o
('small mark') - see figure 9.1a. When single hours were used, the fourth
and last mark was extremely short, nominally equal to only 0.04 h; this
was again known as the hsiao-k'o ('small mark') - see figure 9.1b. It is
doubtful whether early water clocks could define either of the hsiao-k'o
with any precision.
In order to illustrate the division of double hours into k'o, I have
selected an example from chapter 53 of the Yuan-shih. Here it is stated
that the lunar eclipse of AD 1106 Jan 21/22 reached maximum phase 'at 3
 9.2 General remarks on recorded eclipse timings 277

Table 92a Division of double hours and their associated mean local times.
Shih Ok'o 1 k'o 2 k'o 3 k'o 4 k'o 5 k'o 6 k'o 7 k'o 8 k'o
h h h h h h h h h
tzu 23.12 23.36 23.60 23.84 0.08 0.32 0.56 0.80 0.96
ch'ou 1.12 1.36 1.60 1.84 2.08 2.32 2.56 2.80 2.96
yin 3.12 3.36 3.60 3.84 4.08 4.32 4.56 4.80 4.96
mao 5.12 5.36 5.60 5.84 6.08 6.32 6.56 6.80 6.96

ch'en 7.12 7.36 7.60 7.84 8.08 8.32 8.56 8.80 8.96
szu 9.12 9.36 9.60 9.84 10.08 10.32 10.56 10.80 10.96
wu 11.12 11.36 11.60 11.84 12.08 12.32 12.56 12.80 12.96
wei 13.12 13.36 13.60 13.84 14.08 14.32 14.56 14.80 14.96

shen 15.12 15.36 15.60 15.84 16.08 16.32 16.56 16.80 16.96
yu 17.12 17.36 17.60 17.84 18.08 18.32 18.56 18.80 18.96
hsu 19.12 19.36 19.60 19.84 20.08 20.32 20.56 20.80 20.96
hai 21.12 21.36 21.60 21.84 22.08 22.32 22.56 22.80 22.96

marks in the hour of j/u' and ended 'in the initial mark of the hour of hsu\
The first of these intervals corresponds to between 17.72 and 17.96 h local
time (LT), while the second is from 19.00 h to 19.24 h. The alternative
system, using single hours, is exemplified by an entry in chapter 283 of the
Wen-hsien T'ung-k'ao. In this record, it is reported that the solar eclipse
of AD 1052 Nov 24 ended 'at 1 mark in the central half of the hour of
wei\ The equivalent LT is between 14.24 and 14.48 h.
Since the unit known as a k'o represented a time-interval, when a
measurement was expressed to the nearest k'o I have assumed that the
observation was made at any time during that interval. I have thus taken
the average moment as the mid-point of the designated k'o (see Chen,
1983a, 1983b for a detailed discussion). Hence in the above examples,
I shall make the approximate assumptions that the lunar eclipse of AD
1106 reached its maximum at a LT of 17.84 h and ended at 19.12 h while
the solar eclipse of AD 1052 ended at 14.36 h.
On this basis, the equivalent LTs for each double hour and mark are
summarised in table 9.2a, and for each single hour and mark in table 9.2b.
Very occasionally a measurement is quoted to the nearest half-fc'o: for
example '5^ marks in the hour of wu' (AD 1173). In such cases I shall still
adopt the mid-point of the appropriate mark. Measurements expressed in
half-marks are so few in number that it scarcely seems worthwhile assign-
ing them a high weight. No surviving Chinese eclipse records from before
the seventeenth century are expressed to better than the nearest half-k'o.
278 9 Other East Asian observations of solar and lunar eclipses

Table 9.2b Division of single hours and their associated mean local times.

ch'u (initial) half cheng (central) half

Shih Ok'o 1 k'o 2 k'o 3 k'o 4 k'o Ok'o 1 k'o 2 k'o 3 k'o 4 k'o
h h h h h h h h h h
tzu 23.12 23.36 23.60 23.84 23.98 0.12 0.36 0.60 0.84 0.98
ch'ou 1.12 1.36 1.60 1.84 1.98 2.12 2.36 2.60 2.84 2.98
yin 3.12 3.36 3.60 3.84 3.98 4.12 4.36 4.60 4.84 4.98
mao 5.12 5.36 5.60 5.84 5.98 6.12 6.36 6.60 6.84 6.98

ch'en 7.12 7.36 7.60 7.84 7.98 8.12 8.36 8.60 8.84 8.98
szu 9.12 9.36 9.60 9.84 9.98 10.12 10.36 10.60 10.84 10.98
wu 11.12 11.36 11.60 11.84 11.98 12.12 12.36 12.60 12.84 12.98
wei 13.12 13.36 13.60 13.84 13.98 14.12 14.36 14.60 14.84 14.98

shen 15.12 15.36 15.60 15.84 15.98 16.12 16.36 16.60 16.84 16.98
yu 17.12 17.36 17.60 17.84 17.98 18.12 18.36 18.60 18.84 18.98
hsu 19.12 19.36 19.60 19.84 19.98 20.12 20.36 20.60 20.84 20.98
hai 21.12 21.36 21.60 21.84 21.98 22.12 22.36 22.60 22.84 22.98

It is of interest to compare a series of measured local times of sunrise

and sunset expressed in these same units with their computed equivalents.
Since an accurate knowledge of AT is not required, a direct comparison
between observation and computation can be readily made. Maspero
(1939) investigated a series of sunrise and sunset times measured around
AD 600. These are recorded in chapter 17 of the Sui-shu and were for either
Ch'ang-an or Lo-yang (both close to lat +34.5 deg N). They are given at
15-day intervals between the winter solstice and summer solstice and are
expressed in double hours, marks and fen, the latter unit corresponding to
^ of a k'o. All times are quoted to the nearest 5 fen. To give an example,
the time of sunset at the winter solstice was said to be 7 k'o, 50 fen in
the hour of shen, corresponding to 16.88 h; for comparison, the computed
LT is 16.91 h. I have verified the accuracy of Maspero's reductions of
the measurements to local time and his computations. His results reveal
maximum observational errors of 0.09 h and standard errors of 0.06 h,
the latter equivalent to about 15 fen or 1/4 k'o. It is apparent that the
accuracy of the times of sunrise/sunset in the Sui-shu fall considerably
below the expectations of the Sui astronomers. They possibly represent
the best that could be achieved using the water clocks of the period.
Between AD 585 and 768, several solar eclipse timings are expressed
relative to double hours using an unusual terminology. Most of these
 9.2 General remarks on recorded eclipse timings 279

Table 9.3 Anomalous solar eclipse records of uncertain interpretation.

Date AD Ct Measurement Source
585 Jul 31 1 wu-hou 6 k'o Sui-shu, 17
M wei-hou 1 k'o
4 wei-hou 5 k'o

586 Dec 16 M ch'en-hou 2 k'o Sui-shu, 17

4 ju-szu-shih 3 k'o

594 Jul 23 1 wei-hou 3 k'o Sui-shu, 17

721 Sep 26 M wu-cheng-hou 3 k'o Yuan-shih, 53

761 Aug 5 1 ch'en-cheng-hou 6 k'o Chiu-t'ang-shu, 36

T szu-cheng-hou 1 k'o
4 wu-ch'ien 1 k'o

768 Mar 23 M wu-hou 1 k'o Chiu-t'ang-shu, 36

observations incorporate the term hou (after, latter). Examples include

wu-hou 3 k'o (AD 721) and ch'en-cheng-hou 6 k'o (AD 761). On two
occasions, either ju (to enter) or ch'ien (former, before) replaces hou - for
example szu-ju 3 k'o in AD 586. The meaning of all of these various
compounds is somewhat ambiguous and different authorities disagree on
their interpretation. Hence I shall not reduce any measurements employing
these terms. I am grateful to Professors Chen Meidong and Chen Jiujin of
the Institute for the History of Natural Sciences, Academia Sinica, Beijing
for helpful discussion on this issue.
The dates and other details for the anomalous observations between AD
585 and 768 are summarised in table 9.3. This table lists the appropriate
contact, etc. (1 = first contact, M = maximal phase, T = total, 4 = last
contact), the measurement of time and the historical source. It should be
noted that the time of first contact in AD 586 is given simply as ch'en 2
k'o, i.e. 2 marks in the hour of ch'en. This is the sole measurement in the
whole set which uses a standard terminology and therefore the only one
to be retained in subsequent investigation (see section 9.3.1).

9.2.2 Night-watches and their divisions

From the Sung dynasty onwards, lunar eclipses were usually timed in rela-
tion to the twelve double hours, in common with their solar counterparts.
280 9 Other East Asian observations of solar and lunar eclipses

However, in earlier periods (going as far back as the Han dynasty) it was
customary to adopt the night-watches (keng) instead. Day was regarded as
beginning at sunrise (jih-ch'u) and ending at sunset (jih-ju). However, night
was defined as extending only from dusk (hun) to dawn (tan or ming). The
interval from the end of evening twilight to the start of morning twilight
(not sunset to sunrise) was divided into five equal watches. Unlike the
double hours, the watches varied in length with the seasons.
Evening twilight was originally regarded as ending 3 marks (0.72 h) after
sunset, while dawn began a similar time before sunrise. However, from
the Eastern Han (AD 25 onwards) this figure was reduced to l\ marks
(0.60 h). This revised convention remained unchanged for astronomical
purposes at all later periods, despite movement of the capital to different
latitudes (ranging from 30 to 40 deg N). At 0.6 h after sunset or before
sunrise, the Sun attains a zenith distance of approximately 97.5 deg in
central China (latitude 35 deg N). Hence the adopted convention agrees
quite well with the modern definition for the limit of civil twilight - i.e. a
solar zenith distance of 96 deg.
In a section of the calendar treatise of the Sui-shu (chap. 16) we find
the following elementary rules:
Double the number of marks reached at midnight (after twilight); this
gives the number of marks in the (dark) night. Subtract this from 100
marks; what is left is the number of marks in the daytime (including
twilight). Subtract 5 marks from the number of marks in the daytime or
add it to the number of marks in the night. This gives the time in which
the Sun is visible during the day or invisible during the night.

A useful series of explanatory diagrams is given in the Hsin-i-hsiang

Fa-yao ('New Design for an Astronomical Clock') a work composed by
Su Sung in the late eleventh century. These diagrams - one of which
is illustrated by Needham, Wang and Price (1986, p. 203) - show the
length of daylight, twilight, darkness and dawn in marks at the equinoxes,
summer solstice and the winter solstice for the capital city of Pien (Kaifeng,
latitude 34.8 deg N). In each case, the durations of evening and morning
twilight are taken as l\ k'o. At the equinoxes, the duration of daylight is
given as 50 marks and that of night 45 - a total of 100 marks including
dusk and dawn. At the summer solstice, the lengths of daylight and
night are 60 and 35 marks respectively while at the winter solstice the
corresponding figures are 40 and 55 k'o.
Reference to the calendar treatises of the various official histories shows
that the above figures were fairly standard at most periods. These values
lead to minimum and maximum lengths for the five night watches of 7 and
11 marks respectively. For central China (latitude 35 deg N), the extreme
lengths of the keng may be calculated as 6.9 marks (1.65 h) in summer
 9.2 General remarks on recorded eclipse timings 281

and 10.8 marks (2.60 h) in winter - both very close to the traditionally
adopted limits.
Each night watch was further subdivided into five equal graduations,
whose length ranged from about 20 to 30 min, depending on the season.
These divisions carried different names at different periods. Before the
Sui dynasty, they were known as ch'ang ('calls'), after the periodic calls
of the night watchmen. During the Sui, they were renamed ch'ou ('rods')
implying the position of the pointing rod on a water clock. Later, the
term tien ('points'), having a similar significance to rods, was used instead.
However, at all periods there were five equal divisions to a keng.
The subdivisions of a night watch were simply numbered first, second,
third, etc; there was no initial graduation, unlike in the case of marks
within an hour or double hour. By definition, the third division of
the third watch would thus be centred on midnight. Water clocks were
designed to measure time in terms of the watches and their subdivisions,
making allowance for the seasonal variation in the lengths of these units
(Needham, Wang and Price, 1986, chaps. 4 and 6).
Figure 9.2a shows the five night-watches and their divisions at the winter
solstice and figure 9.2b at the summer solstice for a latitude close to +35
deg.
Conversion of a measurement expressed in keng to LT requires a fair
degree of routine calculation since it involves determination of the mo-
ments of sunset and sunrise for the particular latitude. For example, on a
date equivalent to Dec 28/29 in AD 437, the Sung-shu (chap. 12) records
a lunar eclipse observed at Chien-k'ang (Nanjing) as beginning 'at the
fourth call of the second watch'. The equivalent LT may be derived as
follows. Firstly, the times of sunset and sunrise for this latitude (32.0
deg) are calculated as 17.03 h and 6.97 h respectively, taking into account
mean refraction and the solar semi-diameter. (NB horizon profile effects
are ignored - see section 9.7.) Allowing for the durations of both evening
and morning twilight as 0.60 h gives the length of a watch as 2.55 h and
a call as 0.51 h. The LT for the beginning of the second watch may thus
be determined as 20.18 h, while the fourth call would last from 21.71 to
22.22 h. Obviously these are idealised results; in practice, uncertainties
would be introduced by imperfections in timekeeping.
As in the case of k'o, if the time of a lunar eclipse is expressed to
the nearest fifth of a night-watch, I have assumed the mid-point of the
appropriate interval. Thus the beginning of the eclipse of AD 437 is
quoted as occurring at the 'fourth call of the second watch'. Since this
interval would extend from 21.71 to 22.22 h, I have estimated the time of
first contact as approximately 21.97 h.
If the LT of a solar or lunar eclipse is only expressed to the nearest
double or single hour or night watch, the result is probably too imprecise
282 9 Other East Asian observations of solar and lunar eclipses

Midnight

Midday
(a) Winter solstice

Midnight

Midday
(b) Summer solstice
Fig. 9.2 Night watches and their divisions: (a) winter solstice; (b) summer
solstice.
 9.2 General remarks on recorded eclipse timings 283

for a useful determination of AT to be made. Such crude observations

may in some cases only represent casual estimates and thus will not be
considered in this chapter. However, when a measurement is expressed
to either the nearest k'o or fifth of a keng, in principle a fairly accurate
value for the difference between TT and UT may be obtained. Although
one-fifth of a night-watch is typically almost twice as long as a mark, I
have considered it appropriate to assign equal weight to measurements
using either type of unit. Observations in both categories which are
preserved in the dynastic histories and other major sources are likely to
be almost entirely the work of the imperial astronomers at the capital of
the time. As noted in chapter 8, a regular watch for eclipses and other
celestial phenomena was maintained at the capital; an alternative site is
never mentioned in the records. Provincial observers would not normally
possess timing facilities of comparable precision and in any case it is
doubtful whether their measurements would be considered adequate for
inclusion in an official compilation.

9.2.3 Estimates of lunar azimuth

In addition to times, a number of estimates of the azimuth of the Moon
are preserved for eclipses which occurred during the Sui dynasty (late
sixth century AD). Thus at the eclipse of AD 590 Apr 25/26 we find the
following report:
... It was observed that when the Moon first rose from the south-east, it
was already more than half eclipsed. 3 divisions {fen) after it reached the
ch 'en direction, 2 fifteenths remained eclipsed. It gradually recovered. By
the end (wei) of the ch'en direction it was restored to fullness.
[Sui-shu, chap. 17.]
Chinese astronomers divided the horizon into 24 zones of azimuth, each
of width 15 deg. The first of these ranges was centred on due north and
thus extended from azimuth 352.5 to 7.5 deg. Working in a clockwise
direction, the odd-numbered zones - centred on 0, 30, 60 deg, etc. - were
named after the ti-chih or terrestrial branches tzu, ch'ou, yin and so on.
The twelve even-numbered directions - centred on 15, 45, 75 deg, etc. -
were made up of eight of the names of the t'ien-kan or celestial stems (less
wu and chi), together with four of the eight trigrams ch'ien, k'un, sun and
ken - as in the I-ching (Book of Changes). In the above example from AD
590, the ch'en direction covered a range of azimuth from 115 to 127.5 deg.
In most eclipse records of this kind, it is simply noted that the Moon
was 'above' (shang) a particular direction. Since the azimuth of the Moon
typically increases by about 15 deg in an hour, this represents a fairly
rough estimate. In other examples it is stated either that the Moon was at
284 9 Other East Asian observations of solar and lunar eclipses

the 'end' (mo) of a zone or that an observation was made several k 'o after
the Moon reached a certain range of azimuth. These various records will
be discussed in section 9.5.

9.2.4 Accuracy of early eclipse predictions

Early Chinese astronomers were well aware that their eclipse predictions
were crude. For instance, in a section of the calendrical treatise of the
Sui-shu (chap. 18) - compiled around AD 630 - it is remarked that:
Eclipses of the Sun or Moon may begin or end early or late; they can
deviate from normal in either direction. Therefore it is necessary to observe
12| marks (i.e. 3 hours) before and after the predicted time.
In the Sui-shu (chap. 17) and also in the somewhat earlier Sung-shu
(chaps. 12 and 13), the recorded times of a number of eclipses are
compared with their expected times. Discrepancies amounting to several
hours are frequently noted. For example, the report of the lunar eclipse
of AD 437 Jan 8/9 in chap. 12 of the Sung-shu runs as follows:
Yuan-chia reign period, 13th year, 12th month, 16th day, full Moon.
The Moon was eclipsed. The calculated time was the hour of yu (i.e.
17-19 h). The eclipse did not (actually) begin until the start of the hour of
hai (i.e. 21-22 h)...
The above account ends with the measured time of totality expressed
relative to the night watches.
Down the centuries, the quality of eclipse prediction slowly improved.
Thus in the calendar treatise of the Sung-shih (chap. 75) it is recorded that
the astronomer Chou Tsung (c. AD 1064) regarded eclipse predictions
which came within 2k'o (i.e. less than about half an hour) of observation
as ch'in ('close'), those between 2 and 4 k'o as chin ('near') and those
of 5 k'o or more as yuan ('far'). A similar system of assessment was
developed by the astronomers of the Yuan dynasty - see below. Since
Chinese attempts at predicting eclipses were largely restricted to the use
of numerical cycles, little further progress was achieved until the arrival of
the Jesuit astronomers in the seventeenth century. The Jesuits introduced
western methods of eclipse calculation which were superior to indigenous
techniques. Occasionally competitive eclipse prediction was arranged
between them and the native Chinese astronomers, the Jesuits coming off
best! (D'Elia, 1960, p. 41).

9.2.5 Sources of timed observations and terminology

Available sources of eclipse timings in the pre-Jesuit period are few in
number. In the official histories, only the Hou-han-shu Five-Element Trea-
 9.2 General remarks on recorded eclipse timings 285

tise (chap. 28), Sung-shu Calendar Treatise (chaps. 12 and 13), Sui-shu
Calendar Treatise (chap. 17), Chiu-t'ang-shu Astrological Treatise (chap.
36), Sung-shih Calendar Treatise (chaps. 81 and 82) and Yuan-shih Calen-
dar Treatise (chapter 53) contain accurate measurements. Many further
data are preserved in the Wen-hsien T'ung-k'ao (chaps. 283 and 285),
while a few others are contained in the Sung-hui-yao Chi-kao ('Collected
Manuscripts of the Essentials of Sung History'). Of these latter works, the
Wen-hsien T'ung-k'ao has already been mentioned above (see also chapter
8); the Sung-hui-yao Chi-kao was composed during the late Sung (c. AD
1250).
Only an isolated solar eclipse timing expressed to the nearest k'o is
reported in the Hou-han-shu. This dates from AD 193. No further mea-
surements are quoted so precisely until AD 585 (in the Sui-shu), although
a few lunar eclipse times expressed to the nearest fifth of a night-watch
are cited during this interval, beginning with AD 434 (in the Sung-shu).
The most prolific source of surviving measurements of eclipse times
from pre-Jesuit China is chapter 53 of the Yuan-shih. This treatise, entitled
Shou-shih-li I ('Discussion of the Season-Granting Calendar') was written
by Li Ch'ien (AD 1233 - 1311) but was based largely on the work of
the great Yuan astronomer Kuo Shou-ching (AD 1231-1316). It contains
the earliest extensive investigation of historical eclipses (both solar and
lunar) by Chinese astronomers, and includes a detailed discussion of the
observations reported in the Ch'un-ch'iu. In particular, measured timings
of solar eclipses from AD 221 to 1277 and of lunar obscurations from AD
434 to 1280 are compared with the results calculated from two separate
almanacs. These are the contemporary Shou-shih-li ('Season-Granting
Calendar') - devised by Kuo Shou-ching in AD 1281 - and its predecessor
from the Kin dynasty, the Ta-ming-li ('Great Enlightenment Calendar') -
produced by Chao Chi-wei in AD 1182. The relative success of the two
systems in representing observation is evaluated in the Yuan-shih on a
scale of five, depending on the magnitude of the discrepancy: mi ('very
close': within 1 mark); ch'in ('close': 1 mark); ch'i-ch'in ('fairly close': 2
marks); shu ('off': 3 marks) and shu yuan ('far off': more than 3 marks).
Kuo Shou-ching was able to show that in general the newer calendar
better represented observation. However, it is not clear whether in making
these various calculations adequate allowance was made for the different
locations (especially longitudes) of the various observing stations.
The Shou-shih-li I first discusses solar eclipse observations and then
continues with an investigation of reports of lunar obscurations. Although
the original source of data is never specified, the text is careful to state the
dynasty within which the observation was made. Some of the older timings
are only expressed to the nearest shih or half-shih. The earliest solar eclipse
mentioned in this treatise for which a measurement is expressed to the
286 9 Other East Asian observations of solar and lunar eclipses

nearest kyo dates from AD 680. However, lunar eclipse observations which
are timed to the nearest fifth of a keng are cited from AD 434 onwards.
Many of the records are not preserved in other extant works; evidently
Kuo Shou-ching and his colleagues had access to much material which
did not find its way into the official histories of former dynasties.
Figure 9.3 shows a page from a nineteenth century printing of the
Shou-shih-H I. This contains records of solar eclipses from AD 1053 and
1054 together with the results of calculation for each eclipse based on the
Ta-ming-li and Shou-shih-H.
The Wen-hsien T'ung-k'ao also records many solar and lunar eclipses
which are not found in other sources - especially in the period from about
AD 1050 to 1100. This work contains an extensive astronomical treatise,
separate sections of which are devoted to solar eclipses (chap. 283) and
lunar obscurations (chap. 285).
Of the individual eclipse phases noted in the dynastic histories and other
compilations mentioned above, the most commonly reported is the time
of maximum - usually identified by the term shen ('extreme'). In only one
case (AD 761) is the onset of totality (chi) reported for a solar obscuration
but several lunar eclipse accounts note this phase. The start of an eclipse
is often indicated by k'uei-ch'u ('beginning of loss') and the end by fu-man
('restoration of fullness'). Other expressions include shih-sheng ('beginning
of emergence' - after maximum has passed), sheng-kuang ('emergence of
light' - at the end of totality for a lunar eclipse), and fu-yuan ('restoration
of roundness').

9.3 Careful solar eclipse timings recorded in Chinese history

In all, the LTs of about 60 solar eclipse contacts are quoted to the nearest
kyo in the history of China prior to the telescopic era. Because of its very
early date (AD 193 Feb 19), the oldest observation of this kind may well
be of questionable reliability. In particular, it is not clear from the text
whether first contact or maximum phase is referred to. The remaining
observations fall into three distinct chronological groups: between AD
585 and 768; from 1052 to 1277; and after 1572. Throughout this
section, emphasis will be placed on the second group of data. As noted
in section 9.2.1, many of the measurements made in the earliest of the
three periods are difficult to interpret and will not be considered here.
The few surviving measurements which are preserved between 1572 and
1644 - when Jesuit astronomers commenced regular eclipse observations
in China - are still only expressed to the nearest mark. By this period,
AT was so small (of the order of 100 sec) that such rough determinations
are valueless. For a discussion of the accuracy of Jesuit solar eclipse
 9.3 Careful solar eclipse timings recorded in Chinese history 287

Fig. 9.3 Page from a nineteenth century printing of the Shou-shih-li I, containing
records of solar eclipses in AD 1053 and 1054.

observations made at Beijing from 1644 to 1785, see Stephenson and

Fatoohi (1995).
Translations of the few texts in works other than the Yuan-shih which
cite eclipse timings to the nearest k'o are given below. These show an
interesting variety. Many more solar eclipse timings are reported in the
Yuan-shih Calendar Treatise than in the other sources listed above (see
section 9.2.5), but the form of these records is rather stereotyped. Hence
288 9 Other East Asian observations of solar and lunar eclipses

translations of only a few texts from this compilation will be given as

illustrative examples, although a comprehensive summary will be provided
in tabular form (see table 9.4).

9.3.1 Solar eclipse timings recorded in sources other than the Yuan-shih
For each observation investigated in this section the following details are
given: equivalent Julian date, computed magnitude (for reference only),
presumed place of observation (i.e. the appropriate dynastic capital), trans-
lation of the text together with commentary, and the corresponding AT
results. Information in the texts which is irrelevant in the present context
- for example the results of prediction by Chinese astronomers - is not in-
cluded in the various translations. The place of observation is taken as the
appropriate capital: Ta-hsing Ch'eng in AD 586; Pien (from 1052 to 1100);
and Lin-an (1173 and 1245). Geographical co-ordinates of the various sites
are listed in the previous chapter (table 8.3). Eclipse magnitudes - for ref-
erence only - are computed using either equation (4.1) or (8.1), depending
upon whether the appropriate event occurred before or after AD 950.

(1) AD 586 Dec 16 (mag. = 0.84): Ta-hsing Ch'eng

K'ai-huang reign period, 6th year, 10th month, 30th day ting-ch'ou
[14]...It was seen during the observations that when the Sun rose 1 chang
(i.e. roughly 10 deg) above the mountains, it began to be eclipsed; (this
was) at 2 marks in the hour of ch'en. The loss began from the west; it was
two-thirds eclipsed
[Sui-shu, chap. 17.]
The measured LT of beginning corresponds to 7.60 h. The remainder
of the above entry uses an ambiguous terminology (see section 9.2.1) and
will not be considered further.

RESULTS
LT of first contact = 7.60 h, UT = 0.35 h. Computed TT = 1.37 h, thus
AT = 3650 sec.

(2) AD 1040 Feb 15 (mag. = 0.69): Pien

Pao-yuan reign period,firstyear, first month, day ping-ch'en [53], the
first day of the month. The Sun was eclipsed by 6 divisions (fen) and it
was restored at 1 mark in the hour of shen.
[Wen-hsien Tung-k'ao, chap. 283.]
There was no eclipse on or near the stated day in the first year of the
Pao-yuan reign period (which corresponds to AD 1038). Within two years
on either side of the recorded date, only the eclipse of AD 1040 Feb 15
 9.3 Careful solar eclipse timings recorded in Chinese history 289

would be visible at Pien. This event occurred in the third year of the
Pao-yuan reign period, but otherwise both the month and day are correct.
Hence the date 1040 Feb 15 can be confidently assumed.
As will be discussed in section 9.6, the unit of magnitude translated
as 'division' (fen) was equal to one-tenth of the solar (or lunar) apparent
diameter.
The LT of end corresponds to 15.36 h.

RESULTS
LT of last contact = 15.36 h, UT = 7.99 h. Computed TT = 8.79 h, thus
AT = 2850 sec.

(3) AD 1046 Apr 9 (mag. = 0.46): Pien

Ch'ing-li reign period, 6th year, 3rd month, day hsin-szu [18], the first
day of the month. The Sun was eclipsed by 4^ divisions (fen). At 3 marks
in the hour of shen it was restored.
[Wen-hsien T'ung-k'ao, chap. 283.]
The LT of end corresponds to 15.84 h. In chapter 53 of the Yuan-shih,
the time is given as '3 marks in the central half of the hour of shen' -
i.e. 16.84 h. Because of this disagreement, the observation will not be
considered further.

(4) AD 1052 Nov 24 (mag. = 0.11): Pien

Huang-yu reign period, 4th year, 11th month, day jen-yin [39], the first
day of the month. The Sun was eclipsed by more than 2 divisions {fen). At
1 mark in the central half of the hour of wei it was restored.
[Sung-shih, chap. 81.]
The LT of end corresponds to 14.36 h.
RESULTS
LT of last contact = 14.36 h, UT = 6.58 h. Computed TT = 7.39 h, thus
AT = 2900 sec.

(5) AD 1053 Nov 13 (mag. = 0.66): Pien

Huang-yu reign period, 5th year, 10th month, day ping-shen [33], the
first day of the month. At 1 mark in the central half of the hour of wu9 the
Sun was eclipsed by 4^ divisions if en).
[Wen-hsien T'ung-k'ao, chap. 283.]
The LT of maximum corresponds to 12.36 h. In chapter 53 of the
Yuan-shih, the time is given as '1 mark in the hour of wef - i.e. 13.36 h.
Because of this discrepancy between the two sources, the observation will
be rejected.
290 9 Other East Asian observations of solar and lunar eclipses

(6) AD 1054 May 10 (mag. = 0.71): Pien

Chih-ho reign period, 1st year, 4th month, day chia-wu [31], the first day
of the month. The Sun was eclipsed by more than 9 divisions (fen) and the
loss began on the south-west side. It reached its maximum at 1 mark in the
central half of the hour of shen. On this day it was cloudy and rain fell...
[Wen-hsien Tung-k'ao, chap. 283.]

The LT of maximum, which corresponds to 16.36 h, is confirmed by the

record in chapter 53 of the Yuan-shih.

RESULTS
LT of last contact = 16.36 h, UT = 8.65 h. Computed TT = 9.17 h, thus
AT = 1900 sec.

(7) AD 1059 February 15 (mag. = 0.36): Pien

Chia-yu reign period, 4th year, first month, day ping-shen [33], the first
day of the month. The Sun was eclipsed by more than 3 divisions (fen). At
3 marks in the initial half of the hour of wei it was restored.
[Wen-hsien T'ung-k'ao, chap. 283.]

The LT of last contact, which corresponds to 13.84 h, is endorsed by

the record in chapter 53 of the Yuan-shih.

RESULTS
LT of last contact = 13.84 h, UT = 6.47 h. Computed TT = 7.13 h, thus
AT = 2400 sec.

(8) AD 1068 Feb 6 (mag. = 0.55): Pien

Hsi-ning reign period, 1st year, 1st month, day chia-hsu [11], the first
day of the month. The Sun was eclipsed. According to the astronomers, on
this day at 8 marks in the hour of szu the Sun was seen to diminish; the
loss began on the south-west side. After (hou) 5 marks in the hour of wu,
the eclipse reached six divisions (fen). Not until 3 marks in the hour of wei
was it restored to roundness.
[Sung-hui-yao Chi-kao]

In recording the time of maximum, the use of (hou) after the figure 5
will be assumed to be an alternative for shang, implying a moment at the
end of the 5th mark.
The LTs of beginning, maximum and end correspond respectively to
10.96 h, 12.44 h and 13.84 h.
 9.3 Careful solar eclipse timings recorded in Chinese history 291

RESULTS
(i) LT of first contact = 10.96 h, UT = 3.60 h. Computed TT = 4.05 h,
thus AT = 1600 sec.
(ii) LT of maximal phase = 12.44 h, UT = 5.07 h. Computed TT =
5.40 h, thus AT = 1200 sec.
(iii) LT of last contact = 13.84 h, UT = 6.48 h. Computed TT = 6.71
h, thus AT = 850 sec.

(9) AD 1094 Mar 19 (mag. = 0.81): Pien

Yuan-yu reign period, 9th year, 3rd month, day jen-shen [9], the first
day of the month. The Astronomer-Royal reported that the Sun should
have been eclipsed, but on account of cloud cover it was not (fully) seen. It
began (to be visible) at 3 marks in the hour of wei. It was seen through
clouds that the Sun was eclipsed by more than one division (fen) on the
south-west side. At 6 marks in the hour of (wei) it reached its maximum of
7 divisions. Then it became invisible on account of the covering of clouds.
[Wen-hsien T'ung-k'ao, chap. 283.]
Since the eclipse was said to increase from ^ to ^ in 3 marks, it is a
reasonable approximation to assume that first contact occurred \ mark
before the phase of JQ was observed, corresponding to a LT of 13.72 h.
Maximum phase was at 14.56 h, after which the eclipse was lost to view
in the clouds. Evidently the degree of obscuration of the Sun had ceased
to noticeably increase by the time the Sun entered the clouds. The LT of
maximum is confirmed by the report in chapter 53 of the Yuan-shih.

RESULTS
(i) LT of first contact = 13.72 h, UT = 6.20 h. Computed TT = 6.62 h,
thus AT = 1500 sec.
(ii) LT of maximal phase = 14.56 h, UT = 7.04 h. Computed TT =
7.78 h, thus AT = 2650 sec.

(10) AD 1100 May 11 (mag. = 0.34): Pien

Yuan-fu reign period, 3rd year, 4th month, day ting-mao [4], the first day
of the month. The Astronomer-Royal said, 'In the initial half of the hour
of ch'en, the Sun was eclipsed at the north-west by 4 divisions (fen). At 3
marks in the central half of the hour of szu it was restored to fullness...
[Sung-hui-yao Chi-kao.]
The LT of beginning (7-8 h) is only roughly expressed. The LT of end
corresponds to 10.84 h.

RESULTS
LT of last contact = 10.84 h, UT = 3.12 h. Computed TT = 2.84 h, thus
AT = -1000 sec.
292 9 Other East Asian observations of solar and lunar eclipses

(11) AD 1173 Jun 12 (mag. = 0.58): Lin-an

Ch'ien-tao reign period, 9th year, 5th month, the first day of the month.
Both officials and students from this Bureau observed that the Sun was
eclipsed A\ divisions if en). It began from the north-west at 5^ marks in the
hour of wu. The eclipse reached its maximum at the north side at 2 marks
in the initial half of the hour of wei. It was restored to roundness at the
north-east at 1 mark in the initial half of the hour of shen.
[Sung-shih, chap. 81.]

Assuming that '5^ marks in the hour of wif means the middle of
the appropriate mark, the time of first contact is equivalent to 12.32 h.
The LTs of maximum and end correspond respectively to 13.60 h and
15.36 h.

RESULTS

(i) LT of first contact = 12.32 h, UT = 4.30 h. Computed TT = 4.55 h,

thus AT = 900 sec.
(ii) LT of maximal phase = 13.60 h, UT = 5.57 h. Computed TT =
6.12 h, thus AT = 2000 sec.
(iii) LT of last contact = 15.36 h, UT = 7.33 h. Computed TT = 7.72
h, thus AT = 1400 sec.

(12) AD 1202 May 23 (mag. = 0.58): Lin-an

Chia-t'ai reign period, 2nd year, 5th month, day chia-ch'en [41], the first
day of the month. The Sun was eclipsed. The loss began at 1 mark in the
initial half of the hour of wu. It was restored to roundness in the initial
half of the hour of wei... the eclipse reached 3 divisions if en) at 3^ marks
in the initial half of the hour of wu.
[Sung-shih, chap. 81.]

The LT of first contact corresponds to 11.36 h; this is confirmed by the

record in chapter 53 of the Yuan-shih. In the case of last contact, the LT
(13-14 h) is only approximately expressed. Assuming that ' 3 | marks in
the initial half of the hour of wu' means the middle of the appropriate
mark, the LT of maximum phase is equivalent to 11.84 h.

RESULTS

(i) LT of first contact = 11.36 h, UT = 3.28 h. Computed TT = 3.47 h,

thus AT = 700 sec.
(ii) LT of maximum = 11.84 h, UT = 3.76 h. Computed TT - 4.23 h,
thus AT = 1700 sec.
 9.3 Careful solar eclipse timings recorded in Chinese history 293

(13) AD 1245 Jul 25 (mag. = 0.74): Lin-an

Shun-yu reign period, 5th year. According to the calculations of an
astronomical official named Cheng, the Sun should have been eclipsed at 3
marks in the initial half of the hour of wei. It was observed at 4 marks in
the central half of the hour of wei. The magnitude should have been 8
divisions (fen) but it was actually 6 divisions.
[Sung-shih, chap. 82.]
No more than the year is given, but the eclipse of Jul 25 is the only
obscuration of the Sun visible during that time.
The observed LT of beginning corresponds to 14.98 h.

RESULTS
LT of first contact = 14.98 h, UT = 7.05 h. Computed TT = 7.42 h, thus
AT = 1350 sec.

The above investigations are summarised in table 9.4 together with the
analysis of measurements which are recorded only in the Yuan-shih (see
section 9.3.2).

9.3.2 Solar eclipse timings recorded in the Calendar Treatise of the

Yuan-shih
In citing the solar eclipse timings in its Calendar Treatise, the Yuan-
shih draws upon material from a number of earlier dynasties as well as
contemporary data. As already noted, the records in the Shou-shih-H I
follow a very repetitive style. Evidently the original observations were re-
formulated to conform to a standardised pattern, in the process rejecting
all unnecessary detail. Observations of eclipses of the Sun are contained
in a section entitled 'Solar Eclipses from the Three Kingdoms Period
Onwards'. In each individual entry, the word Sun is never mentioned since
the overall title renders it superfluous.
Sexagenary years, rare in Chinese texts, are consistently employed in
addition to regnal years in this treatise, evidently to facilitate counting
intervals. Incidentally, the use of sexagenary years began around 200
BC. The same names were used as for cyclical days (see chapter 8). The
enumeration of the system is such that AD 1 was the 58th year of a cycle.
Hence the cyclical year for any AD date may be found by adding 57 to
the AD year, dividing by 60, then multiplying the remainder by 60. For
example, 1997 is the 14th year (ting-ch'ou) of the present cycle - i.e. the
'Year of the Ox'. (Obviously in the BC period, a correction of one year
would have to be made on account of the absence of a year zero.)
In many cases, a report of an eclipse quoted in the Shou-shih-li I may
also be found in the Astrological Treatise of the appropriate dynastic
294 9 Other East Asian observations of solar and lunar eclipses

history. However, in no instance is a precise time quoted in one of

these alternative sources. Rather surprisingly, the detailed measurements
between AD 585 and 768 which are found in the Calendar Treatise of
the Sui-shu or the Astrological Treatise of the Chiu-t'ang-shu are not
mentioned in the Shou-shi-li I. Just possibly, the Yuan astronomers were
uncertain as to the interpretation of the times quoted in these texts (in
common with modern investigators). Several examples of duplication
between the data in the Shou-shih-li I and Wen-hsien Tung-k'ao or Sung-
shih have already been noted - as well as two cases of discord (AD 1046
and 1053).
The date of the earliest observation for which the time is quoted to the
nearest k'o in chapter 53 of the Yuan-shih is equivalent to AD 680 Nov
27. The text may be translated as follows:
T'ang dynasty, Yung-lung reign period, epochal year keng-ch'en [17],
11th month, day jen-shen [9], the first day of the month. (The Sun) was
eclipsed. It reached its maximum at 4 marks in the hour of szu. (According
to) the Season-Granting System, the eclipse maximum was at 7 marks in
the hour of szu, (whilst according to) the Great Enlightment System the
eclipse maximum was at 5 marks in the hour of szu. The Season-Granting
(System) is off, the Great Enlightment (System) is close.

As noted above (section 9.2.5), a tabular error of 3 marks was regarded

by the Yuan astronomers as 'off' whilst a discrepancy of 1 mark was
described as 'close'. Both of the official T'ang histories simply report the
occurrence of a solar eclipse on the stated day, without mentioning the
time of the contacts.
In common with other T'ang records of eclipses, the place of observation
on this occasion was evidently the capital of Ch'ang-an. The observed LT
of greatest phase corresponds to approximately 10.08 h, a UT of 2.67 h,
leading to a value for AT of 2500 sec.
Several other observations quote the measured times of all three phases,
an example being given by the record of the last eclipse in the series: that
of AD 1277Oct28:
Present dynasty, Chih-yuan reign period, 14th year, 10th month, day
ping-ch'en [53], the first day of the month. (The Sun) was eclipsed.
Beginning of loss in the initial mark of the central half of the hour of wu;
maximum at 1 mark in the initial half of the hour of wei; restoration of
fullness at 2 marks in the central half of the hour of wei...
[Yuan-shih, chap. 53.]
There follows the usual comparison with predictions according to the
Season-Granting and Great-Enlightment systems. By this time Ta-tu
had been the Yuan capital for more than a decade and in the previous
 9.4 Lunar eclipse timings 295

year Kuo Shou-ching had commenced building the Imperial Observatory

there. The approximate observed LTs of beginning, greatest phase and
end correspond respectively to 12.12 h, 13.36 h and 14.60 h., i.e. UTs of
4.10, 5.34 and 6.58 h. Since the computed TTs are 4.18 h, 5.48 h and
6.75 h, AT values of 300, 500 and 600 sec are required to satisfy these
observations.
Since the eclipse records in the Shou-shih-li I follow a highly stereotyped
pattern, no further translations will be given in this section. Calculations
have been made for the following locations: Ch'ang-an (AD 680-702);
Pien (AD 1066-1243); and Ta-tu (AD 1277). A further eclipse timing
is reported on a date corresponding to AD 1260 Apr 12. However, the
observation was made several years before the Yuan capital of Ta-tu was
established. Hence the place of origin of the record (which states that the
eclipse reached its maximum at 2 marks in the central half of the hour of
shen) is very doubtful. The record will thus be rejected.
The various results are summarised in table 9.4, along with the results
from solar eclipse timings analysed in section 9.3.1. In this table are given
the following details, column by column: (1) Julian date (year, month
and day); (2) phase observed (1 = first contact, M = maximum; 4 = last
contact); (3) double hour; (4) (where appropriate) initial (I) or central (II)
half of this double hour; (5) the number of marks within the designated
double or single hour; (6) the equivalent LT in hours and decimals; (7)
the corresponding UT; (8) the computed TT; and (9) the resultant value
of AT (in sec) necessary to satisfy the observation. Full Julian dates are
given in column 1 since most of these are not cited elsewhere in this
section.
NB in AD 1068, maximum phase was recorded 'after' (hou) 5 marks
in the hour of wu. Hence the observation will be presumed to have been
made at the end of this division; this is indicated by a + sign in column 5.
The AT values listed in table 9.4 are plotted in figure 9.4. Following
the usual practice for such diagrams in this book, positive numbers
are adopted for years rather than AD. Nearly all of the observations
represented in this diagram date from between AD 1040 and 1277. Bearing
in mind that each measurement is quoted to the nearest k'o (i.e. some
860 sec), the scatter of points within this date range is considerably larger
than might be expected.

9.4 Lunar eclipse timings

Like the preserved solar eclipse timings discussed above, these observations
fall into three distinct chronological groupings: the date ranges are: from
AD 434 to 596; between AD 948 and 1280; and after 1577. Measurements
in the latest of these periods (which like their solar counterparts are all
296 9 Other East Asian observations of solar and lunar eclipses

Table 9.4 AT values from more reliable solar eclipse timings recorded in
Chinese history.

Date AD Ct Hour Half Mark LT(h) UT(h) TT(h) AT (sec)

586 Dec 16 1 ch'en — 2 7.60 0.35 1.37 3650
680 Nov 27 M szu — 4 10.08 2.67 3.36 2500
691 May 4 M mao — 2 5.60 22.25 22.53 1000
702 Sep 26 M shen — 3 15.84 8.42 9.07 2350

1040 Feb 15 4 shen 1 15.36 7.99 8.79 2850

1052 Nov 24 1 wei II 1 14.36 6.58 7.39 2900
1054 May 10 M shen II 1 16.36 8.65 9.17 1900
1059 Feb 15 4 wei I 3 13.84 6.47 7.13 2400

1066 Sep 22 M wei 2 13.60 5.82 6.29 1700

1068 Feb 6 1 szu — 8 10.96 3.60 4.05 1600
M wu — 5+ 12.44 5.07 5.40 1200
4 wei — 3 13.84 6.48 6.71 850

1069 Jul 21 M ch'en 3 7.84 0.30 0.44 500

1080 Dec 14 M szu — 6 10.56 2.93 2.93 0
1094 Mar 19 1 wei — 2.5 13.72 6.20 6.62 1500
M wei — 6 14.56 7.04 7.78 2650

1100 May 11 4 szu II 3 10.84 3.12 2.84 -1000

1107 Dec 16 1 wei — 2 13.60 5.99 6.28 1050
M wei — 8 14.96 7.35 7.74 1400
4 shen — 6 16.56 8.95 9.27 1150

1173 Jun 12 1 wu 5.5 12.32 4.30 4.55 900

M wei i 2 13.60 5.57 6.12 2000
4 shen i 1 15.53 7.33 7.72 1400
1183 Nov 17 M szu ii 2 10.60 2.39 2.61 800

1195 Apr 12 1 wu i 2 11.60 3.57 3.81 850

1202 May 23 1 wu i 1 11.36 3.28 3.47 700
M wu i 3.5 11.84 3.76 4.23 1700
1216 Feb 19 M shen II 4 16.98 9.20 9.31 400

1243 Mar 22 M szu i 2 9.60 1.67 1.77 350

1245 Jul 25 1 wei II 4 14.98 7.05 7.42 1350

1277 Oct 28 1 wu II 0 12.12 4.10 4.27 300

M wei i 1 13.36 5.34 5.48 500
4 wei II 2 14.60 6.58 6.75 600
 9.4 Lunar eclipse timings 297

-500-

-1000
+600 +700 +800 +900 +1000 +1100 +1200 +1300
Year
Fig. 9.4 AT results derived from Chinese solar eclipse timings.

quoted to the nearest mark until 1644) will not be analysed since by then
AT was very small (see also section 9.3).
During the earliest interval (AD 434-596), there are several useful
observations in the calendar treatises of both the Sung-shu (chaps. 12
and 13) and Sui-shu (chap. 17); each date range is fairly narrow -
respectively from AD 434 to 440 and from AD 585 to 596. In contrast
with the solar eclipse records from the same period, most of the accounts
of lunar obscurations which are to be found in the Sung-shu and Sui-shu
are summarised in the Shou-shih-li I of the Yuan-shih. There is exact
agreement between the two sets of LTs in every case. It should be pointed
out that there are minor differences in terminology between the Shou-shih-
li I and the Sui-shu when describing the same observation. Thus, while the
time-units for the eclipses reported in the Sui-shu are consistently quoted
as 'rods' (ch'ou), the Shou-shih-li I uses the term 'calls' (ch'ang) in AD 592,
'points' (tien) in 595 and 'rods' only in 596. However, these terms all relate
to the same units - i.e. one-fifth of a night-watch (see section 9.2.2).
Between AD 948 and 1280, numerous lunar eclipses are reported. Most
are contained in the Wen-hsien Tung-k'ao (chap. 285) and the Yuanshih
(chap. 53). A very few additional observations are cited in the Sung-shih
Calendar Treatise (chaps. 81 and 82). In our recent paper on the Earth's
past rotation (Stephenson and Morrison, 1995), I had overlooked the lunar
observations recorded in the Wen-hsien Tung-k'ao. I am grateful to Pro-
fessor Xu Zhentao of Purple Mountain Observatory, Nanjing - who lately
spent a year at Durham University - for drawing my attention to them.
298 9 Other East Asian observations of solar and lunar eclipses

Overlap between the Wen-hsien T'ung-k'ao and Yuan-shih is common

over the date range AD 1063 to 1074, but beyond these limits the two
works only record separate eclipses. As in the case of solar obscurations,
the Yuan-shih treatise compares the various measurements with the times
deduced from the Season-Granting and Great Enlightment calendars.
The lunar observations from the various sources may be conveniently
divided into two categories: those where measurements are expressed
in terms of the night watches and those for which double hours and
marks are used instead. (Further classification - e.g. whether or not the
observation is found in the Yuan-shih Calendar Treatise - does not seem
practical owing to textual duplication.) All of the earlier data (prior to
AD 600) come into the first category, along with observations made in
AD 948, 1068, 1168, 1074 and 1185. The remaining records - ranging
in date from AD 1052 to 1280 - are consistently timed to the nearest
k'o. There are too many accounts in both categories to warrant verbatim
translations. Hence in the following pages I have translated only selected
texts to illustrate the type of material available. In each case I have chosen
duplicate entries relating to the same event from different sources. As in
the case of solar eclipses, all reference to prediction in the texts is omitted.
Following the practice adopted throughout this book, all Julian (or
Gregorian) dates for lunar eclipses will be double - e.g. Jun 23/24.

9.4.1 Lunar eclipses timed in fifths of a night-watch

Most records merely assert that a particular phase occurred at a certain
division of a watch. Under these circumstances, the mid-point of the
specified division will normally be assumed (a similar convention was
adopted for solar eclipse timings - see above). However, the eclipse of AD
595 Dec 22/23 was stated to begin 'after' (hou) the fourth rod of the first
watch. In this case the start of the fifth rod will be assumed.
The very earliest lunar eclipse for which an accurate time is specified is
recorded in chapter 12 of the Sung-shu on the night of AD 434 Sep 4/5:
Yuan-chia reign period, 11th year, 7th month, 16th day, full Moon. The
Moon was eclipsed. The calculated time was the hour of mao. The Moon
began to be eclipsed at the second call of the fourth watch, in the initial
half of the hour of ch'ou. The eclipse was total at the fourth call. The
Moon was at the end of the 15th degree of Ying-shih (lunar lodge).
The text ends with an estimate of the RA of the Moon. The Shou-
shih-H I gives a standardised version of the above account, leaving out all
superfluous detail.
Yuan-chia reign period, 11th year, chia-hsu [11], 7th month, day ping-tzu
[13], full Moon. The Moon was eclipsed. Beginning of loss at the second
 9.4 Lunar eclipse timings 299

call of the fourth watch. The eclipse was total at the fourth call of the
fourth watch.

There follows a comparison with the times deduced from the Season-
Granting and Great Enlightment calendars. Interestingly, in reporting the
results of their calculations for this and all subsequent lunar eclipses which
were timed relative to the night-watches, the Yuan astronomers used the
contemporary unit 'point' (tien), while retaining more archaic names in
summarising the observational record itself.
The equivalent LTs of first contact and beginning of totality (taken as
the mid-points of the second and fourth calls of the fourth watch) were
respectively 1.65 h and 2.46 h. These correspond to UTs of 17.65 h and
18.42 h. Comparing these with the computed TTs of 18.14 h and 19.20 h
leads to AT values of respectively 1750 and 2800 sec.
The remaining records in the Sung-shu relate to the eclipses of AD 437
Jan 8/9, 437 Dec 28/29, and 440 Oct 26/27. These have much in common
with the account of the eclipse of AD 434 in the same source. However,
reports from the Sui dynasty tend to be more detailed, noting which part
of the Moon was obscured, as well as specifying the lunar azimuth at
the various phases. A useful example is provided by the account of the
total eclipse of AD 596 Dec 10/11 in the Calendar Treatise of the Sui-shu
(chap. 17):

K'ai-huang reign period, 16th year, 11th month, 16th day, i-ch'ou
[2]... Not until the first rod of the third watch was the Moon seen in the
clouds above the direction ping (roughly SSE, azimuth approximately 165
deg). It was already about 3 fifteenths eclipsed and the loss began from the
east side. Above the direction ting (roughly SSW, azimuth approximately
195 deg), the eclipse was total. Afterwards it reappeared from the SE side.
It was not until the third rod of the fourth watch that it was restored to
fullness; the Moon was then at the end of the direction wei (azimuth
approximately 217.5 deg).

A very brief summary of the above account, giving only the date and
time of last contact is given in chap. 53 of the Yuan-shih:

K'ai-huang reign period, 16th year, ping-ch'en [53], 11th month, day
ping-tzu [13], full Moon. (The Moon) was eclipsed. At the third rod of the
fourth watch it was fully restored.

The measurement of last contact (assumed to relate to the middle of

the third rod) is equivalent to an LT of 2.60 h. Although ignored by
the Yuan astronomers, the observation made near the start of the eclipse
is of comparable precision. Adjusting the eye estimate of ^ (0.20) for
the likely systematic error of observation (see chapter 3) leads to only
300 9 Other East Asian observations of solar and lunar eclipses

about 0.10 of the lunar diameter being obscured when the Moon first
became visible 'through clouds'. This eclipse, which eventually became
total, would reach a phase of 0.10 only about 0.15 h after first contact.
On this basis, the eclipse may be considered to have started 0.15 h
before the middle of the 'first rod of the fourth watch', which occurred
at 22.99 h. The equivalent LT of first contact may thus be estimated as
22.84 h.
If the above LTs are reduced to UT, and subtracted from the computed
TT values, the corresponding AT values may be derived (see table 9.5).
In making the necessary computations for the data in this section, the
places of observation have been taken as Chien-k'ang (AD 434-543), Ta-
hsing Ch'eng (AD 585-596), Pien-chou (AD 948), Pien (AD 1074) and
Lin-an (AD 1185). Pien-chou was the former name for Pien.
In table 9.5 are summarised the various observations for which measure-
ments are specified relative to the night-watches, together with equivalent
LTs (deduced from the computed LTs of sunrise and sunset), equivalent
universal times, and resulting AT values.
In AD 595, first contact was recorded 'after' (hou) the fourth division
of the first watch. Hence the observation will be presumed to have been
made at the end of this division; this is indicated by a + sign in column 4.
The various results listed in table 9.5 (with the exception of the highly
discordant AT value of —2300 sec in AD 1168) are plotted in figure 9.5.
In this diagram, the distribution of dates is quite dissimilar from that in
figure 9.4, with relatively few observations after AD 800. Although the two
results derived from the eclipse of AD 434 are obviously discordant, most
of the AT values between AD 434 and 596 form a fairly self-consistent
set. This is a period when few useful data from other parts of the world
are preserved.

9.4.2 Lunar eclipses timed in double hours and marks

Most of the extant observations in this category are recorded in either
chapter 53 of the Yuan-shih or chapter 285 of the Wen-hsien T'ung-k'ao.
In addition, attention has been drawn in chapter 3 to an isolated report,
dating from AD 1168 Mar 25/26, which is preserved in chapter 82 of the
Sung-shih. The account of this eclipse, which is particularly detailed may
be translated as follows:

Ch'ien-tao reign period, 4th year, 2nd month, 14th day ting-wei [44], full
Moon. The Moon was eclipsed... (predictions)... On this evening when the
Moon rose there was light cloud. Until the fall of darkness it could not be
seen that the Moon had been totally eclipsed. When the 3rd mark of the
initial half of the hour hsu was reached, as expected it was shining and so
 9.4 Lunar eclipse timings 301

Table 9.5 AT values from lunar eclipse timings expressed in terms of the
night watches.

Date A D Ct Watch Div. LT(h) UT(h) TT(h) A T (sec)

434 Sep 4/5 1 IV 2 1.65 17.65 18.14 1750
2 IV 4 2.42 18.42 19.20 2800
437 Jan 8/9 2 I 3 18.95 11.23 13.30 7450
437 Dec 28/29 1 II 4 21.97 14.16 15.83 6000
2 III 1 22.99 15.18 17.00 6550

440 Oct 26/27 1 II 1 20.70 12.54 14.26 6200

M II 3 21.64 13.49 15.74 8100
543 May 4/5 1 III 3 0.00 15.99 17.27 4600
585 Jan 21/22 1 I 1 17.99 10.98 12.65 6000
M I 4 19.49 12.49 14.07 5700
4 II 1 20.49 13.49 15.49 7200

592 Aug 28/29 1 I 3 20.05 12.78 14.42 5900

593 Aug 17/18 1 V 1 3.05 19.82 21.11 4650
595 Dec 22/23 1 I 4+ 19.59 12.38 13.89 5450
M II 3 21.40 17.80 15.32 4050
4 III 1 22.96 15.76 16.73 3500

596 Dec 10/11 1 III 1- 22.86 15.56 16.73 4200

4 IV 3 2.60 19.30 20.63 4800
948 Jan 28/29 1 IV 4 2.94 19.59 20.81 4450
1074 Oct 7/8 1 IV 5 3.19 19.34 19.78 1600
2 V 3 4.56 20.71 20.85 500

1168 Mar 25/26 4 II 2 21.52 13.58 12.93 -2300

1185 Apr 16/17 M III 3 0.00 15.94 15.97 100

it could be known that this eclipse had been total on rising. It was restored
to fullness at the 3rd mark in the central half of the hour of hsu. It was the
2nd point of the 2nd watch.
[Sung-shih, chap. 82.]
This observation is unusual in that it gives the time of last contact
in both systems: hours and marks, and also night-watches and divi-
sions. The LT of third contact corresponds to 19.84 h and that of last
contact (expressed in double hours) to 20.84 h. These two results are
very self-consistent; the computed interval between third and last contact
(independent of AT) was 1.03 h. Converting these to UT (respectively
11.90 h and 12.90 h) and comparing with the computed TTs of 11.90 h
302 9 Other East Asian observations of solar and lunar eclipses

+8000 X

X
X
+7000 -
X
X
+6000 - X *x
X
X
+5000 -

+4000 - $
X

+3000 X

+2000
X
X

+1000 -
X
1
0 1 1 1 1 1 1 X

+400 +500 +600 +700 +800 +900 +1000 +1100 +1200

Year
Fig. 9.5 AT results obtained from lunar eclipse times which were expressed in
terms of the night watches.

and 12.93 h, enable values for AT of 0 and +100 sec to be deduced. The
LT for the end of the eclipse expressed in terms of the night-watches,
equivalent to 21.52 h (rather than 20.84 h), is clearly discordant. Pre-
sumably a scribal error has occurred here; the equivalent value of AT is
-2300 sec.
The numerous Yuan-shih timings expressed to the nearest Wo extend
from AD 1052 to 1280, while the corresponding Wen-hsien T'ung-k'ao
data cover the relatively brief period from AD 1063 to 1099. About 40
individual timings are preserved in each source. Two examples of the type
of observation available are as follows, one from the Sung dynasty (AD
1069 Dec 30/31) and the other from the Yuan itself (1277 May 18/19):

(AD 1069.) Sung dynasty, Hsi-ning reign period, 2nd year, chi-yu [46],
intercalary 11th month, day ting-wei [44], full Moon. (The Moon) was
eclipsed. Beginning of loss at 6 marks in the hour of hai; eclipse maximum
at 5 marks in the hour of tzu; restoration of fullness at 4 marks in the
hour of ch'ou.
[Yuan-shih, chap. 53.]

The above LTs are equivalent to 22.56 h (start), 0.32 h (maximum) and
2.08 h (end). Expressing them in UT and subtracting from the computed
TT values leads to AT results of 2650, 1750 and 800 sec (see table 9.7
below).
 9.4 Lunar eclipse timings 303

(AD 1277.) Present Dynasty, Chih-yuan reign period, 14th year,

ting-ch'ou [14], 4th month, kuei-yu [10], full Moon. (The Moon) was
eclipsed. Beginning of loss at 6 marks in the hour of tzu; the eclipse was
total at 3 marks in the hour of ch'ou; maximum at 5 marks in the hour of
ch'ou; reappearance of light at 7 marks in the hour of ch'ou; restoration of
fullness at 4 marks in the hour of yin.
[Yuan-shih, chap. 53.]
The various measurements may be reduced to the following LTs: 0.56 h
(start); 1.84 h (beginning of totality); 2.20 h (maximum eclipse); 2.68 h
(end of totality); 4.08 h (end). All of these times apart from the LT of
maximum phase may be used in the determination of AT. However, the
moment of greatest phase for a total lunar eclipse cannot be fixed by direct
observation; the recorded time is merely the mean of the determinations
for the beginning and end of totality.
Several of the records in the Wen-hsien T'ung-k'ao are rather detailed
and most give an estimate of magnitude as well as citing the LTs of the
various phases. For example, the record of the partial lunar eclipse of AD
1071 Dec 9/10 may be translated as follows:

Hsi-ning reign period, 4th year, 11th month, day ping-shen [33]. The
Moon was eclipsed. At 2 marks in the hour of mao the eclipse began to be
seen in the west. It commenced at the south-east side. Not until 6 marks in
the hour of mao was the eclipse at its maximum, reaching less than 4^
divisions (fen). It was rather less than 1 deg in Tung-ching (lunar lodge).
Before it became bright it set eclipsed and the restoration was not seen.

The Yuan-shih also reports this eclipse, quoting the same times as above
but ignoring the other observational details. First contact occurred at a
LT of 5.60 h and maximal phase at 6.56 h.
The eclipse of AD 1063 Nov 8/9 is somewhat ambiguously recorded.
This is described in the Wen-hsien Tung-k'ao as follows:

Chia-yu reign period, 8th year, 10th month, day kuei-wei [20]. The
Moon was totally eclipsed. It reached its maximum at 7 marks in the hour
of mao. Then it set eclipsed and was not seen.

There is nothing in the text to indicate that the Moon reappeared

after totality so that the time of maximum may have been incorrectly
interpolated by the observers. Hence the measurement will be rejected.
The same LT of greatest phase is also reported in the Yuan-shih but
without any reference to moonset.
In deriving AT values, the places of observation have been taken as the
Sung capitals of Pien (AD 1052 to 1106) and Lin-an (AD 1168), and the
Mongol capital of Ta-tu (AD 1270 to 1279).
304 9 Other East Asian observations of solar and lunar eclipses

Table 9.6a Analysis of anomalous lunar eclipse records in the Wen-hsien

T'ung-k'ao.

Date AD Ct Hour Mark LT(h) UT(h) TT(h) AT(sec) Remarks

1067 Mar 3/4 1 ch'ou 4 2.08 18.65 21.84 11500 set eel.

1068 Aug 15/16 1 ch'ou 5 2.32 18.73 21.67 10600 set eel.

1069 Dec 30/31 1 hai 1 21.36 13.86 15.80 7000

M tzu 0 23.12 15.62 17.30 6050 —
4 ch'ou 3 1.84 18.34 18.80 1650 —

1073 Oct 18/19 1 ch'ou 4 2.08 18.20 20.74 9150 set eel.
M yin 1 3.36 19.48 22.16 9 650 set eel.

1074 Oct 7/8 1 ch'ou 1 1.36 17.52 19.78 8150 set eel.
2 ch'ou 6 2.56 18.72 20.85 7 650 set eel.

1077 Feb 10/11 1 tzu 3 23.84 16.48 18.44 7050

M tzu 7 0.80 17.44 20.05 9400 —
4 ch'ou 2 1.84 18.48 21.66 11450 —

Most of the observations in the Wen-hsien T'ung-k'ao lead to results

for AT between about 0 and 1000 sec - see table 9.7 below. Values in
this region are in satisfactory agreement with the results obtained from
roughly contemporaneous solar eclipse observations - see table 9.4 and
figure 9.4. However, the data from two discrete periods (AD 1067 to 1069
and also AD 1073 to 1077) are highly discordant, AT values ranging from
around 6000 to 12000 sec. Analysis of these observations is summarised
in table 9.6a. As it happens, four of the records in these two brief intervals
(AD 1067, 1068, 1073 and 1074) state that the Moon set eclipsed, but
this assertion cannot be reconciled with the timed measurements. The
values of AT indicated by the timed contacts, as listed in table 9.6a, lead
to altitudes of the Moon (in the west) at last contact of respectively +5,
+11, +21 and +13 deg. Hence in no case would the Moon set eclipsed.
It appears that serious errors have occurred in the transmission of the
original timings for all four eclipses and this presumably applies to the two
other dates in the table (AD 1069 and 1077). Hence all six observations
will be rejected.
Two of the eclipses cited above (AD 1069 and 1074) are also reported
in chapter 53 of the Yuan-shih, but with considerably differing times from
those in the Wen-hsien T'ung-k'ao. For these dates, the compilers of the
 9.4 Lunar eclipse timings 305

Table 9.6b Comparison between anomalous lunar eclipse records in the Wen-
hsien T'ung-k'ao and counterparts in the Yuan-shih.

Date AD Ct Hour Mark LT(h) AT(sec) Hour Mark LT(h) AT(sec)

1069 Dec 30/31 1 hai 1 21.36 7000 hai 6 22.56 2650
M tzu 0 23.12 6050 tzu 5 0.32 1750
4 ch'ou 3 1.84 1650 ch'ou 4 2.08 800

1074 Oct 7/8 1 ch'ou 1 1.36 8150 IV w 5p 3.19 1600

2 ch'ou 6 2.56 7650 Vw 3p 4.56 500

Yuan-shih treatise evidently relied on a different source from that con-

sulted by Ma Tuan-lin. The two sets of data are compared in table 9.6b.
It should be noted that although most times are in hours and marks, the
Yuan-shih specifies night-watches (w) and points (p) in AD 1074.
In each case the Yuan-shih timings lead to values of AT which are in
good accord with the roughly contemporaneous solar data in figure 9.4.
In particular, the AT figures obtained in AD 1074 are commensurate with
the Moon setting while still eclipsed - as implied in the corresponding
Wen-hsien T'ung-k'ao account. Hence for both AD 1069 and 1074, the
Yuan-shih measurements will be retained in preference to the obviously
faulty information in the Wen-hsien Tung-k'ao.
Observations timed to the nearest mark which are reported in the Yuan-
shih (YS), Sung-shih (SS) and Wen-hsien Tung-k'ao (WH) - as selected
above - are summarised in table 9.7. (NB in the last column of this table,
SC stands for source.) With very few exceptions, all times were expressed
relative to the start of double hours rather than single hours. Of the
several eclipses for which the time of maximum phase (i.e. mid-eclipse)
is recorded, most were partial. For these, it may be confidently assumed
that the quoted time represents a direct measurement. However, three
further eclipses (in the years AD 1088, 1099 and 1277) were total and here
the reported time of mid-eclipse apparently represents an interpolation
between the measurements for second and third contact. On the first
two dates, the times of immersion and emersion are no longer preserved.
Hence it seems permissible to use the interpolated moment of mid-eclipse.
However, in AD 1277, the reported time of this phase is superfluous
since the second and third contact times are available (see also above).
NB all times of greatest phase in table 9.5 (expressed with respect to the
night-watches) relate to partial obscurations of the Moon.
Figure 9.6 is a plot of the AT values listed in table 9.7. All results are
included apart from the two highly discordant data in AD 1099 (AT =
306 9 Other East Asian observations of solar and lunar eclipses

Table 9.7 AT values from lunar eclipse timings expressed in terms of double
hours and marks.
Date AD Ct Hour Half Mark LT(h) UT(h) TT(h) AT(sec) SC
1052 Dec 8/9 1 yin — 4 4.08 20.41 20.99 2100 YS

1069 Dec 30/31 1 hai — 6 22.56 15.06 15.80 2650 YS

M tzu — 5 0.32 16.82 17.30 1750 YS
4 ch'ou — 4 2.08 18.58 18.80 800 YS

1071 Jun 15/16 4 hsu — 5 20.32 12.69 12.82 500 WH

1071 Dec 9/10 1 mao — 2 5.60 21.93 22.02 300 YS,WH

M mao — 6 6.56 22.89 23.15 950 YS,WH

1073 Apr 24/25 1 hai — 1 21.36 13.67 13.92 900 YS,WH

M hai — 6 22.56 14.87 15.26 1400 YS,WH
4 tzu — 4 0.08 16.39 16.60 800 YS,WH

1078 Jul 27/28 2 hsu — 2 19.60 12.05 12.13 300 WH

4 hai — 3.5 21.84 14.29 14.62 1200 WH

1081 May 25/26 4 hsu — 6 20.56 12.87 13.25 1350 WH

1082 Nov 8/9 1 yu 2 17.60 9.75 10.21 1650 WH

M yu — 7 18.80 10.95 11.27 1150 WH
4 hsu — 3 19.84 11.99 12.32 1200 WH

1085 Sep 6/7 1 hsu — 3 19.84 12.15 12.19 150 WH

2 hsu — 7 20.80 13.11 13.13 50 WH
4 tzu — 1 23.36 15.67 15.73 200 WH

1088 Jul 6/7 1 hai — 5 22.32 14.76 14.94 650 WH

M tzu — 6 0.56 17.00 16.75 -900 WH
4 ch'ou — 4 2.08 18.52 18.56 150 WH

1089 Jun 25/26 4 ch'ou — 6 2.56 18.97 19.27 1100 WH

1092 Apr 24/25 1 hai I 1 21.36 13.67 13.87 700 WH

2 hai 7 22.80 15.11 14.80 -1100 WH
4 tzu — 7 0.80 17.11 17.40 1050 WH

1097 Jan 30/31 M hsu — 1 19.36 12.00 12.29 1050 WH

1099 Jun 5/6 2 tzu — 3 23.84 16.17 17.58 5100 WH

3 ch'ou — 2 1.60 17.93 19.00 3850 WH
4 yin — 2 3.60 19.93 20.11 650 WH
 9.5 Azimuth measurements for lunar eclipses 307

Table 9.7 - continued.

Date AD Ct Hour Half Mark LT(h) UT(h) TT(h) AT(sec) SC
1099 Nov30/Decl 1 hai — 4 22.08 14.34 14.66 1150 WH
M tzu — 4 0.08 16.34 16.54 700 WH
4 ch'ou — 4 2.08 18.34 18.41 250 WH

1106 Jan 21/22 M yu — 3 17.84 10.47 10.53 200 WH

4 hsu — 0 19.12 11.75 12.01 950 WH

1168 Mar 25/26 3 hsu I 3 19.84 11.90 11.90 0 SS

4 hsu II 3 20.84 12.90 12.93 100 SS

1270 Apr 7/8 1 ch'ou — 3 1.84 18.08 17.72 -1300 YS

M yin — 0 3.12 19.36 19.23 -450 YS
4 yin \j 4.56 20.79 20.73 -200 YS

1272 Aug 10/11 1 ch'ou — 0 1.12 17.41 17.35 -200 YS

M ch'ou — 6 2.56 18.84 18.63 -750 YS
4 yin — 3 3.84 20.13 19.92 -750 YS

1277 May 18/19 1 tzu — 6 0.56 16.72 16.90 650 YS

2 ch'ou — 3 1.84 18.00 17.99 -50 YS
3 ch'ou — 7 2.80 18.97 19.22 900 YS
4 yin — 4 4.08 20.25 20.31 200 YS

1279 Mar 29/30 1 tzu — 5 0.32 16.60 16.72 450 YS

M ch'ou — 2 1.60 17.84 17.85 50 YS
4 ch'ou — 7 2.80 19.08 18.97 -400 YS

1279 Sep 21/22 1 ch'ou — 5 2.32 18.40 18.59 700 YS

M yin — 0 3.12 19.20 19.45 900 YS
4 yin 4 4.08 20.16 20.31 550 YS

1280 Sep 10/11 4 hsu — 1 19.36 11.50 11.79 1050 YS

5100 and 3850 sec). It seems very likely that scribal errors are responsible
for these anomalous values.

9.5 Azimuth measurements for lunar eclipses

Examples of the use of lunar azimuths in marking eclipse contacts have
already been given earlier in this chapter (e.g. section 9.2.3). Extant
records of eclipses described in this way originate exclusively from the
308 9 Other East Asian observations of solar and lunar eclipses

+3000

-1500
+1050 +1100 +1150 +1200 +1250 +1300
Year
Fig. 9.6 AT results derived from lunar eclipse times expressed in double hours
and marks.

Sui dynasty. Several predictions follow this form, but there are only four
observations: AD 590 Apr 25/26, 590 Oct 18/19, 595 Dec 22/23 and 596
Dec 10/11. The azimuth measurements will be discussed in turn below;
timings or horizon phenomena have already been considered above. All of
the following records are taken from chapter 17 of the Sui-shu. Although
some have been cited earlier in this chapter, they are assembled here for
direct reference.

(1) AD 590 Apr 25/26 (mag. = 0.69): Ta-hsing Ch'eng

K'ai-huang reign period, 10th year, 3rd month, 16th day kuei-mao [40].
According to the almanac... (various predictions)... It was observed that
when the Moon first rose from the south-east, it was already more than
half eclipsed. 3 divisions (fen) after it reached the ch'en direction (azimuth
roughly 115 deg), 2fifteenthsremained eclipsed. It gradually recovered. By
the end (wei) of the ch'en direction (azimuth approximately 127.5 deg) it
was restored to fullness.

The interpretation of the expression '3 divisions (fen) after it reached the
ch'en direction' is obscure. Only the azimuth measurement at last contact
will be considered. Since it is not clear how precise this measurement is, I
have assumed the very end of the zone.
 9.5 Azimuth measurements for lunar eclipses 309

RESULTS
Last contact when Moon at azimuth 127.5 deg. Equivalent LT = 20.72 h
(Apr 25), UT = 13.41 h. Computed TT = 15.09 h, thus AT = 6050 sec.

(2) AD 590 Oct 18/19 (mag. = 0.78): Ta-hsing Ch'eng

K'ai-huang reign period, 10th year, 9th month, 16th day keng-tzu [37].
According to the almanac... (predictions)... The observations now showed
that 2 k'o (i.e. 0.48 h) after the Moon reached the direction wu (172.5 to
187.5 deg), the eclipse began from the east side. Afterwards it moved
towards the south. When it was above the direction wei (202.5 to 217.5
deg), the southern part was four-fifths eclipsed. It gradually recovered. 1^
k'o (0.36 h) after reaching the direction shen (232.5 to 247.5 deg) it was
restored to fullness.
RESULTS
(i) First contact 0.48 h after Moon reached azimuth 172.5 deg. Equivalent
LT = 0.21 h (Oct 19), UT = 16.72 h (Oct 18). Computed TT = 18.10 h,
thus AT - 4950 sec.
(ii) Mid-eclipse when Moon at approximate azimuth 210 deg. Equivalent
LT - 0.88 h (Oct 19), UT - 17.39 h (Oct 18). Computed TT = 19.55 h,
thus AT - 7800 sec.
(iii) Last contact 0.36 h after Moon reached azimuth 232.5 deg. Equiva-
lent LT = 2.12 h (Oct 19), UT = 18.63 h (Oct 18). Computed TT = 21.00
h, thus AT = 8550 sec.

(3) AD 595 Dec 22/23 (mag. = 0.63): Ta-hsing Ch'eng

K'ai-huang reign period, 15th year, 11th month, 16th day keng-wu [7].
According to the almanac... On that night after the 4th rod (ch'ou) of the
first watch, the Moon began to be eclipsed above the direction ch'en (112.5
to 127.5 deg). It began from the south-east side. At the third rod of the
second watch, the Moon was above the direction szu (142.5 to 157.5 deg);
it was about two-thirds eclipsed. It gradually recovered. At the first rod of
the third watch, the Moon was above the direction ping (157.5 to 172.5
deg); it was restored to fullness.
RESULTS
(i) First contact when Moon at approximate azimuth 120 deg. Equivalent
LT - 22.83 h (Dec 22), UT - 15.62 h. Computed TT = 13.90 h, thus AT
~ -6200 sec.
(ii) Mid-eclipse when Moon at approximate azimuth 150 deg. Equivalent
LT - 23.63 h, UT - 16.43 h. Computed TT = 15.32 h, thus AT - -4000
sec.
(iii) Last contact when Moon at approximate azimuth 165 deg. Equiv-
alent LT - 23.86 h, UT - 16.65 h. Computed TT = 16.73 h, thus AT -
+300 sec.
310 9 Other East Asian observations of solar and lunar eclipses

(4) AD 596 Dec 10/11 (mag. = 1.76): Ta-hsing Ch'eng

K'ai-huang reign period, 16th year, 11th month, 16th day, i-ch'ou [2].
According to the almanac... Not until the first rod of the third watch was
the Moon seen in the clouds above the direction p'ing (157.5 to 172.5 deg).
It was already about 3 fifteenths eclipsed and the loss began from the east
side. Above the direction ting (187.5 to 202.5 deg), the eclipse was total.
Afterwards it reappeared from the south-east side. It was not until the
third rod of the fourth watch that it was restored to fullness, the Moon
being then at the end (wei) of the direction wei (azimuth approximately
217.5 deg).
As discussed above (section 9.4.1), I have assumed that the first obser-
vation was made close to first contact. Last contact will be taken to have
occurred when the Moon was at the very end of the azimuth zone wei.
RESULTS
(i) First contact when Moon at approximate azimuth 165 deg. Equivalent
LT - 23.72 h (Dec 10), UT - 16.42 h. Computed TT = 16.73 h, thus AT
~ 1100 sec.
(ii) Mid-eclipse when Moon at approximate azimuth 195 deg. Equivalent
LT - 0.28 h (Dec 11), UT - 16.99 h (Dec 10). Computed TT = 18.68 h,
thus AT - 6100 sec.
(iii) Last contact when Moon at azimuth 217.5 deg. Equivalent LT =
0.65 h (Dec 11), UT = 17.36 h (Dec 10). Computed TT = 20.63 h, thus
AT = 11 750 sec.

It is clear from the large scatter in the derived AT values (ranging from
—6200 sec to +11 750 sec in only 6 years) that this type of observation is
of little value in the study of the Earth's past rotation. Accordingly, the
results obtained in this section will thus not be considered further.

9.6 Estimates of solar eclipse magnitude

The oldest determination of an eclipse magnitude (as a fraction of the
solar diameter) which is preserved from China dates from 134 BC. On
this occasion it was stated that 'more than half of the Sun was obscured.
Over the next few centuries several eclipses were said to be like a hook'
(see chapter 8) - suggesting a magnitude of at least \. However, not
until AD 489 do we find the next recorded estimate of the fraction of
the Sun which was covered by the Moon. Quite possibly, several further
observations were made in the intervening time but if so all have been
lost.
Between AD 489 and 586, eight estimates of solar eclipse magnitude
are preserved, while a further observation (in AD 523) notes the degree
of obscuration of the Sun at sunrise. After AD 586 there is a gap of more
 9.6 Estimates of solar eclipse magnitude 311

than 150 years without any further magnitude estimates. For six of the
observations made between AD 489 and 586, magnitudes are expressed to
the nearest ^ of the solar diameter. Over a similar time-interval (actually
from AD 438 to 596) we find much the same situation for lunar eclipses.
Hence ^ of the diameter of the appropriate luminary seems to have been
a fairly standard unit for specifying eclipse magnitudes at this period.
No further observations of this kind - whether of the Sun or Moon
- are preserved in Chinese history between AD 597 and 767. From
this latter date until AD 1020, there are only isolated magnitude es-
timates (in AD 768, 792, 822, 927, 937 and 994), after which much
more frequent data become available. For those eclipses between AD
768 and 994, no magnitudes are expressed in fifteenths of the solar
or lunar diameter; tenths or other fractions are used instead. From
AD 1020 onwards, most recorded eclipse magnitudes in Chinese history
(whether for the Sun or Moon) are expressed in fen ('divisions') - each
equal to one-tenth of the apparent diameter of the luminary. This con-
vention continued in the Jesuit era. As pointed out in chapter 3, the
practice of quoting magnitude in fifteenths or tenths reflects an inde-
pendent tradition from that adopted in Babylon and elsewhere in the
occidental world; there the standard was the digit or twelfth part of
the disk.
Observations of eclipse magnitude represent crude eye estimates rather
than measurements, while the computed magnitude is not very responsive
to changes in AT. Hence isolated data (as between AD 768 and 994)
or groups of observations from relatively recent epochs are of little value
in the determination of AT. Attention will thus be mainly confined
to the data from the fifth and sixth centuries since they form a useful
chronological set. The very early observation in 134 BC will also be
included - mainly on account of its historical interest.
It is rarely stated whether the northern or southern limb of the Sun was
obscured. Hence in principle each individual observation should lead to
two widely spaced results for AT. Fortunately - as discussed in chapter
3 - of the two possible solutions only one is usually valid since the other
places the Sun below the horizon at the required phase.
On one occasion (AD 490 Mar 7), it was reported in the Wei-shu (chap.
105) that the Sun was ^ eclipsed. However, the text also stated that the
Sun was covered by clouds from time to time. It is thus possible that
the eclipse reached a much larger phase during the period of cloud cover
without the observers being aware of it. Since the magnitude estimate
could well lead to a misleading value for AT, the observation will be
rejected. For the remaining eclipses discussed below, there is nothing in
the text to indicate that visibility of maximal phase was prevented by
cloud - or the Sun reaching the horizon.
312 9 Other East Asian observations of solar and lunar eclipses

(1) BC 134 Aug 19: Ch'ang-an

Yuan-kuang reign period, first year, 7th month, day kuei-wei [20], the
day before the last day of the month. The Sun was eclipsed and it was 8
deg in Yi... At noon the eclipse began from the north-east and reached
more than half of the disk. At the hour of pu (= 15 h) it was restored.
[Han-shu, chap. 27c]
It will be assumed that 'more than half meant somewhere between |
and \ - a mean of 0.59.

RESULTS
For a magnitude of 0.59 near noon at Ch'ang-an, AT = 13 500 sec.
NB the above value of AT would lead to a LT of greatest phase of
12.5 h, in close accord with the text. A result for AT of —350 sec would
also lead to a magnitude of 0.59. However, here the LT of maximum
would be much later: 17.3 h. Hence only the first value for AT need be
considered.

(2) AD 489 Mar 18: P'ing-ch'eng

T'ai-ho reign period, 13th year, 2nd month, day i-hai [12], the first day
of the month. The Sun was 8fifteenthseclipsed.
[Wei-shu, chap. 105.]

RESULTS
No value of AT will produce a visible magnitude smaller than about 0.55
at P'ing-ch'eng. Hence although the observed figure of ^ (i.e. 0.53) may
well be a fair estimate, a result for AT cannot be deduced.
NB P'ing-ch'eng, near modern Ta-t'ung (= Datong) was the capital of
the Northern Wei Dynasty until AD 493, after which Lo-yang became the
imperial residence for this dynasty.

(3) AD 493 Jan 4: Chien-k'ang

Yung-ming reign period, 10th year, 12th month, day kuei-wei [20], the
first day of the month. The calculated time (for the eclipse) was half way
through the hour of wu (11-13 h). It was not until the start of the wei hour
(13-15 h) that the Sun began to be eclipsed. The loss began from the
north-west side. It was eclipsed 4 tenths. At the hour of shen (15-17 h) its
brightness and colour were fully restored.
[Nan-ch'i-shu, chap. 12.]
The generally low precision of eclipse prediction at this early period
has already been discussed (section 9.2.4).
 9.6 Estimates of solar eclipse magnitude 313

RESULTS
No value of AT can render the visible magnitude of this eclipse less than
0.44 at Chien-k'ang. Hence the observation cannot be used to yield a
value for AT.

(4) AD 494 Jun 19: Chien-k'ang

Lung-ts'ang reign period, first year, 5th month, day chia-hsu [11], the
first day of the month. The Sun was one-third eclipsed at the hour of szu
(9-11 h). It was restored at the hour of wu (11-13 h).
[Nan-ch'i-shu, chap. 12.]

RESULTS
No value of AT can render the visible magnitude of this eclipse less than
0.57 at Chien-k'ang. For values of AT which lead to maximal phase
between 9 and 11 h (as reported) the computed magnitude lies between
0.64 and 0.90. Clearly there is a scribal error in the text; the true magnitude
must have been considerably greater than the recorded figure (see also
chapter 8).

(5) AD 512 Jun 29: Lo-yang

Yen-chang reign period, first year, 5th month, day chi-wei [56], the last
day of the month. The Sun was 9fifteenthseclipsed.
[Wei-shu, chap. 105.]

RESULTS
For a computed magnitude of 0.60 at Lo-yang, either AT = 6050 or
35 000 sec.
NB for both of these values the Sun would be above the horizon at
maximum eclipse; however, the latter result is clearly discordant and will
be discarded.

(6) AD 585 Jul 31: Ta-hsing Ch'eng

K'ai-huang reign period, 5th year, 6th month, 30th day... The
observations show that the Sun did not begin to be eclipsed until more
than (shang) 6 marks after (hou) the hour of wu. The loss began from the
north-west side; it (reached a magnitude of) 6fifteenths.Not until 1 mark
after (hou) the hour of wei did it begin to reappear. At 5 marks [after the
hour of wei] it was restored to fullness.
[Sui-shu, chap. 17.]
Most renderings of the recorded times in this and the following entry
are open to conjecture (see section 9.2.1).
314 9 Other East Asian observations of solar and lunar eclipses

Table 9.8 AT values from solar

eclipse magnitude estimates recorded
in Chinese history before AD 750.
Year Est. mag. AT (sec)
-133 >\ 13 500
8
+489 15
3 450
9
+512 15
6050
6
+585 15
6200
2
+586 3
7400

RESULTS

For a computed magnitude of 0.40 at Ta-hsing Ch'eng, AT = 6200 sec.

NB no other value for AT would produce a visible eclipse of magnitude
0.40 at Ta-hsing Ch'eng.

(7) AD 586 Dec 16: Ta-hsing Ch'eng

K'ai-huang reign period, 6th year, 10th month, 30th day ting-ch'ou
[14]...It was seen during the observations that when the Sun rose 1 chang
(i.e. roughly 10 deg) above the mountains, it began to be eclipsed; (this
was) at 2 marks in the hour of ch'en. The loss began from the west; it was
two-thirds eclipsed. At 2 marks after (hou) the hour of ch'en it began to
reappear. At more than (shang) 3 marks from the start (ju) of the hour of
szu it was restored to fullness.
[Sui-shu, chap. 17.]

RESULTS

For a computed magnitude of 0.67 at Ta-hsing Ch'eng, AT = 7400 sec.

NB no other value for AT would produce a visible eclipse of magnitude
0.67 at Ta-hsing Ch'eng.

The above results are summarised in table 9.8, which lists the year (—
or +), recorded magnitude and the derived AT values.
The AT results listed in column 3 of table 9.8 are in fair accord with
one another and with the figures obtained from roughly contemporaneous
eclipse timings (see above). However, since the results are derived from
crude naked-eye estimates, and the AT values obtained from similar
observations made elsewhere (chapters 5 and 13) show a much larger
scatter, they will not be considered further.
 9.7 Sun rising or setting eclipsed 315

9.7 Sun rising or setting eclipsed

Only a very few events in this category are recorded in Chinese history
during the whole of the pre-telescopic period. Possibly many further
observations were made down the centuries but if so the details have not
been preserved. Instances in which the Sun was eclipsed at sunrise or
sunset were reported in the years 35 BC, AD 529, 532, 937 and 1042, but
after the last date there are no similar records until the Jesuit era. To
the above observations may be added that of the eclipse of 89 BC, the
account of which implies that the Sun was fully recovered before sunset.
Translations and discussions of the various records - most of which
are to be found in the astrological treatises of official histories - are given
below. Computed eclipse magnitudes at maximal phase are cited at the
head of each entry; these are for reference only.
The eclipses of both 89 and 35 BC were described as almost complete
(like a hook') and this led to their inclusion in chapter 8 above. However,
for convenience translations will be repeated here. In calculating the LTs
of sunrise, sunset, etc., the effect of horizon profile will be ignored. Apart
from Ch'ang-an (where the two earliest observations were made) all of the
capitals mentioned in this and the following two sections (namely Chien-
k'ang, Lo-yang and Pien) were located in the great plain of north-east
China. The site of Ch'ang-an (also known as Ta-hsing Ch'eng during the
Sui dynasty) is located in a more mountainous region, but the surrounding
terrain in the Wei River valley is fairly level over several tens of kilometres
- especially towards the west. At this early period, AT was so large (~
10000 sec) that minor horizon irregularities are relatively unimportant.
Estimates of the proportion of the Sun obscured at sunrise or sunset
will be considered separately (section 9.9).

(1) BC 89 Sep 29 (mag. = 0.91): Ch'ang-an

Cheng-ho reign period, 4th year, 8th month, day hsin-yu [58], the last
day of the month. The Sun was eclipsed; it was not complete and like a
hook. It was 2 deg in K'ang. At the hour of pu (= 15-17 h) the eclipse
began from the north-west. Towards the hour of sunset it was restored.
[Han-shu, chap. 27.]
It is clear from the text that last contact was observed before the Sun
reached the horizon.

RESULTS

Last contact on Sep 29 before sunset. LT of sunset = 17.99 h. Hence LT

of contact < 17.99 h, UT < 10.59 h. Computed TT = 12.61 h, thus AT
> 7250 sec.
316 9 Other East Asian observations of solar and lunar eclipses

(2) BC 35 Nov 1 (?) (mag. = 0.83): Ch'ang-an

Chien-chao reign period, 5th year, 6th month, day jen-shen [9], the last
day of the month. The Sun was eclipsed. It was not complete and like a
hook; then (yin) it set.
[Han-shu, chap. 27.]
The error in the date of this observation has already been considered
in chapter 8. There it was shown that although the recorded date is
equivalent to Aug 23 in 34 BC, the only viable date for the eclipse is Nov
1 in the previous year. If it is only assumed that the Sun set eclipsed, the
following limits (i) and (ii) may be derived.
RESULTS
(i) First contact on Nov 1 before sunset. LT of sunset = 17.41 h. Hence
LT of contact < 17.40 h, UT < 9.92 h. Computed TT = 12.00 h, thus AT
> 7600 sec.
(ii) Last contact after sunset. LT of contact > 17.40 h, UT > 9.92 h.
Computed TT = 13.94 h, thus AT < 13 500 sec.
Combining these limits yields 7600 < AT < 13 500 sec.
(iii) Alternatively, the text seems to imply that the Sun had the appear-
ance of a hook when it reached the horizon. Hence probably at least
two-thirds of the solar disk would have been obscured at sunset. On this
assumption, a value of AT in the narrower range 10000 < AT < 12000
sec may be deduced, although these limits can only be approximate.
(3) AD 529 Nov 17 (?): Lo-yang
Hung-an reign period, 2nd year, 10th month, day chi-yu [46], the first
day of the month. The Sun rose from beneath the Earth already eclipsed
by 7fifteenths.It began from the south-west side.
[Wei-shu, chap. 105.]
The recorded date cannot be correct; there was no eclipse on or near
Nov 17 in AD 529. During the second year of the Hung-an reign period
(AD 529 Jan 26 to AD 530 Jan 14) there were two solar eclipses but
neither was visible in China. Computations using equation (4.1) show
that within seven years before and after the recorded date only a single
eclipse could have been visible near sunrise at Lo-yang: AD 532 Nov 13.
This event, which is considered immediately below, also occurred in the
10th month of a year but on the day hsin-yu [58]. It is possible that this
latter eclipse was mistakenly reported twice in the same history - both
on the correct date and also three years beforehand. However, it may
be calculated that for any value of AT which renders the eclipse of AD
532 visible at sunrise at the Wei capital of Lo-yang, the magnitude at
maximum phase cannot have exceeded 0.3 there. This is significantly less
than the ^ (i.e. 0.47) said to have been obscured in the earlier year.
 9.7 Sun rising or setting eclipsed 317

(4) AD 532 Nov 13 (mag. = 0.21): Lo-yang

Tai-chang reign period, 1st year, 10th month, day hsin-yu [58], the first
day of the month. The Sun rose eclipsed from beneath the Earth. It began
from the south-west side.
[Wei-shu, chap. 105.]

In contrast to the previous entry, the recorded date of this event is

exactly correct. The reference to the eclipse beginning from the south-west
side possibly suggests that the phase was still small at sunrise. However,
no estimate of the degree of obscuration of the solar disk is given, which
may imply that only the last stages of the eclipse were visible. If so, the
position angle of first contact may have merely been interpolated. Under
these circumstances, it seems best to assume only that the Sun rose at
some time between first and last contact.

RESULTS
(i) First contact on Nov 13 before sunrise. LT of sunrise = 6.83 h. Hence
LT of contact < 6.83 h, UT < 23.14 h (Nov 12). Computed TT = 23.81
h, thus AT > 2400 sec.
(ii) Last contact after sunrise. LT of contact > 6.83 h, UT > 23.14 h
(Nov 12). Computed TT = 1.05 h (Nov 13), thus AT < 6850 sec.
Combining these limits yields 2400 < AT < 6850 sec.

(5) AD 937 Feb 14 (mag. = 0.52): Pien

T'ien-fu reign period, 2nd year, first month, day i-mao [52]. The Sun was
eclipsed... It was 3 tenths eclipsed... When the Sun rose in the east it was 3
tenths eclipsed. It gradually (re-)emerged. At the hour of mao it was
restored to fullness.
[Chiu-wU'tai-shih, chap. 139.]

The text seems to imply that the phase was already declining when the
Sun rose (i.e. between greatest phase and last contact). On the day in
question, sunrise would occur at Pien at 6.48 h - a time already towards
the end of the hour of mao.

RESULTS
(i) Maximum before sunrise. LT of sunrise = 6.48 h. Hence LT of mid-
eclipse < 6.48 h, UT < 23.12 h (Feb 13). Computed TT = 23.25 h, thus
AT > 500 sec.
(ii) Last contact after sunrise. LT of contact > 6.48 h, UT > 23.12 h
(Feb 13). Computed TT = 0.10 h (Feb 14), thus AT < 3500 sec.
Combining these limits yields 500 < AT < 3500 sec.
318 9 Other East Asian observations of solar and lunar eclipses

(6) AD 1042 Jun 20 (mag. = 0.41): Pien

Ch'ing-li reign period, 2nd year, 6th month, day jen-shen [9], the first
day of the month. The Sun was eclipsed by 5 divisions (fen). At 6 marks in
the hour of yu (18.56 h) the eclipse was 2 divisions and then the Sun set
beneath the Earth and was invisible.
[Wen-hsien Tung-k'ao, chap. 283.]
The above account clearly indicates that the Sun set after maximal
phase. Sunset would actually occur at 19.22 h, so that the recorded LT is
about 2/3 of an hour in error.

RESULTS
(i) Maximum before sunset. LT of sunset = 19.22 h. Hence LT of mid-
eclipse < 19.22 h, UT < 11.60 h. Computed TT = 11.36 h, thus AT >
-850 sec.
(ii) Last contact after sunset. LT of contact > 19.22 h, UT > 11.60 h.
Computed TT = 12.17 h, thus AT < 2050 sec.
Combining these limits yields -850 < AT < 2050 sec.

The AT limits obtained in this section are summarised in table 9.9,

along with the results from lunar eclipses to be derived in section 9.8.

9.8 Moon rising or setting eclipsed

Between AD 438 and 590, six instances where the Moon rose or set
eclipsed are noted in Chinese history. Subsequently, no similar accounts are
preserved until the late eleventh century, when a further ten observations
(between AD 1042 and 1081) are extant. After this latter date, only a single
report in this category (from AD 1168) is accessible until the Jesuit period.
The earliest records mentioned above are cited in the dynastic histories,
but those from the eleventh century are to be found in chapter 285 of the
Wen-hsien Tung-k'ao. Finally, the eclipse of AD 1168 is reported in the
Sung-shih. For all of these events - as in the case of the other eclipses
discussed in this chapter - there is no valid reason for assuming a place
of observation other than the capital of the time.
The accounts of the eclipses of AD 571 Oct 18/19, 1063 Nov 8/9, 1074
Oct 7/8, 1081 May 25/26 and 1168 Mar 25/26 all seem to imply that
the Moon rose eclipsed whilst in total shadow. As discussed in chapter
3, such observations are of dubious reliability since the totally eclipsed
Moon is probably not visible when on the horizon. Hence they will not be
considered further. An additional record from Ta-hsing Ch'eng (AD 590
Apr 25/26), which implies that the Moon rose more than half eclipsed,
will also be rejected since the eastern horizon as seen from this site is
somewhat mountainous - see, for example, entry (7) in section 9.6.
 9.8 Moon rising or setting eclipsed 319

A full translation of each of the remaining records is given below,

together with a commentary and the calculated AT limits. Although only
a few records lead to critical limits for AT, the observations are of interest
since they give a good indication of the variety of lunar eclipse reports
which are preserved in Chinese history. NB as emphasised in chapter 3,
the computed LT of moonrise or moonset for any given place and date is
a function of AT.
Estimates of the fraction of the Moon eclipsed at moonrise or moonset
will be discussed separately in section 9.9.

(1) AD 438 Jun 23/24 (mag. = 1.13): Chien-k'ang

Yuan-chia reign period, 15th year, 5th month 15th day, full Moon. When
the Moon first rose it was already eclipsed. Its brightness had already
regained one-quarter (of itself). It was 16 deg in (Nan-)tou (lunar lodge).
[Sung-shu, 12.]
Since the eclipse was actually total, it seems clear that by the time that
the Moon rose, the phase was declining.

RESULTS
(1) Third contact before moonrise on Jun 23. LT of moonrise = 19.13 h.
Hence LT of contact < 19.13 h, U T < 11.19 h. Computed TT = 12.53 h,
thus A T > 4800 sec.
(ii) Last contact after moonrise. LT of moonrise = 19.19 h. Hence LT
of contact > 19.19, U T > 11.25 h. Computed TT = 13.71 h, thus AT <
8850 sec.
Combining these limits yields 4800 < A T < 8850 sec.

(2) AD 503 Jun 25/26 (mag. = 1.08): Lo-yang

Ching-ming reign period, 4th year, 5th month, day ting-mao [4]. The
Moon was in (Nan-)tou (lunar lodge). It rose eclipsed from beneath the
Earth. It was eclipsed 12 fifteenths.
[Wei-shu, 105.]
There is no mention of this eclipse being total, suggesting that the Moon
rose after the end of totality.

RESULTS
(i) Third contact before moonrise on Jun 25. LT of moonrise = 19.20 h.
Hence LT of contact < 19.20 h, UT < 11.69 h. Computed TT = 13.08 h,
thus AT > 5000 sec.
(ii) Last contact after moonrise. LT of moonrise = 19.23 h. Hence LT
of contact > 19.23, UT > 11.73 h. Computed TT = 14.45 h, thus AT <
9800 sec.
Combining these limits yields 5000 < AT < 9800 sec.
320 9 Other East Asian observations of solar and lunar eclipses

(3) AD 513 Jun 4/5 (mag. = 1.35): Lo-yang

Yen-ch'ang reign period, 2nd year, 4th month, day chi-hai [36]. The Sun
was in Tsui (lunar lodge) and the Moon in Wei. It (i.e. the Moon) rose
from beneath the Earth with 3fifteenthsremaining (eclipsed). It gradually
became full.
[Wei-shu, chap. 105.]
The recorded date actually corresponds to Jun 25, but taking the cyclical
day as i-hai [16] rather chi-hai gives Jun 5 instead. The written characters
i and chi are very similar and confusion between them is fairly common.
Tsui (the 20th lunar lodge) and Wei (the 6th lodge) were roughly 180
deg apart. Since the eclipse was actually total, the phase evidently was
already waning at moonrise.
RESULTS
(i) Third contact before moonrise on Jun 4. LT of moonrise = 19.17 h.
Hence LT of contact < 19.17 h, UT < 11.59 h. Computed TT = 12.22 h,
thus AT > 2250 sec.
(ii) Last contact after moonrise. LT of moonrise = 19.22 h. Hence LT
of contact > 19.22, UT > 11.65 h. Computed TT = 13.40 h, thus AT <
6300 sec.
Combining these limits yields 2250 < AT < 6300 sec.
(4) AD 514 May 24/25 (mag. = 0.92): Lo-yang
Yen-ch'ang reign period, 3rd year, 4th month, day kuei-szu [30]. The
Moon was in Wei (lunar lodge). It rose eclipsed from beneath the Earth. It
was eclipsed 14 fifteenths.
[Wei-shu, 105.]
The observed degree of obscuration (0.93) is very close to the calculated
magnitude. Comparing this entry with others in the same source (see
above) suggests that on this occasion moonrise occurred around the time
of maximal phase. However, it is also possible that the Moon rose only
slightly eclipsed and eventually reached a magnitude of 14/15. Because
of this uncertainty, it seems best to assume only that moonrise took place
between first contact and approximately maximum eclipse.
RESULTS
(i) First contact before moonrise on May 24. LT of moonrise = 19.02 h.
Hence LT of contact < 19.02 h, UT < 11.41 h. computed TT = 11.35 h,
thus AT > -250 sec.
(ii) Maximum phase after moonrise. LT of moonrise = 19.09 h. Hence
LT of contact > 19.09 UT > 11.48 h. Computed TT = 13.02 h, thus AT
< 5500 sec.
 9.8 Moon rising or setting eclipsed 321

Combining these limits yields -250 < AT < 5500 sec.

Obviously the upper limit can only be very rough.

(5) AD 1042 Jul 5/6 (mag. = 1.11): Pien

Ch'ing-li reign period, 2nd year, 6th month, day ting-hai [24]. The Moon
was eclipsed by 6 divisions. Then it set eclipsed and the restoration was
not seen.
[Wen-hsien T'ung-k'ao, chap. 285.]

This eclipse was total, and it seems clear from the text that the Moon
set before the total phase had been reached.

RESULTS

(i) First contact before moonset on Jul 6. LT of moonset = 4.80 h. Hence

LT of contact < 4.80 h, UT < 21.23 h (Jul 5). Computed TT = 20.76 h,
thus AT > -1700 sec.
(ii) Second contact after moonset. LT of moonset = 4.87 h. Hence LT
of contact > 4.87 h, UT > 21.30 h (Jul 5). Computed TT = 22.02 h, thus
AT < 2600 sec.
Combining these limits yields -1700 < AT < 2600 sec.

(6) AD 1067 Mar 3/4 (mag. = 1.43): Pien

Chih-ping reign period, 4th year, 2nd month, day chia-wu [31]. The
Moon was eclipsed. It began to be seen at 4 marks in the hour of ch'ou in
the west. It was was 15 deg in / (lunar lodge). At 6 marks in the hour of
ch'ou the eclipse reached its maximum. It was more than 8 divisions. It set
eclipsed in the direction yu and was not seen.

[Wen-hsien T'ung-k'ao, chap. 285.]

As in the previous example, it is apparent from the text that the Moon
set before the onset of totality.

RESULTS

(i) First contact before moonset on Mar 4. LT of moonset = 6.15 h. Hence

LT of contact < 6.15 h, UT < 22.71 h (Mar 3). Computed TT = 21.84 h,
thus AT > -3150 sec.
(ii) Second contact after moonset. LT of moonset = 6.17 h. Hence LT
of contact > 6.17 h, UT > 22.73 h (Mar 3). Computed TT = 22.94 h,
thus AT < 750 sec.
Combining these limits yields —3150 < AT < 750 sec.
322 9 Other East Asian observations of solar and lunar eclipses

(7) AD 1068 Aug 15/16 (mag. = 038): Pien

Hsi-ning reign period, first year, 7th month, day i-yu [22]. The Moon
was eclipsed at 5 marks in the hour of ch'ou. It was 10 deg in Wei (lunar
lodge). It began to be seen at the north-eastern side and was eclipsed by
2 \ divisions. Then it set eclipsed and was not seen.
[Wen-hsien T'ung-k'ao, chap. 285.]
The estimated fraction of the lunar diameter covered at moonset is
significantly less than the computed magnitude at greatest phase. Hence it
may be inferred that the Moon set before the eclipse reached its maximum.
RESULTS
(i) First contact before moonset on Aug 16. LT of moonset = 5.32 h.
Hence LT of contact < 5.32 h, UT < 21.73 h (Aug 15). Computed TT =
21.67 h, thus AT > -200 sec.
(ii) Maximal phase after moonset. LT of moonset = 5.38 h. Hence LT
of contact > 5.38 h, UT > 21.79 h (Aug 15). Computed TT = 22.73 h,
thus AT < 3400 sec.
Combining these limits yields -200 < AT < 3400 sec.
(8) AD 1071 Jun 15/16 (mag. = 0.95): Pien
Hsi-ning reign period, 4th year, 5th month, day chi-hai [36]. The Moon
was eclipsed. At 1 mark in the hour of hsu it began to be seen at the
south-east. It rose eclipsed; it was not round but 6 divisions (were
covered). It was within the degrees of Tung-ching (lunar lodge). After 5
marks it was restored.
[Wen-hsien T'ung-k'ao, chap. 285.]
At the time, the Sun (not the Moon) would be in Tung-ching, which is
situated in Gemini. It seems clear from the text that the Moon rose when
the phase was declining. For example, there is no mention of any increase
in phase despite the fact that the estimated obscuration at moonrise (0.60)
is much less than the maximum (0.95). Furthermore, the reported period
of visibility after moonrise (5 marks or 1.2 h) is only about one-third of
the computed total duration.
RESULTS
(i) Maximum phase before moonrise on Jun 15. LT of moonrise = 19.22
h. Hence LT of contact < 19.22 h, UT < 11.59 h. Computed TT = 11.09
h, thus AT > -1800 sec.
(ii) Last contact after moonrise. LT of moonrise = 19.29 h. Hence LT
of contact > 19.29, UT > 11.66 h. Computed TT = 12.82 h, thus AT <
4200 sec.
Combining these limits yields -1800 < AT < 4200 sec.
 9.8 Moon rising or setting eclipsed 323

(9) AD 1071 Dec 9/10 (mag. = 0.43): Pien

Hsi-ning reign period, 4th year, 11th month, day ping-shen [33]. The
Moon was eclipsed. At 2 marks in the hour of mao (the eclipse) began to
be seen in the west. It commenced at the south-east side. Not until 6
marks in the hour of mao was the eclipse at its maximum, reaching less
than 4^ divisions. It was rather less than 1 deg in Tung-ching (lunar lodge).
Before it became bright it set eclipsed and the restoration was not seen.
[Wen-hsien Tung-k'ao, chap. 285.]
The estimated fraction of the Moon in shadow (0.45) is very close to
the computed magnitude. Nevertheless, it is apparent from the text that
the eclipse was declining when the Moon set.

RESULTS

(i) Maximum before moonset on Dec 10. LT of moonset = 7.17 h. Hence

LT of contact < 7.17 h, UT < 23.51 h (Dec 9). Computed TT = 23.15 h,
thus AT > -1150 sec.
(ii) Last contact after moonset. LT of moonset = 7.22 h. Hence LT of
contact > 7.22 UT > 23.56 h (Dec 9). Computed TT = 0.28 h (Dec 10),
thus AT < 2600 sec.
Combining these limits yields — 1 1 5 0 < A T < 2600 sec.

(10) AD 1073 Oct 18/19 (mag. = 0.56): Pien

Hsi-ning reign period, 6th year, 9th month, day i-mao [52]. The Moon
was eclipsed. At 4 marks in the hour of ch'ou the eclipse began to be seen
at the north-east side. At 1 mark in the hour of yin it was at its maximum,
reaching 6 divisions. Then it began to be restored. When it was 3 divisions
short of roundness it set eclipsed and the restoration was not seen.
[Wen-hsien T'ung-k'ao, chap. 285.]
It is clear that the Moon set between greatest phase and last contact.

RESULTS

(i) Maximum before moonset on Oct 19. LT of moonset = 6.54 h. Hence

LT of contact < 6.54 h, UT < 22.67 h (Oct 18). Computed TT = 22.16 h,
thus AT > -1850 sec.
(ii) Last contact after moonset. LT of moonset = 6.59 h. Hence LT of
contact > 6.59 h, UT > 22.72 h (Oct 18). Computed TT = 23.58 h, thus
AT < 3100 sec.
Combining these limits yields — 1 8 5 0 < A T < 3 1 0 0 sec.
324 9 Other East Asian observations of solar and lunar eclipses

Table 9.9 AT limits from horizon eclipses.

AT Range (sec)
Year* Type* Rise/set c UI/
-88 Sun rise 7250 —
-34 Sun set -10000 -12000

+438 Moon rise 4800 8 850

+503 Moon rise 5000 9 800
+513 Moon rise 2250 6 300
+514 Moon rise -250 - 5 500
+532 Sun rise 2400 6 850

+937 Sun rise 500 3 500

1042 Sun set -850 2050
1042 Moon rise -1700 2 600
1067 Moon set -3150 750
1068 Moon set -200 3 400

1071a Moon rise -1800 4200

1071b Moon set -1150 2600
1073 Moon set -1850 3100
1081 Moon rise -850 2 750
a
The year ( - or +).
b
Whether the eclipse was of the Sun or Moon.
c
Whether the eclipse occurred near the rising or setting of the appropriate
luminary.
d
The derived lower limit to AT in sec (LL).
e
The derived upper limit to AT in sec (UL).

(11) AD 1081 Nov 19/20 (mag. = 1.63): Pien

Yuan-feng reign period, 4th year, 10th month, day hsin-szu [18]. The
Moon was eclipsed. At dusk it rose eclipsed by 7 divisions. It was in the
degrees of Pi (lunar lodge). After 2 marks it was restored.
[Wen-hsien Tung-k'ao, chap. 285.]
The Moon evidently rose considerably after the end of totality.

RESULTS

(i) Third contact before moonrise on Nov 19. LT of moonrise = 17.04 h.

Hence LT of contact < 17.04 h, UT < 9.23 h. Computed TT = 8.99 h,
thus AT > - 8 5 0 sec.
(ii) Last contact after moonrise. LT of moonrise = 17.07 h. Hence LT
 9.9 Estimates of the proportion of the Sun or Moon obscured 325

1 1 1 1 1 1 I 1 I 1 1
_ V
+12000 Key -

Upper ATlimit V
+10000 - /s V
Lower ATlimit /\ -

>•

+8000
\s
V
_ +6000 V
-

o
CD
fr +4000 --
X,
< V _
-

+2000 V

/\
0

-2000
1 1 1 1 1 1 1 i i i 1
-4000
-100 0 + 1 0 0 +200 +300+400 +500 +600 +700 +800 +900+1000+1100
Year
Fig. 9.7 AT limits obtained from sunrise/sunset and moonrise/moonset eclipse
observations.

of contact > 17.07 h, UT > 9.26 h. Computed TT = 10.03 h, thus AT <

2750 sec.
Combining these limits yields —850 < AT < 2750 sec.

The above results, along with those obtained from the solar obscurations
discussed in section 9.7, are listed in table 9.9.
The above AT limits are shown in figure 9.7. Although most limits are
clearly redundant, a few others are more critical. However, it should be
noted that the upper limit to AT set by the observation in +514 (5500
sec) is only approximate.

9.9 Estimates of the proportion of the Sun or Moon obscured at its rising
or setting
In all of the following cases, the fact that the Sun or Moon rose or
set whilst eclipsed has already been considered in sections 9.7 or 9.8.
Hence only abbreviated translations will be cited below. In each case
the magnitude at greatest phase is given for reference. As discussed in
chapter 3 (see especially figure 3.23), corrections to the proportion of the
lunar diameter covered at moonrise or moonset will be made in order
to allow for systematic errors in estimating magnitude; these are most
significant for small eclipses. No such amendment will, however, be made
for the two solar observations.
326 9 Other East Asian observations of solar and lunar eclipses

(1) AD 438 Jun 23/24 (Moon: mag. = 1.13): Chien-k'ang

Yuan-chia reign period, 15th year, 5th month 15th day, full Moon.
When the Moon first rose it was already eclipsed. Its brightness had
already regained one-quarter (of itself)...
[Sung-shu, 12.]
The (adjusted) proportion of the lunar diameter covered at moonrise
will be taken as 0.78 - rather than 0.75.

RESULTS
LT of moonrise on Jun 23 = 19.18 h. For 0.78 of lunar diameter to be
obscured at moonrise (phase decreasing), AT = 5800 sec.

(2) AD 503 Jun 25/26 (Moon: mag. = 1.08): Lo-yang

Ching-ming reign period, 4th year, 5th month, day ting-mao [4]. The
Moon... rose eclipsed from beneath the Earth. It was eclipsed 12 fifteenths.
[Wei-shu, 105.]
Since there is no mention of this eclipse being total, the Moon apparently
rose after the end of totality. The (adjusted) proportion of the lunar
diameter covered at moonrise will be taken as 0.84.

RESULTS
LT of moonrise on Jun 25 = 19.23 h. For 0.84 of lunar diameter to be
obscured at moonrise (phase decreasing), AT = 6000 sec.

(3) AD 513 Jun 4/5 (Moon: mag. = 1.35): Lo-yang

Yen-ch'ang reign period, 2nd year, 4th month, day chi-hai [36]...It (i.e.
the Moon) rose from beneath the Earth with 3fifteenthsremaining
(eclipsed). It gradually became full.
[Wei-shu, chap. 105.]
As the eclipse was total, when the Moon rose the phase evidently was
already on the decline. The (adjusted) proportion of the lunar diameter
covered at moonrise will be taken as 0.12.

RESULTS
LT of moonrise on Jun 4 = 19.24 h. For 0.12 of lunar diameter to be
obscured at moonrise (phase decreasing), AT = 5750 sec.

(4) AD 514 May 24/25 (Moon: mag. = 0.92): Lo-yang

Yen-ch'ang reign period, 3rd year, 4th month, day kuei-szu [30]. The
Moon... rose eclipsed from beneath the Earth. It was eclipsed 14 fifteenths.
[Wei-shu, 105.]
 9.9 Estimates of the proportion of the Sun or Moon obscured 327

RESULTS
The observed phase (0.93: adjusted value approximately 0.98) is very close
to - and actually exceeds - the computed magnitude. Hence it is not
possible to derive a useful value for AT from this observation.
(5) AD 937 Feb 14 (Sun: mag. = 0.52): Pien
T'ien-fu reign period, 2nd year, 1st month, day i-mao [52]...When the
Sun rose in the east it was 3 tenths eclipsed. It gradually (re-)emerged...
[Chiu-wu-tai-shih, chap. 139.]

RESULTS
LT of sunrise on Feb 14 = 6.50 h. For 0.30 of solar diameter to be
obscured at sunrise (phase decreasing), AT = 2100 sec.

(6) AD 1042 Jun 20 (Sun: mag. = 0.41): Pien

Ch'ing-li reign period, 2nd year, 6th month, day jen-shen [9], the first
day of the month. The Sun was eclipsed by 5 divisions. At 6 marks in the
hour of yu (18.56 h) the eclipse was 2 divisions (i.e. 2 tenths) and then the
Sun set...
[Wen-hsien T'ung-k'ao, chap. 283.]

RESULTS
LT of sunset on Jun 20 = 19.24 h. For 0.20 of solar diameter to be
obscured at sunset (phase decreasing), AT = 1200 sec.
(7) AD 1042 Jul 5/6 (Moon: mag. = 1.11): Pien
Ch'ing-li reign period, 2nd year, 6th month, day ting-hai [24]. The Moon
was eclipsed by 6 divisions (i.e. 6 tenths). Then it set eclipsed ...
[Wen-hsien Tung-k'ao, chap. 285.]
The proportion of the lunar diameter covered at moonset will be taken
as 0.60. From figure 3.23, negligible amendment is necessary at this
particular magnitude.

RESULTS
LT of moonset on Jul 6 = 4.81 h. For 0.60 of lunar diameter to be
obscured at moonset (phase increasing), AT = 700 sec.
(8) AD 1067 Mar 3/4 (Moon: mag. = 1.43): Pien
Chih-ping reign period, 4th year, 2nd month, day chia-wu [31]. The
Moon was eclipsed... at 4 marks in the hour of ch'ou... It was more than 8
divisions (i.e. 8 tenths) at 6 marks in the hour of ch'ou... It set eclipsed...
[Wen-hsien Tung-k'ao, chap. 285.]
The (adjusted) proportion of the lunar diameter covered at moonset
will be taken as 0.83.
328 9 Other East Asian observations of solar and lunar eclipses

RESULTS
LT of moonset on Mar 4 = 6.14 h. For 0.83 of lunar diameter to be
obscured at moonset (phase increasing), AT = 100 sec.

(9) AD 1068 Aug 15/16 (Moon: mag. = 038): Pien

Hsi-ning reign period, first year, 7th month, day i-yu [22]. The Moon
was eclipsed at 5 marks in the hour of ch'ou... It was eclipsed by 2\
divisions (i.e. l\ tenths)...Then it set eclipsed...
[Wen-hsien Tung-k'ao, chap. 285.]
The (adjusted) proportion of the lunar diameter covered at moonset
will be taken as 0.18.

RESULTS
LT of moonset on Aug 16 = 5.31 h. For 0.18 of lunar diameter to be
obscured at moonset (phase increasing), AT = 950 sec.

(10) AD 1071 Jun 15/16 (Moon: mag. = 0.95): Pien

Hsi-ning reign period, 4th year, 5th month, day chi-hai [36]. The
Moon...rose eclipsed; it was not round but 6 divisions (i.e. tenths) (were
covered)... After 5 marks it was restored.
[Wen-hsien T'ung-k'ao, chap. 285.]
The (adjusted) proportion of the lunar diameter covered at moonrise
will be taken as 0.60.

RESULTS
LT of moonrise on Jun 15 = 19.29 h. For 0.60 of lunar diameter to be
obscured at moonrise (phase decreasing), AT = 1300 sec.

(11) AD 1071 Dec 9/10 (Moon: mag. = 0.43): Pien

Hsi-ning reign period, 4th year, 11th month, day ping-shen [33]. The
Moon was eclipsed... Not until 6 marks in the hour of mao was the eclipse
at its maximum, reaching less than 4^ divisions... Before it became bright
it set eclipsed...
[Wen-hsien T'ung-k'ao, chap. 285.]

RESULTS
The estimated phase (0.45: adjusted value 0.42) is so close to the computed
magnitude that it is not possible to derive a useful value for AT from
this observation. Even quite large changes in AT would produce little
alteration in the computed phase at moonset.
 9.10 Other records of lunar eclipses 329

(12) AD 1073 Oct 18/19 (Moon: mag. = 0.56): Pien

Hsi-ning reign period, 6th year, 9th month, day i-mao [52]. The Moon
was eclipsed... At 1 mark in the hour of yin it was at its maximum,
reaching 6 divisions. Then it began to be restored. When it was 3 divisions
(i.e. tenths) short of roundness, it set eclipsed...
[Wen-hsien T'ung-k'ao, chap. 285.]
The (adjusted) proportion of the lunar diameter covered at moonset
will be taken as 0.24.

RESULTS
LT of moonset on Oct 19 = 6.55 h. For 0.24 of lunar diameter to be
obscured at moonset (phase decreasing), AT = 1850 sec.

(13) AD 1081 Nov 19/20 (mag. = 1.63): Pien

Yuan-feng reign period, 4th year, 10th month, day hsinszu [18]. The
Moon... rose eclipsed by 7 divisions (i.e tenths)... After 2 marks it was
restored.
[Wen-hsien Tung-k'ao, chap. 285.]
The (adjusted) proportion of the lunar diameter covered at moonrise
will be taken as 0.72.

RESULTS
LT of moonrise on Nov 19 = 17.07 h. For 0.72 of lunar diameter to be
obscured at moonrise (phase decreasing), AT = 100 sec.
Table 9.10 summarises the AT values obtained in this section.

The results in table 9.10 are plotted in figure 9.8. The range of indi-
vidual values is best illustrated by the set of observations in the narrow
interval between AD 1042 and 1081 (i.e. AT between 100 and 1850 sec).
As comparison with figure 9.6 shows, the scatter is similar to that for
contemporaneous timed data. This diagram also includes the few AT
results obtained in section 9.6 from estimates of solar eclipse magnitude
at greatest phase (see table 9.8).

9.10 Other records of lunar eclipses

At least six lunar eclipses are known to be alluded to on the oracle bones
of the Shang dynasty. However, as yet there is no general agreement
on the dates of these events, largely because the extant inscriptions are
incomplete.
Xu et a\. (1989) and Xu et al. (1995) give translations of all the known
Shang records of lunar eclipses. From these I have selected the following
330 9 Other East Asian observations of solar and lunar eclipses

Table 9.10 AT values derived from eclipses in which the phase

was estimated on the horizon.
Yearfl Type* Rise/set.c Mag. est/ AT (sec)*
3
+438 Moon rise 4 5800
12
+503 Moon rise 15 6000
3
+513 Moon rise 15
5750
3
+937 Sun rise 10 2100

2
1042 Sun set 10 1200
6
1042 Moon set 10 700
8
1067 Moon set 10 100
2.5
1068 Moon set 950
To
6
1071 Moon rise 10 1300
3
1073 Moon set 10 1850
7
1081 Moon set 10 100
a
Year.
b
Type of eclipse (Sun or Moon).
c
Whether the eclipsed object was rising or setting.
d
Original estimate of phase.
e
Computed value of AT.

two examples:
(i) The divination on the day kuei-wei [20] was performed by Cheng:
'Will there be no disaster in the next ten days?' On the third day
i-yu [22], at night, an eclipse of the Moon was reported. In the 8th
month.

(ii) Day jen-shen [9], at night, the Moon was eclipsed.

The first of these texts is the only surviving Shang record of an eclipse
which gives both the cyclical day and the month. As many inscriptions
of a non-astronomical nature reveal, it was the custom to place the lunar
month at the end of a text. Unfortunately little is known about how the
calendar was regulated at this early period or at what stage during each
night the date was changed. Hence dating of this and other Shang eclipse
observations is problematical. For instance, Tung Tso-pin (1945) derived
a series of dates between 1362 and 1218 BC, while Dubs (1947) preferred
an interval roughly 150 years later. Recently, on the assumption that all
of the known records originate from around the same time, Pang et al.
(1988b) deduced a set of dates between about 1100 and 1080 BC.
 9.10 Other records of lunar eclipses 331

+8000 1 1 1 1 1 1
o
+7000 - Key -
Horizon estimate of phase *
o
+6000
-. t Solar magnitude estimate o
-

+5000

+4000 - -
o
+3000 -

+2000 - *

* *
+1000 - * -
*

1 1 1 i i i **
+400 +500 +600 +700 +800 +900 +1000 +1100
Year
Fig. 9.8 AT values derived from estimates of eclipsed proportion of Sun and
Moon when on horizon and solar eclipse magnitudes.

Since an unambiguous date has yet to be derived for any of the Shang
lunar eclipse texts, and individual accounts give little or no information on
the time of night when an observation was made, the various inscriptions
are at present of negligible value in the study of the Earth's past rotation.
Further research on this issue would seem very desirable.
Although there are numerous references to total lunar eclipses through-
out Chinese history (mainly using the character chi), allusions to the colour
of the Moon when completely obscured are very rare. The earliest identifi-
able observation dates from AD 499. At the total lunar eclipse of Sep 5/6
in that year, the Astrological Treatise (chap. 12) of the Nan-ch'i-shu noted
that 'its colour was all red'. The same source, in describing an eclipse of
the Moon alleged to occur on a date corresponding to AD 498 May 18
stated that 'its colour was red like blood'. There was no eclipse on or
near the stated day and the true date cannot be isolated. No comparable
descriptions are found in subsequent Chinese history until the sixteenth
century. The allusion to blood in AD 498 is particularly interesting since
numerous medieval European accounts of the totally eclipsed Moon make
similar comparisons (see chapter 11).
Among the many occultations of stars and planets noted in Chinese
history, only one (in AD 712) was said to have involved the eclipsed Moon.
This is recorded in the Astrological Treatise (chap. 36) of the Chiu-t'ang-
332 9 Other East Asian observations of solar and lunar eclipses

12000 - V Key o Solar

• Lunar
11000
V ATupper limit
10000
/\ ATlower limit
9000
8000 -
7000 • •
g 6000 -
2 5000
4000 -
3000
; 'I
2000 -
• °°
1000 - o
1 1 1 1 1 1 I
0
-100 0 +100 +200 +300 +400 +500 +600 +700
-1000 Year
Fig. 9.9 AT results and more critical limits obtained in this chapter from Chinese
observations of solar and lunar eclipses between the dates —100 and +750.

Key o Solar
+3000 • Lunar
o o
• o V ATupper limit
+2500 _
o /\ A7lower limit
\y •
+2000 0
o #
0
+1500 o o
o
£ +1000 •
°o ° o •
>
+500
° o o
o • • \
# ^
ti
-500
#
-1000
i i i
1000 1050 1100 1150 1200 1250
Year

Fig. 9.10 AT results and more critical limits obtained in this chapter from
Chinese observations of solar and lunar eclipses between +1000 and +1300.
 9.11 Conclusion 333

shu. The following observation is described on a date corresponding to

AD 712 Sep 19/20:
Hsien-t'ien reign period, first year, 8th month, 14th day. At night the
Moon was eclipsed and it was complete. A star entered (ju) within the
spirit (p'o) of the Moon.

Evidently the 'spirit' of the Moon was its dimly lit disk during the total
phase (mag. = 1.39). It would seem unlikely that the 'star' was a meteor;
there was a special term for such a rapidly moving object - liu-hsing
('flowing star'). The term ju ('to enter') does not necessarily imply any
movement of the occulted object, only apparent motion relative to the
Moon. This was one of several standard terms to describe the occultation
of a star or planet by the Moon.
The longitude and latitude of the Moon (corrected for parallax) at
the middle of the total phase were respectively 356.4 deg and —0.2 deg.
Correcting for precession and converting to RA and dec, the approximate
celestial co-ordinates at the epoch J2000 are 13.24 h, +5.5 deg. There was
no planet in the vicinity of the eclipsed Moon on the night in question,
although the rather faint star o Vir (mag. +4.8) was very close. This was
presumably the star alluded to in the text.

9.11 Conclusion
Chinese records are obviously a prolific source of valuable observations
of both solar and lunar eclipses. The various AT results and more critical
limits obtained in this chapter (apart from the azimuth data and solar
magnitudes which are discarded) are plotted in figures 9.9 and 9.10. Figure
9.9 covers the date range from -100 to +750 and figure 9.10 from +1000
to +1300. There are no useful observations in the intervening interval
(actually from +703 to +936).
 10
Records of eclipses in
ancient European history

10.1 Introduction
Compared with the careful observations of similar age which are recorded
on the Late Babylonian astronomical texts, many of the eclipse records in
ancient Greek and Roman history come as something of an anticlimax.
Although numerous descriptions of both solar and lunar obscurations are
preserved in these sources, commencing as early as the seventh century BC,
most accounts are too vague to be suitable for investigating the Earth's
past rotation. The majority of writings which mention eclipses are literary
rather than technical, and include historical works, biographies and even
poems. Late nineteenth and early twentieth century authors such as Nevill
(e.g. 1906a), Ginzel (1899), Cowell (e.g. 1906b), and Fotheringham (e.g.
1920b) paid much attention to these observations. However, this was
largely because little other material was available at the time.
As noted in chapter 3, the mainstay of investigations made around the
beginning of the present century was undoubtedly untimed observations
of large solar eclipses. Attempts to date the various records and identify
the places of observation proved an almost irresistible challenge to Fother-
ingham and his contemporaries, and much effort was expended in these
pursuits. Considerable interest was also shown in using ancient eclipses to
date historical events. (For a recent summary, see Stephenson, 1993.)
There seems little doubt that many records in Greek and Roman history
relate to eclipses which were either total or fell not much short of this
phase. Thus Fotheringham (1908) noted that although total obscurations
of the Sun are slightly less frequent than their annular counterparts on the
Earth's surface as a whole, the majority of the solar eclipses mentioned
in the Greek and Roman classics were in fact generally total. The
implications are that where these events were observed, most exceeded the
magnitude of a typical central annular eclipse (i.e. greater than about 0.95).
However, such a statistical conclusion, although interesting, is insufficient
to allow useful information on AT to be obtained from individual records.

334
 10.1 Introduction 335

The most valuable observations for this purpose are those which either
affirm that the central phase (whether annular or total) was seen or
expressly deny this - by stating that the Sun was reduced to a crescent.
Careful timings of the contacts for only a single solar eclipse - dating
from AD 364 - are extant from ancient Europe, and indeed there are no
further examples from this part of the world until as late as the fourteenth
century (see chapter 11). Although lunar eclipses are also recorded in
abundance in ancient European literature, most are of no value in the
determination of AT since they merely note the occurrence of an eclipse.
Several Greek timings dating from between 201 BC and AD 125, and also
one instance of the Moon rising partially obscured, are reported in a single
source: Ptolemy's Almagest. In addition to this material, an observation
of the Moon rising eclipsed in 331 BC quoted by Pliny is also worth
considering.
In this chapter, ancient European records of solar and lunar eclipses
which (at least in principle) are of value for studying Earth's past rotation
will be considered under the following categories:
(i) Solar eclipses analysed by Fotheringham in his 1920b paper (section
10.2).
(ii) Other descriptions of large solar eclipses in which such details as
darkness or the visibility of stars is noted (10.3).
(hi) Timed lunar eclipses reported in the Almagest (10.4).
(iv) The moonrise eclipse of 331 BC (10.5).
The geographical co-ordinates of the principal places mentioned in this
chapter are listed in table 10.1.
Figure 10.1 shows the remarkably self-consistent AT results obtained
from the Late Babylonian timings which were assigned double weight in
chapters 5 and 6. These observations are mainly of lunar eclipses, along
with a few solar events. The abcissa is graduated in negative and positive
years, rather than their BC equivalents; on the scale of the diagram, any
differences are negligible. Also included in this diagram is the range in
AT set by the highly reliable observation of a total solar eclipse in 136
BC. The linear fit to the timed data (a useful first approximation) has the
equation
AT = 9520-1612T, (10.1)
where T is measured in centuries from the epoch 0 (i.e. 1 BC). This
equation is denoted by a dashed line in figure 10.1. Where appropriate,
both this diagram and equation (10.1) will be utilised to help establish
dates of the ancient European eclipses discussed in this chapter.
336 10 Records of eclipses in ancient European history

Table 10.1 Principal sites at which eclipses were

recorded in ancient Europe.

Place Latitude (deg) Longitude (deg)

Alexandria +31.22 -29.92
Athens +37.98 -23.73
Chaeroneia +38.40 -22.90
Palermo +38.13 -13.30
Paros +37.07 -25.10

Rhodes +36.43 -28.23

Rome +41.90 -12.50
Sparta +37.08 -22.42
Syracuse +37.07 -15.30
Thasos +40.77 -24.70

Thebes +38.32 -23.32

I "1 1 1 1

Key
^_ •
• Lunar eclipse times
20000- ^ - . -
~~ ~- ^ • ° Solar eclipse times
I Total solar eclipse

15000-
o • - ^ •

10000- -

5000
-700 -600 -500 -400 -300 -200 -100 +100
Year
Fig. 10.1 Best fitting straight line to the AT results obtained from the Late
Babylonian timings which were assigned double weight in chapters 5 and 6.

10.2 Solar eclipses analysed by Fotheringham

Although the series of observations which Fotheringham discussed in his
1920b paper is probably no more reliable than any of the other ancient
data investigated in this chapter, it has attracted such attention among
geophysicists - notably Munk and MacDonald (1960, pp. 188-191) and
Dicke (1966) - that it seems appropriate to consider it first.
Fotheringham analysed ten observations of either real or supposed
central eclipses of the Sun together with timings of the various phases of a
 10.2 Solar eclipses analysed by Fotheringham 337

2.4
/
2.2 ^ - ,
2.0

1.8
^ /
1.6 /
s// A

c
w
1.4

1.2
[ If $ \ / yyj^ •
4^K
1.0 :
/ ^
0.8

0.6"
\\
0.4, \
0.2'
\ ^
' "9.2"9A 9.6 9.8' 10.0 10.2 10.4 10.6 "10.8 11.0 11.2 11.4 11.6 11.8 12.0
Moon
Fig. 10.2 Fotheringham's (1920b) diagram showing the solar and lunar acceler-
ations on UT derived from ancient eclipses.

single partial obscuration. Most of these are recorded in Greek literature,

and will be considered here. However, the two earliest observations -
which have already been considered in chapters 4 and 5 - were from
Babylonia (c. 1000 BC: uncertain identification) and Assyria (763 BC).
From each of the various eclipse records which he analysed, Fother-
ingham derived linear equations relating the lunar and solar accelerations
(C* and cf) on UT and plotted them in a diagram which has since become
well known to geophysicists. A copy of this diagram is shown in figure
10.2. (For a discussion of the relation between the accelerations of the
Moon and Sun on a UT framework and their relation to changes in the
length of the day on TT, see chapter 1.)
Fotheringham selected a small triangular area of solution space,
bounded by the limits obtained from three eclipses (—762, —128 and
+71: i.e. 763 BC, 129 BC and AD 71) as the most likely result. This area
is shown shaded in figure 10.2. However, Fotheringham's technique was
heavily dependent on the reliability of each individual observation which
he used. Rejection of even a single observation would have considerably
enlarged his preferred area. As discussed in chapter 4, there is nothing
338 10 Records of eclipses in ancient European history

in the record to suggest that the Assyrian observation in 763 BC relates

to a total eclipse; a large partial obscuration of the Sun would probably
suffice. Hence Fotheringham's assumption of totality in that year gives
the oblique line for this date in figure 10.2 an exactness which is quite
artificial.
Only four of the eclipses which Fotheringham analysed (763, 431, 310
BC and AD 364) can be accurately dated by historical means alone. The
dates of the remainder can only be derived with the aid of astronomical
computation. In several instances the precise place of observation is
in doubt, while frequently the magnitude at greatest phase is open to
conjecture. Newton (1970) reconsidered this material in some depth
(along with other ancient data), but whereas Fotheringham probably had
too high an opinion of the reliability of ancient European observations,
Newton was perhaps over-critical. A middle course would seem to be
preferable and this will be adopted here.
In the remainder of this section, each of the ancient European ob-
servations analysed by Fotheringham will be discussed in chronological
order. Years will be expressed on the BC/AD system. For reference,
the computed magnitude in the central zone will be given in parentheses
after each individual Julian date (apart from the timed solar eclipse of
AD 364). Where it is possible to derive meaningful AT limits or discrete
values, these will be quoted towards the end of each entry.

(1) BC 648 Apr 6 ? (total, 1.07): Paws or Thasos?

Archilochus, one of the earliest Greek poets after Homer and Hesiod,
makes a clear reference to a very large solar eclipse in one of his poems:

Nothing can be surprising any more or impossible or miraculous, now

that Zeus, father of the Olympians has made night out of noonday, hiding
the bright sunlight (apokruphas phaos heliou lampontos), and ... fear has
come upon mankind. After this, men can believe anything, expect
anything. Don't any of you be surprised in future if land beasts change
places with dolphins and go to live in their salty pastures, and get to like
the sounding waves of the sea more than the land, while the dolphins
prefer the mountains.
[Archilochus, fragment 122; trans. Barron and Easterling (1985, p. 127).]
A more accurate rendering of the phrase apokruphas phaos heliou lam-
pontos, here translated as 'hiding the bright sunlight', would be 'hiding the
light of the gleaming Sun' - much as in the quotation cited by Fother-
ingham. Combining this expression with the description of the loss of
daylight 'night out of noonday' strongly suggests that the eclipse was either
total or fell only slightly short of this phase. Evidently the phenomenon
 10.2 Solar eclipses analysed by Fotheringham 339

left a singularly profound impression on its beholders, as illustrated by

Archilochus' subsequent use of hyperbole.
Fotheringham regarded the date BC 648 as the only possible one and
he assumed that the eclipse was witnessed at either Paros or Thasos -
islands respectively in the extreme south and north of the Aegean - where
Archilochus is known to have spent most of his life. However, both the
date and place of observation require further consideration.
Although the dates of both the birth and death of Archilochus are
unknown, two of his poems provide useful chronological information (see
Rankin, 1977, chap. II). In one of these, Archilochus mentions King Gyges
of Lydia (c. 687-652 BC) and in the second he alludes to the destruction
of the city of Magnesia by Treres (c. 655-650 BC). That Archilochus was
a contemporary of King Gyges (as asserted by Herodotus, I, 12) is clear
from the context of his poem which begins:

I have no interest in the property of golden Gyges. Envy has never

taken hold of me...

It is known that Archilochus was born on Paros. He later moved

to Thasos where he served as a soldier (possibly a mercenary). After
occasional travels elsewhere in the region (including Sparta) he eventually
returned to Paros, where he was later killed helping to defend the island
against invasion. Archilochus often mentions Paros or Thasos by name in
his poetry. The locations of both islands (and also Sparta) are shown in
figure 10.3.
Newton (1970, p. 92) regarded the solar phenomenon to which
Archilochus refers as merely a literary eclipse'. However, that the poet
himself saw the eclipse is suggested by the following remarks of Barron
and Easterling (1985, pp. 117-119):

(Archilochus) is the first Greek writer to take his material from what he
claims to be his own experience and emotions, rather than from the stock
of traditions... It would be absurd to claim that a poet composing songs
for performance in a small community in which everyone knew everyone
else would not exploit the audience of that society and its relationships.

Based on the above considerations, I shall assume that the eclipse was
either total or nearly so at either Paros or Thasos. Reference to the global
eclipse maps of von Oppolzer (1887) indicates that between 700 and 610
BC only the following solar obscurations could have been total in this
region: BC 691 Jul 28, 657 Apr 15, 648 Apr 6, 646 Sep 8 and 637 Aug
29. Several annular eclipses would also have beeen visible in the Aegean
during this same period (in 689, 662, 661, 651, 650, 641, 635 and 633 BC)
but in each case the magnitude within the central zone was between about
340 10 Records of eclipses in ancient European history

Fig. 10.3 The locations of the islands of Paros and Thasos (and also Sparta),
where Archilochus is known to have lived.

0.93 and 0.96. None of these events could have caused the impressive loss
of daylight described so vividly by Archilochus.
On historical grounds, it is probably impossible to distinguish between
the five dates of total eclipses noted above so that astronomical computa-
tion must be resorted to - see table 10.2.
None of the eclipses listed in table 10.2 could have been total around
noon on either island - contrary to what Archilochus implies - but in each
case the computed time of totality was within 2 or 3 hours of midday,
the Sun being high in the sky (altitude between about 40 and 50 deg). In
principle, any of these circumstances would probably satisfy the poetic ref-
erence. However, not all of the indicated AT ranges are plausible. For the
eclipses of 691, 646 and 637 the appropriate limits are in marked discord
with the results derived from roughly contemporaneous Babylonian obser-
vations (see figure 10.1). Computations for these same three eclipses based
on equation (10.1) effectively eliminate them from further consideration.
In each case, small magnitudes are indicated at both Paros and Thasos (be-
tween 0.62 and 0.77), while all three local times of maximum phase would
be late in the afternoon (after about 17.0 h) with low solar altitude (10 or
20 deg). The calculated circumstances elsewhere in the Aegean would be
very similar. Almost certainly, a choice must rest between 657 and 648 BC.
 10.2 Solar eclipses analysed by Fotheringham 341

Table 10.2 AT ranges necessary for seventh century BC eclipses to attain

totality at either Paros or Thasos.

A T Range (sec)
a
Year B C Mag.* 7
Place c
LI/ UL e LT (hY Alt. (deg)s
691 1.06 Paros 25 900 26 800 15.2 45
Thasos 26900 27 850 14.9 49

657 1.07 Paros 17050 18000 10.3 52

Thasos 16050 17150 10.7 52

648 1.07 Paros 19450 21050 9.3 41

Thasos 17 800 19 350 10.1 45

646 1.04 Paros 27150 28050 15.2 38

Thasos 28 800 29 700 14.5 44

637 1.05 Paros 25100 25 950 15.0 42

Thasos 25 850 26700 14.7 45

a
Year BC.
b
Magnitude in the central zone.
c
Assumed place of observation (either Paros or Thasos).
d
Lower limit to AT in sec (LL) to render the eclipse total at the selected place.
e
The corresponding upper limit to AT (UL).
f Computed local time of greatest phase to the nearest 0.1 h, as deduced from the mean
AT of columns 4 and 5.
g
Calculated solar altitude in deg at this moment.

In table 10.3 are listed the magnitude, local time of greatest phase
and altitude at Paros and Thasos for the eclipses of 657 and 648
BC, calculated using equation (10.1) (AT values respectively 20100 and
19950 sec).
From table 10.3, it would appear that 648 BC, the date favoured by
Fotheringham, is the better choice for the eclipse of Archilochus - whether
the place of observation was Paros or Thasos (or elsewhere in the region).
However, a date of 657 BC cannot be entirely eliminated; as is evident
from figure 10.1, data points at this early period are too sparse to rule
out the AT range necessary to achieve totality at either Paros or Thasos
(16050 < AT < 18 000 sec).
The poem can thus be confidently dated to within a decade (i.e. between
657 and 648 BC), but possibly no better than this. Although this result
is of some importance to the historian, because of the uncertain date the
record is of little value in the investigation of the Earth's past rotation.
342 10 Records of eclipses in ancient European history

Table 10.3 Local circumstances for the eclipses of 657 and 648
BC at Paros and Thasos computed using equation (10.1).

Year BC Place Mag. LT(h) Alt. (deg)

657 Paros 0.94 9.8 47
Thasos 0.88 9.8 46

648 Paros 0.97 9.9 46

Thasos 1.07 9.9 44

(2) BC 585 May 28 ? (total 1.07): Asia Minor

Herodotus, who has been aptly described as the 'Father of History', alludes
to the interruption of a battle in Asia Minor by a marked loss of daylight:
After this, seeing that Alyattes would not give up the Scythians to
Cyraxes at his demand, there was war between the Lydians and the Medes
five years... They were still warring with equal success, when it chanced, at
an encounter which happened in the sixth year, that during the battle the
day was turned to night. Thales of Miletus had foretold this loss of
daylight to the Ionians, fixing it within the year in which the change did
indeed happen. So when the Lydians and the Medes saw the day turned to
night, they ceased from fighting, and both were the more zealous to make
peace. Those who reconciled them were Syennesis the Cicilian and
Labnetus the Babylonian
[Herodotus, I, 74; trans. Godley (1975, vol. I, pp. 91-93).]
The first century AD writer Pliny makes no mention of the battle, but
in alluding to Thales he makes it clear that the celestial phenomenon
mentioned by Herodotus was understood to be a solar eclipse. Pliny also
specifies the year when it occurred:
The original discovery (of the cause of eclipses) was made in Greece by
Thales of Miletus, who in the fourth year of the 48th Olympiad (585/4
BC) foretold the eclipse of the Sun that occurred in the reign of Alyattes,
in the 170th year after the foundation of Rome (584/3 BC).
[Pliny, Naturalis Historia, II, 53; trans. Rackham (1938, vol. I, p. 203).]
Although Herodotus (c. 484-425 BC) wrote some 150 years after the
events which he describes, his account of the various historical events
occurring at this time is fairly detailed - for example he is able to give
the names of the intercessors who made peace between the two sides.
However, Herodotus also records a similar fall of darkness a century later
- in his own time - which cannot be explained by an eclipse. A translation
of Herodotus' description of this event is as follows:
 10.2 Solar eclipses analysed by Fotheringham 343

The army (of Xerxes) then wintered, and at the beginning of spring was
ready and set forth from Sardis to march to Abydos. When they had set
forth, the Sun left his place in the heavens and was unseen, albeit the sky
was without clouds and very clear, and day was turned into night.
[Herodotus, VII, 37; trans. Godley, (1971, vol. II, p. 351).]
On this latter occasion, no suitable eclipse can be identified. The date
when Xerxes left Sardis is fixed as 480 BC by Herodotus' reference to the
Olympic Games being held in the summer of the same year (VII, 206).
However, the only large eclipse visible in Asia Minor around this time
was that of BC 478 Feb 17 and this was only annular - as little as 0.90
of the Sun being covered in the central zone. Hence either the darkness
was of atmospheric origin or Herodotus has mistakenly synchronised two
quite separate events.
It thus seems that Herodotus' account of the battle between the Lydians
and the Medes being interrupted by the onset of darkness by day may
also be of doubtful reliability, but further discussion seems worthwhile.
As noted above, the year 170 AUC given by Pliny corresponds to 584/3
BC Allowing some uncertainty in the date, Fotheringham selected the
eclipse of BC 585 May 28. Judging from the charts of von Oppolzer
(1887), no other eclipse could have turned the day into night in Asia
Minor - where the battle occurred - between about 602 and 558 BC,
and this is confirmed by detailed investigation using equation (10.1). Two
annular eclipses during this period (588 and 581 BC) were probably fairly
large in Asia Minor but in each central zone the Moon would cover no
more than 0.96 of the Sun. Hence these events could scarcely produce the
effects described by Herodotus.
The calculated date of 585 BC is remarkably close to that indicated by
Pliny. However, the precise place where the battle was fought is unknown;
as Fotheringham stressed, it could have been almost anywhere in Asia
Minor. Because of the uncertainty in the interpretation of the record
and the place of observation, it would seem hardly feasible to use the
observation in the investigation of AT.
Whether the eclipse was indeed predicted by the Greek astronomer
Thales of Miletus (who died at an advanced age in 548 BC) may well
be doubted. It appears that the first person to give the true explanation
of eclipses was Anaxagoras (500 - 428 BC) - rather than Thales (Heath,
1932, pp. xxiii and 27). At this early period, Babylonian astronomers,
although probably furnished with centuries of past observations, attained
very poor success in anticipating eclipses for a given location (see chapter
3). To quote Neugebauer (1975, p. 604),
In the early days of classical studies one did not assume that in the sixth
century BC a Greek philosopher had at his disposal the astronomical and
344 10 Records of eclipses in ancient European history

mathematical tools necessary to predict a solar eclipse. But then one could
invoke the astronomy of the 'Chaldeans' from whom Thales could have
received whatever information was required. This hazy but convenient
theory collapsed in view of the present knowledge about the chronology of
Babylonian astronomy in general and the lunar theory in particular. It is
now evident that even three centuries after Thales no solar eclipse could be
predicted to be visible in Asia Minor - in fact not even for Babylon. There
remains another vague hypothesis: the prediction by means of cycles (if
need be again available upon request from Babylon). But unfortunately
there exists no manageable cycle of solar eclipses visible at a given
locality...
A recent revival of the idea that Thales indeed predicted an eclipse -
inferring the most likely date as BC 582 Sep 21 - is due to Panchenko
(1994). However, it should be emphasised that the track of totality on
this occasion did not reach further north than latitude +34.0 deg at any
point on the Earth's surface. Thus even on the southern coast of Asia
Minor (lat. = +36.0), the magnitude cannot have been greater than 0.93,
regardless of the value of AT adopted. In fact, using equation (10.1) (AT
= 18 950 sec), the magnitude at any point in the peninsula cannot have
exceeded 0.85.
In brief, the assertions by Herodotus and Pliny probably owe their
origin to the various legends which accumulated around the personality
of Thales.

(3) BC 463 Apr 30 ? (total, 1.06): Thebes

In the odes of Pindar, who is generally regarded as the greatest lyric poet
of ancient Greece, there are two allusions to the effects of a very large
solar eclipse:
(i) Beam of the Sun! O thou that seest from afar, what will thou be
devising? O mother of mine eyes! O star supreme, reft from
(kleptomenon) us in the daytime (en amerai)! Why hast thou
perplexed the power of man and the way of wisdom, by rushing
forth on a darksome track? Art thou bringing on us some new and
strange disaster? Yet by Zeus, I implore thee, thou swift driver of
steeds! Do thou, O queen! Change this world-wide portent into
some painless blessing for Thebes ...
But art thou bringing a sign of some war, or wasting of produce,
or an unspeakably violent snow-storm, or fatal faction, or again,
some overflowing of the sea on the plain, or frost to bind the earth,
or heat of the south wind streaming with raging rain? Or wilt
thou, by deluging the land, cause the race of men to begin anew? I
in no wise lament whate'er I shall suffer with the rest!
[Pindar, Paean, ix; trans. Sandys (1978, pp. 547-549).]
 10.2 Solar eclipses analysed by Fotheringham 345

(ii) God can cause unsullied light to spring out of black night. He can
also shroud in a dark cloud of gloom the pure light of day.
[Pindar, fragment 142; trans. Sandys (1978, p. 577).]

Both of these texts may refer to one and the same event, but the first,
entitled T o the Thebans', is particularly important since it makes specific
reference to the city which was both the birthplace of Pindar and his
main domicile. In an article specially devoted to this eclipse (1920c),
Fotheringham gives a fuller discussion than in his better known 1920b
paper. However, in both works he considered only the eclipse of BC 463
Apr 30. Whether there were other large eclipses visible in Greece during
Pindar's lifetime needs to be carefully examined.
Pindar was born in a village adjacent to Thebes in either 522 or 518 BC,
the latter date being more likely (Race, 1986, p. 1). Since he is believed to
have attained the age of 80, the most probable year of his death is 438
BC. It is possible to date many of his poems with confidence, not least
because some are written in honour of the victors at various Olympic
Games. Unfortunately Pindar's ninth paean (excerpted above) does not
come into this category. Scholars are agreed that he wrote his earliest ode
in 498 BC (while still a young man) and his last in 446 (Race, ibid.).
Bowra (1964, pp. 83-4 and pp. 378-9) had no doubt that Pindar saw
the eclipse, remarking that there is a sense of urgency in his paean. It
evidently left a deep impression on him, and although he seems to have
heard that the phenomenon was visible at other places too ('worldwide'),
he regarded it as a major portent for Thebes. In describing the apparent
theft of the Sun and the resulting darkness, Pindar clearly refers to either
a complete solar eclipse or one in which the Sun all but disappeared.
Although the former alternative would seem more in keeping with his
description and subsequent anxiety, Fotheringham (1920b) pointed out
that use of the present tense of the verb translated as 'reft from' (i.e.
kleptomenon) is consistent with incomplete action.
From von Oppolzer (1887), only four eclipses could have been large in
Greece between 506 BC and the death of Pindar - a conclusion supported
by computation using equation (10.1). The dates are: BC 493 Nov 24 (to-
tal), 488 Sep 01 (annular), 478 Feb 17 (annular) and 463 Apr 30 (total). All
were thus during his active lifetime. Since any attempt to judge his degree
of maturity when he produced Paean ix is subjective, it is not possible to
separate these four dates on literary grounds. (I am grateful to Professor
C. Cary of Royal Holloway College, Egham for helpful discussion on this
matter.) However, astronomical techniques prove more useful.
The eclipse of 478 BC can be fairly easily eliminated from further
consideration. On this occasion, no more than 0.89 of the solar diameter
would have been obscured (even in the zone of annularity) so that the
346 10 Records of eclipses in ancient European history

event could not have caused darkness; it might well have passed unnoticed.
For totality at Thebes in 493 BC, a value of AT in the range 22150 < AT
< 22 850 sec would be required. However, this narrow zone is in marked
conflict with several of the roughly contemporary Babylonian limits in
figure 10.1. Calculation using equation (10.1) (AT = 17450 sec) indicates
a magnitude of only 0.82 at Thebes exactly at noon.
Both the eclipses of 488 and 463 BC are more promising candidates.
Although the former event was also annular, the Moon was sufficiently
close to the Earth to cover almost 0.99 of the solar diameter. Values of
AT which render the eclipse central at Thebes (17900 < AT < 18450 sec)
are quite plausible, as reference to figure 10.1 shows. However, they yield
a time of greatest phase just at sunrise (local time 5.4 h). Although the
darkness would probably be heightened under these circumstances, it is
debatable whether a sunrise eclipse would be described as 'in the daytime'
(en amerai). In 463 BC, the AT limits required to produce totality at
Thebes (18 100 < AT < 20600 sec) are also fairly reasonable. Here, the
corresponding local time would be 13.5 h with the solar altitude 58 deg -
very much a daytime event.
From the above arguments, it seems highly likely that the eclipse of
Pindar is that of 463 BC, but since a date of 488 BC cannot be en-
tirely eliminated, the observation recorded by Pindar cannot be effectively
utilised to set limits to the value of AT.

(4) BC 431 Aug 3 (annular, 0.98): Athens

In his History of the Peloponnesian War, the contemporary Greek historian
Thucydides notes three eclipses: two of the Sun and one of the Moon. Only
the first solar obscuration appears to have been unusually large, but this is
especially interesting since the account is the earliest in European history
to mention the visibility of stars during an eclipse. Among the records
of other civilisations, only a single Chinese observation, dating from 444
BC, makes an earlier allusion to stars seen under these circumstances (see
chapter 8). Thucydides' account may be translated as follows:
The same summer, at the beginning of a new lunar month (the only
time by the way at which it appears possible), the Sun was eclipsed after
noon. After it had assumed the form of a crescent (menoeides), and some
of the stars had come out, it returned to its natural shape.
[Thucydides, II, 28; trans. Livingstone (1949, p. 109).]
This same eclipse is probably alluded to by both Cicero (De Republica,
I, 16) and Plutarch (Life of Pericles, XXXV) in a story concerning Pericles,
a pupil of Anaxagoras. It is related that a solar eclipse occurred during
the Peloponnesian War, bringing on darkness and creating fear. However,
using his cloak, Pericles demonstrated the cause of the eclipse, thus
 10.2 Solar eclipses analysed by Fotheringham 347

dispelling the alarm. The details of the tale as given by Cicero and Plutarch
differ considerably, but both writers lived fully 500 years after the event.
According to Plutarch the story was told in the schools of philosophy.
Only the contemporary account by Thucydides merits consideration here.
It is well established that the Peloponnesian War between Athens and
Sparta began in 431 BC and lasted for 27 years. Thucydides (II, 1) notes
that the opening battle occurred in the spring, towards the end of the year
when Pythodorus was archon (chief magistrate) at Athens. Complete lists
of the Athenian archons are preserved from 481 to 292 BC (Bickerman,
1980, pp. 138-9) and these indicate that Pythodorus was archon in 432/1
BC. The eclipse itself occurred only a few months after the start of the
war, and thus the date must be BC 431 Aug 3. This was the largest eclipse
visible in Greece for several years around this time.
Thucydides was an Athenian. Born probably in the early 450s, he
assumed command of a section of the Athenian forces in 424 BC, the
eighth year of the war (IV, 104). The other two eclipses which he mentions
are of little astronomical value, but the solar obscuration acts as a useful
chronological aid. Occurring in the summer of the eighth year, this is
described by Thucydides (IV, 52) as 'small'. It may be confidently dated as
BC 424 Mar 21. Using equation (10.1) (AT = 16 350 sec), the computed
magnitude at Athens was 0.72. Only one other solar eclipse was visible in
Greece over a period of ten years beginning in 430 BC This took place
on 426 Nov 4, during the tenth year of the war, and could scarcely be
designated a 'summer' event. (NB it was Thucydides' practice to divide
each year into two conventionalised seasons, 'summer' and 'winter', each
lasting about six months.) Finally, the lunar eclipse was said to happen in
the summer of the 19th year (VII, 50). Of the two lunar eclipses in BC 413
(Mar 4 and Aug 27), the latter better fits the events discussed at the time.
As in the record from the first year of the war, Thucydides notes that
the solar eclipse in the eighth year occurred at the time of new Moon,
while he remarks that the lunar obscuration happened at full Moon. This
may illustrate his knowledge of the true cause of eclipses, which had only
recently been explained by the Athenian Anaxagoras (500-428 BC).
It seems quite likely that Thucydides saw the eclipse of 431 BC himself,
although this cannot be proved. He seems to have had a special interest
in such phenomena, remarking (I, 23) that solar eclipses were unusually
numerous during the Peloponnesian War. In 431 BC, Thucydides was
presumably in his home city of Athens. He makes no mention of any
travels from there until the eighth year of the war (IV, 104), while in
the year after the eclipse he tells us (II, 48) that he caught the plague in
Athens. We know (IV, 105) that Thucydides had the right of working gold
mines in Thrace (lat. c. +41 deg, long. c. —26 deg). However, there is
nothing in his text to indicate that he was there when the eclipse occurred.
348 10 Records of eclipses in ancient European history

For values of AT which render the eclipse of 431 BC large in Greece,

greatest phase would occur in the late afternoon, and thus long after
midday. However, Thucydides probably only gave a very general indi-
cation of the time of day. Applying equation (10.1) (AT = 16450 sec),
the eclipse reached its maximum of 0.86 at 17.4 h, the solar altitude
then being 19 deg. In Thrace, the circumstances would be very similar,
although the computed magnitude would be rather greater (0.91). Thucy-
dides describes only a very large partial eclipse (the term menoeides means
'crescent-shaped'), suggesting that the ring phase was not witnessed.
Even if the eclipse had been central (magnitude 0.98), Thucydides'
reference to the visibility of several stars is difficult to explain. Although
Venus (mag. —3.4), 19 deg to the east of the Sun, would no doubt have
been fairly prominent, no other planet or star should have been detectable.
In the late afternoon, both Jupiter (41 deg to the west of the Sun) and
Sirius (62 deg to the south-west) would be below the horizon. Mercury,
fairly close to Venus, would be faint (mag. +1.3). As Fotheringham noted,
after Venus the brightest star or planet above the horizon would be Vega
(mag. +0.4). Perhaps some allowance for exaggeration should be made
here.

RESULTS
For a partial eclipse at Athens, AT < 11 800 or > 12 650 sec.
NB only a narrow zone of avoidance is thus indicated. In Thrace the
corresponding AT limits are AT < 14100 or > 14750 sec, but assumption
of an observation from there is purely speculative.

(5) BC 310 Aug 15 (total 1.05): Sicily

In his Library of History, the first century BC historian Diodorus Siculus
(i.e. Diodorus of Sicily) records two eclipses. The earlier of these (364 BC)
is reported without any details, but the description of the later event is
remarkably vivid. This occurred after the tyrant Agathocles, with a fleet
of sixty ships, had made his escape from a blockade at Syracuse harbour
by the Carthaginians.

Agathocles, who was already at the point of being overtaken and

surrounded, gained unhoped for safety as night closed in. On the next day
there occurred such an eclipse of the Sun that utter darkness set in and the
stars were seen everywhere; wherefore Agathocles' men, believing that the
prodigy portended misfortune for them, fell into even greater anxiety
about the future. After they had sailed for six days and the same number
of nights, just as day was breaking, the fleet of the Carthaginians was
unexpectedly seen not far away... When Libya came into sight....
[Diodorus, XX, 5-6; trans. Geer (1954, vol. X, pp. 155-157).]
 10.2 Solar eclipses analysed by Fotheringham 349

Diodorus noted (XX, 3) that the above events occurred during the
archonship of Hieromemnon in Athens, corresponding to 310/09 BC
(Bickerman, 1980, p. 139). Hence there can be no question that the correct
date is BC 310 Aug 15. Diodorus is careful to name the Athenian archon
for each year at the appropriate point in his history. From von Oppolzer,
no other eclipse could have been total in Sicily between 346 and 303 BC
- a result confirmed by calculations based on equation (10.1). Nothing
short of totality could have caused 'utter darkness', with the stars being
seen 'everywhere'. It is unfortunate that the precise place of observation
cannot be determined; otherwise this would have been the most valuable
record of a total solar eclipse from ancient Europe.
Whether Agathocles sailed northwards or southwards after escaping
from Syracuse is not known, although he was probably still off the coast
of eastern Sicily when the eclipse occurred. There seems to be no reason
for doubting Diodorus' assertion that the eclipse occurred on the day after
the escape. Values of AT (around 15000 sec) which produce totality in
the neighbourhood of Syracuse indicate that the eclipse happened early
in the morning (local time 7.4 h). Hence although the time of day when
the escape took place cannot be ascertained, Agathocles would probably
have been at sea for less than 24 hours (perhaps considerably less) when
the Sun was obscured.
The voyage from Syracuse to the Libyan coast (possibly in the neigh-
bourhood of Cape Bon) took six days and six nights. There is little
likelihood of the fleet calling at some other port en route for most of
Sicily was in enemy (i.e. Carthaginian) hands (Diodorus, XX, 3). If
the ships travelled northwards (through the straits of Messina), the total
distance covered to reach Libya would be about 600 km; if southwards
only about 400 km. Hence the average speed - assuming continuous
travel - would be respectively about 4 or 3 km per hour. Captain
R. Hardy of South Tyneside College assures me that such a low mean
rate is quite reasonable for primitive sailing vessels. Agathocles' ships
were equipped with oars, which would almost certainly be used dur-
ing the chase. If we assume even double the above rate between the
escape from Syracuse and the eclipse, the fleet may only have covered
some 100 km northward or 70 km toward the south - significantly less
if the ships escaped from the harbour when the day was already well
advanced.
The two extreme positions (roughly 100 km to the north or 70 km to
the south of Syracuse) are thus perhaps (lat. =+37.8 deg, long. =—15.3
deg) and (lat. =+36.6 deg, long. =—15.0 deg). In figure 10.4, these are
indicated respectively by NL and SL.
Fotheringham derived limits of: (lat. =+38.42 deg, long. =—15.50 deg)
and (lat. =+36.58 deg, long. =—15.00 deg). The southern location
350 10 Records of eclipses in ancient European history

Fig. 10.4 The eclipse of Agathocles (BC 310 Aug 15). The edges of the computed
belts of totality for specific AT limits are shown.

is similar to that derived here. However, the distance to his suggested

northern site (150 km) would seem excessive.
During totality, Venus (mag. —3.4), Mercury (mag. —1.2), Jupiter (mag.
—1.2), and Sirius (mag. —1.4) would all be well placed for visibility.

RESULTS
For totality at Syracuse itself, 14 350 < AT < 16 500 sec, which is in
excellent accord with the results obtained from Babylonian data (see
figure 10.1). If the fleet of Agathocles had reached the estimated northern
limit of 100 km from Syracuse, as long as the southern limit of totality
extended as far south as this point, the required conditions would be
fulfilled. A value of AT > 13 350 sec would thus be needed. Equally, if
the fleet was 70 km from Syracuse in a southerly direction, it would be
necessary for the northern limit of totality to extend at least as far north
as this. A value of AT < 17200 sec is thus indicated. Combining these
results, we find that a reasonable range for AT would appear to be 13 300
< AT < 17200 sec. However, it is most unfortunate that more definite
results cannot be obtained.
 10.2 Solar eclipses analysed by Fotheringham 351

In figure 10.4 the southern limit of totality for AT = 13 300 sec and the
northern limit for AT = 17 200 sec are shown by full lines. Also depicted
(by dashed lines) are the limits of totality computed with a mean value for
AT of 15 250 sec (i.e. the average of 13 300 and 17200 sec) to illustrate
the true geographical width of the track.

(6) BC 129 Nov 20 ? (total, 1.01): Hellespont

Hipparchus, the eminent Greek astronomer of the second century BC,
is renowned for his skill in both observation and theory. One of his
numerous achievements was to calculate the distance of the Moon using
two simultaneous observations of an unidentified solar eclipse. This was
seen to be fully total in the Hellespont, while at Alexandria | of the solar
diameter was obscured. From these observations, he deduced that the
distance of the Moon was approximately 67 \ Earth radii, a result which
compares well with the true mean of 60.25 in the same units.
Most of Hipparchus' writings survive only in quotations and para-
phrases by other ancient authors. A brief account of his use of the eclipse
is given by Ptolemy in the Almagest. The fullest description is by Pappus
of Alexandria (c. AD 320), who was probably the last great geometer
of ancient Greece. Further information (without mentioning Hipparchus
by name) is supplied by the first century BC astronomer Cleomedes.
Translations of these three accounts are as follows:

(i) Hipparchus tries to demonstrate the Moon's distance by guessing

at the Sun's. First he supposes that the Sun has the least
perceptible parallax, in order to find its distance, and then he uses
the solar eclipse which he adduces; at one time he assumed that
the Sun has no perceptible parallax, at another that it has a
parallax big enough [to be observed]. As a result, the ratio of the
Moon's distance came out different for him for each of the
hypotheses he put forward; for it is altogether uncertain in the case
of the Sun, not only how great its parallax is, but even whether it
has any parallax at all.
[Ptolemy, Almagest, V, 11; trans. Toomer (1984, pp. 243-244).]

(ii) So Hipparchus, being uncertain concerning the Sun, not only how
great a parallax it has but whether it has any parallax at all,
assumed in his first book of 'On Sizes and Distances' that the
Earth has the ratio of a point and center to the Sun [i.e. the Sun's
sphere]. And at one time using the eclipse he adduced, he assumed
that it had the least parallax, and at another time a greater
parallax. Hence the ratios of the Moon's distances came out
different. For in Book 1 of 'On Sizes and Distances' he takes the
following observation: an eclipse of the Sun, which in the
352 10 Records of eclipses in ancient European history

Hellespontine region was an exact eclipse of the whole Sun, such

that no part of it was visible, but at Alexandria by Egypt
approximately four-fifths of the diameter was eclipsed. By means
of the above he shows in Book 1 that, in units of which the radius
of the Earth is one, the least distance of the Moon is 71, and the
greatest 83. Hence the mean is 77... Then again he himself in Book
2 of 'On Sizes and Distances' shows from many considerations
that, in units of which the radius of the Earth is one, the least
distance of the Moon is 62, the mean 671 and the Sun's distance
490. It is clear that the greatest distance of the Moon will be 721.
[Pappus, Commentary on the Almagest, V, 11;
trans. Toomer (1974).]

(iii) Moreover, such an observation has been made in the case of an

eclipse of the Sun. Once when the Sun was wholly eclipsed in the
Hellespont, it was observed in Alexandria to be eclipsed except for
the fifth part of its diameter, which is, according to the sight,
except for two digits and a little more... Now since it is 5000 stades
from Alexandria to Rhodes and just so many from thence to the
Hellespont, necessarily a digit will be seen at Rhodes; besides,
proceeding hence to the Hellespont, this will also decrease in
proportion, since when the Hellespont is reached, it will entirely
vanish
[Cleomedes, De Motu Circularis Corporum, II, 3; trans, auct]
Toomer (1974) explains that the reason for Hipparchus obtaining two
'mean' results for the lunar distance of 77 and 67^ Earth radii is that in
Book 1 he assumed a solar parallax of zero (i.e. placing the Sun at an
infinite distance away) while in Book 2 he adopted a parallax of 7 arcmin.
There is nothing in any of the above records to suggest that Hipparchus
witnessed the eclipse, or even that it occurred during his lifetime. The
date of his treatise is unknown. On historical evidence alone, all that can
be established is that he was able to gain access to two careful records of
the same eclipse from the region of the Hellespont and from Alexandria.
Since Alexandria was founded in 331 BC, no date earlier than this can
be considered. In principle, the eclipse might have occurred at any time
between 331 BC and the death of Hipparchus (c. 120 BC) - an interval
of more than two centuries.
Fotheringham (1909 and 1920b) considered only the dates BC 310 Aug
15 (i.e. the eclipse of Agathocles) and 129 Nov 20, rejecting the former
since he was of the opinion that a magnitude estimate from Alexandria
at such an early period was unlikely. However, more recent investigations
by Newton (1970, pp. 104-110) and Toomer (1974) have discussed several
alternative dates, ranging from 310 to 125 BC These will all be considered
below, along with other possible identifications.
 10.2 Solar eclipses analysed by Fotheringham 353

It should be emphasised that four-fifths (the estimate from Alexandria)

is an unusual way of expressing an eclipse magnitude. As discussed in
chapter 3, the Greeks - following the Babylonians - normally used digits
or twelfths of the solar disk. However, Newton (1970, p. 134) made an
interesting suggestion. A magnitude of 9 digits of area corresponds closely
to 9.5 digits of diameter, as is evident from Ptolemy's conversion table
(Almagest, VI, 8) - see also chapter 3. When converted to a fraction, 9.5
digits is virtually equal to | (actually 0.79). This may well be the true
explanation of the value used by Hipparchus.
Although the eclipse might well have taken place before Hipparchus'
own time, it is useful to summarise details about his life. The following
information is derived mainly from Dicks (1960, pp. 3ff.).
Hipparchus was born in Bithynia in the north-west of Asia Minor,
probably around 194 BC. He was a citizen of Nicaea. It is known from
Ptolemy that Hipparchus made observations of the risings and settings
of stars in Bithynia (Dicks, 1960, p. 3). Subsequently, when he was aged
about thirty, he left Bithynia for Rhodes, where he seems to have spent the
rest of his life. What is probably Hipparchus' earliest known observation
at Rhodes (of the autumnal equinox) dates from 162 BC (Almagest, III,
1). His latest surviving measurements (of the positions of the Sun and
Moon) were made in 127 BC (Almagest, VI, 5). He probably died about
120 BC, although the exact year of his death is unknown.
Totality of the eclipse investigated by Hipparchus is very carefully de-
scribed by Pappus in the following words: 'an exact (akribos) eclipse of
the whole Sun, such that no part of it was visible'. That the eclipse was
indeed complete is confirmed by Cleomedes. It is noteworthy that both
writers omit superfluous details such as darkness or the appearance of
stars; only the knowledge that the eclipse was complete was important to
Hipparchus. Pappus asserts that the observation was made 'in the Helle-
spontine region' (en tois peri ton Hellesponton topois). Rather than simply
meaning the strait itself (now known as the Dardanelles), this denotes one
of the climata or belts of latitude 400 stades (approximately 0.6 deg) wide,
within which celestial phenomena did not appear to change appreciably
(Dicks, 1960, pp. 154-162). For all practical purposes, inhabitants of the
same clima were considered to be in the same latitude.
In his Commentary on Aratus, his only fully extant work, Hipparchus
states that 'in the Hellespontine regions' (en tois peri ton Hellesponton
topois - precisely the expression used by Pappus), the longest day is 15
hours. Ignoring refraction and the semidiameter of the Sun, this gives
a latitude of +41.07 deg. For comparison, Ptolemy obtained 40;56 (i.e.
40.93) deg. There is obviously a considerable error in both results since
the latitude of the Hellespont strait itself ranges from +40.0 to +40.4
deg. From Strabo - the first century BC geographer who derived much
354 10 Records of eclipses in ancient European history

of his information from the work of Hipparchus - it is possible to fix the

latitude limits of the Hellespontine clima with fair accuracy. Towards the
south, Alexandria in the Troad (lat. = +39.8 deg) was included in the
same clima, but the zone did not quite reach as far north as Lysimachia
(lat. = +40.6 deg), a little beyond the eastern end of the strait (Strabo,
Geography, II, 5.40). A latitude range between about +39.8 and +40.4 deg
thus seems most likely.
Hipparchus considered Alexandria in Egypt (at which the magnitude of
| was estimated) to be 10 deg south of the Hellespont and thus close to
latitude 31 deg N - a result closely confirmed by Ptolemy, who gave 30;58
(or 30.97) deg. These values are quite near the true figure of +31.22 deg.
In his calculation of the the lunar distance, Hipparchus needed to
assume that both places where the eclipse was observed lay under the
same meridian. Following Eratosthenes, he believed that Alexandria in
Egypt (long. = -29.9 deg), Rhodes (-28.2 deg), the Troad (-26.2 deg),
Lysimachia (—26.6 deg) and Byzantium (—29.0 deg) were on the same
meridian of longitude (Strabo, II, 5.7). Hence it is reasonable to infer that
the observation in the Hellespont was made in the vicinity of the strait
itself, rather than much to the east or west (but still in the same latitude
zone). As illustrated in figure 10.5, Hipparchus' assumption entails er-
rors of several degrees in longitude but in ancient times determination of
this parameter presented serious difficulties. In alluding to the magnitude
of the eclipse at Rhodes, as well as at the Hellespont and Alexandria,
Cleomedes is making an indirect reference to Hipparchus. However, this
detail is purely incidental.
From von Oppolzer, several eclipses could possibly have been total in
the Hellespont region between 331 BC and 120 BC These are BC 310 Aug
15, 303 Apr 2, 282 Aug 6, 228 Sep 7, 221 May 5, 190 Mar 14, 188 Jul 17,
183 Oct 19, 176 Jun 6, 174 Oct 10, 163 Mar 15, 136 Apr 15 and 129 Nov
20. After the last date there is no further candidate until BC 116 Aug 29
- fully ten years after Hipparchus' last known observation. Newton (1970,
p. 109) also considered the eclipse of BC 125 Sep 7, but this was only
annular. In the central zone, 99.7 per cent of the Sun would have been
covered by the Moon, leaving a narrow ring of light still visible. This is
not consistent with the careful description of totality by Pappus. Toomer
(1974) made calculations for the eclipse of BC 217 Feb 11. However, the
track of totality on this occasion could not have reached further north
than about 34 deg N latitude anywhere on the Earth's surface. Hence,
regardless of the precise value of AT adopted, the eclipse could not have
been central anywhere near the Hellespont.
In table 10.4, I have listed for each of the eclipses cited above - apart
from 217 BC - the computed magnitudes (using equation (10.1)) at both
the Hellespont and Alexandria and also the approximate ranges of AT
 10.2 Solar eclipses analysed by Fotheringham 355

Fig. 10.5 Hipparchus' assumed meridian through Alexandria and Byzantium.

necessary to produce totality at the Hellespont (annularity in the case of

125 BC). For this purpose I have taken the mean co-ordinates of the
Hellespont as: lat. =+40.1 deg, long. =—26.5 deg. The indicated lower
limit (LL) and upper limit (UL) to AT are estimated to the nearest 100
sec in each case. It will be noted that in 125 BC, the AT limits to achieve
annularity are effectively identical. This is because the apparent lunar and
solar diameters were almost equal.
Reference to figure 10.1 shows that only three of the dates in table
10.4 can be considered viable: BC 310 Aug 15, 190 Mar 14 and 129
Nov 20. For each of the three selected eclipses, the AT ranges derived
for the Hellespont are very close to the mean straight line through the
Babylonian timed data. However, on all other occasions, the results are
extremely discordant.
In order to attempt to distinguish between the dates 310, 190 and
129 BC, it is necessary to consider both the historical and astronomical
circumstances in more detail.
A date as early as 310 BC seems most unlikely for a variety of reasons.
What Hipparchus required for his calculation was a reliable - i.e. 'exact'
- observation of totality in the Hellespont region combined with a careful
magnitude estimate (|) for the same eclipse at Alexandria. He felt able to
356 10 Records of eclipses in ancient European history

Table 10.4 Computed magnitudes of selected eclipses be-

tween 331 and 120 BC at the Hellespont and Alexandria
and AT limits to produce a central eclipse at the Hellespont.

Hellesp. Alex. AT Range (sec)

Year B C Mag. Mag. LL UL
310 1.06 0.76 8 600 14900
303 0.66 0.93 7900 9 300
282 0.84 0.55 7 300 8400
228 0.60 0.34 5 300 5 800
221 0.63 0.82 3 900 4000
190 0.99 0.92 11400 12400
188 0.85 0.57 17200 17400
183 0.81 0.66 5100 6700
176 0.60 0.76 3 500 3 700
174 0.82 0.66 10000 10200
163 0.80 0.51 20900 22200
136 0.72 0.86 5 200 6100
129 1.01 0.80 12100 12400
125 0.88 0.96 17 500 17 500

put sufficient trust in both observations to use them in a determination of

the lunar distance, and - as we know in retrospect - they were sufficiently
accurate to enable him to obtain a reasonably sound result. However,
there is nothing in the available historical records to suggest that he could
have obtained effective data long before his own time.
Hipparchus was in possession of numerous Babylonian observations
(see chapter 4). By comparison, he seems to have had access to only
scattered groups of Greek data, other than his own measurements. Apart
from an isolated solstice determination by Meton in 432 BC (which
Hipparchus rejected as unreliable), the earliest Greek observations which
Hipparchus is known to have used date from between 295 and 272 BC
These were nearly all by the Alexandrian astronomer Timocharis and
were exclusively of conjunctions of either the Moon or Venus with certain
stars (Almagest, VII, 3; X, 4). The next set of early Greek data which
Hipparchus analysed were of measurements of planetary position, ranging
in date from 272 to 241 BC (Almagest, IX, 7; IX, 10). Here, the observer
cannot be established with certainty. Finally, Hipparchus investigated a
series of four observations of lunar eclipses made at Alexandria between
201 and 174 BC (Almagest, IV, 11 and VI, 5; see also section 10.4 below).
Once again, the observer is unknown.
 10.2 Solar eclipses analysed by Fotheringham 357

Judging from his analysis of these various observations, as outlined in

the Almagest, if Hipparchus had possessed records of many further Greek
observations, he would have had no particular motive for dividing them
into discrete chronological groupings. This suggests that he had little else
but the above to draw on. In particular, although the first and third sets
of observations were from Alexandria, Toomer (1974) doubts whether
there was ever a 'school of astronomy' there at this period. He remarks
that we can name very few people in Alexandria who made astronomical
observations or calculations during this interval.
The solar eclipse of 310 BC occurred more than a century before the
first eclipse record which definitely originates from Alexandria and fully
15 years before the first known astronomical observation made at this city.
Further, unless a special watch for solar eclipses was maintained there, the
eclipse of 310 BC might well have escaped notice. Computations using the
mean value for AT to produce totality at the Hellespont, as listed in table
10.4 (AT = 11 800 sec), indicate that at maximum phase the solar altitude
would be as much as 40 deg so that if the sky was clear the radiance of
the Sun would be dazzling. As it happens, similar calculations imply that
the Sun would be much lower in the sky both in 190 BC (altitude 22 deg)
and 129 BC (13 deg).
If we turn now to the observation at the Hellespont, as noted above,
the descriptions of totality at this location by both Pappus and Cleomedes
are remarkably precise and devoid of any irrelevant details such as the
visibility of stars. Whatever its date, the 'eclipse of Hipparchus' is the
only known example from ancient Europe which clearly describes the
complete disappearance of the Sun. Most early writers were much more
concerned with relating the effects of the darkness which accompanied a
large eclipse. Hipparchus required - and clearly obtained - more than
this. The eclipse of 310 BC occurred several generations before his own
era. Hence for any report of this event from the Hellespont he would
probably have to rely on an account in some historical or literary work.
If this resembled the material discussed elsewhere in this chapter, he
would have no justification for assuming that the whole of the Sun was
obscured - only that the phase must have been very large. Many further
early records, compiled by Ginzel (1899), are of even poorer quality than
this.
Only two other ancient writers are known to have obtained access
to separate accounts of the same eclipse from long before their own
time. In each case the reports relate to the lunar obscuration of 331
BC. Pliny (Natural History, II, 180) noted that at Arbela (lat. 35.9 deg
N, long. 44.3 deg E), the eclipse of Sep 20 in that year occurred at
the second hour of night whilst in Sicily the rising Moon was obscured
(see section 10.5 below). On the other hand, Ptolemy (Geography, I, 4)
358 10 Records of eclipses in ancient European history

stated that the same event was seen at Arbela in the fifth hour but took
place at Carthage in the second hour. Neither author gives any other
information. Just how Pliny and Ptolemy synchronised the dates of the
separate observations which they describe - i.e. at Arbela and another
location - is not known. Nevertheless, it is clear that at least one of their
sources was unreliable. Hipparchus was himself well aware of the value
of independent observations of the same lunar eclipse in determining the
difference between the longitudes of two sites (Strabo: Geography, I, 1.12
- see also chapter 13). However, there is no evidence that he was ever
able to put this technique to effective use because 'the total absence in
antiquity of any scientific organisation deprived the whole method of its
practical importance' (Neugebauer, 1975, p. 667).
In summary, the likelihood of Hipparchus obtaining two unusually
reliable descriptions of the same eclipse from at least 150 years before he
made his calculation of the lunar distance must be minimal. The eclipse
of 310 BC can thus almost certainly be discounted, and in the subsequent
discussion only the dates 190 and 129 BC will be considered. In both
cases the historical circumstances are much more favourable.
Toomer (1974) made a strong case for a date of 190 BC for the eclipse
of Hipparchus. He remarked that this fell in the middle of the period from
which Hipparchus obtained four reports of lunar eclipses from Alexandria
(201 to 174 BC) and thought that it was possible that the same person
made all four observations - as well as estimating the magnitude of the
solar eclipse. With regard to the record of totality at the Hellespont,
Toomer noted that the historian Livy (XXXVII, 9) gives details of a
Roman naval attack on the towns of the Hellespont strait which occurred
in the spring of 190 BC - about the time that the eclipse took place. If
the fleet had indeed witnessed totality, the commander, Livius Salinator,
could well have sent a detailed report back to Rome.
Toomer was able to further deduce that if Hipparchus had access to a
record of totality from the Hellespont in 190 BC and a magnitude estimate
of | at Alexandria, he could easily have obtained a result for the lunar
distance very close to that cited by Pappus. Toomer considered that this
would not be possible in either 310 or 129 BC. In making his calculations,
Toomer assumed that on account of the limited capabilities of ancient
mathematics Hipparchus would have had to suppose that the Sun was
in the meridian at both places at the time of the eclipse. However, this
would be a very poor approximation. For values of AT (around 12000
sec) which produce totality at the Hellespont, greatest phase at both sites
would occur fully 3.5 hours before noon. Since all four lunar eclipses
observed from Alexandria around this period were carefully timed (see
section 10.4), it seems reasonable to suppose that if the solar eclipse of
190 BC was indeed seen there, the time of day would also be recorded.
 10.2 Solar eclipses analysed by Fotheringham 359

Under these circumstances, Hipparchus could conceivably have made at

least some allowance for the local time of observation.
As it happens, calculation shows that if the eclipse of 190 BC was fully
total at the Hellespont, the magnitude at Alexandria must necessarily have
been between 0.86 and 0.90. Hence the width of the unobscured portion
of the solar limb would have been almost twice the amount assumed by
Hipparchus. As a result, if he had made even a rough correction for the
time of day he would have obtained a bad value for the lunar distance.
Similar considerations for the eclipse of 129 BC (AT around 12250
sec) yield a local time of around 15.9 h at both locations. However,
if the eclipse was total at the Hellespont, the magnitude at Alexandria
would be 0.80 - exactly the observed figure. This would lead to a sound
result for the lunar distance if Hipparchus had made even an approximate
allowance for the time of day. The eclipse of 129 BC has the attraction
that it occurred during Hipparchus' active lifetime. Hence if he observed it
to be fairly large at Rhodes he would have ample opportunity to enquire
(i) just where - towards the north - it was total and (ii) if the magnitude
had been estimated further south - i.e. as far from the central zone as
possible. However, the situation is hypothetical; we cannot establish on
historical grounds whether the eclipse of 129 BC took place before or
after Hipparchus wrote his treatise.
Since we can only speculate on the method of analysis used by Hip-
parchus and whether or not he might have seen the eclipse in question,
it is probably impossible to decide between the dates 190 and 129 BC
Unfortunately, these dates are too far apart to enable deduction of a useful
AT range by combining the separate AT intervals on the assumption that
totality was definitely observed at the Hellespont region on one or other
of these two dates.

(7) AD 29 Nov 24? (total 1-02): Nicaea?

Eusebius, the great theologian and historian of the fourth century AD,
preserves a description of a major solar eclipse by the second century
writer Phlegon of Tralles. Very little of Phlegon's works now survive.

And Phlegon also who compiled the Olympiads writes about the same
things in his 13th book in the following words: 'In the fourth year of the
202nd Olympiad (AD 32-33), an eclipse of the Sun took place greater than
any previously known, and night came on at the sixth hour of the day, so
that stars actually appeared in the sky; and a great earthquake took place
in Bithynia and overthrew the greater part of Nicaea'.
[Phlegon, Olympiades, fragment 17 - quoted by Eusebius, Chronicon;
trans. Fotheringham, (1920b).]
360 10 Records of eclipses in ancient European history

Both the date and place of observation present difficulties. Fothering-

ham, following Kepler, identified the eclipse as that of AD 29 Nov 24
(the first year of the 202nd Olympiad). This was the only large eclipse
visible in the eastern Mediterranean for many years around this time. The
computed time of maximal phase in this region (approximately 11.0 h
using equation (10.1)), is in fair accord with that recorded by Phlegon
(i.e. the sixth hour). Although the date thus seems firmly established,
the place of observation is unknown. Phlegon, a citizen of Tralles (lat.
= 37.85 deg, long. = —27.85 deg) wrote more than a century after the
event. In selecting Nicaea (lat. = 40.42 deg, long. = —29.60 deg), where
the earthquake occurred, as the possible place where the eclipse was seen,
Fotheringham considered this inference 'weaker than any of the the other
presumptions' which he had used.
Computation using equation (10.1) (AT = 9050 sec) indicates that the
eclipse of AD 29 was indeed fully total at Nicaea (at 10.8 h), while at
Tralles the phase would be virtually total (0.997) at 11.0 h. Quite possibly
the eclipse recorded by Phlegon may have been witnessed in this region
but the precise place of observation is purely a matter of speculation. A
reliable solution for AT thus cannot be made.

(8) AD 71 Mar 20? (total: 1.01): Chaeroneia?

The Greek philosopher and biographer Plutarch gives a vivid account of
a total eclipse in one of his dialogues entitled The Face on the Moon. In
this same work he also makes a brief reference to the corona.
(i) (Lucius) smiled thereat and said... 'Now grant me that nothing that
happens to the Sun is so like its setting as a solar eclipse. You will
if you call to mind this conjunction recently which, beginning just
after noonday, made many stars shine out from many parts of the
sky and tempered the air in the manner of twilight. If you do not
recall it, Theon here will cite us Mimnermus and Cydias,
Archilochus and Stesichorus besides, and Pindar, who during
eclipses bewail "the brightest star bereft" and "at midday night
falling" and say that the beam of the Sun <is sped> the path of
shade'.
[Plutarch, The Face on the Moon, 931D-E;
trans. Cherniss and Helmbold (1957, vol. XII, pp. 117-119).]

(ii) Even if the Moon, however, does sometimes cover the Sun entirely,
the eclipse does not have duration or extension; but a kind of light
is visible about the rim which keeps the shadow from being
profound and absolute.
[Plutarch, The Face on the Moon, 932E;
trans. Cherniss and Helmbold (1957, vol. XII, p. 121).]
 10.2 Solar eclipses analysed by Fotheringham 361

The descriptions of eclipses by Archilochus and Pindar mentioned in

(i) have already been discussed above. Regrettably, the relevant works by
the poets Mimnermus, Cydias and Stesichorus - all of whom lived c. 600
BC - no longer survive.
Cherniss and Helmbold (1957, vol. XII, p. 11) argued that if Plutarch
was indeed referring to the corona in (ii), his description is 'remarkably
tame'. They suggested instead that the account is more likely to refer
to an annular eclipse. However, surprising as it often seems to modern
astronomers who have witnessed total obscurations of the Sun, the corona
does not appear to have left much of an impression on observers in
ancient and medieval times. Before AD 1600, only one other account of a
total eclipse (AD 968) definitely mentions the corona, even though many
detailed descriptions of total eclipses are preserved in medieval European
and Arabic chronicles (see chapters 11 and 12). The record from AD
968, which originates from Constantinople, likens the corona to 'a certain
dull and feeble glow, like a narrow headband, shining around the extreme
portion of the edge of the disk' (see chapter 11). Plutarch's description is,
in fact, not too dissimilar from this much later record. By comparison,
during a central annular eclipse the unobscured portion of the Sun is
dazzling in brightness. The ring phase could scarcely be described as
keeping 'the shadow from being profound and absolute' since often there
is hardly any noticeable reduction in daylight.
Fotheringham, and later Cherniss and Helmbold, regarded the first
statement - (i) above - as a reference to a real event (rather than some
product of Plutarch's imagination as suggested by Newton (1970, pp.
115-117). Muller (1975), who had personally witnessed several total
eclipses from various sites, thought that account had 'the definite flavour
of personal experience and eye-witness description... the probability that
this is a real record is very high'. There are sound reasons for believing
that the eclipse was authentic. Plutarch is the only known ancient writer
to allude to the corona, and he is unique among classical authors in noting
a fall in temperature during a total eclipse. His whole account, which also
describes the appearance of stars in different parts of the sky, is so original
that it seems highly likely that either Plutarch himself or one of his close
associates was an eye-witness to the events which he so vividly describes.
In Plutarch's dialogue, the eclipse was the basis of an intellectual - if
somewhat entertaining - discussion. Most of the characters identified
in the text are known to have been associates of Plutarch himself (see
Cherniss and Helmbold (1957, pp. 3ff.)).
On the basis of a careful consideration of the content of Plutarch's nu-
merous dialogues, Fotheringham considered it 'a reasonable presumption
that the eclipse was total at either Delphi or Chaeroneia'. He further
regarded AD 71 Mar 20 as the only viable date. However, although there
362 10 Records of eclipses in ancient European history

can be little doubt that a total obscuration of the Sun is described above
- a lesser phase could scarcely cause many stars to 'shine out from many
parts of the sky' - both the date and place of observation require careful
discussion.
Plutarch was born at Chaeroneia in Boeotia around AD 46 and died
after AD 119. Although he was normally resident in Chaeroneia through-
out his life, he is known to have travelled throughout central Greece. He
also paid several official visits to Rome and lectured on philosophy there.
Plutarch had close links with the Athenian Academy, while from about
AD 95 he held a priesthood for life at Delphi - not far from Chaeroneia.
The beginning of Plutarch's dialogue entitled The Face on the Moon is
lost and, with it, any indication of date or place. Fotheringham pointed
out that on this occasion, Plutarch's brother Lamprias - who also resided
at Chaeroneia - presided over the assembled company. However, Lu-
cius, to whom the reference to the eclipse is attributed, appears in two
other dialogues based in Rome. Both of these details should probably
be regarded as circumstantial. Plutarch's many dialogues are usually set
in various places in central Greece, but sometimes in Rome - the places
with which he himself was familiar. In view of the fact that the 're-
cent conjunction' was so clearly remembered, it seems highly likely that
totality was witnessed at one of these locations. The place of observa-
tion of the eclipse will thus be assumed to be either central Greece or
Rome. More remote places, such as Alexandria - which Plutarch is only
known to have visited once - would seem much less likely. In particular,
Plutarch seldom appeals to his Alexandrian experiences in his numerous
dialogues.
During the lifetime of Plutarch, only four eclipses could have been
total in the central or eastern Mediterranean: AD 59 Apr 30, 71 Mar
20, 75 Jan 5 and 83 Dec 27 - as may be judged from the charts of von
Oppolzer (1887). In all probability, a choice must be made between one
of these alternatives. The eclipse of AD 59 occurred when Plutarch was
only about 13 years old and thus would no longer be 'recent' when he
wrote his dialogue. However, although less likely than the other selected
dates, it cannot be discounted on these grounds alone.
The magnitude, local time and solar altitude for each eclipse at
Chaeroneia (taken as representative of central Greece) and Rome - as
computed from equation (10.1) - and the AT limits (LL and UL) in sec-
onds necessary to produce totality at either place are listed in table 10.5.
With reference to table 10.5, no value of AT will produce a central
eclipse at either Chaeroneia or Rome in AD 59 or at Rome in AD 83.
As clearly depicted by Ginzel (1899), the track of totality in AD 59 ran
almost parallel to the equator throughout much of the Mediterranean.
Regardless of the precise value of AT at the time, the eclipse can never
 10.2 Solar eclipses analysed by Fotheringham 363

Table 10.5 Investigation of solar eclipses during the life of Plutarch.

AT Range (sec)
Year AD Place Mag. LT Alt. LL UL
59 Chaeroneia 0.92 15.3 +40 — —
Rome 0.81 14.3 +49 — —
71 Chaeroneia 1.01 11.0 +48 9 500 9 600
Rome 0.80 10.1 +40 6250 6 350

75 Chaeroneia 0.86 16.1 +6 12000 12 200

Rome 0.94 15.3 +11 8000 8 200
83 Chaeroneia 0.80 14.2 +21 3 250 3 650
Rome 0.71 13.1 +23 — —

have been total north of latitude 36.3 deg N. The central zone would thus
pass far to the south of Rome and the Italian peninsula and also a little
to the south of the Grecian peninsula. Even at Sparta (the scene of other
dialogues by Plutarch), which was much better placed than Chaeroneia
in AD 59, the magnitude cannot have exceeded 0.96, which is far from
sufficient to produce the effects described.
For values of AT which indicate totality at Rome in AD 83, the Sun
would set before greatest phase was reached. At Chaeroneia, a highly
discordant range of AT (around 3500 sec) would be needed to produce
totality and here maximum eclipse would occur only a few minutes before
sunset - with the solar altitude no more than 3 deg. Thus the description
'beginning just after noonday' would be far from appropriate.
The circumstances in AD 75 are also unfavourable. Computations using
the AT results listed in table 10.5 indicate that totality at both Rome and
Chaeroneia would again occur in the late afternoon. At Rome, the local
time would be 15.7 h with a solar altitude of only 7 deg (compared
with the meridian altitude of 25 deg). In Chaeroneia, the local time
would be 15.2 h and the altitude of the Sun 14 deg (meridian altitude
29 deg). Since it is likely that the eclipse would not be noticed until it
was already well advanced, neither of these situations could be described
as close to midday. Although the AT range required to achieve totality
at Rome (around 8100 sec) is in fair agreement with figure 10.1, that for
Chaeroneia (some 12 100 sec) is discordant; AT had already decreased to
between 11210 and 12 120 sec by 136 BC, as indicated by the Babylonian
observation of a total solar eclipse at that date.
The eclipse of AD 71 is a much more promising candidate. Reference
to figure 10.1 shows that the AT limits necessary to secure totality at
364 10 Records of eclipses in ancient European history

Chaeroneia (close to 9550 sec) are quite acceptable. However, for Rome,
a value for of only around 6300 sec would be needed. Anticipating the
results from the analysis of the careful timed observations at Alexandria
in AD 364 - see the immediately following entry - a figure for AT of as
much as 8300 sec is indicated at this late date.
The only difficulty with the eclipse of AD 71 as seen from Chaeroneia
is that for totality at this site greatest phase would occur around 11.0 h,
rather than 'just after noonday'. However, the Sun would then be almost
at its height - altitude 48 deg, or only 3 deg less than the meridian altitude.
There is no suggestion in the record that time was carefully measured; to
the casual bystander this eclipse would be regarded as occurring close to
midday. Although the duration of totality would not exceed 40 sec, this
would be sufficient to render several stars visible, with Venus (mag. —3.5)
prominent, 37 deg to the east of the Sun.
In summary, a date for the 'eclipse of Plutarch' of AD 71 Mar 20 seems
very likely. However, since the historical details are somewhat tenuous
and the above date has been derived only as the result of extensive
astronomical computations, it would seem scarcely justifiable to use the
eclipse to make any deductions on the value of AT.

(9) AD 364 Jun 16: Alexandria

In his commentary on the Almagest, Theon of Alexandria (c. AD 370),
records a solar eclipse which he observed during the afternoon.

... the exact ecliptic conjunction which we have discussed, and which
took place according to the Egyptian calendar in the 1112th year from the
reign of Nabonassar, 2 | equal or equinoctial hours after midday on the
24th of Thoth, and according to the Alexandrian calendar reckoned by
simple civil days in the 1112th year of the same reign, 2 | equal or
equinoctial hours after midday on the 22nd of Payni... And moreover we
observed with the greatest certainty the time of the beginning of contact,
reckoned by civil and apparent time, as 2 | equinoctial hours after midday,
and the time of the middle of the eclipse as 3 | hours, and the time of
complete restoration as \\ hours approximately after the said midday on
the 22nd of Payni.
[Theon of Alexandria, 332; trans. Fotheringham, (1920b).]
This account follows very much the style of the eclipse records in
the Almagest itself (see chapter 4 and section 10.4 below). Fotheringham
converted the recorded date of the conjunction to the Julian calendar (AD
364 Jun 16) and this corresponds exactly with that of a tabular solar eclipse
(e.g. as listed by von Oppolzer, 1887). For a more recent discussion of
the eclipse, see Tihon (1976-77). There seems no reason for doubting that
Alexandria, where Theon lived, was the place of observation - especially
 10.3 Other ancient reports of large solar eclipses 365

since the first person plural is used in referring to the measurements.

Fotheringham regarded the event as 'the only ancient eclipse of the Sun
for which an astronomically observed time is recorded'. This remark is no
longer valid in view of the many similar Babylonian reports which have
since become available - see chapter 5. However, it is still the only solar
eclipse for which careful measurements of time are available from ancient
Europe.
The local times of the various phases are all conveniently expressed in
equal hours after midday.

RESULTS

(i) Local time of first contact on Jun 16 = 14.83 h, UT = 12.79 h.

Computed TT = 15.09 h, thus AT = 8100 sec.
(ii) LT of mid-eclipse (i.e. greatest phase) = 15.64 h, UT = 13.76 h.
Computed TT = 16.01 h, thus AT = 8300 sec.
(iii) LT of last contact = 17.39 h, UT = 14.46 h. Computed TT = 16.79
h, thus AT = 8400 sec.
These three results are remarkably self-consistent - evidence of Theon's
care. (NB at Alexandria the computed magnitude = 0.38.)
In summary, of the material investigated by Fotheringham in his 1920b
paper, the last observation - the timed eclipse of AD 364 - is the most
reliable. In almost every other case there are severe difficulties of inter-
pretation.

10.3 Other ancient reports of large solar eclipses

There are many further allusions to solar eclipses in ancient Greek and
Latin writings which were not considered by Fotheringham in his 1920b
paper. Most of these were discussed by Ginzel (1899). However, only in
a few instances is there any reason for believing that an unusually large
obscuration of the Sun occurred at a known place and on an unambiguous
date. Sources which relate events occurring centuries beforehand are often
unreliable. In particular, Plutarch recorded several solar eclipses which
were seen long before his own time. He had a habit of including allusions
to darkness which were not mentioned in accounts of the same event by
earlier writers (notably in 431 and 364 BC). Perhaps he was influenced by
his experiences of the total eclipse of AD 71 (see section 10.2).
From the large array of material, I have selected only the following
three records since they seem to indicate a particularly large eclipse of the
Sun and the place and date are at least reasonably well established.
366 10 Records of eclipses in ancient European history

(1) BC 394 Aug 14 (annular, 0.94): near Chaeroneia

This eclipse is recorded as partial both by the contemporary writer
Xenophon in his Hellenica and by Plutarch in his Life of Agesilaus.
When the eclipse occurred, the army of King Agesilaus II of Sparta was
on its way to engage in battle with the Thebans near Chaeroneia, having
marched southwards through Thrace.

(i) Next day he (Agesilaus) crossed the mountains of Achaea Phthiotis

and for the future continued his march through friendly territories
until he reached the confines of Boeotia. Here at the entrance of
that territory, the Sun seemed to appear in a crescent shape
[Xenophon, Hellenica IV, 3, 10; trans. Dakyns (1892, vol. II, p. 54).]

(ii) Agesilaus now marched through the pass of Thermopylae,

traversed Phocis, which was friendly to Sparta, entered Boeotia,
and encamped near Chaeroneia. Here a partial eclipse of the Sun
occurred... After advancing as far as Coroneia and coming in sight
of the enemy...
[Plutarch, Life of Agesilaus, XVII;
trans. Perrin (1961, vol. V, p. 47).]
The date of the arrival of Agesilaus in Boeotia is fixed by Diodorus
Siculus (XIV, 82) as the archonship of Diophantos, the second year of the
96th Olympiad, and thus 395/4 BC (Bickerman, 1980, p. 138).
Xenophon accompanied Agesilaus on his campaign and thus probably
witnessed the eclipse himself. He does not give the place of observation
more precisely than the border of Phocis and Boeotia. However, the
subsequent battle took place in the plain of Coroneia, just within the
border of Boeotia. Plutarch, who clearly relies mainly on Xenophon's
narrative, names the place as Chaeroneia, his own home town. He
evidently deduced this from a consideration of the route followed by
Agesilaus. From the map published by Stier and Kirsten (1956, p. 19),
Chaeroneia was almost on the border of Boeotia and Phocis and about
10 km to the NW of the site of the battle. The eclipse must thus have
been seen within a few kilometres of Chaeroneia.
Calculations using equation (10.1) (AT = 15 200 sec) indicate that at
Chaeroneia the eclipse would reach a magnitude of 0.92 at 9.6 h with a
solar altitude of 52 deg. Although the Sun would be high in the sky at the
time, Agesilaus would be marching in a south-easterly direction, facing
into the Sun, so that the eclipse would be more easily noticeable.

RESULTS
For a partial eclipse at Chaeroneia, AT < 12 710 or > 14250 sec.
 10.3 Other ancient reports of large solar eclipses 367

(2) BC 188 Jul 17 (total, 1.01): Rome

The great Roman historian Livy (Titus Livius) in his work entitled 'From
the Founding of the City' records an unusual darkness in the city of Rome
soon after the Ides of March.

Then, when Marcus Valerius Messala and Gaius Livius Salinator had
been inaugurated as consuls on the Ides of March, they consulted the
senate... Before the new magistrates (i.e. the consuls) departed for their
provinces, a three-day period of prayer was proclaimed in the name of the
College of Decemvirs at all the street corner shrines because in the
daytime, between about the third and fourth hours, darkness had covered
everything (tenebrae obortae fuerant).
[Livy, XXXVIII 36, 4; trans. Sage (1936, vol. XI, pp. 117-119).]
Marcus Valerius Messala and Gaius Livius Salinator were consuls in
566 AUC or 188/7 BC (Bickerman, 1980, p. 147). Although Livy (c. 60
BC - AD 17) wrote long after the eclipse, he researched his material with
considerable care. Livy offers no explanation for the darkness, but the
eclipse of Jul 17 in that year was certainly extremely large in Rome. As is
well known, the Roman calendar was in chaos around this period so that
it is quite plausible that the Ides of March could have taken place in the
summer.
Although a more accurate rendering of the phrase tenebrae obortae
fuerant is 'darkness had arisen', it seems clear from the declaration of the
three-day period of prayer at shrines throughout the city of Rome that
the darkness was awe-inspiring. If the eclipse interpretation is correct,
the lack of reference to such an event could possibly be explained by its
occurrence on an overcast day. Computations using equation (10.1) (AT
= 12 500 sec) indicate that at Rome the eclipse was virtually total (mag.
= 0.995) at 6.3h, the solar altitude then being 17 deg. Since sunrise was
at 4.48 h, greatest phase would actually be reached during the second
(seasonal) hour of the day, but it is clear that the recorded time is only
intended to be very approximate.
Although the eclipse of 188 BC was almost certainly the cause of the
darkness by day in Rome, the circumstances are too vague to warrant
deduction of any AT limits from the record.

(3) AD 484 Jan 14 (total 1.03): Athens

A very large eclipse of the Sun occurring in AD 484 was regarded as an
omen of the death of the eminent Greek philosopher, Proclus. This event
is described by Marinus Neapolitanus in his Life of Proclus. Marinus
became head of the Athenian School of Philosophy in AD 485 on the
death of Proclus.
368 10 Records of eclipses in ancient European history

A year before his death there were various omens. There was an eclipse
of the Sun which was so pronounced as to turn day into night and the
darkness was deep enough for the stars to become visible; it occurred in
the eastern horn of the sign of Capricorn. And the almanacs predicted
another eclipse that would occur after the first year. They say that such
events that are observed to happen in the heavens are indicative of things
that happen on the earth; so that these eclipses clearly foretold us of the
privation and departure as it were of the light of philosophy.
[Marinus, Life of Proclus, chap. 37; trans. Rosan (1949, p. 34).]
According to Marinus (chap. 36), Proclus died in Athens 124 years
after the accession of Emperor Julian (AD 361) and in the year when
Nicagoras the Younger was the archon at Athens (AD 484/5). The exact
date of Proclus' death is specified as the 17th day of the Athenian month
Munychion and the 17th day of the Roman month April - i.e. AD 485
Apr 17.
The only large eclipse visible in Greece around the stated date occurred
on AD 484 Jan 14. The Sun (long. = 295 deg) would indeed be in
Capricorn at the time. If we calculate on the basis of equation (10.1) (AT
= 5500 sec) for Athens, the eclipse would reach greatest phase (mag. 0.997)
at 7.2 h, with the Sun barely above the eastern horizon (alt. = 0.8 deg). It
is unfortunate that the record does not specify whether or not the eclipse
was total. However, such ambiguity is not unexpected. In January, the
Sun remains invisible for more than half an hour after sunrise at Athens,
since it is obscured by Mt Hymettus (1 km high and about 8 km ESE of
the city). Thus only the last stages of the eclipse would actually be seen.
As it happens, no star or planet brighter than about mag. 0 would be
visible during this eclipse. The altitude of Venus would be very similar to
that of the Sun so that the planet would also be hidden by Mt Hymettus
when the eclipse was at its height. Both Jupiter and Sirius would be far
below the western horizon. Mercury (mag. +0.3), 26 deg to the west of
the Sun would be well placed for visibility, but although both Mars and
Saturn would both be above the horizon, they would be faint (magnitudes
respectively +1.5 and +0.8). Hence, although the solar altitude was so
low, a very large magnitude would be required to render several stars
visible. Unfortunately, because of the lack of direct reference to the degree
of obscuration of the Sun, it is not possible to use the observation to set
any firm limits to the value of AT at this epoch.

10.4 Greek lunar eclipse observations recorded in Ptolemy's Almagest

The timed Babylonian observations of eclipses recorded in Ptolemy's
Mathematike Syntaxis, or Almagest, have already been discussed in detail
in chapter 4. Similar measurements by Greek astronomers between 201
 10.4 Greek lunar eclipse observations in the Almagest 369

BC and AD 145 (all relating to eclipses of the Moon) are also reported
in various sections of the Almagest (IV, 9; IV, 11; and VI, 5). These data
were considered by Fotheringham (1920a) and later by Newton (1970, pp.
100 ff.), but it seems desirable to make a full re-investigation here.
Five of the Greek observations were made at Alexandria, while one
(141 BC) is from Rhodes. Very probably, the observer at Rhodes was
Hipparchus himself. Ptolemy (Almagest, III, 1) alludes to two further
lunar eclipses in 146 and 135 BC which Hipparchus presumably observed,
but no times are preserved.
As in the case of the Babylonian eclipses recorded by Ptolemy, the dates
of the Greek observations are expressed in terms of the 365-day Egyptian
year. This enabled the exact number of days between any two selected
epochs to be readily calculated. Some years are numbered from the era of
Nabonassar of Babylon (747 BC), but on several occasions the Kallippic
cycle is used. This cycle was named after the fourth century BC Athenian
astronomer Kallippos, and was closely equal to four Metonic cycles - each
of 19 years. The first Kallippic cycle began in 330 BC. Ptolemy makes
reference to the two subsequent cycles, starting in 254 and 178 BC. Each
commenced around the time of the summer solstice.
In the various translations given below, which are all quoted from
Toomer (1984), the Egyptian month numbers are given in Roman nu-
merals after the month names. Toomer converted each date to the Julian
calendar, using negative years in the BC period. In every case, the reduced
date is in exact accord with that of a tabular eclipse.
In 201 BC, the Moon was seen to rise already eclipsed. Analysis of
this observation follows the pattern outlined in chapter 7; note that the
computed local time of moonrise is a function of AT, as discussed in
earlier chapters.

(1) BC 201 Sep 22/23 (mag. = 0.73): Alexandria

We will pass to the second set of three eclipses he (Hipparchus) set out,
which he says were observed in Alexandria. He says that the first of these
occurred in the 54th year of the Second Kallippic Cycle, Mesore [XII] 16
in the Egyptian calendar [—200 Sep 22]. In this eclipse the Moon began to
be obscured half an hour before it rose, and its full light was restored in
the middle of the third hour [of night]...
[Almagest, IV, 11; trans. Toomer (1984, p. 214).]
The time of onset may have been estimated from the degree of ob-
scuration of the Moon when it rose. Both the statement that the Moon
rose eclipsed and the measurement that it ended 2^ hours after sunset
(i.e. the middle of the third hour) are of value in the determination of
AT. It is evident from the magnitude of the time-interval between sunset
and last contact that the Moon rose before maximal phase was reached.
370 10 Records of eclipses in ancient European history

Use of ordinal numerals in expressing the time of last contact implies the
use of seasonal hours and this will be assumed here. However, as the
eclipse took place so close to the equinox (which would occur on Sep 26),
whether seasonal or equinoctial hours is used is unimportant. Since the
seaport of Alexandria lies at the head of the flat plain of the Nile Delta,
no allowance for horizon profile is necessary.

RESULTS
(1) First contact on Sep 22 before moonrise. LT of moonrise = 18.02 h,
hence LT of contact < 18.02 h, U T < 15.93 h. Computed TT = 18.85 h,
thus A T > 10500 sec.
(ii) Maximal phase after moonrise. LT of moonrise = 18.06 h, hence
LT of contact > 18.06 h, U T > 15.97 h. Computed TT = 20.39 h, thus
A T < 15 900 sec.
Combining these limits, 10 500 < A T < 15 900 sec.
(hi) Last contact at 2.5 seasonal hours after sunset. LT of sunset =
18.12 h, hence length of night = 11.76 h, 1 seasonal hour = 0.98 h. LT of
last contact = 20.57 h, U T = 18.48 h. Computed TT = 21.92 h, thus A T
= 12400 sec.

(2) BC 200 Mar 19/20 (mag. = 139): Alexandria

... He says that the next eclipse occurred in the 55th year of the same
cycle, Mechir [VI] 9 in the Egyptian calendar [—199 Mar 19], that it began
when 51 hours of night had passed, and was total...
[Almagest, IV, 11; trans. Toomer, p. 214.]
The terminology again suggests the use of seasonal hours, but so close
to the equinox the precise units are unimportant.
RESULTS
First contact on Mar 19 at 5.33 seasonal hours after sunset. LT of sunset
= 17.99 h, hence length of night = 12.02 h, 1 seasonal hour = 1.00 h. LT
of first contact = 23.32 h, UT = 21.48 h. Computed TT = 0.76 h (Mar
20), thus AT = 11 800 sec.

(3) BC 200 Sep 11/12 (mag. = 1.59): Alexandria

... He says that the third eclipse occurred in the same (55th) year of the
Second Cycle, on Mesore [XII] 5 in the Egyptian calendar [—199 Sep 11]
and that it began when 6 | hours of the night had passed, and was total.
He also says that mid-eclipse occurred at about 8 | hours of night, that is
2^ seasonal hours after midnight
[Almagest, IV, 11; trans. Toomer, p. 215.]
The time of mid-eclipse was presumably derived from measurements
made at second and third contact.
 10A Greek lunar eclipse observations in the Almagest 371

RESULTS
(i) First contact on Sep 12 at 6.67 seasonal hours after sunset. LT of
sunset = 18.30 h, hence length of night = 11.40 h, 1 seasonal hour = 0.95
h. LT of first contact = 0.64 h, UT = 22.60 h (Sep 11). Computed TT =
2.12 h (Sep 12), thus AT = 12650 sec.
(ii) Mid-eclipse at 8.33 seasonal hours after sunset. LT of mid-eclipse =
2.21 h, UT = 0.17 h. Computed TT = 3.94 h, thus AT = 13 550 sec.

(4) BC 174 Apr 30/May 1 (mag. = 0.62): Alexandria

In the seventh year of Philometor, which is the 574th from Nabonassar,
on Phamenoth [VII] 27/28 in the Egyptian Calendar [-173 May 0/1],
from the beginning of the eighth hour till the end of the tenth in
Alexandria, there was an eclipse of the Moon which reached a maximum
obscuration of 7 digits from the north, so mid-eclipse occurred 2^ seasonal
hours after midnight, which corresponds to l\ equinoctial hours....
[Almagest, VI, 5; trans. Toomer, p. 283.]

RESULTS
(i) First contact on May 1 at 7 seasonal hours after sunset. LT of sunset
= 18.62 h, hence length of night = 10.76 h, 1 seasonal hour = 0.90 h. LT
of first contact = 0.90 h, UT = 22.82 h (Apr 30). Computed TT = 1.88 h
(May 1), thus AT = 11000 sec.
(ii) Last contact at 10 seasonal hours after sunset. LT of contact = 3.59
h, UT = 1.51 h. Computed TT = 4.50 h, thus AT = 10750 sec.
NB Ptolemy's deduction that 2j seasonal hours was equal to 2 | equinoc-
tial hours is only approximate. A better equivalence would be l\ seasonal
hours.
(5) BC 141 Jan 27/28 (mag. = 0.26): Rhodes
... Again, in the thirty-seventh year of the Third Kallippic Cycle, which
is the 607th from Nabonassar, Tybi [V] 2/3 in the Egyptian Calendar
[-140 Jan 27/28], at the beginning of the fifth hour [of night] in Rhodes,
the Moon began to be eclipsed; the maximum obscuration was 3 digits
from the south. Here then, the beginning of the eclipse was 2 seasonal
hours before midnight, which corresponds to l\ equinoctial hours....
[Almagest, VI, 5; trans. Toomer, p. 284.]

RESULTS
First contact on Jan 27 at 4 seasonal hours after sunset. LT of sunset =
17.07 h, hence length of night = 13.86 h, 1 seasonal hour = 1.16 h. LT of
first contact = 21.69 h, UT = 20.09 h. Computed TT = 22.47 h, thus AT
= 8550 sec.
NB 2 seasonal hours corresponded better to l\ equinoctial hours than
A-
372 10 Records of eclipses in ancient European history

(6) AD 125 Apr 5/6 (mag. = 0.15): Alexandria

The first eclipse we used is the one observed in Babylon... The second
eclipse we used is the one observed in Alexandria in the ninth year of
Hadrian, Pachon [IX] 17/18 in the Egyptian Calendar [125 Apr 5/6], 3§
equinoctial hours before midnight. At this eclipse too the Moon was
obscured | of its diameter from the south.
[Almagest, IV, 9; trans. Toomer, p. 206.]
In the subsequent account it is implied that the time refers to mid-eclipse.
The date of the 'first eclipse' corresponds to 491 BC (see chapter 3).

RESULTS
Mid-eclipse on Apr 5 at 3.60 equinoctial hours before midnight, hence LT
= 20.40 h, UT = 18.45 h. Computed TT = 21.32 h, thus AT = 10 350 sec.

10.5 The moonrise eclipse of BC 331 Sep 20/21

Several Greek and Roman writers record that an eclipse of the Moon
happened about the time of a battle between the army of Alexander the
Great and the Persian forces at Arbela (near the site of Nineveh, lat. 35.9
deg N, long. 44.3 deg E). Only Pliny the Elder (Gaius Plinius Secundus)
notes that on the same evening the Moon was seen to rise eclipsed in
Sicily. The observation at Arbela is of considerable historical importance
since it enables the exact date of the battle to be fixed. However, only
the Sicilian report is of any value in determining AT. In a section of his
Natural History discussing the effect of the curvature of the Earth (and, in
particular, of terrestrial longitude) on the visibility of eclipses, Pliny gives
the following account:

Consequently inhabitants of the East do not perceive evening eclipses of

the Sun and Moon, nor do those dwelling in the West see morning eclipses,
while the latter see eclipses at midday later than we do. The victory of
Alexander the Great is said to have caused an eclipse of the Moon at
Arbela at 8 p.m. (noctis secunda hora) while the same eclipse in Sicily was
when the Moon was just rising (exoriens)... this was because the curve of
the globe discloses and hides different phenomena for different localities.
[Pliny, Natural History, II, 72; trans. Rackham (1938, vol. I, p. 313).]
In the same passage, Pliny also describes a relatively recent solar eclipse
(AD 59 April 30) that was visible in Campania (southern Italy, long. 15
deg E) between the 7th and 8th hours, while in Armenia (long. 45 deg E)
it occurred between the 10th and 11th hours.
Writing c. AD 70, and thus fully four centuries after the lunar obscu-
ration, Pliny does not specify his source. As noted above (section 10.2),
 10.5 The moonrise eclipse of BC 331 Sep 20/21 373

Table 10.6 AT results obtained from ancient

Greek timings of lunar and solar eclipses.

Year Type Ct AT (sec)

-200 lunar 1 12400
-199a 1 11800
-199b 1 12650
-199b M 13 550

-173 1 11000
-173 4 10750
-140 1 8 550
+125 M 10 350

+364 solar 1 8100

+364 M 8 300
+364 4 8400

Ptolemy (Geography, I, 4) stated that the eclipse was seen at Arbela in the
fifth hour but took place at Carthage in the second hour.
The date of the battle is given by Arrian (Anabasis, II, 7.6) as during
the month Pyanopsion when Aristophanes was archon at Athens. Aristo-
phanes held the position of archon in 331/0 BC (Bickerman, 1980, p. 139).
Pyanopsion was the fourth month in the Athenian calendar so that, since
the first month of the year began in midsummer, the date of the lunar
eclipse can be firmly established as BC 331 Sep 20/21. According to
Plutarch (Life of Alexander, XXXI) the eclipse preceded the battle by 11
days. Hence the battle itself would take place on Oct 1.
It is a pity that the precise place of observation in Sicily is not recorded.
I have taken as centres for calculation the major cities of Syracuse in
the east of Sicily and Panormus (i.e. Palermo, lat. = 38.13 deg, long. =
—13.30 deg) in the west of the island. How far advanced the eclipse was
at moonrise is not stated. Hence it can only be assumed that the Moon
rose at some time between first and last contact. Of course, no allowance
for horizon profile can be made.

RESULTS Syracuse

(i) First contact on Sep 20 before moonrise. LT of moonrise = 18.12 h,

hence LT of contact < 18.12 h, UT < 17.02 h. Computed TT = 20.77 h,
thus AT > 13 500 sec.
(ii) Last contact after moonrise. LT of moonrise = 18.20 h, hence LT
374 10 Records of eclipses in ancient European history

Table 10.7 AT limits obtained from ancient Euro-

pean untimed observations of lunar and solar eclipses.

AT Range (sec)
Year Type LL UL
-430 solar 12 650 11800
-393 solar 14250 12710
-330 lunar 13 000 25050
-309 solar 13 300 17 200
-200 lunar 10 500 15 900

Key
1 Central eclipse
! Partial eclipse
16000 V o Solar eclipse time
• Lunar eclipse time
14000 • A Lower limit (untimed)
A
: v Upper limit (untimed)
_ 12000
A •
& 10000
• §
8000

6000

4000

2000

-500 -400 -300 -200 -100 0 100 200 300 400 500
Year
Fig. 10.6 AT values and limits derived from the more reliable ancient European
eclipse observations.

of contact > 18.20 h, UT > 17.11 h. Computed TT = 0.07 h (Sep 21),

thus AT < 25 050 sec.
Combining these limits yields 13 500 < AT < 25 050 sec.

RESULTS Panormus
(i) First contact before moonrise. LT of moonrise = 18.12 h, hence LT of
contact < 18.12 h, UT < 17.16 h. Computed TT = 20.77 h, thus AT >
13 000 sec.
(ii) Last contact after moonrise. LT of moonrise = 18.21 h, hence LT
of contact > 18.21 h, UT > 17.25 h. Computed TT = 0.07 h (Sep 21),
thus AT < 24 550 sec.
 10.6 Conclusion 375

Combining these limits yields 13 000 < AT < 24 550 sec. Taking the
widest possible limits for both places yields 13 000 < AT < 25050 sec.
An interesting account of the eclipse by Alexander's troops at Arbela is
recorded by Curtius in his History of Alexander. This gives one of the few
known allusions in ancient history to the red colour of the totally eclipsed
Moon.
But about the first watch the Moon in eclipse, hid at first the brilliance
of her heavenly body, then all her light was sullied and suffused with the
hue of blood.
[Curtius, IV, 10, 1; trans. Rolfe (1946, vol. I, p. 253).]
Curtius relates that Alexander's soldiers regarded the phenomenon as
a bad omen and accordingly Alexander ordered his Egyptian soothsayers
to give their opinion. They foretold the defeat of the Persians, and
Alexander's troops, heartened by these disclosures, proved victorious in
the subsequent combat.

10.6 Conclusion
In table 10.6 are summarised the AT results obtained from the various
timed measurements for both lunar and solar eclipses made at Alexandria
and Rhodes between -200 (i.e. 201 BC) and +364 (AD 364). As in earlier
chapters, negative dates are preferred to their BC equivalents.
The few usable results obtained from the untimed observations discussed
in this chapter are listed in table 10.7. In this table, since the solar eclipses
of —430 and —393 were described as partial, a range of AT is excluded.
This is indicated by the lower limit being numerically greater than the
upper limit. For example, in —393 either AT < 12 700 or > 14250 sec.
The results listed in both tables are plotted in figure 10.6. Clearly the
timed data form a useful set, only the observation from Rhodes in —140
(probably by Hipparchus!) being discordant. As will be seen later, the
three measurements by Theon of Alexandria in +364 lie in a gap in which
few other useful observations are preserved from elsewhere.
 11
Eclipse records from medieval Europe

11.1 Introduction
For the purpose of this chapter, the term 'medieval' will be loosely inter-
preted to mean the period between the close of the classical age in Europe
(c. AD 500) and the beginning of the telescopic era. It thus includes
the Renaissance. Numerous observations of both solar and lunar eclipses
were recorded in Europe during this period, especially after AD 1000.
In most cases, the date is accurately reported and the precise place of
observation is known. However, with only a few exceptions, times were
only crudely estimated to the nearest hour or so. Thus most observations
may effectively be regarded as untimed.
As discussed in chapter 3, untimed lunar eclipses are of negligible value
for the determination of AT (unless it is clearly implied that the Moon rose
or set eclipsed). Most of this chapter will be devoted to the investigation
of solar obscurations in which the Sun was either totally or very largely
covered. In section 11.9, a few careful timings of solar eclipses from the
early fourteenth century will also be analysed. Finally, in section 11.10,
an eighth century report of an occultation of the planet Jupiter by the
eclipsed Moon - a very rare event - will be discussed.

11.2 Historical sources

European reports of solar eclipses (as well as lunar obscurations) from
the Middle Ages and Renaissance are mainly found in historical works,
notably chronicles; only a few observations are reported in astronomical
treatises. During this period the chronicle was a major literary form
and numerous works of this kind were compiled at a wide variety of
monasteries and towns scattered throughout much of Europe. Most texts
were composed in Latin, although the vernacular was frequently used -
especially after about AD 1400. Often chronicles extended over several
centuries, sometimes being continued by scribes of widely differing inter-

376
 11.2 Historical sources 317

ests. Many annalists were in the habit of including references to eclipses

and other striking celestial phenomena (e.g. bright comets, meteors, and
displays of the aurora borealis) along with matters of local and national
concern. Despite the rarity of total solar eclipses at any one place, nu-
merous reports of these events are preserved since there was such a large
number of potential observing sites in Europe. In fact, more than two-
thirds of all known medieval accounts of complete obscurations of the
Sun originate from Europe.
As will be evident from some of the records translated below, few
chroniclers (in common with many educated people today) had much
understanding of the cause of eclipses. Further, most observers would
have no advance warning when such an event was about to take place
and might not notice an eclipse until it was already well under way.
Nevertheless, medieval accounts are often remarkably detailed, sometimes
giving a careful description of the complete disappearance of the Sun or
noting the effects of the loss of daylight on people and animals.
A large number of European chronicles have been published in their
original language by editors such as Muratori (1723-) and Pertz (1826-).
Although I have consulted several manuscript sources (in the Vatican
Library), the search for records of specific phenomena occurring at widely
spaced intervals is a very tedious operation. Hence I have relied almost
exclusively on the printed texts. The compilations of Muratori (Scriptores
Rerum Italicarum) and Pertz (Monumenta Germaniae Historica, Scriptores)
will be subsequently abbreviated to SRI and MGH respectively. Other
sources consulted will be cited in full - including secondary sources where
a primary source is relatively inaccessible.
Celoria (1877a, 1877b) and Ginzel (1884a, 1884b, 1918) made ex-
tensive searches of published chronicles for accounts of solar eclipses,
whether of large or small magnitude. They were able to uncover numer-
ous records which they quoted in extenso, leaving them in their original
language. Other literature searches were made about half a century ago
by Thorkelsson (1933), who concentrated on Icelandic chronicles, and
Vyssotsky (1949), who restricted his attention to Russian annals. Recent
careful work by Dr G. R. Levi-Donati of Perugia, Italy and by Dr Marek
Zawilski of the Planetarium and Astronomical Observatory, Lodz, Poland
have brought to light a number of important records of eclipses from the
fourteenth and fifteenth centuries. Levi-Donati's principal papers on this
theme were published in 1987 and 1989, while Zawilski's compilation -
revised in 1994 - is as yet unpublished. I have no doubt that further
literature searches would reveal additional important observations, but
any such effort would be a major undertaking and progress seems likely
to be slow.
Newton (1972b), who provided useful historical notes about many
378 11 Eclipse records from medieval Europe

chronicles, translated most of the records cited by Celoria and Ginzel.

However, several of his quotations are incomplete and his translations
occasionally lack technical accuracy. In this chapter, I have either re-
translated or cited reliable translations of all of the medieval European
texts known to me which either (a) assert that a central eclipse was
witnessed or (b) specifically deny totality or the ring phase.
Chronicles of medieval Europe can be divided into two main types:
regional and local. Regional annals, typified by the Anglo-Saxon Chronicle
and many other British chronicles, drew on a wide area. These often noted
eclipses, but usually the place of observation is very uncertain. In contrast
to the regional annals, monastic and town chronicles - especially those of
continental Europe - were mainly concerned with their own locality. It is
thus usually quite reasonable to assume that eclipses noted in these works
were observed at the centre where the chronicle was compiled. Sometimes
this same place is specifically mentioned in the report of an eclipse, while
occasionally the annalist tells us that he was an eyewitness. In general, if
an obscuration of the Sun was particularly large, descriptions from even
quite neighbouring locations differ in so many details as to be obviously
independent. Annals of monasteries and towns provide the main source
of the observations discussed in this chapter.
A few untimed reports of total and annular eclipses from the sixteenth
and early seventeenth century are contained in astronomical treatises by
Clavius, Kepler, and others. These observations will also be considered,
along with a few unusually careful timings of solar eclipse contacts by the
fourteenth century French astronomer Johannes de Muris.

11.3 Chronological discussion

The Venerable Bede (early eighth century) was the first historian to sys-
tematically number years relative to the Christian Era, and this practice
was adopted by most medieval European chroniclers. However, Spanish
historians often used the inception of the Julian calendar (38 BC) instead,
while both Byzantine and Russian annalists counted years from the as-
sumed era of the creation, i.e. Sep 1 in BC 5509. Michael the Syrian, who
was Patriarch of Antioch in the twelfth century, recorded many celestial
phenomena in his chronicle - including total eclipses of the Sun in AD
812 and 1176. He followed the practice of the Eastern Orthodox Church
by counting years from the 'era of Alexander', actually a standardised
Seleucid era; on this scheme, year 1 was taken to commence on Oct 1 in
312 BC. (NB This date is very close to the Seleucid era on the ancient
Macedonian calendar - i.e. Oct 7 in 312 BC.)
The Julian calendar, with its associated Kalends (Kalendae), Nones
(Nonae) and Ides (Idus), was widely adopted in medieval Europe. In the
 11.3 Chronological discussion 379

ancient Roman calendar, the Ides occurred around the middle of the lunar
month, while the Nones was so-called because it was the ninth day before
the Ides. The Kalends was the first of each month. In every month of
the year except March, May, July and October the Nones occurred on
the 5th day and the Ides on the 13th. However, in the four months just
mentioned, both the Nones and Ides took place two days later than usual
(i.e. on the 7th and 15th).
Dates in a month were counted retrospectively from each of these three
reference days. The day before the Kalends (i.e. the last day of the previous
month) was referred to as Pridie Kalendas. Since inclusive counting was
used, the day before the Pridie was known as the 'third day before the
Kalends' (usually abbreviated to / / / Kalendas) and so on. For example,
Christmas Day was the eighth day before the Kalends of January (VIII
Kalendas Januarias). Similar rules applied for the Nones and Ides, the
day before the Nones being Pridie Nonas and before the Ides Pridie Idus.
However, in these cases the date specified occurred in the same month as
the Nones or Ides. Thus the date of the eclipse of AD 840 May 5 was
recorded as / / / Nonas Maias - see section 11.6.1.
The operational rules of the Julian calendar are summarised in table
11.1. The general layout of this table is based on table 12.82.1 in the
Explanatory Supplement to the Astronomical Almanac (Seidelmann, 1992).
It should be noted that the details listed for the month of February are for
a common year (not divisible by 4), when February would have 28 days.
Under these circumstances, Feb 28 would be Pridie Kalendas Martias,
while in a bissextile or leap year Feb 28 would be /// Kalendas Martias
and Feb 29 Pridie Kalendas Martias.
In medieval Italy, days were often counted forwards from the beginning
of a month (intrante) or backwards from the end (stante), depending on
which was nearer. However, some chroniclers anticipated our present-day
system of counting straight through the month. The weekday is often
recorded in European chronicles. Sunday was the first day of the week;
this was normally termed Dies Dominica. However, the other weekdays
were usually simply rendered as feria 2 (Monday), feria 3 (Tuesday), and
so on - see table 11.2.
Most recorded dates when converted to the Julian calendar (or the
Gregorian calendar from AD 1582) are in exact accord with the tabular
dates of eclipses which according to calculation were large in Europe. In
particular, for all of the records translated in this chapter, the weekday
is invariably given correctly. Saint's days and other Christian festivals
were often used to mark days; the equivalent Julian dates are specified in
individual entries below.
Although the months of the Syrian calendar - as used by Michael the
Syrian - were corrupted versions of Babylonian names (see chapter 4),
380 11 Eclipse records from medieval Europe

Table 11.1 The Julian calendar.

January April February March
August June May
December September July
November October
1 Kalendae Kalendae Kalendae Kalendae
2 IV Nonas IV Nonas IV Nonas VI Nonas
3 III Nonas III Nonas III Nonas V Nonas
4 Pridie Nonas Pridie Nonas Pridie Nonas IV Nonas
5 Nonae Nonae Nonae III Nonas

6 VIII Idus VIII Idus VIII Idus Pridie Nonas

7 VII Idus VII Idus VII Idus Nonae
8 VIIdus VIIdus VII Idus VIII Idus
9 VIdus VIdus VIdus VII Idus
10 IV Idus IV Idus IV Idus VIIdus

11 III Idus III Idus III Idus VIdus

12 Pridie Idus Pridie Idus Pridie Idus IV Idus
13 Idus Idus Idus III Idus
14 XIX Kalendas XVIII Kalendas XVI Kalendas Pridie Idus
15 XVIII Kalendas XVII Kalendas XV Kalendas Idus

16 XVII Kalendas XVI Kalendas XIV Kalendas XVII Kalendas

17 XVI Kalendas XV Kalendas XIII Kalendas XVI Kalendas
18 XV Kalendas XIV Kalendas XII Kalendas XV Kalendas
19 XIV Kalendas XIII Kalendas XI Kalendas XIV Kalendas
20 XIII Kalendas XII Kalendas X Kalendas XIII Kalendas

21 XII Kalendas XI Kalendas IX Kalendas XII Kalendas

22 XI Kalendas X Kalendas VIII Kalendas XI Kalendas
23 X Kalendas IX Kalendas VII Kalendas X Kalendas
24 IX Kalendas VIII Kalendas VI Kalendas IX Kalendas
25 VIII Kalendas VII Kalendas V Kalendas VIII Kalendas

26 VII Kalendas VI Kalendas IV Kalendas VII Kalendas

27 VI Kalendas V Kalendas III Kalendas VI Kalendas
28 V Kalendas IV Kalendas Pridie Kalendas V Kalendas
29 IV Kalendas III Kalendas IV Kalendas
30 III Kalendas Pridie Kalendas III Kalendas

31 Pridie Kalendas Pridie Kalendas

 11A Observations of large solar eclipses 381

Table 11.2 Days of the week in medieval

Latin chronicles.
Name Equivalent
Dies Dominica Sunday
Feria 2 Monday
Feria 3 Tuesday
Feria 4 Wednesday

Feria 5 Thursday
Feria 6 Friday
Feria 7 Saturday

they had exact parallels with the months of the Julian calendar. However,
the first month of the year corresponded with October. Only two instances
of the use of this calendar will be found in the present chapter (in sections
11.6 and 11.8). The Syrian calendar was often used by medieval Arab
astronomers - see chapter 13.
Seasonal hours (12 to the day and 12 to the night), following the ancient
Greek and Roman practice, were commonly used throughout medieval
times. On this system, the first hour commenced with sunrise, while the
seventh started at noon. Very occasionally, hours were counted from
sunset or midnight instead. As already emphasised above, estimates of
time tended to be informal, no matter which system was used.

11.4 Interpretation of observations of large solar eclipses

In medieval Latin texts, the phrase eclipsis solis is frequently used to
describe a solar eclipse. The term eclipsis is a derivation from the Greek
ekleipsis. Expressions such as sol obscuratus est ('the Sun was obscured')
or sol defectus est ('the Sun failed') are also common. Medieval accounts
of large solar eclipses often mention the occurrence of darkness or the
visibility of stars. It is important to stress here that neither feature is
sufficient by itself to establish totality. Sensations of loss of daylight are
particularly subjective, while the appearance of stars is often noted at
eclipses which were only annular on the Earth's surface. An account of
the eclipse of AD 1191 Jun 23 in a contemporary English chronicle would
seem to suggest a particularly large magnitude:

1191. In the month of June on Sunday, the Vigil of the Nativity of St

John the Baptist (Jun 23), the 9th day before the Kalends of July, on the
27th day of the Moon, at the 9th hour of the day, the Sun was eclipsed
382 11 Eclipse records from medieval Europe

and it lasted for three hours; the Sun was so obscured that darkness arose
over the Earth and stars appeared in the sky. And when the eclipse
withdrew, the Sun returned to its original beauty.
[Gesta Regis Henrici II et Ricardi I; Stubbs (1867, I, p. 339).]
The date of the eclipse, including the weekday, is correctly given. The
precise place of observation in England cannot be established, but it can
be calculated that even where the ring phase was visible no more than 0.95
of the solar diameter was obscured by the Moon. Although the allusion
to darkness and stars would seem to imply a major eclipse, the diminution
in daylight would actually be fairly small.
It is of interest to investigate the astronomical circumstances relating to
those annular eclipses in which stars were reported by medieval European
chroniclers. In addition to 1191, the dates of these events are as follows:
AD 891 Aug 8 (Constantinople), 990 Oct 21 (Constantinople), 1033 Jun
29 (Beze, France), 1147 Oct 26 (Gembloux, Belgium), and 1310 Jan 31
(Avignon, France). Only a single star was noticed in 1147, but on each
of the other occasions several stars were said to be seen. In 1033, it was
stated that 'stars shone behind and in front of (i.e. to the east and west
of) the Sun' (Annales Besuenses: MGH, V, 19). The following eyewitness
description of the eclipse of 1310 Jan 31 from Avignon is unusually
careful:

1310. On the last day of January at the 8th hour of the day at Avignon
there was an eclipse of the Sun, and it was eclipsed in an extraordinary
manner, and was notably sparkling. There appeared as if at nightfall a
single star, a second was the opinion of the crowd (vulgi). Then a
remarkable semicircle was seen and it lasted until past the ninth hour.
[Ptolomaei Lucensis Hist, eccles. lib. XXIV; SRI, XI, 1232.]
Since annularity is not alleged, the proportion of the Sun actually
covered may have been appreciably less than in the central zone. This
remark could also apply to all the other observations whose dates are
given above.
In table 11.3a are listed the following details for each of the above six
eclipses: year, place of observation, magnitude in the central zone, com-
puted magnitude at the place of observation, local time and solar altitude
at greatest phase. Calculated details are obtained using a preliminary
expression for AT - as derived from equation (8.1). It will be noted that
only in one case (AD 1033) can more than about 95 per cent of the solar
diameter have been obscured - even where the ring phase was visible. On
this occasion, the zone of annularity ran almost parallel to the equator in
the longitudes of France, and did not reach further north than latitude
+46.4 deg - see figure 11.1. It may be deduced that the magnitude at Beze
 11.4 Observations of large solar eclipses 383

Table 11.3a Medieval annular eclipses at which stars were seen.

Year AD Place Cent. Mag. LT (h) Alt. (deg)

891 Constantinople 0.95 0.93 11.6 64
990 Constantinople 0.92 0.84 12.9 35
1033 Beze 0.99 0.96 11.6 65
1147 Gembloux 0.95 0.92 10.2 21
1191 England 0.95 0.95 10.9 57
1310 Avignon 0.94 0.84 13.7 27

30° W 15° W 0° 15° E 30° E 45° E 75° E

60° N
60° N

55° N
55° N

50° N
50° N

45° N
45° N

40° N
40° N

35° N 35° N

30° N 30° N

15° E 45° E
Fig. 11.1 Map showing belt of annularity at the solar eclipse of AD 1033 Jun
29 in relation to Beze and Cluny (after Schroeter, 1923).

(lat = +47.47 deg) cannot have exceeded 0.96, regardless of the precise
value of AT. At this same eclipse, the Sun was said to become like the
crescent Moon as seen from Cluny - see section 11.7.
Table 11.3b lists the magnitude and elongation from the Sun of Mercury,
Venus and Jupiter and the elongation of the bright star Sirius (mag. —1.5)
on each of the above dates. On every occasion, Saturn was fainter than
384 11 Eclipse records from medieval Europe

Table 11.36 Bright planets and stars at medieval annular eclipses.

Mercury Venus Jupiter Sirius
Year AD Mag. Elong. Mag. Elong. Mag. Elong. Elong.
891 +0.5 27°E -4.2 40°E -1.5 68°E 53°W
990 -0.6 4°E -3.6 14°E -2.3 131°E 121°W
1033 +3.2 3°W -3.3 28°E -1.7 94°E 13°W
1147 -0.8 4°W -3.6 42°W -2.1 123°W 125°W
1191 +2.8 5°E -3.5 1°E -2.2 126°W 6°W
1310 +1.5 10°E -3.4 28°E -1.5 3°E 132°E

mag. 0 while Mars was fainter than mag. + 1 ; thus neither of these planets
are included.
Venus would be above the horizon on each occasion. It seems likely
that provided the sky was clear, this planet would be seen at five of the
six eclipses (in 1191 it may have been too close to the Sun for visibility).
Mercury may not have been detected on any date: on the two occasions
(AD 990 and 1147) that the planet was fairly bright, it was very near the
Sun. Jupiter would be above the horizon in 1033, 1147 and 1310, and
Sirius in 891, 1033 and 1191. Since more than one 'star' appears to have
been noticed on each date apart from 1147, it seems likely that Jupiter
and/or Sirius was also seen. Although the precise identification of the stars
reported may be open to conjecture, the records provide strong evidence
that even inexperienced observers could discern stars at an eclipse which
falls far short of totality.
Terms such as maximus ('very great'), terribilis, or horribilis are some-
times used to describe an eclipse but their meaning is vague. The first
of these might well relate to a very large partial eclipse rather than a
complete obscuration of the Sun. Expressions such as eclipsis universalis
or eclipsis generalis are also frequently found in medieval annals. The
most likely interpretation here is that the phenomenon was seen over a
wide area - as indicated by reports from travellers - rather than that the
whole Sun was covered; expressions of this kind are seldom accompanied
by allusions to darkness or the other characteristic features of totality. The
phrase tenebrae in toto orbe is also ambiguous; in principle, it can either
be rendered 'darkness over the whole world' or 'darkness in the whole
disk (of the Sun)'. The former alternative - a hyperbole similar to the
equally common term tenebrae in universa terra ('darkness over the whole
Earth') - would seem to be more likely. This could simply mean that the
whole sky was darkened, giving the impression that the phenomenon was
worldwide.
 11.4 Observations of large solar eclipses 385

In view of the above remarks, I have not assumed totality unless

the complete disappearance of the Sun is clearly described. This means
rejecting a number of unusually vivid records, such as the following from
Toledo which relates to the eclipse of AD 1239 Jun 3:

The Sun was obscured on Friday at the 6th hour of the day, and it
lasted for a while between the 6th and 9th hours and it lost all its strength
(toda su fuerza) and there was as though night. There appeared many
stars, and then the Sun grew bright again of its own accord, but for a long
time it did not regain the strength that it usually has. Era 1277 (Julian =
AD 1239).
[Anales Toledanos Segundos; in Spanish; Florez (1747), XXIII, 409.]
Although the above account relates that the Sun lost all its strength,
this could simply mean that what remained of it could be viewed without
dazzling the eyes - rather than necessarily implying totality. Other vivid
descriptions in which either totality is not clearly asserted or the place of
observation is in doubt are cited in section 11.8.
Some accounts of large eclipses may well have been inspired by the
Passion narrative in the Synoptic Gospels. Indeed several texts closely
resemble the wording in the Vulgate (Latin) versions of these gospels.
In particular, it is well known that St Luke (XXIII, 45) attributes the
darkness at the Crucifixion to a solar eclipse (Sawyer, 1972). The fact that
this darkness lasted three hours may account for some of the excessive
durations of darkness found in a number of medieval records of large
solar eclipses - extreme examples being three hours (Coimbra, AD 1239)
and four hours (Reichersberg, AD 1241) - see section 11.6.1 below. A
possible alternative explanation is provided by the following report of the
eclipse of 1860 Jul 18, as observed in Sudan by Bey (1860):

But at the moment of totality, all became silent and dumb. Neither a
cry nor a rustling, nor even a whisper (was heard), but everywhere there
was anxiety and consternation. To everyone the two minutes of the eclipse
were like two hours. I do not exaggerate or imagine any of these details.
Several people whom I questioned after the eclipse regarding the duration
of totality replied that it had lasted for two hours...

Evidently the shock caused by the onset of sudden and intense darkness
can cause the unsuspecting observer to temporarily lose all sense of the
passage of time.
In marked contrast to accounts of this kind, some medieval reports of
total solar eclipses merely state that the Sun was completely obscured -
without including any descriptive remarks. For instance in AD 1485, the
Annales Mellicenses from Melk (Austria) only report that
386 11 Eclipse records from medieval Europe

On the 16th day of March at the 4th hour there was a complete (plena)
eclipse of the Sun.
[MGH, IX, 524.]
Because of the lack of supporting evidence in such cases, I have felt that
it is unwise to make use of the AT limits derived from such brief records.
It seems quite possible here that the annalist summarised a more detailed
description and inferred totality even when this was not originally alleged.
When describing the appearance of the Sun at eclipses which were
generally annular, medieval European accounts often use ambiguous ter-
minology. As noted in chapter 3, even at a considerable distance from
a zone of annularity the angle between the cusps of the crescent is
fairly small. In consequence, it is not always possible to decide whether
a complete or broken ring is referred to. For instance, records from
Cologne (Annales Colonienses Maximi: MGH, XVII, 822) and Hirschau
(Trithemii Annales Hirsaugienses: MGH I, 512) both state that at the
eclipse of AD 1207 Feb 27 'many testified that they saw a human
head in the Sun'. These two reports probably share a common ori-
gin. Nevertheless, a rather similar metaphor occurs in the description
of the eclipse of AD 1147 Oct 25 from Gembloux (Continuatio Gem-
blacensis Sigeberti Chronicon: MGH VI, 389). It is tempting to assume
that there is an oblique reference to the ring phase in such cases, but
a very large partial eclipse might well suffice instead. When assessing
medieval European records of annular eclipses, I have preferred to adopt
a very cautious view, bearing in mind that the observers were not as-
tronomers.

11.5 Place co-ordinates

The eclipse observations discussed in this chapter were made at a wide
variety of locations. Table 11.4 lists the geographical co-ordinates (in
degrees and decimals) of the various places (towns, monasteries, etc.)
which are referred to. I have followed the convention adopted throughout
this book that northern latitudes and westerly longitudes are positive. In
most cases I have given the modern names of towns. A useful reference
work which gives the modern equivalents of early European place names
is the Orbis Latinus, edited by Plechl (1972). If the name of a city has an
established English equivalent - such as Florence, Prague or Vienna - I
have adopted this. For historical reasons, I have preferred to use Antioch
(rather than the modern name of Antakya) and also Constantinople
(rather than Istanbul). The former city was also called Antakya in the
medieval Islamic period (see also chapter 13). Obviously, no naming
scheme can be fully satisfactory. I have determined nearly all geographic
 11.6 Eclipses which were central at a known location 387

co-ordinates from careful measurements on 1:50000 maps. I am grateful

to Mr B. Allaker - formerly of the Map Library, University of Newcastle
upon Tyne - for much assistance in the use of the excellent map collection
which he supervised.

11.6 Eclipses which were central at a known location

In this section, both total and annular eclipses are considered. In accor-
dance with my comments in section 11.4, I have divided the records into
two categories: those which give a detailed account of a central eclipse
(section 11.6.1) and those less reliable reports which simply mention the
occurrence of the total phase (11.6.2). As it happens, all entries in the
latter category date from either AD 1433 or 1485. Although AT limits
will be calculated in each case, only the results obtained in section 11.6.1
will be considered further.
Each entry discussed below begins with the calculated Julian date
(including the weekday). This is followed by a statement of the type of
eclipse (together with the computed magnitude in the central zone for
reference) and the place of observation. A full translation of the relevant
part of each text is given, together with its source. Most records are in
Latin; otherwise the language is specified. Unless stated to the contrary,
all translations both here and elsewhere in this chapter are by the author
- with occasional help from colleagues. Following the translation are brief
comments, together with the derived range of AT. Finally, I have given the
computed local time and solar altitude at maximum phase based on the
provisional AT formula adopted earlier - equation (8.1). These data are
for reference only, and do not influence the selection of the material in any
way. It should be noted that if two or more accounts of the same eclipse are
preserved from different locations, I have listed the places alphabetically.

11.6.1 Detailed descriptions

(1) AD 840 May 5 [Wednesday] (total mag. = 1.08): Bergamo (Italy)
In the third year of the Indiction, the Sun was hidden from this world
and stars appeared in the sky as if it were midnight, on the third day
before the Nones of May (May 5) during the Litanies of Our Lord. There
was great distress, and while the people beheld it, many thought that this
age would last no longer. But while they were contemplating these simple
things, the Sun shone again and trembling as it were began to escape from
its former shade.
[Andreas Bergomatis Chronicon; MGH, III, 235.]
The Indiction was a 15-year fiscal period, the first of which commenced
in AD 312. The third year of the appropriate Indiction began on AD 839
388 11 Eclipse records from medieval Europe

Table 11.4 Geographic co-ordinates for medieval European sites.

Place Latitude (deg) Longitude (deg)

Antioch 36.20 -36.17
Arezzo 43.47 -11.90
Augsburg 48.37 -10.90
Avignon 43.93 -4.80
Belvoir 52.90 -0.73

Bergamo 45.40 -9.67

Bergen 60.38 -5.33
Beze 47.47 -5.27
Braunschweig 52.25 -10.50
Brauweiler 49.87 -7.75

Celle 52.63 -10.06

Cesena 44.13 -12.25
Cluny 46.42 -4.65
Coimbra 40.22 +8.42
Constantinople 41.02 -28.98

Erfurt 50.97 -11.03

Evreux 49.06 -1.18
Farfa 42.22 -12.70
Florence 43.77 -11.25
Foligno 42.95 -12.72

Forcalquier 43.93 -5.77

Gembloux 50.57 -4.70
Hamburg 53.55 -10.00
Heilsbronn 49.35 -10.83
Karlstejn 49.95 -14.20

Kerkrade 50.87 -6.07

Kobrin 52.27 -24.37
Krakow 50.05 -19.92
Lanercost 54.95 +2.72
Liege 50.63 -5.58

Magdeburg 52.17 -11.63

Malmesbury 51.60 +2.10
Marola 44.50 -10.50
Marseille 43.30 -5.37
Melk 48.23 -15.35
 11.6 Eclipses which were central at a known location 389

Table 11.4 - continued.

Place Latitude (deg) Longitude (deg)
Mytho 53.70 -25.10
Montpellier 43.62 -3.88
Naples 40.83 -14.25
Neider Alteich 48.87 -12.70
Novgorod 58.50 -31.33

Nurnberg 49.45 -11.08

Orkney 58.98 +2.97
Perugia 43.12 -12.38
Prague 50.10 -14.42
Pskov 57.80 -28.43

Reichersberg 48.07 -13.37

Rome 41.90 -12.48
St Blasien 47.77 -8.13
St Germain 48.87 -2.33
Salzburg 47.80 -13.05

Siena 43.32 -11.33

Split 43.50 -16.45
Stade 53.62 -9.48
Toledo 39.87 +4.03
Vienna 48.20 -16.37

Vigeois 45.38 -1.52

Voormezele 50.82 -2.87
Vysehrad 50.07 -14.42
Wroclaw 51.08 -17.00
York 53.97 +1.08

Zbraslav 49.97 -14.40

Zwiefalten 48.23 -9.45

Sep 1. This vivid eye-witness record is unusual in that it not only alludes to
the complete disappearance of the Sun but also notes its emergence after
the total phase. The 'trembling' may be a reference to the formation of
Baily's beads after totality (Muller, 1975, p. 8.47). Although the chronicle
of Andreas of Bergamo covers much of Lombardy (N Italy), he was a
contemporary and may well be reporting his own observation.
390 11 Eclipse records from medieval Europe

RESULTS

For totality at Bergamo, 1610 < AT < 6800 sec.

In northern Italy, the zone of totality ran almost parallel to the equator
so that the observation at Bergamo is satisfied by a very wide range of
AT.
(NB local time of maximum phase = 13.4 h; solar altitude = 57 deg.)

(2) AD 968 Dec 22 [Tuesday] (total mag. = 1.03): Constantinople

When the Emperor was waging war in Syria, at the winter solstice there
was an eclipse of the Sun such as has never happened apart from that
which was brought on the Earth at the Passion of our Lord on account of
the folly of the Jews... The eclipse was such a spectacle. It occurred on the
22nd day of December, at the 4th hour of the day, the air being calm.
Darkness fell upon the Earth and all the brighter stars revealed
themselves. Everyone could see the disc of the Sun without brightness,
deprived of light, and a certain dull and feeble glow, like a narrow
headband, shining round the extreme parts of the edge of the disc.
However, the Sun gradually going past the Moon (for this appeared
covering it directly) sent out its original rays and light filled the Earth
again... At the same time I myself was also staying in Constantinople,
undertaking a liberal course of study...
[Leonis Deaconis Historiae, lib. IV, cap. 11; Niebuhr (1828, p. 72);
in Greek, with a rendering by the editor into Latin.]

Niebuhr, in his commentary, gives the year as AD 968. Apart from

the eclipse of AD 968 Dec 22 (the month and day of which are correctly
given in the text), no other winter obscuration of the Sun could have
been total at Constantinople for several decades. Leo, who was aged 38
at the time, includes the above eye-witness description in his history of
Constantinople. He stresses that he was there when the eclipse occurred;
the reference to Syria is merely an aside. Leo's splendid account contains
the earliest known reference to the corona which is definitely datable. It
is also one of the remarkably few surviving allusions to this phenomenon
before the eighteenth century. From Leo's description of the corona as a
narrow band, it would appear that the Sun was near minimum activity at
the time. Later historians of Constantinople (such as Cedrenus) mention
this same eclipse but in much less detail - see Ginzel (1884a). Presumably
they are merely drawing on Leo's account.
In the central and eastern Mediterranean, the track of totality ran
almost parallel to the equator (see figure 3.4) and hence this eclipse would
be total at Constantinople for a wide range of AT. Since the zone of
totality came no further south than 40.1 deg N at any point on the Earth's
surface, an observation from Syria is out of the question.
 11.6 Eclipses which were central at a known location 391

RESULTS
For totality at Constantinople, 1580 < AT < 5400 sec.
(NB local time of maximum phase = 11.4 h; solar altitude = 25 deg.)

(3) AD 968 Dec 22 [Tuesday] (total mag. = 1.03): Farfa (Italy)

969. The Sun was in darkness (fuit in tenebris) on the 22nd day of the
the month of December.
[Annales Farfenses, MGH, XI, 589.]

Although brief, this record from the chronicle of the monastery of Farfa
still seems to assert that the Sun was completely obscured. There is clearly
an error of one year in the date; both the month and day are correct. No
other eclipse would be large in Italy for many years before or after AD
968.
As for Constantinople, a wide range of AT would render this eclipse
total at Farfa.

RESULTS
For totality at Farfa, -3200 < AT < 2600 sec.
(NB local time of maximum phase = 9.8 h; solar altitude = 1 8 deg.)
Combining the results from the Constantinople and Farfa observations
yields 1580 < AT < 2600 sec.

(4) AD 1124 Aug 11 [Monday] (total, 1.05): Novgorod (Russia)

(6632.) In the month of August on the 11th day, before the evening
service, the Sun began to diminish and perished completely. Great fright
and darkness were everywhere. And the stars appeared and the Moon
(sic). And the Sun began to augment and became full again and everyone
in the town was very glad.
[Novorodskaya I Letopic; in Russian; trans. Vyssotsky (1949, p. 8).]
A copy of the manuscript version of the above text is illustrated in
figure 11.2.
The year 6632 - numbered from the Byzantine era - corresponds to
AD 1123/4. Other chronicles of Novgorod also give similar descriptions
of the eclipse. The 'town' mentioned in the text can be no other than
Novgorod itself, the Russian capital of the time.

RESULTS
For totality at Novgorod, 960 < AT < 2700 sec.
(NB local time of maximum phase = 14.4 h; solar altitude = 37 deg.)
392 11 Eclipse records from medieval Europe

v•

Fig. 11.2 Account of the total solar eclipse of AD 1124 Aug 11 from Novgorod
as reported in the Novorodskaya I Letopic. (Courtesy: Dr Marek Zawilski.)

(5) AD 1133 Aug 2 [Wednesday] (total 1.07): Augsburg (Germany)

(1133.) Duke Frederick... set fire to the town of Augsburg and killed
many of its citizens... An eclipse of the Sun occurred on the 4th day before
the Nones of August at midday for about one hour, such as is not seen in
a thousand years. Eventually the whole sky was dark like night, and stars
were seen over almost the whole sky. At length the Sun, emerging from the
darkness, appeared like a star, afterwards in the form of a new Moon;
finally it assumed its original form.
[Honorii Augustodensis: Summa Totius et Imagine Mundi; MGH X, 131.]

This is the last entry in the chronicle of Honorius, who at the time
was Presbyter of Augsburg. He may well have seen the eclipse himself;
certainly his account is based on an eye-witness description.

RESULTS
For totality at Augsburg, -60 < AT < 1150 sec.
(NB local time of maximum phase = 12.5 h; solar altitude = 57 deg.)

(6) AD 1133 Aug 2 [Wednesday] (total 1.07): Heilsbronn (Germany)

In the year of the Incarnation of our Lord 1133 ... on the 4th day
before the Nones of August (Aug 2), the 4th day of the week (Wednesday)
when the day was declining towards the ninth hour, the Sun in a single
moment became as black as pitch, day was turned into night, very many
(plurimae) stars were seen, objects on the ground appeared as they usually
do at night...
[Notae Halesbrunnenses; MGH, XVI, 13.]
 11.6 Eclipses which were central at a known location 393

The Notae Halesbrunnenses consists of no more than three entries by an

unknown hand on the last page of another chronicle. All relate to major
natural events: an earthquake (AD 1117), eclipse (1133) and a locust
plague (1338). In view of the title of this brief series of records, there is
no particular reason for doubting that it originates from the monastery
of Heilsbronn.
The description of the eclipse, whose date and weekday (Wednesday)
are correctly assigned, is clearly that of an eye-witness. Several accounts
from various parts of Europe mention that an unusual number of stars
were visible on this occasion, while various English annals - although
not alleging totality - remark that 'stars were seen around the Sun'.
In particular, Venus (mag. —3.5) and Mercury (mag. —0.5) would be
prominent on opposite sides of the Sun (respectively 29 deg W and 13 deg
E), while several of the brighter stars would be well placed for visibility.

RESULTS
For totality at Heilsbronn, 320 sec < AT < 1570 sec.
(NB local time of maximum phase = 12.5 h; solar altitude = 56 deg.)

(7) AD 1133 Aug 2 [Wednesday] (total 1.07): Reichersberg (Austria)

1133. That great eclipse of the Sun occurred on the 4th day before the
Nones of August, the 27th day of the Moon, the 13th year of the
Indiction. After midday, between the 7th and 8th hours, an eclipse of the
Sun was seen in Leo ... Very many (plurimae) stars were seen near the Sun;
the hearts of many were transfixed, despairing of the light. The Sun, as if it
did not exist, was entirely concealed; for about half an hour it was like
night. The face of the world was sad, terrible, black, wonderful...
[Chronicon Magni Presbyterii; MGH, XVII, 454.]

The annalist Magnus, who was attached to the monastery of Reichers-

berg, died in 1195. In the above account, he correctly states that the Sun
(whose longitude was 136 deg) was in Leo.

RESULTS
For totality at Reichersberg, 540 sec < AT < 1740 sec.
(NB local time of maximum phase = 12.8 h; solar altitude = 56 deg.)

(8) AD 1133 Aug 2 [Wednesday] (total 1.07): Salzburg (Austria)

1133. In this year on the 4th day before the Nones of August in the heat
of midday the Sun suddenly disappeared (subito sol disparuit) and a little
afterwards it seemed terribly darkened over (obtenebratus) like sackcloth of
hair (saccus cylicinus); and stars also appeared in the sky.
[S. Rudperti Salisburgensis Annales Breves; MGH, IX, 758.]
394 11 Eclipse records from medieval Europe

The Annales Breves is just one of the chronicles of the monastery of St

Peter in Salzburg. It is considered to be original after AD 1060. Newton
(1972b, p. 266), reading saccus cyclinus for the saccus cylicinus of the text,
inferred that the eclipsed Sun looked like 'a round bag' and both he and
Muller (1975, pp. 8.52 and 8.53) regarded this as an allusion to the corona.
This is no more than a supposition, but even if the original expression
saccus cylicinus is retained, both the sudden disappearance of the Sun and
its dark appearance are indicative of totality.

RESULTS
For totality at Salzburg, 280 < AT < 1460 sec.
(NB local time of maximum phase = 12.8 h; solar altitude = 57 deg.)
Combining the AT ranges deduced from the observations in AD 1133
yields 540 < AT < 1150 sec.

(9) AD 1147 Oct 26 [Sunday] (annular, 0.96): Brauweiler (Germany)

1147. On Sunday, the 7th day before the Kalends of November (Oct 26),
a solar eclipse occurred at the 3rd hour and persisted until after the 6th.
This eclipse stood fixed and motionless for a whole hour, as noted on the
'clock' (horologio)... During this hour a circle (circulus) of different colours
and spinning rapidly (maximo rotatu) was said to be in the way.
[Annales Brunwilarenses; MGH, XVI, 727.]
Once again, the recorded date (including the weekday) is accurate.
Immediately after the word horologio, the text is damaged. Although a
little obscure, this report from the monastery of Brauweiler seems to make
a direct reference to the ring phase; the obstacle in the way of the Sun (i.e.
the Moon) was said to be circular. The duration of annularity is grossly
exaggerated, possibly because the attention of the observers was intently
fixed on the Sun; the 'clock' may have been a sundial.

RESULTS
For annularity at Brauweiler, 300 sec < AT < 1190 sec.
(NB local time of maximum phase = 10.9 h, solar altitude = 23 deg.)

(10) AD 1176 Apr 11 [SundayJ (total, mag. = 1.06): Antioch

In this year 1487 (Seleucid), on New Sunday, the 11th of the month of
Nisan (= April), at daybreak, at the end of Office, that is, after the reading
of the Gospel, the Sun was totally obscured; night fell and the stars
appeared; the Moon itself was seen in the vicinity of the Sun. This was a
sad and terrifying sight, which caused many people to lament with
weeping; the sheep, oxen and horses crowded together in terror. The
darkness lasted for two hours; afterwards the light returned. Fifteen days
after, in this month of Nisan at the decline of Monday, at dusk, there was
 11.6 Eclipses which were central at a known location 395

an eclipse of the Moon in the part of the sky where the eclipse of the Sun
had taken place...
[Chronicle of Michael the Syrian, Book XX, chap. 3; in Syriac;
trans, from the rendering into French by Chabot (1905, vol. Ill, p. 367).]
The Syriac text, as printed in Chabot's edition, is illustrated in fig-
ure 11.3.
The year 1487 Seleucid is equivalent to AD 1175/6. For both the solar
and lunar eclipses, the recorded dates are accurate - equivalent to AD
1176 Apr 11 and 26 respectively. The principality of Antioch had been
founded by European Christians in AD 1098 in territory captured from
the Muslims during the First Crusade. It endured until 1268. Michael, who
had been Patriarch of Antioch (now known as Antakya) since AD 1166,
was writing as a contemporary. In view of the ecclesiastical references in
the eyewitness account of the solar eclipse, he probably experienced the
phenomenon himself. There is no particular reason for assuming a place
of observation other than Antioch.

RESULTS
For totality at Antioch, -190 < AT < 1600 sec.
(NB local time of maximum phase = 7.1 h; solar altitude = 20 deg.)

(11) AD 1185 May 1 [Wednesday] (total, 1.07): Novgorod

Several chronicles of Novgorod give similar descriptions of this eclipse,
but the following accounts are especially detailed:
(i) (6693.) On the first day of the month of May, on the day of the
Saint Prophet Jeremiah, on Wednesday, during the evening service,
there was a sign in the Sun. It became very dark, even the stars
could be seen; it seemed to men as if everything were green, and
the Sun became like a crescent of the Moon, from the horns of
which a glow similar to that of red-hot charcoals was emanating. It
was terrible to see this sign of the Lord.
[Lavrentievskaya Letopis; in Russian; trans. Vyssotsky (1949, p. 9).]

(ii) (6693.) On the first day of the month of May, during the ringing of
the bells for the evening service, there was a sign in the Sun. It
became very dark for an hour or longer and the stars were visible
and to men everything seemed as if it were green. The Sun became
like a crescent of the new Moon and from its horns a glow like a
roasting fire was coming forth and it was terrible to see the sign of
the Lord. Then the Sun cleared and we were happy again.
[Novgorodskaya II Letopis; in Russian;
trans. Vyssotsky (1949, p. 10).]
396 11 Eclipse records from medieval Europe

Fig. 11.3 Copy of the Syriac text showing Michael the Syrian's account of the
total solar eclipse of AD 1176 Apr 11 and a total lunar eclipse 14 days later as
seen at Antioch. (Chabot, 1905.)
 11.6 Eclipses which were central at a known location 397

The year 6693 (Byzantine) corresponds to AD 1185/6. The above

records, although both originating from Novgorod, differ to some extent.
The 'roasting fire' or 'glow similar to that of red-hot charcoals' emanating
from the horns of the crescent are almost certainly allusions to the
chromosphere. This only becomes visible when an eclipse is on the
verge of totality. In this context, it is intriguing that Captain Bullock -
see Ranyard (1879, p. 114) - described the total solar eclipse of AD 1868
Aug 18 in closely similar terms:

About two seconds before (emersion), and at the position where the Sun
reappeared, there suddenly broke out a thin rim of beaded fire like
intensely glowing coals, which I first took to be the Sun's limb, and was
hesitating whether or not I should call 'time' when the Sun itself burst out
in its splendour.

Although the eclipse of AD 1185 is not described as total, the most

likely interpretation of the records in the Novgorodskaya chronicle is that
the emergence of an unusually bright chromosphere signalled the end of
the darkness of totality.
In the longitudes of western Russia, the track of totality ran almost
parallel to the equator and hence this eclipse would be fully complete at
Novgorod for an extremely wide range of AT (see also Fig. 3.13).

RESULTS

For totality at Novgorod, -2200 < AT < 10 500 sec.

(NB local time of maximum phase = 16.7 h; solar altitude = 25 deg.)

(12) AD 1239 Jun 3 [Friday] (total 1.08): Arezzo (Italy)

While I was in the city of Arezzo, where I was born, and in which I am
writing this book, in our monastery, a building which is situated towards
the end of the fifth latitude zone (clima), whose latitude from the equator
is 42 and a quarter degrees and whose westerly longitude is 32 and a third,
one Friday, at the 6th hour of the day, when the Sun was 20 deg in
Gemini and the weather was calm and clear, the sky began to turn yellow
and I saw the whole body of the Sun covered step by step and it became
night. I saw Mercury close to the Sun, and all the animals and birds were
terrified; and the wild beasts could easily be caught. There were some
people who caught birds and animals, because they were bewildered. I saw
the Sun entirely covered for the space of time in which a man could walk
fully 250 paces. The air and the ground began to become cold; and it (the
Sun) began to be covered and uncovered from the west.
[Ristoro d'Arezzo: Delle composizione del mondo, lib. I, cap. XVI;
in Italian; quoted by Celoria (1877a).]
398 11 Eclipse records from medieval Europe

This is just one of several detailed accounts of the total eclipse of AD

1239 from southern Europe: Portugal, Spain, France, Italy and Croatia -
see below.
Although the dates of birth and death of the philosopher Ristoro
d'Arezzo are not known, he completed his book entitled Delle composizione
del mondo in AD 1282. This is the oldest surviving scientific treatise in
Italian. Although he gives no date for the eclipse, it may be readily
identified as 1239 Jun 3 (a Friday). The solar longitude was then 79 deg
(i.e. 19 deg in Gemini) - almost exactly the value given in the text. The
only other large eclipse in Italy for many years occurred on Oct 6 (a
Sunday) in AD 1241.
Ristoro's estimate of his latitude is more than a degree in error (correct
figure 43.47). Possibly the longitude was measured from Baghdad.
This vivid account is the earliest known which gives a meaningful
estimate of the duration of totality; near the central line the computed
duration of totality is around 5 min 45 sec, during which time a man
might perhaps walk 250 Roman (double) paces - Muller (1975, p. 8.62).
On this occasion, Mercury (mag. +1.0) would be 21 deg west of the
Sun, but the much more brilliant planet Venus (mag. —3.5) would be
only 0.8 deg from the solar limb. Presumably Ristoro confused the two
planets; he would be aware that Mercury is normally closer to the Sun
than Venus.
In the longitudes of Italy, the track of totality ran almost parallel to the
equator, and hence this eclipse would be total at Arezzo and other Italian
sites (see below) for a wide range of AT. This is unfortunate, in view of
both the quantity and quality of the available observations.

RESULTS
For totality at Arezzo -4350 < AT < 3600 sec.
(NB local time of maximum phase = 13.2 h; solar altitude = 64 deg.)

(13) AD 1239 Jun 3 [Friday] (total, 1.08): Cesena (Italy)

1239. On Friday at the beginning of June, after the 9th hour, the Sun
was covered with darkness and it became completely black. It remained
like this for the space of an hour, and the Moon was in front of it. Almost
all of the stars were manifestly seen in the sky and this appeared plainly to
everyone. There was also a certain fiery aperture (foramen ignitum) in the
Sun's disc on the lower part. The Moon itself was on the 29th day. Night
arose over the whole Earth. In verse:
'In the year one thousand, two hundred and thirty-nine
When June was beginning; on the third day:
The Sun was obscured, with its disc covered with darkness,
In full daylight the Sun became without light.
 11.6 Eclipses which were central at a known location 399

For a whole hour the Sun was dead and remote from us,
This marvel happened on the sixth day of the week'.
[Annales Caesenates; SRI, XIV, 197.]
The 'fiery aperture' almost certainly refers to a major prominence on
the Sun, one of the very few such observations in the entire pre-telescopic
period.

RESULTS
For totality at Cesena -3300 < AT < 2700 sec.
(NB local time of maximum phase = 13.3 h; solar altitude = 64 deg.)

(14) AD 1239 Jun 3 [Friday] (total, 1.08): Coimbra (Portugal)

On the 3rd day before the Nones of June (Jun 3), on the same day that
Christ suffered, namely the 6th day of the week (Friday), and at the same
time that darkness occurred over the whole Earth at the Passion of our
Lord, namely from the 6th to the 9th hours of the era 1237, there occurred
a sign such has never happened since the Passion of our Lord until the
present day. There was indeed night between the 6th and 9th hours and
the Sun became as black as pitch and the Moon (sic) and many stars
appeared in the sky. Then the receding of the darkness of night was
followed by the receding and recovering of the Sun's original clarity. Many
men and women assembled in the Church of the Holy Cross in
Coimbra... everywhere the rays of the Sun penetrated into some hole
(foramen).
[Chronicon Conimbricense, III; Florez (1747, vol. II, p. 336).]
Here we have a vivid eyewitness account in one of the chronicles of
the city of Coimbra, then capital of Lusitania (Portugal). The Julian year
is mistakenly given as 1237 rather than 1277, the scribe having written
MCXXXVII rather than MCLXXVII. However, since the month and day
are correct, the date is clearly AD 1239 Jun 3. Possibly the 'hole' refers to
the prominence seen at Cesena, but the interpretation is obscure.

RESULTS
For totality at Coimbra, -500 < AT < 1450 sec.
(NB local time of maximum phase = 11.2 h; solar altitude = 70 deg.)

(15) AD 1239 Jun 3 [Friday] (total 1.08): Florence

1238. On the 3rd day of June, the whole of the Sun was obscured at the
sixth hour and it remained obscured for several hours and from day it
became night and the stars appeared; so that many people ignorant of the
course of the Sun and the other planets marvelled greatly...
[Storie Fiorentina, VI; in Italian; quoted by Celoria (1877a).]
400 11 Eclipse records from medieval Europe

Note the error of a year. A slightly briefer account, giving the correct
year, is to be found in the Istoria Fiorentina (SRI, VIII, 967).

RESULTS

For totality at Florence -4100 < AT < 3050 sec.

(NB local time of maximum phase = 13.2 h; solar altitude = 65 deg.)

(16) AD 1239 Jun 3 [Friday] (total, 1.08): Montpellier (France)

The King (James the Conqueror) entered the city of Montpellier on
Thursday the 2nd of June of the year 1239; and on the next day, Friday,
between midday and the ninth hour, the King writes that the Sun was
eclipsed in a way people did not remember ever having seen before,
because it was entirely covered by the Moon and the day grew so dark
that one could see the stars in the sky.
[Zurita, Anales de la Corona de Aragon, Lib. Ill, cap. 36; in Spanish;
quoted by Celoria (1877a).]
There is a similar account in Latin, noting that the eclipse caused
sudden darkness, in the Indices rerum ab Aragoniae Regibus gestarum; this
is also cited by Celoria (1877a).

RESULTS

For totality at Montpellier, -6050 < AT < 1460 sec.

(NB local time of maximum phase = 12.5 h; solar altitude = 69 deg.)

(17) AD 1239 Jun 3 [Friday] (total 1-08): Siena (Italy)

1239, on Friday at the 6th hour, the Sun began to be obscured as if by a
veil and was covered in a clear sky. At the ninth hour it was totally
obscured, whence it gave no light; and as if a dark night arose with the
result that a starry sky was seen, as on a clear night. People lit lamps in
houses and shops. After some space of time it gradually became uncovered
and restored to Earth, with the result that before the evening hour it was
restored to its brilliance.
[Archivio del Duomo di Siena, 106; SRI, XV, pp. 25-26.]
Note the excessive time-interval between first sighting of the eclipse and
the onset of totality. This should have been no more than about one
hour.

RESULTS

For totality at Siena, -4700 < AT < 3600 sec.

(NB local time of maximum phase = 13.2 h; solar altitude = 65 deg.)
 11.6 Eclipses which were central at a known location 401

(18) AD 1239 Jun 3 [Friday] (total, 1.08): Split (Croatia)

At the same time, AD 1239 on the third day from the beginning of the
month of June, a wonderful and terrible eclipse of the Sun occurred, for
the entire Sun was obscured, and the whole of the clear sky was in
darkness. Also stars appeared in the sky as if during the night, and a
certain greater star shone beside the Sun on the western side. And such
great fear overtook everyone, that just like madmen they ran about to and
fro shrieking, thinking that the end of the world had come. However, it
was a Friday, the 30th day of the (lunar) month. And although the same
defection of the Sun appeared throughout the whole of Europe, it was not
however spoken of in Asia and Africa.
[Ex Thomae Historia Pontificum Salonitanorum et Spalatinorum;
MGH, XXIX, 584-585.]
This chronicle of events before and during the time of Archbishop
Roger (AD 1249-1266) was compiled by Thomas, Archdeacon of Split
(= Spalato) who died in 1268. Only two years after the eclipse, the same
writer records another great eclipse, but although on this occasion (AD
1241 Oct 6) he makes no mention of the complete disappearance of the
Sun, his account seems worth quoting in full. The time-interval between
two such major events is remarkably short.
In this same year also, namely 1241 from the Incarnation, on the 6th
day from the beginning of October, on Sunday, the Sun was again eclipsed
and all the air was darkened. There was great terror among everyone, just
as in that eclipse which happened three years previously, as we have
attested above...
[Ex Thomae Historia Pontificum Salonitanorum et Spalatinorum;
MGH, XXIX, 585.]

RESULTS
For totality at Split in AD 1239, -3200 < AT < 4600 sec.
(NB local time of maximum phase = 13.7 h; solar altitude = 61 deg.)
In summary, only the observation from Coimbra provides reasonably
narrow limits for AT in AD 1239 (i.e. -500 < AT < 1450 sec). However,
the upper limit is supported by the result for Montpellier (AT < 1460).
The various observing sites mentioned above for this eclipse are shown in
figure 11.4 along with the computed belt of totality for AT = 0.
In addition to the detailed records discussed above, Levi-Donati (1989)
has drawn attention to a Latin inscription carved on a pillar at Marola, a
village in the Appenine Mountains near Florence. This inscription, which
is illustrated in figure 11.5, may be translated as follows:
1239. On the 3rd day before the Nones of June the Sun died at the
ninth hour.
402 11 Eclipse records from medieval Europe

Fig. 11.4 The principal observing sites for the total solar eclipse of AD 1239
Jun 3.

Although less reliable than the accounts cited previously, totality at

Marola is assured for any value of AT in the range —2950 < AT < 1610
sec. This fully encompasses the range indicated by the careful observation
of totality at Coimbra (—500 < AT < 1450 sec). The computed local time
of maximum (13.2 h) is actually towards the end of the 7th hour, but the
quoted time is likely to be no more than a rough estimate.

(19) AD 1241 Oct 6 [Sunday] (total, 1.05): Reichersberg

1241. On the day before the Nones of October (Oct 6), the Sun while it
was bright was suddenly covered with wonderful blackness a little after
midday. As a result, no part of it could be seen and stars were seen as if
during the night for about four hours.
[Chronicon Magni Presbyteri Continuatio; MGH, XVII, 528.]
This continuation of the chronicle of Magnus of Reichersberg (see
above) is contemporary with the events it describes.

RESULTS
For totality at Reichersberg, 450 sec < AT < 1400 sec.
(NB local time of maximum phase = 13.0 h; solar altitude = 32 deg.)

(20) AD 1241 Oct 6 [SundayJ (total 1.05): Stade (Germany)

1241... There was an eclipse of the Sun on the Octave of St. Michael
(Oct 6), namely the day before the Nones of October, on Sunday some
 11.6 Eclipses which were central at a known location 403

Fig. 11.5 Latin inscription from AD 1239 recording the 'death of the Sun',
carved on a pillar at Marola in Italy. (Courtesy: Dr G.R. Levi-Donati.)

time after midday. Stars appeared and the Sun was completely hidden
from our sight. Yet the sky was so clear that no clouds appeared in the air.
[Annales Stadenses, MGH, XVI, 368.]
This section of the chronicle of the monastery of St. Maria in Stade was
compiled by Albertus who became Abbot there in 1232. Note Albertus'
use of the first person plural, implying that he was an eye-witness. The
date is precisely correct, including the weekday. Since the Feast of St
Michael was on Sep 29, the Octave one week later (i.e. 8 days in inclusive
counting) would be on Oct 6.
404 11 Eclipse records from medieval Europe

RESULTS
For totality at Stade, 630 sec < AT < 1630 sec.
(NB local time of maximum phase = 12.6 h; solar altitude = 28 deg.)
Combining the AT results from Reichersberg and Stade, 630 < AT <
1400 sec. These limits are considerably closer together than those obtained
from the much more extensive observations in AD 1239, largely because
the track of totality in 1241 was steeply inclined to the equator in Central
Europe.

(21) AD 1263 Aug 5 [Sunday] (annular, 0.95): Orkney (Scotland)

The eve of St Olaf was on a Sunday... When King Hacon lay in
Rognvaldsvoe, a great darkness (myrkur) overtook the Sun, so that a little
ring (hringur litill) was bright round it on the outside, and that lasted a
while of the day... On St. Lawrence's day king Hacon sailed out of
Rognvaldsvoe over the Pentland firth.
[Haconar Saga; trans. Dasent (1894, pp. 346-7).]
The year can be firmly established from the context as AD 1263, while
the date of the eclipse can be fixed as between the eve of St Olaf (Jul 29)
- a Sunday - and St Lawrence's day (Aug 10).

RESULTS
For annularity in Orkney, -3500 < AT < -2150 sec.
(NB local time of maximum phase = 13.6 h; solar altitude = 43 deg.)
The above AT range is in marked conflict with roughly contemporane-
ous results, which indicate a value around +1000 sec. The eclipse could
therefore have only been partial in Orkney. Even if it had been central
there, no more than 0.95 of the solar diameter would have been covered
so that the reference to 'great darkness' is much exaggerated.

(22) AD 1267 May 25 [ Wednesday] (total 1.02): Constantinople

At that time the Moon obscured the Sun when it was in the 4th part
(degree) of Gemini, at the 3rd hour before midday on the 25th day of May
in the year 6775 (AD 1267). It was a total eclipse of about 12 digits or
points. Also, such darkness arose over the Earth at the time of mid-eclipse
that many stars appeared. No doubt this portended the very great and
destructive calamities which were soon to be vented on the Romans by the
Turks.
[Nicephori Gregorae Byzantinae Historiae, Lib. IV, cap. 8; Migne (1865,
CXLVIII, 246); in Greek, with a rendering into Latin by the editor.]
Nicephorus Gregoras, who himself lived at Constantinople, wrote about
a century after the event. He does not directly name the place of obser-
vation but the above detailed account is included in his history of Con-
staninople. In reporting the partial eclipse of AD 1330 Jul 16 (Lib, IX,
 11.6 Eclipses which were central at a known location 405

cap. 12), Nicephorus is careful to state that this was seen in the parallel of
Constantinople (rather than the city itself), and specifies certain regions
(e.g. Thrace) where observations were made. Although he assigns the
correct date for the eclipse of AD 1267, Nicephorus is several degrees in
error in his solar position; at the time of the eclipse the Sun would be in
longitude 70.5, the 11th degree of Gemini.

RESULTS
For totality at Constantinople, -850 < AT < 820 sec.
(NB local time of maximum phase = 11.5 h; solar altitude = 70 deg.)

(23) AD 1406 Jun 16 [Wednesday] (total, 1.06): Braunschweig (Germany)

1406. In this year there was an eclipse of the Sun so that the Sun
stopped shining (vorgingk or schyn) before the Prime of the day (i.e. the
Office held c. 6 a.m.) on St. Vitus' day (Jun 15); it was so dark that people
could not recognise one another.
[Bothonis Chronicon Brunsvicensis picturatum; in German;
quoted by Ginzel (1884a).]

RESULTS

For totality at Braunschweig, —100 < AT < 1450 sec.

(NB local time of maximum phase = 6.8 h; solar altitude = 26 deg.)

(24) AD 1406 Jun 16 [Wednesday] (total 1.06): Hamburg

In the year 1406, on the day of the Annunciation of the Virgin Mary
(= Jun 16), in the morning at the 6th hour, the Sun was entirely (ganz)
covered, and (there was) such a great eclipse that the people believed that
it was the end of the world.
[Adami Trazigeri Chronicon Hamburgensis; in German;
quoted by Zawilski (1994).]

RESULTS

For totality at Hamburg, - 8 3 0 < A T < 740 sec.

(NB local time of maximum phase = 6.8 h; solar altitude = 26 deg.)

(25) AD 1406 Jun 16 [Wednesday] (total, 1.06): Liege

1406. In that year on Jun 16, between the 6th and 7th hours before
midday, there was a total eclipse of the Sun, with the result that the dawn,
which was bright and clear, became as dark as at midnight.
[Chronique Latine de Jean de Stavelot; quoted by Zawilski (1994).]
This contemporary chronicle is specifically concerned with Liege.
406 11 Eclipse records from medieval Europe

RESULTS
For totality at Liege, -550 < AT < 900 sec.
(NB local time of maximum phase = 6.5 h; solar altitude = 22 deg.)

(26) AD 1406 Jun 16 [Wednesday] (total 1.06): Magdeburg (Germany)

1406. In the same year on St Vitus' day, at the beginning of the 9th
hour in the morning, there was an eclipse here in the land, that is a
darkening of the Sun, so that the Sun stopped shining (schin vorging)
before the Prime of the day; it was so dark that there was no light and
people could not recognise one another... People thought that the end of
the world was coming.
[Magdeburger Schoppenchronik; in German; quoted by Zawilski (1994).]
As noted above, in citing the entry from Braunschweig, St Vitus' Day
is on Jun 15 rather than Jun 16. There are other similarities (as well as
significant differences) between this account and that from Braunschweig.
The two places are about 100 km apart so that inter-communication
would be relatively easy.

RESULTS
For totality at Magdeburg, +170 < AT < 1700 sec.
(NB local time of maximum phase = 6.9 h; solar altitude = 26 deg.)
Combining the various limits deduced above yields 170 < AT < 740
sec.

(27) AD 1415 Jun 7 [Friday] (total 1.07): Neider Alteich (Germany)

1415. A dark eclipse of the Sun occurred with the result that the Sun
entirely lost its light for twice the duration of the Misere-mei (Psalm 51) at
the 6th hour of the day on the 7th day of June.
[Notae Altahenses; MGH, XVII, 424.]

RESULTS
For totality at Neider Alteich, -650 < AT < 970 sec.
Here we have the second of only two medieval allusions to the duration
of totality (see also AD 1239, Arezzo). Near the central line, the total
phase would last for 3 min 40 sec.
(NB local time of maximum phase = 7.5 h; solar altitude = 27 deg.)

(28) AD 1415 Jun 7 [Friday] (total 1.07): Prague

1415. At that same time, on the 7th day of June, which was the sixth
day of the week, after the Feast of S. Boniface (Jun 5), at the the 11th
hour of the day, the whole Sun was eclipsed. As a result, the Mass could
 11.6 Eclipses which were central at a known location 407

not be celebrated without lights... (This happened) on account of the

death of Master Jan Hus (which occurred) soon afterwards...
[Magister Laurentius de Brezina. De gestis et variis accidentibus
regno Boemiae, II, part 1, 333; quoted by Ginzel (1884a).]
At this point in his chronicle, the author, who was a contemporary, is
relating events which happened at Prague. Jan Hus, the prominent Czech
religious reformer, was burnt at the stake on 6 July in AD 1415.

RESULTS
For totality at Prague, -1160 < AT < 670 sec.
(NB local time of maximum phase = 7.1 h; solar altitude = 28 deg.)

(29) AD 1415 Jun 7 [Friday] (total 1.07): Wroclaw (Poland)

An eclipse of the Sun on the 6th day of the week (Friday). In the same
year 1415, on the Octave of Corpus Christi, at the 12th hour (after sunset),
at about the relighting of the Moon, there was a total eclipse of the Sun.
This stood still for close to half an hour and there was terror and alarm
among many people.
[Chronica Sigismundi Rosiczi; Scriptores rerum Silesicarum,
vol. 12; quoted by Zawilski (1994).]
The movable Feast of Corpus Christi, the first Thursday after Trinity,
fell on May 30 in AD 1415 so that the Octave would be on Jun 6.
From Zawilski, the annalist Sigmund Rositz used notes preserved in
Wroclaw; in particular the immediately preceding and following entries
in his chronicle relate to (non-astronomical) events occurring in this city.

RESULTS
For totality at Wroclaw, -1190 < AT < 720 sec.
(NB local time of maximum phase = 7.3 h; solar altitude = 30 deg.)
When the various AT ranges obtained for the eclipse of AD 1415 are
combined, —650 < AT < 670 sec.

(30) AD 1431 Feb 12 [Monday] (total 1-04): Foligno (Italy)

1431. Memorandum, that on the 12th day of February at the 21st hour,
which was the Monday of the Carnival, the Sun was so greatly obscured
that it became as dark as night and the eye of the Sun became black like
charcoal. And on that day the Moon turned (i.e. was new)...
[Pietruccio di Giacomo degli Unti: Memoriale; in Italian;
quoted by Levi-Donati (1987).]
This contemporary chronicle is concerned with events in the city of
Foligno. The local time was evidently expressed relative to sunset. Feb 12,
408 11 Eclipse records from medieval Europe

the day of the eclipse, was indeed a Monday. The time was apparently
measured from sunset on the previous evening; see also the following
entry.

RESULTS
For totality at Foligno, 10 < AT < 870 sec.
(NB local time of maximum phase = 15.3 h; solar altitude = 20 deg.)

(31) AD 1431 Feb 12 [Monday] (total, 1.04): Perugia (Italy)

1431. On February 12 at about the 21st or 22nd hour, the Sun was
completely obscured and in front of the Sun was placed a black circle like
a little wheel (rotella). It became as dark as night and the sky revealed the
stars. The birds went to roost as they usually do at night. Everyone was
feeling ill at ease as a result of this event. It began half an hour before the
Sun was covered over. It gradually lost its light even to the hour stated
above...
[Antonio dei Veghi: Diario dalVanno 1423 al 1491; in Italian;
quoted by Levi-Donati (1987).]
Antonio dei Veghi lived at Perugia. A copy of the manuscript version
of the above text is illustrated in figure 11.6.
Once again the local time was apparently estimated from sunset.

RESULTS
For totality at Perugia, -160 < AT < 700 sec.
(NB local time of maximum phase = 15.3 h; solar altitude = 20 deg.)
Combining the results from the observations at Foligno and Perugia
leads to a value for AT at the epoch AD 1415 in the range 10 < AT <
700 sec.

(32) AD 1485 Mar 16 [Wednesday] (total, 1.05): Augsburg (Germany)

In the year of salvation 1485, in the month of January, according to the
ancient custom, the consuls of Augsburg ...were elected. On the 16th day
of March, at the 3rd hour, during meal-time, the Sun was totally eclipsed.
This produced such horrid darkness on our horizon for the space of half
an hour that stars appeared in the sky. Crazed birds fell from the sky and
bleating flocks and fearful herds of oxen unexpectedly began to return
from their pastures to their stables.
[Achilli Pirmini Gassari: Annales Augustburgenses;
quoted by Zawilski (1994).]
In the longitudes of Central Europe, the track of totality ran almost
parallel to the equator and hence this eclipse would be total at Augsburg
for a wide range of AT.
 11.6 Eclipses which were central at a known location 409

•A l
'rTcfl
1
j

<

Fig. 11.6 A copy of the manuscript account of the total solar eclipse of AD 1431
Feb 12 in the Diario dalVanno 1423 al 1491 by Antonio dei Veghi. (Courtesy: Dr
G. R. Levi-Donati.)

RESULTS

For totality at Augsburg, -5500 < AT < 780 sec.

(NB local time of maximum phase = 16.5 h, solar altitude = 16 deg.)
The time of day, expressed relative to noon, was only crudely estimated.

(33) AD 1560 Aug 21 [Wednesday] (total, 1.05): Coimbra

I shall cite two remarkable eclipses of the Sun, which happened in my
own time and thus not long ago. One of these I observed about midday at
Coimbra in Lusitania in the year 1559 (sic), in which the Moon was placed
between my sight and the Sun with the result that it covered the whole
Sun for a considerable length of time. There was darkness in some manner
greater than night; neither could one see clearly where one stepped. Stars
appeared in the sky and (marvellous to behold) the birds fell down from
the sky to the ground in terror of such horrid darkness...
[Clavius (1593, p. 508).]
Clavius has mistaken the year; the eclipse of AD 1560 Aug 21 was the
only one total in Portugal for many years around this time. Clavius, who
was then aged 23, was studying at the University of Coimbra.

RESULTS

For totality at Coimbra, -490 < AT < 220 sec.

NB for a duration of totality greater than one minute (i.e. 'a considerable
length of time'), this range should be narrowed to —475 < AT < 205 sec.
This latter interval would seem more appropriate.
(NB local time of maximum phase = 11.3 h; solar altitude = 57 deg.)
410 11 Eclipse records from medieval Europe

Differential radial scale: 1 second of arc = i

True lunar limb

Mean lunar limb

Fig. 11.7 The lunar limb profile (exaggerated by a factor of 60) at the solar
eclipse of AD 1567 Apr 9. (After Duncombe, 1973).

(34) AD 1567 Apr 9 [Wednesday] (annular-total, 1.003): Rome

...The other (eclipse) I saw in Rome in the year 1567 also about
midday in which although the Moon was placed between my sight and the
Sun it did not obscure the whole Sun as previously but (a thing which
perhaps never before occurred at any other time) a certain narrow circle
was left on the Sun, surrounding the whole of the Moon on all sides.
[Clavius (1593, p. 508).]
This is a continuation of the previous account (i.e. from AD 1560).
Clavius had now moved to Rome, where he was teaching mathematics at
the Collegio Romano.
The eclipse was neither total nor annular at Rome; the angular diam-
eters of both Sun and Moon as seen from Rome were virtually identical.
Even where the eclipse was central, part of the solar photosphere would
be visible down the deep clefts at the Moon's edge - see figure 11.7, which
is based on the lunar limb charts of Duncombe (1973). The approxi-
mate topocentric librations - as defined by Duncombe - are respectively:
L = - 5 deg, B = 0 deg.
 11.6 Eclipses which were central at a known location 411

RESULTS
Making a careful allowance for limb profile, Stephenson et al. (1997) have
estimated a value of AT in the narrow range 145 < AT < 165 sec from
this observation.
(NB local time of maximum phase = 12.2h; solar altitude = 59 deg.)

(35) AD 1601 Dec 24 [Monday] (annular, 0.90): near Bergen (Norway)

Those who had seen the above mentioned eclipse described it with very
great astonishment: 'But in the same form, the Sun had encompassed the
Moon centrally within its own circumference in such a way that its own
light, having been diffused up to the edge on all sides, more or less shone
out equally by 1 \ digits' (i.e. 1 \ twelfths of the apparent solar diameter).
[Longomontanus (1622).]
The above account of the ring phase, by fishermen on the sea coast near
Bergen, was communicated to Longomontanus by Andreas Fosse, Bishop
of Bergen, who was also a mathematician. The width of the annulus was
considerably overestimated, probably on account of irradiation.

RESULTS
For annularity at Bergen, —1460 < AT < 3170 sec.
By this late period, only a few years before the telescopic era, AT
can be estimated fairly accurately as around 100 sec using extrapolation
from the results obtained from telescopic observations. For any acceptable
value of AT, Bergen would lie near the centre of this very wide zone of
annularity. Hence the observation does not enable a useful result to be
derived, although the fishermen deserve commendation for providing one
of the few careful descriptions of the ring phase from the whole of the
pre-telescopic period.
Annularity would last for more than eight minutes. On account of
the low solar altitude (only 3 deg at mid-eclipse), the ring phase would
probably be very distinct.
(NB local time of maximum phase = 13.8 h; solar altitude = 3 deg.)

(36) AD 1605 Oct 12 [Wednesday] (total, 1.03): Naples?

I add a fourth argument (that the heavens are subject to change) from
the eclipse of the Sun which we saw in the month of October. At Naples,
indeed, the whole Sun appeared hidden, but this was entirely the same as
that which Plutarch himself saw 1500 years ago; whose words from page
318 of my Optics are as follows: 'A certain splendour shines forth around
the circumference, preventing the shadow from becoming profound and
too great'. At Naples, indeed, this happened in the above mentioned year
(AD 1605): the whole Sun was accurately hidden, but at any rate it did
not last for a very long time. In the centre, where the Moon was, there was
412 11 Eclipse records from medieval Europe

Table 11.5 AT limits deduced from

medieval European observations of
central solar eclipses.

AT Range (sec)
Year LL UL
+840 1610 6800
+968 1580 2 600
1124 960 2 700
1133 540 1150
1147 300 1190

1176 -190 1600

1185 -2200 10500
1239 -500 1450
1241 630 1400
1267 -850 820

1406 170 740

1415 -650 670
1431 10 700
1485 -5 500 780
1560 -475 205

1567 145 165

1601 -1460 3170

the appearance as if of a black cloud; all around it was a red and fiery
splendour, of equal breadth on all sides, which occupied a good part of the
sky. Out of the region of the Sun, towards the north, the sky was entirely
dark, and as profound as night; stars, however, were not seen.
[Kepler (1605).]

RESULTS
For totality at Naples, 1400 < AT < 2590 sec.
Totality on this occasion cannot have been observed from the city of
Naples; the indicated range of AT is much too high at such a recent epoch;
as in the previous entry (AD 1601), AT can be estimated as around 100
sec from telescopic observations. This result gives a magnitude of only
0.97 in the city of Naples. However, the eclipse would certainly be total in
the extreme south of Italy (Calabria), which was then part of the kingdom
of Naples. The precise place of observation cannot be established; Kepler
 11.6 Eclipses which were central at a known location 413

3000 -

7? 2000 --

1000-

+800 +900 +1000 +1100 +1200 +1300 +1400 +1500 +1600

Year
Fig. 11.8 AT limits obtained from central eclipses recorded in medieval Europe.

is surprisingly vague on this issue. On the other hand, he gives a careful

description of the corona, which had not been previously referred to
directly since AD 968 (see above).
(NB local time of maximum phase = 14.8 h, solar altitude = 30 deg.)

Table 11.5 summarises the AT limits derived in this section, apart from
the anomalous results from AD 1263 and 1605. When more than one
observation was made at the same eclipse, I have combined the various
individual AT ranges, as already noted above. These limits are plotted in
figure 11.8.

11.6.2 Brief reports of total eclipses

As noted above, although AT limits will be derived from the following
brief records, they will not be included in subsequent analysis since the
reliability of the observations is in doubt.

(1) AD 1433 Jun 17 [Wednesday] (total 1.07): Celle (Germany)

1433. At Celle in the Duchy of Braunschweig, after midday the whole
(totus) Sun was eclipsed by the intervention of the Moon in the Tail of the
Dragon (Cauda Draconis)
[Heinrich Bunting: Chronologia Catholica; quoted by Zawilski (1994).]
414 11 Eclipse records from medieval Europe

RESULTS

For totality at Celle, 530 < AT < 2270 sec.

The Sun would then be in longitude 94 deg (in Cancer). The allusion
to the Tail of the Dragon' is obscure.
(NB local time of maximum phase = 16.1 h; solar altitude = 35 deg.)

(2) AD 1433 Jun 17 [ Wednesday] (total, 1.07): Nurnberg (Germany)

1433. Afterwards, on the 17th day of June, there was a transformation
(verwandelung) of the Sun, which entirely (gancz) lost its light, and this
happened on St Ullerius' day (Jun 17). At 12 hours and 4 minutes after
(sunrise) the change reach its greatest. On Sunday after the transformation
occurred, before Vespers, it began to rain and it rained day and night until
St John's day.
[Chronik aus Kaiser Sigmundus Zeit bis 1434: Nurnberg; in German;
quoted by Zawilski (1994).]

The precision with which the time is quoted suggests that it was calcu-
lated from tables rather than observed; this may also apply to the allusion
to totality.
Rain started to fall on Sunday Jun 21 and continued unabated until the
following Wednesday (Jun 24 - i.e. the Nativity of St John the Baptist).

RESULTS
For totality at Nurnberg, -530 < AT < 1070 sec.
(NB local time of maximum phase = 16.3 h; solar altitude = 34 deg.)

(3) AD 1433 Jun 17 [Wednesday] (total 1-07): Karlstejn (Czech Repub-

lic)
1433. In this same year on the 4th day of the week (Wednesday), after
St Vitus' Day (Jun 15), at about the 21st hour (after sunset), the Sun was
eclipsed totally in its body. And in the same week and for two weeks
afterwards there were great inundations of water.
[Kronika Bartoska z Drahonic; Fontes rerum Bohemicarum, vol. 5;
quoted by Zawilski (1994).]

This record, which contains only minimal astronomical details, is prob-

ably from Karlstejn, where Bartosek was writing his chronicle (Zawilski,
1994).

RESULTS
For totality at Karlstejn 420 < AT < 2040 sec.
(NB local time of maximum phase = 16.5 h; solar altitude = 32 deg.)
 11.6 Eclipses which were central at a known location 415

(4) AD 1485 Mar 16 [ WednesdayJ (total, 1.05): Melk (Austria)

1485. On the 16th day of March at the 4th hour there was a complete
eclipse of the Sun.
[Annales Mellicenses; MGH, IX, 524.]
The value of this brief record in the annals of the monastery of Melk
is reduced by the lack of any descriptive details.

RESULTS
For totality at Melk, -4400 < AT < 2100 sec.
(NB local time of maximum phase = 16.8 h; solar altitude = 13 deg.)

(5) AD 1485 Mar 16 [Wednesday] (total 1-05): Salzburg (Austria)

1485. In this same year, The Sun suffered a total eclipse on St
Gertrude's Day (Mar 16) during Lent.
[Chronicon Salisburgense; quoted by Zawilski (1994).]

RESULTS
For totality at Salzburg, -1000 < AT < 2160 sec.
(NB local time of maximum phase = 16.6 h; solar altitude = 15 deg.)

(6) AD 1485 Mar 16 [Wednesday] (total, 1.05): Vienna

1485. Mar 16. And on that day the Sun was totally eclipsed. As I
observed, the eclipse in fact occurred an hour earlier (than expected).
[Chronicle of the Viennese physician Johannes Tichtel;
quoted by Zawilski (1994).]

RESULTS
For totality at Vienna, -4500 < AT < 1780 sec.
(NB local time of maximum phase = 16.9 h; solar altitude = 12 deg.)

(7) AD 1605 Oct 12 [Wednesday] (total, 1.03): Marseilles

Wendelin at Forcalquier in Provence saw the whole Sun hidden apart
from a very narrow thread towards the north, which he ascribed to the
illuminated atmosphere. He added that namely in Marseilles the whole
Sun (totum solem) appeared obscured with dense darkness.
[Riccioli (1665, Book II, chap. 2); quoted by Zawilski (1994).]
The observation of a partial eclipse at Forcalquier will be considered in
section 11.7.

RESULTS
For totality at Marseilles, 650 < AT < 1700 sec.
(NB local time of maximum phase = 13.7 h; solar altitude = 34 deg.)
416 11 Eclipse records from medieval Europe

The indicated range of AT at this late epoch is far outside the figure
of approximately 100 sec derived from extrapolation of the telescopic
data. Presumably the sparse information which Wendelin appears to have
obtained at second hand from Marseilles was unreliable.

11.7 Eclipses which were definitely partial, yet very large

Each of the following observations suggests that the observer was not far
outside the zone of totality or annularity.

(1) AD 1033 Jun 29 [Friday] (annular, 0.994): Cluny (France)

AD 1033... On the 3rd day before the Kalends of July (Jun 29), the sixth
day of the week (Friday) and the 28th of the Moon, an eclipse or defection
of the Sun occurred from the 6th hour to the 8th; it was extremely terrible.
Indeed, the Sun became sapphire in colour, showing in its upper part the
appearance of the crescent Moon on the 4th day from its relighting...
[Glabri Rudolphi Cluniacenses monachi Historiarum sui temporis;
quoted by Ginzel (1884a).]
Glaber, a monk at Cluny, was a contemporary writer. His chronicle
extends to 1044.
This is one of the few early accounts of a large partial solar eclipse
to indicate whether the upper or lower portion of the Sun was obscured
at maximum phase. In the longitudes of France, the narrow track of
annularity ran almost parallel to the equator and in fact did not quite
reach as far north as the latitude of Cluny (+46.42 deg) on any part of
the Earth's surface (northern limit 46.35 deg) - see figure 11.1. Hence no
value of AT will render the eclipse central at this site; calculation merely
confirms the observation of a large partial eclipse in which the upper
portion of the solar disc remained unobscured. At Cluny, the computed
magnitude cannot have exceeded 0.992. Because of irradiation, the width
of the solar crescent would probably appear much enhanced so that it
seemed to resemble the crescent Moon considerably after new.
(NB local time of maximum phase = 11.7h ; solar altitude = 66 deg.)
Several other accounts of this same eclipse from France and Belgium
also assert that the Sun became like the crescent Moon. However, the var-
ious sites are still further north than Cluny, so that the eclipse magnitudes
must have been smaller.

(2) AD 1133 Aug 2 [Wednesday] (total 1.07): Vysehrad (Prague)

1133. On the 4th day before the nones of August (Aug 2), an eclipse of
the Sun appeared in a wonderful manner. This defection gradually
diminished so much that a crown like a crescent Moon proceeded to the
 11.7 Large eclipses which were definitely partial 417

south part, afterwards turning round to the east, henceforth to the west. At
length it was transformed to its original state.
[Canonici Wissegradensis continuatio Cosmae; MGH, IX, 938.]
This continuation of the chronicle of Cosmas of Prague between 1125
and 1142 was by an anonymous canon of Vysehrad - a suburb of the city
of Prague. He had an unusual interest in astronomy, and recorded many
celestial phenomena (especially aurorae) which he presumably witnessed.
Nevertheless, it is clear from the descriptions which he gives that he
possessed only a very limited knowledge of astronomy. There can be little
doubt that the above account relates to an eclipse which was not quite
total. The author is mainly concerned with describing the changing aspect
of the solar crescent and does not suggest that the Sun ever completely
disappeared. However, the text is garbled. If the southern part of the Sun
had remained unobscured, the crescent would have appeared to rotate
from the south-east to the south-west.

RESULTS
For a partial eclipse at Vysehrad in which the southern limb of the Sun
remained visible, AT > 2750 sec.
The above result is in conflict with those derived from several observa-
tions of totality made elsewhere in Europe at the same eclipse (see section
11.6); these all indicate that AT was less than about 1500 sec (see above).
However, if it be supposed that the northern part of the Sun remained
uncovered at Vysehrad, this discord would be removed; a value of AT
< 1470 sec would be indicated. A possible explanation is that the Sun
was viewed by reflection in water (a known practice in medieval times)
but the author omitted to allow for this when recounting his experiences.
However, this is no more than conjecture.
(NB local time of maximum phase = 12.8 h; solar altitude = 55 deg.)

(3) AD 1147 Oct 26 [Sunday] (annular, 0.95): Braunschweig (Germany)

1147. On the feast of St Simon and St Jude (Oct 28), the Sun was
obscured with the result that it resembled a sickle.
[Bothonis Chronicon Brunsvicensis picturatum; in German;
quoted by Ginzel (1884a).]
Note the error of two days in the date of the eclipse both in this and
the following entry.

RESULTS
For a partial eclipse at Braunschweig, either AT < 1680 or > 2640 sec.
(NB local time of maximum phase = 10.8h; solar altitude = 21 deg.)
418 11 Eclipse records from medieval Europe

V
u
7 8
Fig. 11.9 Sketches of the eclipsed Sun in AD 1153 Jan 26 as found in the
chronicles of Erfurt (Germany). (Holder-Egger, 1899.)

(4) AD 1147 Oct 26 [Sunday] (annular, 0.96): Magdeburg (Germany)

1141. In this same year on the 5th day before the Kalends of November
(Oct 28), about midday there was an eclipse of the Sun and the Earth was
covered with horrible darkness. Indeed a circle like a crescent was seen...
[Annales Magdeburgenses; MGH, XVI, 188.]

RESULTS
For a partial eclipse at Magdeburg, either AT < 1920 sec or > 2870 sec.
(NB local time of maximum phase = 10.9 h; solar altitude = 22 deg.)
Combining the limits from the two observations of partial eclipses in AD
1147 leads to either AT < 1680 sec or > 2870 sec. Much more restricted
AT limits at this date are yielded by the observation of annularity from
Brauweiler - see section 11.6.
(5) AD 1153 Jan 26 [Monday] (annular, 0.93): Erfurt (Germany)
1153. A sign appeared in the Sun on the 7th day before the Kalends of
February (Jan 26) in this manner...
[Annales Sancti Petri Erphesfurdenses; MGH, XVI, 21.]
 11.7 Large eclipses which were definitely partial 419

This and several other Erfurt chronicles contain sketches showing the
Sun as a crescent with the cusps pointing upwards (Holder-Egger, 1899,
XLVII, 19, 56, 57 and 178) - see figure. 11.9.

RESULTS
For a partial eclipse at Erfurt with the southern limb of the Sun uncovered,
AT < - 5 0 sec.
This result is clearly discordant when compared with those derived from
a number of roughly contemporaneous observations - which indicate a
value for AT of at least 500 sec (see section 11.6). If the sketches are
taken to represent the Sun as seen by reflection in water (as in the case of
the observation at Vysehrad in 1133), a more reasonable result of AT >
1370 sec is indicated instead. However, this is no more than speculation;
unfortunately, none of the surviving accounts of the eclipse from Erfurt
(all of which are very brief) state which limb of the Sun was obscured.
(NB local time of maximum phase = 12.9 h; solar altitude = 21 deg.)

(6) AD 1178 Sep 13 [Wednesday] (total 1.05): Vigeois (France)

In the year from the Incarnation of Our Lord 1178, on the 4th day of
the week (Wednesday), the Ides of September (Sep 13), the 28th day of the
Moon, on a clear day, about the 5th hour the Sun suffered an eclipse; and
its disc began to be covered from the east (sic) until it was like a two- or
three-day-old Moon. The star Venus was seen to the north. After 6 (i.e. the
6th hour) the brightness returned from the east (sic), in the order in which
it was blackened, until the Sun was fully illuminated. Then we saw each
others' faces as they appear beside a glowing furnace.
[Ex Chronico Gaufredi Vosiensis; Bouquet (1781, p. 447).]
Gaufredus, the author of this eyewitness description, was prior of the
monastery of Vosium (now Vigeois) at the time of the eclipse. Although
he gives a careful description of a partial eclipse, he has mistaken the limb
of the Sun at which it began to be covered and uncovered; in each case
this would be the western edge. The reference to the apparent reddening
of the observers' faces perhaps suggests that Gaufredus and his colleagues
had been looking at the Sun for too long!

RESULTS
For a partial eclipse at Vigeois, either AT < 1140 sec or > 2320 sec.
(NB local time of maximum phase = 11.6 h; solar altitude = 46 deg.)

(7) AD 1312 Jul 5 [Wednesday] (total 1-01): Lanercost (England)

AD 1312. On the third of the Nones of July (Jul 5), on the vigil of the
octave of the Apostles Peter and Paul, was a new Moon (the thirtieth
lunation), and an eclipse of the Sun about the first hour of the day, and
420 11 Eclipse records from medieval Europe

the Sun appeared like a horned Moon, which was small at first and then
larger, until about the third hour it recovered its proper and usual size;
though sometimes it seemed green, but sometimes of the colour which it
usually has.
[The Chronicle of Lonercost \ trans. Maxwell (1913, p. 198).]
Unlike most eclipse records from English monastic annals, this seems
original. The chronicle, which is contemporary, is frequently concerned
with local events. Since the Feast of the Apostles Peter and Paul was on
Jun 29, the Vigil of the Octave would be 6 days later or Jul 5. This eclipse
was recorded as producing intense darkness in Iceland (see section 11.8).

RESULTS
For a partial eclipse at Lanercost, either AT < 8700 sec or > 8900 sec.
Since other roughly contemporaneous observations indicate a value for
AT of less than about 1000 sec, the observation of a partial eclipse,
although confirmed, is redundant.
(NB local time of maximum phase = 6.7 h; solar altitude = 24 deg.)

(8) A.D 1330 Jul 16 [Monday] (total, 1.02): Zbraslav (Czech Republic)
1330. In this same year on the Ides of July (Jul 15) at the 8th hour of
the day, the Sun was so greatly obscured that of its great body only a
small extremity like a three-night-old Moon was seen.
[Chronicon Aulae regiae; quoted by Ginzel (1884a).]
The town in which the monastery of Aula Regia is situated is now
known as Zbraslav; this is located about 10 km to the south of Prague.

RESULTS
For a partial eclipse at Zbraslav, either AT < 890 sec or > 1210 sec.
(NB local time of maximum phase = 17.2 h; solar altitude = 23 deg.)

(9) AD 1333 May 14 [Friday] (annular, 0.997): Coimbra (Portugal)

In the (Julian) year 1371, on the 14th day of the month of May, there
was an eclipse of the Sun and the Sun became so diminished that it
resembled nothing so much as the new Moon, very small in appearance. It
increased in size and returned to its normal state; and as it grew it went
through many colours, in such a way that the day was very dark, deprived
of its brightness. This occurred at the hour of midday, and the Sun
remained thus for one hour and a half.
[Chronicon Conimbricense; in Portuguese;
Florez (1747, vol. XXIII, p. 344).]
As noted in an earlier record from Coimbra - AD 1239 (see section
11.6.1) - the year is reckoned from the first year of the Julian calendar.
 11.7 Large eclipses which were definitely partial All

RESULTS
For a partial eclipse at Coimbra, either AT < 4150 sec or > 4250 sec.
Because the belt of annularity was so narrow, only a very restricted
range of AT is disallowed. Many roughly contemporaneous observations
indicate a value of AT at this date of considerably less than about 1000
sec so that the above limits, although not in discord with other results,
are redundant.
(NB local time of maximum phase = 14.2 h; solar altitude = 55 deg.)

(10) AD 1354 Sep 17 [Wednesday] (total, 1.004): Perugia (Italy)

In this year on 17 September that novelty appeared. The Sun became
dark on a Wednesday at about the third hour and it lasted for the space of
two hours. Above the Sun and Moon, which were joined together - that is,
the Moon was covering the Sun - there appeared a very large star with
fiery rays like a torch... Many people viewed the rays of the small Sun by
reflection in a mirror or in clear water. And the rays of the Sun were so
small and so dark, on account of the Moon covering the Sun, that there
did not remain unobscured as much as 3fingersof the Sun... Everyone
appeared deathly pale.
[Memorie di Perugia dall'anno 1351 al 1438; in Italian;
quoted by Levi-Donati (1987).]
This account of a partial eclipse is clearly much exaggerated. However,
the estimate of the magnitude in digits is interesting. The 'very large star
with fiery rays like a torch' seen above the Sun was possibly a bright
comet, although East Asian records make no mention of such an object.
It was not Venus, which was about 18 deg to the east of the Sun and in
mid-morning would be at a rather lower altitude.

RESULTS
For a partial eclipse at Perugia either AT < 4200 sec or > 4320 sec.
As in AD 1333, the belt of totality was extremely narrow and this leads
to a very narrow excluded zone, only 120 sec wide.
(NB local time of maximum phase = 9.4 h; solar altitude = 34 deg.)

(11) AD 1605 Oct 12 [Wednesday] (total, 1.03): Forcalquier (France)

Wendelin at Forcalquier in Provence saw the whole Sun hidden apart
from a very narrow thread towards the north, which he ascribed to the
illuminated atmosphere.
[Riccioli (1665, Book II, chap. 2); quoted by Zawilski (1994).]
Presumably Wendelin attributed the narrow thread of light to scattering
in the supposed lunar atmosphere.
422 11 Eclipse records from medieval Europe

Table 11.6 AT limits deduced from

medieval European observations of large
partial solar eclipses.

AT Range (sec)
Year LL UL
1147 2870 1680
1178 2320 1140
1330 1210 890
1333 4250 4150
1354 4320 4200
1605 — 1060

RESULTS
For a partial eclipse at Forcalquier in which the southern portion of the
solar disc was obscured, AT < 1060 sec. At this late date, immediately
before the telescopic era, this result - although sound - is by no means
critical.
(NB local time of maximum phase = 13.7 h; solar altitude = 35 deg.)

The AT limits derived in this section - apart from the discordant results
from AD 1133 and 1153 - are listed in table 11.6. In each case, a set of
AT values is excluded. Hence in this table, as in table 8.8, the lower limit
of one acceptable range of AT is actually higher than the upper limit
which bounds another acceptable range of AT.

11.8 Detailed records which are deficient in some way

Although alleging a very large eclipse, the following records are lacking
in important details. Either (i) they do not necessarily imply totality or
annularity or (ii) the place of observation is in doubt. They will thus
not be used to determine AT. I have included the observations because
they give some additional indication of the remarkable variety of the
descriptions which are preserved in medieval European texts - as well as
further illustrating the effects of a major solar eclipse on an unsuspecting
observer. In each case I have used a preliminary value of AT (determined
from equation (8.1)) to deduce the approximate magnitude and local time
of greatest phase at the assumed place of observation.

(1) AD 733 Aug 14 [Friday] (annular, 0.98): northern England

In the year 733 an eclipse of the Sun occurred on the 19th day before
the Kalends of September (i.e. Aug 14), about the third hour of the day,
 11.8 Detailed records which are deficient in some way 423

with the result that almost the whole of the Sun's disc seemed to be
covered by a black and horrid shield.
[Bedae Continuatio; ed. Plummer (1896, vol I, p. 361)]
This accurately dated account is one of the earliest observations of an
eclipse recorded in an English chronicle. However, it is not clear from the
text whether the ring phase was witnessed. Possibly a very large partial
obscuration of the Sun would suffice.
Calculations based on equation (8.1) (AT = 3100 sec) indicate that in
northern England (lat. = 55 deg, long. = +1.5 deg) - the most likely
source of information for the continuation of Bede's history - the eclipse
would reach a magnitude of 0.91 at about 8.8 h. Although small, this
may have been sufficient to produce the observed effects. The track of
annularity itself did not reach further north than latitude 52.7 deg N at
any point on the Earth's surface and would cross S. England.

(2) AD 812 May 14 [Friday] (total 1.04): Syria?

In the year 1123 (Seleucid), on the 14th of Ayyar (= May) there was a
total eclipse of the Sun from the ninth to the 11th hours. The darkness
was as profound as night; the stars were seen and people lit torches. The
Sun eventually reappeared over about an hour.
[Chronicle of Michael the Syrian, Book XII, chap. 7; in Syriac; translated
from the rendering into French by Chabot (1905, vol. Ill, p. 26).]
This is one of the earliest detailed accounts of a total solar eclipse from
any part of the world. The event occurred more than three centuries
before the chronicler's own era. Although the date is accurately recorded
(1123 Seleucid = AD 811/2), the place of observation is unfortunately
not specified.

(3) AD 1133 Aug 2 [Wednesday] (total, 1.07): Kerkrade (Netherlands)

In the year from the Incarnation of our Lord 1133, darkness occurred
over the whole Earth at the time of an eclipse of the Sun about midday. It
lasted for almost a whole hour of the day, the Moon being on the 17th
day, on the 4th day before the Nones of August (Aug 2). For then, just as
at night, stars appeared, and the birds in the sky flew away, and the
ground was wet with dew. Men were struck with very great terror; they
thought that the last day had come.
[Annales Rodenses, MGH, XVI, 710.]
Although the date on the Julian calendar is correctly given, it seems
likely that the reported '17th day' of the Moon's age is a scribal error for
the 27th day. This is one of the very few early records to mention the
formation of dew resulting from the fall in temperature during a large
eclipse. Calculations using equation (8.1) (AT = 1200 sec) indicate that
424 11 Eclipse records from medieval Europe

at Kerkrade the magnitude would be about 0.995 at 12.0 h - close to the

reported time.
(4) AD 1133 Aug 2 [Wednesday] (total 1.07): Zwiefalten? (Germany)
1133. On the third day before the Nones of August... an eclipse of the
Sun occurred. Concerning this same eclipse, we read in a Zwiefalten
parchment. After meal-time there, at the beginning of midday, the Sun was
suddenly covered with darkness (so it is written); people working in the
houses, since they could not see to work, hurried outside in order to see,
but things had suddenly changed. Then, with no clouds appearing, the disc
of the Sun was very clearly without light. After half an hour, a small
portion (particulam) began to appear, like the one-day old Moon.
Afterwards the Sun gradually emerged, eventually with its usual splendour.
[Footnote to Bertholdi Zwifaltensis Chronicon; MGH, X, 120;
quoted by Zawilski (1994).]
The above passage, which is not in the Bertholdi Zwifaltensis Chronicon
itself, is from the Annales Suevici, a chronicle compiled around AD 1590 by
Martinus Crusius. Crusius, who frequently abstracted from the Bertholdi
Zwifaltensis Chronicon, came across the parchment describing the eclipse
while searching for material for his own chronicle. However, it would
appear that Bertholdus, who was a monk at the monastery of Zwiefalten
at the time of the eclipse and became abbot there in AD 1139, was unaware
of this document. The parchment may well have been acquired by the
monastery from some other source long after the time of Bertholdus.
With AT = 1200 sec (equation (8.1)), the computed magnitude at
Zwiefalten is 0.975 at 12.4 h.
(5) AD 1140 Mar 20 [Wednesday] (total 1.06): Malmesbury (England)
1140. During this same year in Lent, on the 13th day before the
Kalends of April (March 20), at the 9th hour, on the 4th day of the week
(Wednesday), there was an eclipse of the Sun throughout the whole of
England, as I have heard. With us, certainly, and with all our neighbours,
the solar eclipse was so remarkable that men seated at tables, as almost
everywhere they were at the time, for it was Lent, feared the primeval
chaos; but soon learning what it was, they went outside and saw stars
around the Sun. It was thought and said by many, not mistakenly, that the
King (Stephen) would not continue to reign for a year without loss.
[Willelmi monachi Malmesburiensis Historia Novella, lib. II;
Potter (1955, pp. 42-43).]
William was librarian at the monastery of Malmesbury. His 'New
History', which covers the interval from AD 1126 to 1142, is an important
source for the period. In February of 1141, less than a year after the
eclipse, King Stephen was imprisoned following the defeat of his army.
 11.8 Detailed records which are deficient in some way 425

From equation (8.1) (AT = 1200 sec) it may be computed that the
eclipse was fully total at Malmesbury at 14.6 h. This evidently was the
time when men were dining, probably after the midday service (Sext).

(6) AD 1230 May 14 [Tuesday] (total 1.05): Belvoir? (England)

In the same year (1230) an extraordinary eclipse of the Sun occurred, in
the very early morning immediately after sunrise, on the day before the
Ides of May (May 14) in Rogationtide, namely the third day of the week
(i.e. Tuesday). As a result, the workers in thefieldsand many others,
leaving their morning's work on account of the excessive darkness, decided
to return to bed and go back to sleep. But at length, after the space of one
hour, to the astonishment of many, the Sun regained its usual brightness.
[Roger of Wendover: Flores Historiarum; Hewlett (1887, vol. II, p. 384).]
Around this time, Roger of Wendover was prior of Belvoir in Le-
icestershire. Although his Flores Historiarum ('Flowers of History') is a
chronicle of events in the whole of England, he may have witnessed the
phenomenon personally. Calculations based on equation (8.1) (AT = 900
sec) indicate that this eclipse would be total throughout much of eastern
England (including Belvoir) at sunrise, the local time being about 4.1
h. This was indeed in the 'very early morning' and emphasises the long
working day of agricultural labourers during the summer months.

(7) AD 1312 Jul 5 [Wednesday] (total 1.01): Eastern Iceland

1312. A solar eclipse occurred in the month of March; there was such
great darkness in the eastern fjords that one could not find one's way
about, neither on the water nor indeed on land. Afterwards came a great
mortality of men.
[Oddaverja annal; in Icelandic;
trans, into German by Thorkellson (1933).]
The month is incorrectly given as March. Since the eastern fjords
cover almost the whole of the east coast of Iceland, the exact place of
observation cannot be located, but the mean co-ordinates are: lat. ~ +65
deg, long. ~ +14 deg. With AT = 800 sec (from equation (8.1)), the
eclipse would be indeed total here at 6.3 h.

(8) AD 1415 Jun 7 [Friday] (total 1.07): Kobrin/Mytho (Belarius)

1415. An astonishing eclipse of the Sun. When King Vladislav was
riding from Kobrin to Mytho, on the 6th day of the week, after the Octave
of Corpus Christi (= Oct 6) at the 3rd hour of the day, a notable eclipse
of the Sun appeared. As a result of this unexpected and unfamiliar event,
King Vladislav and his followers were at first astonished and bewildered,
later full of reverence. For it was so notable that the birds, terrified by the
sudden darkness, fell to the ground, and the stars were shining as if by
426 11 Eclipse records from medieval Europe

night. The same King Vladislav was forced to stop on account of the
darkness and could not proceed until the obscuration of the Sun passed.
[Joannis Dlugossi: Annales sen Chronicae inditii Regni Poloniae;
quoted by Zawilski (1994).]
Despite the description of the sudden fall of darkness, the complete
disappearance of the Sun is not directly alluded to. The chronicler, Jan
Dlugosz, who was himself born in 1415, is clearly quoting an eye-witness
account. Kobrin and Mytho (near Lida) are about 100 km apart. It
can only be presumed that the observation was made somewhere between
these towns.
With AT = 400 sec (from equation (8.1)), the eclipse would be total at
both Kobrin and Mytho at 8.0 h.

(9) AD 1476 Feb 25 [Sunday] (total, 1.04): Pskov (Russia)

(6984.) During the same month (i.e. February), on the 25th day, in
Butter week, at the second hour of the day, as the Sun was rising and
people were going to markets and elsewhere in the town, it suddenly began
to grow darker and the darkness lasted for a little while, less than an hour;
people could not see one another on the market nor anywhere else in the
town, and they were frightened; and again God gave us light as before.
[Pskovskaya I Letopis; in Russian*; trans. Vyssotsky (1949, p. 20).]
Here we have another very large eclipse at sunrise. The year 6984
is expressed relative to the Byzantine era, and thus corresponds to AD
1475/6. From equation (8.1) (AT = 300 sec), totality would occur soon
after sunrise (6.9 h), the solar altitude then being only about 2 deg.

(10) AD 1544 Jan 24 [Thursday] (total, 1.004): Neider Alteich

In the year of Our Lord 1544, on the Feast of St. Timothy the Apostle
(i.e. Jan 24), there was an eclipse of the Sun before breakfast at the 9th
hour, and there was darkness over the whole world (in universa terra) as if
it was entirely night, with the result that people could not see one another
well. Such an eclipse was not in the memory of men.
[Notae Altahenses; MGH, XVII, 426.]

Using the extrapolated AT value of 120 sec based on early telescopic

data (rather than equation (8.1)) for this relatively recent date leads to a
magnitude at Neider Alteich of 0.997 at 9.5 h. Evidently the recorded local
time was expressed relative to midnight. In order for this eclipse to have
been total (which is not specifically alleged) at Neider Alteich —290 < AT
< —40 sec. These values are in discord with roughly contemporaneous
figures.
 11.9 Timed solar eclipse contacts 427

11.9 Timed solar eclipse contacts

Goldstein (1979) has drawn attention to a number of timed eclipse contacts
measured by the astronomer Levi ben Gerson in southern France between
AD 1321 and 1339, and also by Johannes de Muris in Northern France
in the years 1333 and 1337. By this period, AT would be fairly small
(only 750 sec according to equation (8.1)), and it seems doubtful whether
timings of eclipses made with the unaided eye can yield a useful result for
this parameter. However, the measurements by de Muris, although few in
number, seem much more careful than those of ben Gerson and I have
considered them in detail below.
The observations by Johannes de Muris were discovered in a Latin
manuscript by Beaujouan (1974). The main objective of de Muris in
making the measurements seems to have been to compare them with
calculations based on the Alfonsine tables. He also made several determi-
nations of the meridian altitude of the Sun. I have translated the following
accounts from the texts published by Beaujouan.

(1) AD 1333 May 14: Evreux (France)

Three brothers and I, in the presence of the Queen of Navarre, observed
the beginning of this eclipse at Evreux near St Germain. And the altititude
of the Sun at the moment of the beginning of the eclipse was close to 50
deg, and the altitude at the end of the eclipse was 33 deg. The magnitude
of the eclipse, according to our estimate, was 10 digits... And thus the
eclipse began after our midday by 2 hours and 20 minutes....

Johannes de Muris noted that the eclipse took place about 17 minutes
earlier than as predicted by the Alfonsine tables. This same eclipse was
observed to be partial in Coimbra - see section 11.7.

RESULTS
(1) Assuming a solar altitude of 49.5 deg (in the west) at first contact,
equivalent LT = 14.40 h, U T = 14.24 h. Computed TT = 13.89 h. Hence
A T = - 1 2 5 0 sec.
(ii) For a solar altitude of 33 deg (in the west) at last contact, equivalent
LT = 16.18 h, U T = 16.03 h. Computed TT = 16.30 h. Hence AT = 1000
sec.

(2) AD 1337 Mar 3: St Germain des Pres (France)

In the current year of our Lord 1337, on the 3rd day of March after
sunrise, a Monday, we saw the beginning of the eclipse, with the altitude
of the Sun 10 deg. And already part was sensibly eclipsed, whereby we
concluded that the edges of the luminaries could have touched at the
altitude of 9 deg. Similarly we saw the exit of the Moon from contact with
the Sun, as far as was possible, with the Sun at an altitude of 27 deg and
428 11 Eclipse records from medieval Europe

about 30 min... (The magnitude) was only 5 digits. In this experiment

there were ten of us present and several had good astrolabes.
Johannes de Muris remarked that the eclipse occurred about 16 minutes
earlier than expected from the Alfonsine tables.

RESULTS
(i) For a solar altitude of 9 deg (in the east) at first contact, equivalent LT
= 7.20 h, UT = 7.22 h. Computed TT = 7.49 h. Hence AT = 950 sec.
(ii) For a solar altitude of 29 deg (in the east) at last contact, equivalent
LT = 9.61 h, UT = 9.63 h. Computed TT = 9.71 h. Hence AT = 300 sec.
The AT result obtained from the first contact measurement in AD 1333
is in poor agreement with the other values derived in this section. Although
it may well be that the recorded altitude of 'nearly 50 deg' at first contact
on this occasion represents a scribal error, the other measurements are too
few in number for it to be rejected as discordant. It would thus appear
that the whole set of results is inadequate to obtain a refined value for
AT. Incidentally, the two magnitude estimates are remarkably accurate.
Using equation (8.1) (AT = 600 sec), the calculated magnitude in AD
1333 would be 0.82 - very close to the observed figure of 0.83. Similarly,
in AD 1337, the calculated magnitude would be 0.42 - identical with the
observed value. However, at this late epoch it is most unlikely that these
unaided-eye estimates would themselves yield meaningful results for AT.
Rather more than a century later, the German astronomers Regiomon-
tanus and Bernard Walther also made careful measurements of eclipse
times (mainly of lunar obscuration) at Nurnberg. These observations ap-
pear to be of similar precision to those of Johannes de Muris. For details,
see Zinner (1990). However, by this period AT would be so small (less
than about 200 sec) that the observations are of mainly historical interest.
Accordingly, they will not be analysed here.

11.10 An eighth century occultation of Jupiter by the eclipsed Moon

Unique in ancient and medieval history is an account of an occultation
of the planet Jupiter by the eclipsed Moon in AD 755. The report of
this compound event, which is due to the twelfth century English annalist
Symeon of Durham, may be translated as follows:
AD 756... Also the Moon was covered over with the redness of blood
on the 8th day before the Kalends of December (Nov 24), in the 15th day
of its age, that is, at full Moon. And the darkness gradually decreasing, it
returned to its original light. Further, astonishingly, a bright star following
that same Moon, and passing through it, preceded the luminary by as
great a space as it had followed it before it was obscured.
[Symeonis Monachi Historia Regum; Arnold (1885, p. 41).]
 11.10 An occultation of Jupiter by the eclipsed Moon 429

66°.5 66°.O 65°.5 65°.O 64°.5 Long

Jupiter
-0.5°
Last
contact

-1.0°-

Fig. 11.10 Calculated position of the Moon relative to the planet Jupiter at each
phase of the total lunar eclipse of AD 755 Nov 23/24.

The dates of the birth and death of Symeon, who was the precentor of
Durham Cathedral in the early twelfth century, are uncertain. However,
he is known to have written his history between 1104 and 1129. He seems
to have taken his eighth century material from a contemporary chronicle,
which because of the frequent references to York is sometimes referred to
as the 'York Annals' (Blair, 1963).
Since celestial bodies appear to move from east to west on account of the
diurnal rotation of the Earth, the text may be understood to imply that the
bright star was first seen to the east of the Moon and afterwards appeared
to the west of it. There was a lunar eclipse on Nov 11 in AD 756 but this
was only partial and there was no bright star or planet nearby (Jupiter
would be about 14 deg to the east of the Moon on that date). However,
on the evening of Nov 23/24 in AD 755 the totally eclipsed Moon was
very close to the planet Jupiter and this must be the correct identification.
Although fairly rare in other civilisations, many medieval European
accounts of total lunar eclipses allude to the blood-red colour of the
Moon; possibly the annalists were inspired by the Old Testament reference
in Joel II, 31 - repeated in Acts II, 20.
Calculations based on equation (8.1) (AT = 2950 sec) show that from
the viewpoint of York, Jupiter would be occulted by the Moon during the
closing stages of the eclipse of AD 755 Nov 23/24, as the record asserts
(see figure 11.10).
430 11 Eclipse records from medieval Europe
I I
3200
*l

2800 Key
!
Central eclipse
I
2400 I ; Partial eclipse

2000

1600

1200 I
I
800
|

400
I i
I
0 I
I
. *! ! i
400
+800 +900 +1000 +1100 +1200 +1300 +1400 +1500 +1600
Year

Fig. 11.11 AT limits obtained from European observations of both central and
partial solar eclipses. Discordant results are indicated by an asterisk.

The computed local times of the eclipse contacts are as follows: begin-
ning at 16.9 h; immersion at 18.0 h; emersion at 19.5 h; and end at 20.7
h. The Moon would start to occult Jupiter around 19.8 h local time (soon
after emersion) and the occultation would end at 20.7 h (just before last
contact). Jupiter would appear to pass about midway between the lunar
centre and its upper limb. Although the observation is too imprecise for a
value of AT to be deduced, it is of considerable historical interest because
of its uniqueness.

11.11 Conclusion
In figure 11.11 are plotted the AT limits obtained in this chapter from
both central solar eclipses (section 11.6) and partial eclipses (section 11.7).
The anomalous results obtained from the partial eclipse observations in
1133 (Vysehrad) and 1153 (Erfurt) - are indicated by asterisks. Apart from
these, it can be seen that there is generally good mutual accord between
the various AT limits, especially bearing in mind that the observers were
not astronomers. It is possible to trace a gradual decline in AT from
about 2000 sec around the epoch +900 to less than 200 sec near +1600.
 12
Solar and lunar eclipses recorded
in medieval Arab chronicles

12.1 Introduction
Arab observations of solar and lunar eclipses which are of value in the
study of the Earth's past rotation originate exclusively from the medieval
period. These records are found in two main sources: (i) chronicles, and (ii)
compendia on astronomy. As might be expected, the observations reported
in chronicles are essentially qualitative; measurements of any kind are
fairly rare. Eclipses and other celestial phenomena (such as bright comets
and meteor showers) were mainly noted on account of their spectacular
nature. Yet chronicles contain some important astronomical records. The
present chapter will be devoted to eclipses described in these works. Arab
compendia on astronomy contain many careful measurements of the times
of the various phases for both solar and lunar eclipses. These observations
will be discussed in chapter 13.
The Arabic names for the Moon and Sun are respectively al-Qamar
and al-Shams. In general, medieval Arab chronicles use the term khusuf
al-Qamar for an eclipse of the Moon and kusuf al-Shams for an eclipse of
the Sun. These designations are still in use today, both among astronomers
and the general public. However, medieval Muslim astronomers mostly
used kusuf for both types of eclipse, adding the appropriate term for the
Moon or Sun. The word khusuf means 'sinking', but in describing a lunar
eclipse it came to mean a failing of the Moon's light. By contrast, kusuf
means a 'cut' - i.e. in the solar (or lunar) limb. Totality of an eclipse was,
and still is, usually indicated by expressions such as tamm, kamil or kulliyy.
Dates of eclipses are normally carefully recorded in Arab chronicles,
but when the local time of occurrence is given, this is nearly always
crudely expressed - rarely to better than the nearest hour. As remarked
in previous chapters, such informal estimates of time are of negligible
value in the determination of AT. Several detailed accounts of total solar
eclipses are preserved from the medieval Arab world and these merit
careful investigation. There are also a few allusions to very large partial

431
432 12 Solar and lunar eclipses recorded in medieval Arab chronicles

eclipses of the Sun, but no references to central annular obscurations.

Apart from the records of large solar eclipses (both total and partial), the
only viable observations for the determination of AT are instances where
the Sun or Moon rose or set eclipsed. It should be emphasised that the
occasional estimates of the magnitudes of partial solar eclipses are too
rough to be of value.

12.2 Chronicles and chroniclers

Arab chronicles cover much the same period as their European coun-
terparts: mainly from about AD 800 to 1500. However, whereas vast
numbers of European annals have been collected together and published
(see chapter 11), relatively few Arabic texts of this nature have appeared
in print. This is probably the main reason why only a relatively small
number of reports of eclipses from the medieval Arab world have so far
come to light. To date, searches for eclipse observations have concentrated
exclusively on published chronicles. It may well be that many further ob-
servations are contained in manuscripts, but at present it is impossible to
assess the potential of this material - much of which is probably widely
scattered.
Until very recently, researchers had shown little interest in the eclipse
records in Arab chronicles. However, in * 1989, the author along with
Dr Said S. Said and Dr Wafiq S. Rada, published the results of an extensive
literature search through printed annals for references to eclipses of the
Sun (Said et a/., 1989); this work will subsequently be abbreviated to SSR.
Many records of lunar eclipses have since been uncovered in these and
other printed sources (Stephenson and Said, 1997). All of the observations
discussed in this chapter are taken from these two compilations. Printed
editions of the Arab chronicles referred to below are listed in the appendix
at the end of this chapter (section 12.10), rather than placed in the main
list of references. I have not considered it necessary to include translations
of the names of various book titles, some of which are quite lengthy.
As was also true of medieval European annalists, interest in celestial
phenomena among Arab chroniclers evidently varied very much from one
individual to another. Several writers seem to have had a special interest
in astronomical matters, recording many eclipses and other occurrences.
However, other chroniclers passed over such events in silence. Sometimes
an eclipse was regarded as foretelling the death of a ruler. Not all such
predictions proved to be successful! Thus, in reporting the lunar eclipse of
AD 1487 Feb 7/8 the contemporary annalist Ibn Iyas relates the following:

(892 AH.) In (the month of) Safar, the Moon's body was eclipsed and
the Earth was darkened. It remained in eclipse for about 50 degrees (i.e. 3
 12.2 Chronicles and chroniclers 433

Table 12.1 The principal Arab chroniclers recording eclipses of the

Sun and Moon.
Chronicler AD dates Main domicile
al-Tabari 839-923 Baghdad
Ibn Hayyan 987-1076 Cordoba
Ibn al-Jawzi 1116-1201 Baghdad
Ibn al-Athir 1160-1233 Mosul
Abu Shama 1199-1266 Damascus

al-Maqrizi 1367-1442 Cairo

al-'Asqalani 1372-1449 Cairo
Ibn Taghri Birdi 1411-1470 Cairo
Ibn Iyas 1448-1524 Cairo
Ibn Tulun 1475-1546 Damascus
al-'Umari 7-1816+ Mosul

hours 20 min). People were saying that the demise of the Sultan was
coming near. Nothing of what they said happened and the Sultan stayed
(in power) for a long time after that.
[BadaV al-Zuhur fi Waqa'V al-Duhur (vol III, p. 238);
trans. Stephenson and Said (1997).]

The elderly Sultan (al-Malik al-Ashraf Qaytabay) lived for a further

nine years after the eclipse.
Ibn Iyas had the unusual habit for an annalist of recording time intervals
- both astronomical and in everyday life - in degrees (each equivalent to
4 minutes), as in the above account. The computed duration of the above
event is 51 deg (3 h 23 m) so that on this occasion Ibn Iyas' figure is
remarkably accurate.
The inclusive dates and main residence of each of the annalists who
recorded the eclipses which are investigated in this chapter are listed in
table 12.1. Exact dates are known for all of these men except al-'Umari,
for whom neither the date of birth nor death is established; he is only
known to have died some time after AD 1816.
In Arab as well as European chronicles, eclipse observations are cited
alongside mainly non-astronomical matters, purely in order of occurrence.
Generally, Arab annalists were in the habit of reporting material from
a wide geographic area; few of their writings are comparable with the
town and monastic chronicles of medieval Europe. Hence unless the
location where an eclipse was seen is clearly stated or there is good reason
for believing that the writer witnessed the event himself, the true place
434 12 Solar and lunar eclipses recorded in medieval Arab chronicles

Table 12.2 Geographic co-ordinates of places of observation for

eclipses noted in Arab chronicles.

Place Latitude (deg) Longitude (deg)

Aleppo 36.23 -37.17
Baghdad 33.33 -44.43
Cairo 30.05 -31.25
Cizre 37.35 -42.22
Cordoba 37.88 +4.77

Cueva de la Mora 40.33 +3.92

Damascus 33.50 -36.32
Granada 37.17 +3.58
Mosul 36.35 -43.13
Olmos 40.15 +3.98

of observation may be in doubt. Fortunately, some annalists are quite

specific regarding these details.
The modern names and geographic co-ordinates (in degrees and deci-
mals) of the various sites where the eclipses mentioned in this chapter were
observed are listed in table 12.2. As elsewhere in this book, the adopted
sign convention is that longitudes west of the Greenwich meridian are
positive, as are latitudes north of the equator.

12.3 Date conversion

Eclipse dates in Arab chronicles and astronomical compendia are mainly
expressed in terms of the Islamic lunar calendar. On this calendar, years
are numbered from al-Hijrah - the migration of the Prophet Muhammad
from Mecca to Medina in AD 622. Years reckoned from this epoch are
customarily designated by AH (Anno Hegirae). The Islamic year consists
of 12 lunar months, each of either 29 or 30 days. From the time of the
Prophet himself, intercalation has never been practised so that as a period
of 12 lunar months is about 11 days shorter than the orbital period of the
Earth around the Sun, the beginning of each year regresses through the
seasons in approximately 33 years. In particular, a Julian or Gregorian
century is closely equal to 103 Islamic years.
Whenever possible, each lunar month should commence with the sight-
ing of the lunar crescent. At all significant centres of population a watch
is kept for the young crescent at the close of the 29th day of every month.
Should the crescent prove to be invisible - for example on account of
cloud - an extra day is added to that month. Since meteorological condi-
 12.3 Date conversion 435

Table 12.3 Lunar months of the Islamic year.

No. Month Days
1 Muharram 30
2 Safar 29
3 Rabi' al-Awwal 30
4 Rabi' al-Akhir 29
5 Jumada al-Ula 30
6 Jumada al-Ukhra 29

7 Rajab 30
8 Sha'ban 29
9 Ramadan 30
10 Shawwal 29
11 Dhu al-Qa'dah 30
12 Dhu al-Hijjah 29/30

tions are a major factor in determining crescent visibility, it is possible for

adjacent communities to begin the month one day apart. Observation of
the crescent is especially important in the last five months of the Islamic
year. These are Sha'ban, Ramadan and Shawwal (to fix the month of
Fasting - i.e. Ramadan), and also Dhu al-Qa'dah and Dhu al-Hijjah (to
determine the dates of Pilgrimage to Mecca).
Although date conversion can be effected using computation of the
visibility of the crescent based on a series of observational boundary con-
ditions (see for example the details in SSR), it is more convenient to employ
tables based on a 'standardised' Islamic calendar. The 12 lunar months of
the Islamic year along with their tabular lengths are listed in table 12.3.
It should be noted that the following month names are occasionally
confused in Arabic literature: (i) Rabi' al-Awwal (often abbreviated to
Rabi I) and Rabi' al-Akhir (Rabi II); (ii) Jumada al-Ula (Jumada I)
and Jumada al-Ukhra (Jumada II). Because the first part of the name is
identical in each case, a careless scribe might easily make a mistake. NB
the terms al-Awwal and al-Akhir (both masculine) mean respectively first
and last, as do al-Ula and al-Ukhra (both feminine).
In both this and the succeeding chapters, interconversion of dates be-
tween the Islamic and Julian calendars has been achieved using a computer
program modelled on the well-known tables of Freeman-Grenville (1977).
These tables are based on the following assumptions:
(i) The first day of Muharram in the year in which the Hijra took place
was Friday July 16 in AD 622. (NB some authorities prefer the
alternative of July 15.)
436 12 Solar and lunar eclipses recorded in medieval Arab chronicles

(ii) The Islamic year consists of alternate lunar months of 30 and 29

days duration in the first eleven months (from Muharram to Dhu
al-Qa'dah) and either 29 or 30 days in the twelfth month (Dhu
al-Hijjah). The Islamic year thus contains either 354 or 355 days.
(iii) The number of days assigned to the twelfth month follows a regular
pattern based on a 30-year cycle commencing in AD 622: it contains
30 days in the years 2, 5, 7, 10, 13, 16, 18, 21, 24, 26 and 29 of every
cycle and 29 days in other years. The total number of days in a
cycle of 360 months is thus 10631 - an average length for the lunar
month of 29.530 56 days. This compares very favourably with the
true synodic month length of 29.530 59 days.
As is true for any lunar calendar based on first visibility of the crescent
(for example that of ancient Babylon), fully accurate conversion of dates
to the Western calendar can never be guaranteed unless independent
information is supplied - for example the weekday. However, errors
seldom exceed a single day, which is unimportant for events as rare as
eclipses.
Many Arab accounts of eclipses specify the day of the week. From
Sunday (the first day of the week), to Thursday (the fifth day of the week),
each day has a name which simply indicates the appropriate number. For
example Sunday is yawrn al-Ahhad, meaning 'Day One'. Friday is called
yawm al-Jumu'ah, meaning the 'Day of Assembly'. Saturday is known as
yawm al-Sabt - meaning the 'Day of Rest'; this is presumably derived
from the Hebrew Sabbath (Shabbat). When the weekday is specified, it
is usually possible to derive the exact date on the Julian or Gregorian
calendar and in such cases there is usually precise accord with calculated
eclipse dates.
One of the earliest Arab eclipse records containing accurate dates
originates from AD 882. This is cited by the contemporary Baghdad
writer al-Tabari as follows:
(269 AH.) In (the month of) Muharram in this year, the Moon was
eclipsed on the night of the 14th (day) and set eclipsed. The Sun was
eclipsed at the time of sunset on Friday, when two nights remained to the
completion of Muharram, and set eclipsed. So in this month there were
both lunar and solar eclipses.
[al-Tabari, Tarikh al-Tabari: Tarikh al-Rusul wa'l Muluk, vol IX, p. 613;
trans. SSR.]
In the Islamic calendar, lunar eclipses consistently take place on or
about the 14th day of a month and solar eclipses around the 28th day.
When the two dates in the above example are converted to the Julian
calendar using the scheme outlined above, they correspond respectively
 12.3 Date conversion 437

Table 12.4 Daily prayer times in Islam.

Prayer Time
Fajr Dawn
Zuhr Noon
'Asr Afternoon
Maghrib Sunset
'Isha Nightfall

to Aug 2/3 and Aug 17 in AD 882. In particular, Aug 17 was indeed a

Friday. On each of these dates it may be calculated that an eclipse did
occur, indicating that al-Tabari's chronology is very accurate.
The Islamic day begins at sunset, roughly 6 hours before the start
of the civil day (i.e. local midnight). Hence in converting the recorded
date of a lunar eclipse which began before local midnight to the Julian
(or Gregorian) calendar, one day must be subtracted to allow for this
discrepancy. In the case of a lunar obscuration commencing after midnight
- and for all solar eclipses - the Islamic and civil dates are identical so
that this correction is unnecessary. The practice adopted throughout this
book of using double dates for all lunar eclipses avoids such confusion.
Local times of eclipses are occasionally expressed relative to the five
Islamic prayer times. These are listed in table 12.4. For example, in AD
1377 it is recorded that an obscuration of the Sun lasted from Zuhr (the
noon prayer) until after 'Asr (the afternoon prayer). Similarly, in AD 1399
a lunar eclipse was said to begin after Isha (the nightfall prayer) and to
last until midnight.
Theoretically, the dawn prayer begins when the Sun's depression below
the eastern horizon is about 18 deg while the nightfall prayer starts when
the Sun has reached a similar depression below the western horizon. As
it happens, these angular values correspond to the limit of astronomical
twilight in modern convention; when the Sun is beyond this limit there
is negligible sunlight glow. In high latitudes, of course, the Sun never
attains a depression of 18 deg during the summer months, but in the
low latitudes occupied by most Muslim countries this problem does not
arise. The sunset prayer closely follows sunset - normally a well-defined
reference moment. However, to fix the time of the noon and afternoon
prayers may require the use of a sundial, as found at many mosques.
The noon prayer begins a little after the Sun crosses the meridian (in
principle, when the length of the shadow of a gnomon begins to show a
noticeable increase after the noon minimum). By comparison, according
to majority opinion the afternoon prayer commences when the shadow
438 12 Solar and lunar eclipses recorded in medieval Arab chronicles

length equals its noon length plus the height of the gnomon itself. For a
discussion of these various definitions, see Said and Stephenson (1991).
One early eclipse (occurring in AD 1241) is dated relative to the Coptic
calendar which numbers years from the persecution of Christians by the
Roman Emperor Diocletian in AD 284. This scheme, which has twelve
months of 30 days followed by 5 or 6 intercalary days, is still adopted
by the Coptic church today. The twelve month-names are derived from
ancient Egyptian names. Each year, which normally begins on Aug 29 of
the Julian calendar, commences with the month Tut.

12.4 Records of total and near-total solar eclipses

In this section I shall only discuss those few records for which the place
of observation is fairly well established and either totality or a very large
partial eclipse is clearly described. At the head of each entry below is
given the tabular Julian date of the eclipse and the calculated weekday.
These are followed by the type of eclipse, computed magnitude in the
central zone and place of observation.
All translations of Arabic texts in this and the following three sections
(12.5 to 12.7) are by SSR, with minor emendations. After each record
are given a few brief comments, followed by the indicated AT range. For
general reference, the local time of maximum phase and the solar altitude
at that time are calculated using equation (8.1).

(1) AD 912 Jun 17 [Wednesday] (total, mag. = 1.06): Cordoba

(299 AH.) In this year, the Sun was eclipsed and all (jamV) of it
disappeared on Wednesday when one night remained to the completion of
(the month of) Shawwal. The stars appeared and darkness covered the
horizon. Thinking it was sunset, most of the people prayed the Maghrib
(Sunset) Prayer. Afterwards, the darkness cleared and the Sun reappeared
for half an hour and then set.
[Ibn Hayyan: al-Muqtabis fi Tarikh al-Andalus, vol. Ill, p. 147.]
The recorded date (including the weekday), is correct. Although Ibn
Hayyan wrote rather more than a century after the event, this section of
al-Muqtabis is a chronicle of the reign of 'Abd Allah b. Muhammad, the
Umayyad Caliph, in Cordoba and frequently mentions events in the city.
Ibn Hayyan himself was born in Cordoba and spent most of his life there.
It thus seems highly likely that the place of observation was Cordoba.

RESULTS
For totality at Cordoba, 880 < AT < 2600 sec.
(NB computed local time of maximum phase = 19.2h, solar altitude =
+ 3 deg.)
 12.4 Records of total and near-total solar eclipses 439

(2) AD 1061 Jun 20 [Wednesday] (total mag. = 1.06): Baghdad

(453 AH.) On Wednesday, when two nights remained to the completion
of (the month of) Jumada al-Ula, two hours after daybreak, the Sun was
eclipsed totally. There was darkness and the birds fell whilst flying. The
astrologers claimed that one-sixth of the Sun should have remained
(uneclipsed) but nothing of it did so. The Sun reappeared after four hours
and a fraction (of an hour). The eclipse was not in the whole of the Sun in
places other than Baghdad and the provinces.

[Ibn al-Jawzi: al-Muntazam fi Tarikh al-Muluk wa'l-Umam, vol. Ill, p. 221.]

A copy of the printed Arabic text of a page containing this record is

illustrated in figure 12.1.
The Baghdad chronicler Ibn al-Jawzi wrote a century after the eclipse.
He is, however, careful to specify the precise place of observation and his
account is clearly based on an eyewitness description. The date which he
gives is exactly correct.

RESULTS

For totality at Baghdad, 800 < AT < 2140 sec.

(NB local time of maximum phase = 7.7 h; solar altitude = 33 deg.)

(3) AD 1176 Apr 11 [Sunday] (total mag. = 1.06): Cizre

(571 AH.) In this year the Sun was eclipsed totally and the Earth was in
darkness so that it was like a dark night and the stars appeared. That was
the forenoon (Duha) of Friday the 29th of (the month of) Ramadan at
Jazirat Ibn 'Umar (Ibn 'Umar's island), when I was young and in the
company of my arithmetic teacher. When I saw it I was very much afraid;
I held on to him and my heart was strengthened. My teacher was learned
about the stars and told me, 'Now you will see that all of this will go
away', and it went quickly.
[Ibn al-Athir: al-Kamilfi al-Tarikh; vol. IX, p. 138.]
For a copy of the printed Arabic text of this account, see figure 12.2.
This eclipse was also observed to be total in Antioch (= Antakya) - as
discussed in chapter 11. Although the year, month and day of the month
are correctly given in the above account, Ibn al-Athir has mistaken the
weekday. Nevertheless, he is careful to tell his readers precisely where he
was when he saw the eclipse; this is now known as Cizre, a Turkish town
on the frontier with Syria. At the time of this event, Ibn al-Athir was
aged 16. He gives a particularly appealing account of the effect of a total
eclipse on a young and unsuspecting eye-witness.
440 12 Solar and lunar eclipses recorded in medieval Arab chronicles

JJtfidiJi) jkii i.wS I* **\ 4; *•>. U Uii:i J j i 6! iljlj

aJli i j U 4_ife J l l j i UjT V U J*l|

Fig. 12.1 Printed Arabic text showing the report of the solar total eclipse of AD
1061 Jun 20 by Ibn al-Jawzi (Baghdad).

RESULTS
For totality at Cizre, 600 < AT < 2350 sec.
(NB local time of maximum phase = 7.1 h; solar altitude = 25 deg.)
Combining the above limits with those set by the observation of the
same eclipse at Antioch (i.e. —190 < AT < 1600 sec) yields a revised
range 600 < AT < 1600 sec. The southern edge of the computed track
of totality calculated with AT = 600 sec (passing through Cizre) and the
northern edge calculated with AT = 1600 sec (passing through Antioch)
are delineated in figure 12.3. For any value of AT in the range 600 < AT
< 1600 sec, the eclipse would be total at both places in accordance with
the records.
 12.4 Records of total and near-total solar eclipses 441

j> j i l v^i*) ^lI X^? ^-« If- j * J ^ t U^-O SJJJU)) ^L^U j > ^ *l w jt^T IK* V V - r ^

oUK* J j ij'jjl^wC*^ 4*1" 4J^^j i ^ i j

1
V'J • *•-* C/s~*J ^-^ ^jbjl J j ^ »j^ O&j ^j^- ^ f e Uljsrjlfj ojjf^ £jJU 4:
Ji jlSj jj**3l 4J! j>>yj jLr la ^Jl J^W J-^j^ ^ ^ W11/
0. \j&4 oj^U a J j J^» jLU o- MI 4^j j ^ Ul/tj J. J oar ^ 1 Jl

-1 > JuJ

Fig. 12.2 Printed Arabic text showing the report of the solar total eclipse of AD
1176 Apr 11 by Ibn al-Athir (Cizre).
442 12 Solar and lunar eclipses recorded in medieval Arab chronicles

Fig. 12.3 Edges of the computed track of totality in AD 1176 for the two AT
limits set by the observations at Cizre and Antioch (southern limit for AT = 600
sec, northern limit for AT = 1600 sec).

(4) AD 1433 Jun 17 [Wednesday] (total mag. = 1.06): Cairo

(836 AH.) On Wednesday the 28th of (the* month of) Shawwal, the Sun
was eclipsed by about two-thirds in the sign Cancer more than one hour
after the 'Asr (Afternoon) Prayer. As the Sun was setting, the eclipse
started to clear. During the eclipse there was darkness and some stars
appeared.
[Al-Maqrizi: al-Suluk fi Ma'rifat Duwal al-Muluk, vol. IV, p. 892.]
Al-Maqrizi was a contemporary writer living in Cairo. It is possible that
he witnessed the eclipse himself. An eclipse of magnitude only | certainly
would not bring out 'some of the stars' but the effect of irradiation may
well have made the degree of obscuration of the Sun seem smaller than
it actually was. Hence the record may best be interpreted as denying
totality. Both Venus and Jupiter were well placed to the east of the Sun,
which was indeed in Cancer (longitude 94 deg), but unless the eclipse was
extremely large, only Venus would be likely to be seen.

RESULTS
For a partial eclipse at Cairo, AT < —4000 or > —2460 sec.
At this relatively recent epoch, extrapolation of telescopic results indi-
cates that AT would be fairly small. In consequence, the above limits are
of little significance. They will not be considered further.
(NB local time of maximum phase = 18.2 h; solar altitude = 9 deg.)
 12.5 Other records of large solar eclipses 443

Table 12.5 AT limits derived from medieval

observations of total solar eclipses in Arabic
chronicles.
AT Range (sec)
Year LL UL
+912 880 2600
1061 800 2140
1176 600 1600

The AT ranges derived in this section are summarised in table 12.5.

This table gives the year (+ rather than AD), lower limit (LL) and upper
limit (UL) to AT for each date. Since there were two separate observations
of totality in 1176 (at Antioch - see chapter 11 - and Cizre), the two sets
of AT limits are combined.

12.5 Other records of large solar eclipses

Although the following records of large eclipses are deficient in some way
(e.g. the place of observation is uncertain or the precise magnitude is in
doubt), they seem interesting enough to render them worth quoting in full.
They will not, however, be used to fix AT. For each observation, I have
used equation (8.1) to calculate the apparent magnitude and local time of
greatest phase.

(1) AD 939 Jul 19 [Friday] (total, mag. = 1.07): Olmos/Cueva del la

Mora (Spain)
(327 AH.) The Caliph al-Nasir li Din Allah advanced (northwards from
Cordoba) heading for his holy battle (jihad) until he reached Toledo on
Thursday, when seven nights (sic) remained to the completion of (the
month of) Ramadan. He stayed there for six days and left on Thursday,
when two nights remained to the completion of Ramadan, for Welmish
fortress and on Friday to Khalifah Castle. During the forenoon (Duha) of
that day (Friday) the Sun was eclipsed totally and its disk became dark
except for a slight portion as seen by eye.
[Ibn Hayyan, al-Muqtabis fi Tarikh al-Andalus, vol V, p. 434.]

Ibn Hayyan is here quoting from the contemporary writer al-Razi, who
probably resided at Cordoba. It would appear that the eclipse was wit-
nessed while the Caliph and his army were travelling between Welmish
and Khalifah. I learn from Dr Edward Cooper of London Guildhall
444 12 Solar and lunar eclipses recorded in medieval Arab chronicles

University that the present-day names for Welmish and Khalifah are re-
spectively Olmos and Cueva de la Mora. Olmos is situated approximately
30 km north of Toledo, while Cueva de la Mora is some 20 km further
north than Olmos.
It is difficult to reconcile the two descriptions (i) 'the Sun was eclipsed
totally' and (ii) 'its disk became dark except for a slight portion as seen
by eye'. Possibly the Sun was reduced to a small point of light or a bright
prominence appeared at the solar limb. However, the true phase must
remain in doubt. The value of the observation is further weakened by the
uncertainty in the precise place of observation.
Calculations using equation (8.1) (AT = 1980 sec) indicate that at both
Olmos and Cueva de la Mora the eclipse would be fully total at 7.2 h.
(2) AD 993 Aug 20 [SundayJ (total, mag. = 1.06): Cairo?
(383 AH.) In this year the Sun was eclipsed totally at the end of (the
month of) Jumada al-Ukhra. It was so dark that the stars appeared and
people could not see the palms of their hands. The eclipse cleared at the
end of the day.
[Al-Maqrizi: IttVaz al-Hunafa bi Akhbar al-A'imma
al-Fatimiyin al-Khulafa, vol. I, p. 280.]
Although al-Maqrizi spent most of his life in Cairo, he wrote more than
four centuries after the eclipse. Hence the place of observation must be
regarded as very doubtful; there is nothing in the text itself to indicate a
Cairo source. Although al-Maqrizi appears to quote from an eye-witness
description of totality, careful observations made in Cairo and recorded by
the contemporary astronomer ibn Yunus indicate that the eclipse was only
partial there and ended in mid-morning (see chapter 13). Al-Maqrizi may
have obtained his information from a source much to the east of Cairo.
Calculations using equation (8.1) (AT = 1750 sec) indicate that at Cairo
the magnitude would be 0.96 at 8.9 h.
(3) AD 1176 Apr 11 (total 1.06): River Orontes
(570 AH.) In the last days of (the month of) Shawwal, I remember we
crossed the River al- 'Asi (Orontes) on our return (from Hamah to
Damascus). The Sun was eclipsed and it became dark in the day time.
People were very frightened and stars appeared. Then we arrived at
Hams...
[Al-Katib al-Isfahani, as quoted by Abu Shama:
Kitab al-Rawdataim fi Akhbar al-Dawlatain, vol. I, p. 250.]
Al-Katib al-Isfahani (AD 1125-1201) accompanied Saladin (the Sultan
Yusuf b. Ayyub Salah al-Din) on many campaigns. However, the date
which he gives is incorrect. This corresponds to within a day or two
 12.5 Other records of large solar eclipses 445

of AD 1175 May 21 but there was no solar eclipse visible at the time.
The nearest eclipse which was large in Syria occurred almost a year later,
AD 1176 Apr 11 (28 Ramadan in 571 A.H.). Thus we must assume a
serious dating error here, as was suggested by Lane-Poole (1926, p. 143).
Although the place of observation can be accurately deduced - the point
where the road from Hamah to Hams crosses the River Orontes (Syria:
lat. = 37.35 deg, long. = —42.18 deg) - the text fails to describe the
complete disappearance of the Sun.
According to computation using equation (8.1) (AT = 1050 sec), at the
Orontes crossing the eclipse would be fully total at 7.6 h.

(4) AD 1241 Oct 6 [Sunday] (total, mag. = 1.05): Nile Delta

On the 9th of the month of Babah of the year 958... in the reign of
al-Malik al-Salih Najm al-Din Ayyub (King Ayyub the Good) and under
the presidency of Patriarch Cyril, something strange and wonderful
occurred in the world such that everyone who saw it or heard about it was
astonished. Namely, the Sun was gradually darkened until it became
completely dark. Also the day became dark like night. Some people saw
the stars and people lit lamps and people were very frightened and
prayed... The darkness cleared instantly and then the Sun appeared as
usual and lit up the world, and lamps were extinguished. The duration of
the solar eclipse was about one hour - from the middle of the 8th to the
middle of the 9th hour...
[Synaxarium Alexandrinum]

Observations of this same eclipse were also made in Europe - see chap-
ter 11. The Synaxarium Alexandrinum is a contemporary Coptic religious
yearbook written in Arabic. It is mainly a compilation of the lives of mar-
tyrs, saints, etc. of the Coptic church. The eclipse is the only astronomical
event mentioned in the entire work. The year, 958 relative to the Diocle-
tian era, corresponds to AD 1241/2. Babah, the second month of the year,
is a version of the ancient Egyptian month Phaophi. This month began on
Sep 28 so that Babah 9 would be equivalent to Oct 6. Hence the recorded
date is exactly correct. The stated local time corresponds to between about
13.5 and 14.5 h. According to Ginzel (1918), the place of observation was
either Cairo (the residence of the Coptic patriarchs) or elsewhere in the
Nile Delta - where the other Coptic communities were mainly to be found.
However, this is an expansive area, extending roughly 200 km north-south
and east-west. The Synaxarium itself was compiled under the direction of
Amba Michael, Bishop of Atrib, a town in the Nile Delta. A translation
of this yearbook into Latin has been published by Forget (1963).
Because of the uncertainty in the precise place of observation, firm limits
to the value of AT (necessary at such a late date) cannot be set.
446 12 Solar and lunar eclipses recorded in medieval Arab chronicles

Using equation (8.1) (AT = 850 sec), the eclipse would be total at Cairo
and throughout much of the Nile Delta at 15.1 h.
(5) AD 1431 Feb 12 [Monday] (total mag. = 1.04)
Two separate accounts of this eclipse - which was also observed to be total
in Italy (see chapter 11) - are preserved by contemporary Arab writers:
al-Maqrizi and al-Asqalani, both of whom lived in Cairo.
(i) (834 AH.) In (the month of) Jumada al-Ukhra, the astrologers
warned that the Sun would be eclipsed and in Cairo there were
callings to the people that they should pray and do good deeds.
However, the eclipse did not occur and those who gave the
warnings were denounced. Then news arrived from al-Andalus
(Islamic Spain) of the occurrence of an eclipse there covering all of
the Sun's body except one-eighth of it. That was after midday (nisf
al-nahar) on the 28th of the month.
[Al-Maqrizi: al-Sulukfi Marifat Duwal al-Muluk, vol. IV, p. 855.]

(ii) (834 AH.) In (the month of) Jumada al-Ula it was known that the
calendar experts agreed that the Sun was to be eclipsed on the 28th
of the month after the Zawal (i.e. after the Sun had crossed the
meridian). The Sultan and the people were prepared for it and were
watching the Sun until it set but nothing of it had changed at all.
[Al-Asqalani: Inba al-Ghumr bi 'Bna' al-Vmr, vol. VIII, p. 179.]
The correct lunar month is Jumada al-Ula rather than the following
month Jumada al-Ukhra. The most likely place of observation in Spain -
where the eclipse was said to be fairly large - is the region of Granada,
the capital of the Muslim territories there. Using equation (8.1) (AT =
400 sec), it may deduced that the magnitude at Granada was 0.90 at 13.8
h. Despite the negative observation in Cairo, similar calculations indicate
that as much as 0.43 of the solar diameter would be covered there in
the late afternoon (16.8 h). The failure of so many potential observers
to notice it is indeed surprising, since neither text makes any mention of
unfavourable weather.
(6) AD 1433 Jun 17 [Wednesday] (total mag. = 1.06): Aleppo
There are two records of this eclipse from Aleppo.
(i) (836 AH.) On the 28th of (the month of) Shawwal, the Sun was
eclipsed after the Asr (Afternoon) Prayer and continued until the
time of sunset. It cleared up after the conclusion of the eclipse
prayer, which I led in the Great Mosque. Then the Sun set and we
prayed the Maghrib (Sunset) Prayer in the mosque. When the
eclipse prayer was concluded, I sent a witness to ascend the
 12.5 Other records of large solar eclipses 447

minaret of the mosque to see if the Sun had cleared. He returned,

saying that it had cleared completely.
[Al-'Asqalani: Inba' al-Ghumr hi yBna al-Vmr, vol. VIII, p. 280.]
The editor of the above text refers to a gloss (marginal note) in one of
the manuscripts of the chronicle of al-'Asqalani, presumably written by a
student or companion of his. In this note, we find the following statement:
(ii) The eclipse was dense and it became dark such that we thought
that (the time for) the Maghrib (Sunset) Prayer had arrived. Then
we reckoned that it was still afternoon. I looked at the Sun and
found that it was eclipsed and that the eclipse was great {'azim).
We accompanied the author (i.e. al-'Asqalani) to the Great Mosque
and prayed after him until it cleared.

We know from his chronicle that at the time of the eclipse al-'Asqalani
was in Aleppo, having travelled there from Cairo with the Sultan. The
'Great Mosque' is a well known mosque in Aleppo, which confirms the
location. Although the writer of the gloss (ii) gazed directly at the Sun,
he did not give a clear description of the degree of obscuration. His
remark that the eclipse was 'great' could either imply totality or a very
large partial phase. From equation (8.1) (AT = 380 sec), the magnitude at
Aleppo was 0.998 (i.e. on the verge of totality) at 18.4 h, the solar altitude
then being only 9 deg.

(7) AD 1463 May 18 [Wednesday] (annular, mag. = 0.96): Cairo

(i) (867 AH.) In the month of Sha'ban, the Sun was eclipsed and the
eclipse was excessive (fahish) from late forenoon (after Duha)
almost to the afternoon such that the Earth was darkened in the
eyes of the people.
[Ibn Iyas: BadaV al-Zuhur fi Waqa'V al-Duhur, older edition,
vol II, p. 75.]
The above record is taken from an edition of Ibn Iyas dating from
1894. A more recent edition of the same work, published between 1960
and 1975, gives the following account:
(ii) (867 AH.) In the month of Sha'ban, there was a complete (tamm)
eclipse of the Sun such that the Earth was darkened. The eclipse
continued for about 40 degrees (i.e. 2 hours and 40 minutes).
[Ibn Iyas: BadaT al-Zuhur fi Waqa'V al-Duhur,
new edition, vol. II, p. 447.]
A contemporary writer, Ibn Iyas lived at Cairo. In his chronicle, he
often omitted to state the day of the month (which on this occasion would
be the 28th of Sha'ban).
448 12 Solar and lunar eclipses recorded in medieval Arab chronicles

Why the above accounts differ so much is unknown. In describing the

eclipsed Sun, the term 'excessive' in the older edition could mean either
extremely large or total, but the newer edition implies totality. As it
happens the eclipse was only annular, no more than 0.96 of the solar
diameter being covered where the central phase would be visible. At
Cairo, the computed magnitude based on equation (8.1) (AT = 300 sec)
is only 0.85 at 12.5 h. The calculated duration there is as much as 55 deg
(3h 40 m), considerably more than that reported by Ibn Iyas. To achieve
annularity at Cairo (if this is a plausible interpretation of the record), a
range of AT between 2360 and 3030 sec would be required. This is an
order of magnitude too great at such a recent epoch. Both this and the
following observation recorded by Ibn Iyas (which also alleges totality)
would appear to be unreliable.

(8) AD 1491 May 8 [Sunday] (annular, mag. = 0.94): Cairo

(896 AH.) In the month Jumada al-Ukhra, the Sun was eclipsed totally.
The Sun stayed eclipsed for about 30 degrees (i.e. two hours).
[Ibn Iyas: Bada'V al-Zuhur fi WaqaV al-Duhur, vol. Ill, p. 282.]
The calculated date of this event is equivalent to the 28th of the month
Jumada al-Ukhra. As in the previous entry, although this eclipse was
described by Ibn Iyas (a contemporary) as total, it was in fact generally
annular. The record is lacking both in descriptive detail and any reference
to the place of observation.
The configuration of this eclipse track is such that only impossibly
high values of AT (between about 28000 and 30000 sec) could lead to
annularity at Cairo. Using equation (8.1) (AT = 280 sec), only 0.40
of the solar diameter would be covered at maximum phase around 17.1
h. Possibly Ibn Iyas is quoting the results of prediction rather than
observation. As it happens, the computed duration of the eclipse at Cairo
is 2h 0m (30 deg), identical to the reported figure.
(9) AD 1513 Mar 7 [Monday] (annular, mag. = 0.998): Damascus
(918 AH.) In the afternoon of Monday the 29th of (the month of) Dhu
al-Hijjah, the Sun was eclipsed for 13 deg (i.e. 52 minutes). There was
darkness and some shopkeepers lit lamps in their shops. In the mean time
there were clouds. After the Jumu'ah (Friday) Prayer in the Umayyad
Mosque, a khutbah (sermon) was delivered about the eclipse and the
eclipse prayer was performed.
[Ibn Tulun: Mufakahat al-Khullanfi Hawadith al-Zaman, part I, p. 375.]
The contemporary writer Ibn Tulun lived in Damascus, the site of
the Umayyad Mosque. In the narrow central zone the loss of daylight
would be considerable. However, the degree of obscuration of the Sun
 12.6 Sun eclipsed near sunrise or sunset 449

at Damascus is not directly specified. As noted elsewhere in this chapter,

it was not unusual for Muslims to assemble at the mosque during a
significant eclipse in order to recite a special prayer - a practice which
still continues. At the eclipse of AD 1513, the computed magnitude would
be as much as 0.98 at 15.6 h (assuming a value for AT of 240 sec). The
computed duration of the eclipse is 2h 35 m (38 deg). This is much greater
than the recorded figure of 13 deg, but cloud may have considerably
limited the period of visibility.

12.6 Sun eclipsed near sunrise or sunset

There are a number of examples in Arab chronicles where the Sun is
said to have risen or set whilst eclipsed. Whether such observations are
as reliable as those made by astronomers is doubtful. In particular, the
precise place of observation is seldom stated directly. Additionally, if
a chronicler makes a simple statement such as 'the Sun was eclipsed at
sunrise' - as was the case in AD 1174 - it is conceivable that he may
simply mean that the event occurred in the early morning. Until AD
1398 there are only two records which assert that the Sun rose or set
whilst eclipsed (AD 882 and 1178). By AD 1398, AT would be so small
(probably less than about 300 sec) that it is debatable whether the rather
crude observations found in chronicles are of much value in this context.
In view of these cautionary remarks, I have only analysed the observa-
tions in AD 882 and 1178. In each case the author was a contemporary,
living in Baghdad. The following translations are by SSR. I have made
no allowance for horizon profile in computing local times of sunrise or
sunset: Baghdad is situated in a very flat plain near sea-level. Although
a further observation in which the Sun set after the end of totality was
reported from Cordoba in AD 912 (see section 12.4), this will not be
utilised to determine limits to AT. The horizon towards the north-west of
Cordoba is quite hilly, and thus materially affects the local time of sunset
in summer.

(1) AD 882 Aug 17 [FridayJ: Baghdad

(269 AH.) In (the month of) Muharram in this year... the Sun was
eclipsed at the time of sunset on Friday, when two nights remained to the
completion of Muharram, and set eclipsed
[Al-Tabari: Tarikh al-Tabari: Tarikh al-Rusul wa'l Muluk, vol. IX, p. 613.]
The complete record, which notes eclipses of both Sun and Moon, is
cited in section 12.3.
Al-Tabari settled in Baghdad around AD 875 and spent the rest of
his life there. Hence Baghdad is the most likely place of observation.
450 12 Solar and lunar eclipses recorded in medieval Arab chronicles

Unfortunately, how far the eclipse was advanced by sunset is not clear
from the record. Hence it will only be assumed that the Sun reached the
horizon between first and last contact.

RESULTS
(1) First contact before sunset at Baghdad on Aug 17. LT of sunset =
18.58 h. Hence LT of contact < 18.58 h, U T < 15.65 h. Computed TT =
15.99 h, thus A T > 1250 sec.
(ii) Last contact after sunset. LT of contact > 18.58 h, U T > 15.65 h.
Computed T T = 17.60 h, thus A T < 7000 sec.
Combining these limits yields 1250 < A T < 7000 sec.

(2) AD 1178 Sep 13 [Wednesday]: Baghdad

(574 AH.) The Moon was eclipsed after the last third of the night in the
middle of (the month of) Rabi' al-Awwal, and stayed in that state until it
set after sunrise. Also the Sun was eclipsed on Wednesday the 29th of
Rabi' al-Awwal in the afternoon and stayed like that until near sunset.
[Ibn al-Jawzi: al-Muntazam fi Tarikh al-Muluk wa'l-Umam, vol. X, p. 283.]
The lunar observation will be considered in section 12.7. In the case of
the solar eclipse, the text seems to imply that the eclipse ended before the
Sun had set.

RESULTS
Last contact before sunset at Baghdad on Sep 13. LT of sunset = 18.10
h. Hence LT of contact < 18.10 h, UT < 15.02 h. Computed TT = 14.12
h, thus AT > -3250 sec.
This single limit is in keeping with all realistic values of AT and will
not be considered further. However, the inference that the eclipse ended
before sunset is confirmed.

The results deduced in this section are listed in table 12.6 along with
the AT limits deduced from lunar observations in section 12.7. It should
be noted that the AT range derived from the solar eclipse of AD 882 is
much narrower than for the lunar eclipse in the same year.

12.7 Moon eclipsed near moonrise or moonset

Records of lunar eclipses occurring near moonrise or moonset present
similar difficulties to the solar observations discussed in the last section.
In particular, Ibn al-Athir (AD 1160-1233) notes the occurrence of three
moonrise or moonset eclipses which took place more than two centuries
before his own time. The dates of these events are AD 951, 966 and
969. Since Ibn al-Athir wrote so long afterwards, it is no more than a
 12.7 Moon eclipsed near moonrise or moonset 451

presumption that Mosul, where he spent most of his life, was the place
of observation. In the following century, the Nestorian Bishop Elias of
Nisibis, who frequently abstracted from Arab chronicles, made a list of as
many as 17 observations of lunar eclipses which were 'made in our own
time'. The chronicle of Elias, which is in Syriac, has been translated into
French by Delaporte (1910). The eclipses which Elias cites range in date
from AD 1005 to 1030; he died in 1049. In several instances the Moon
was said to rise or set eclipsed. Elias may well have witnessed some of
these events at Nisibis (now known as Nusaybin: lat. = +37.08 deg, long.
= —41.18 deg) but in no case is the source of the original record preserved
by him. Even if we could be sure that Nisibis was the place of observation,
the fact that this site is in a mountainous area would necessitate careful
allowance for horizon profile.
If the above observations are rejected, there are only two seemingly
useful records before AD 1400 - the rather arbitary cutoff date adopted
in section 12.6. In both cases (AD 882 and 1178) the author was a
contemporary. As in the previous section, no allowance for horizon
profile is necessary since the two events were reported from Baghdad.
Following the usual policy, a double date (e.g. Aug 2/3) is given for
lunar eclipses. The calculated weekday refers to the day beginning at
sunset - i.e. the second of the pair of dates. In each case the computed
magnitude of the eclipse is also given for reference.

(1) AD 882 Aug 2/3 [Friday] (mag. = 1.43): Baghdad

(269 AH.) In (the month of) Muharram this year... the Moon was
eclipsed on the night of the 14th (day) and set eclipsed.
[Al-Tabari: Tarikh al-Tabari: Tarikh al-Rusul wa'l Muluk,
vol. IX, p. 613.]
An observation of a solar eclipse, recorded two weeks later by al-Tabari,
has already been discussed in section 12.6. There is no mention of totality,
but since the record is so brief, it cannot be assumed that the Moon set
before the onset of the total phase. Hence it will merely be supposed that
the Moon reached the western horizon between first and last contact.

RESULTS

(i) First contact before moonset at Baghdad on Aug 3. LT of moonset =

5.13 h. Hence LT of contact < 5.13 h, UT < 2.23 h. Computed TT = 1.07
h, thus AT > -4200 sec.
(ii) Last contact after moonset. LT of moonset = 5.30 h. Hence LT of
contact > 5.30 h, UT > 2.41 Computed TT = 1.07 h, thus AT < 7550 sec.
Combining these limits yields -4200 < AT < 7550 sec.
452 12 Solar and lunar eclipses recorded in medieval Arab chronicles

Table 12.6 AT limits from solar and lunar eclipses

occurring near rising or setting.

AT Range (sec)
Year Type LL UL
+882 Moon -4200 7550
+882 Sun 1250 7000
1178 Moon -5650 3450

(2) AD 1178 Aug 29/30 [Wednesday] (mag. = 0.46): Baghdad

(574 AH.) The Moon was eclipsed after the last third of the night in the
middle of (the month of) Rabi' al-Awwal, and stayed in that state until it
set after sunrise...
[Ibn al-Jawzi: al-Muntazam fi Tarikh al-Muluk wa'l-Umam, vol. X, p. 283.]
An observation of a solar eclipse recorded in this same passage has
already been discussed in the previous section. Ibn al-Jawzi was living in
Baghdad at the time. His remark that the Moon set after sunrise suggests
unusually careful observation.

RESULTS
(i) First contact before moonset at Baghdad on Aug 3. LT of moonset =
5.65 h. Hence LT of contact < 5.65 h, UT < 2.65 h. Computed TT = 1.08
h, thus AT > -5650 sec.
(ii) Last contact after moonset. LT of moonset = 5.74 h. Hence LT of
contact > 5.74 h, UT > 2.74 h. Computed TT = 3.70 h, thus AT < 3450
sec.
Combining these limits yields —5650 < AT < 3450 sec.
NB the computed LT of sunrise is 5.63 h, so that on any reasonable
value for AT the Moon would set a little after sunrise, as the text alleges.

The few AT results derived in this and the previous section are assem-
bled in table 12.6.

12.8 Selected records of total lunar eclipses

In this section, a few examples of records of total lunar eclipses are given,
as an illustration of the kind of material available in Arabic chronicles.
Translations are by Stephenson and Said (1996). It should be emphasised
that the observations are of no value for determining AT.
Most Arab records of lunar eclipses, although often distinguishing be-
tween total and partial obscurations, contain little descriptive information.
 12.8 Selected records of total lunar eclipses 453

One of the most detailed accounts of an eclipse in which the Moon com-
pletely disappeared is given by the fifteenth century Cairo annalist Ibn
Taghri Birdi. This relates to the event of AD 1461 Jun 22/23:
(865 AH.) On the night of Tuesday, the 14th of (the month of)
Ramadan, the whole of the Moon's body was eclipsed and disappeared in
the eclipse for about seventy degrees. The stars in the sky became as
though it was the 29th of the month (i.e. on a moonless night). Perhaps the
occurrence of such an eclipse is very rare.
[Ibn Taghri Birdi: Hawadith al-Duhur fi Mada 7 -Ayyam wa'l Shuhur,
extract from vol. VII, part 4, pp. 741-2.]
Ibn Taghri Birdi gives the correct weekday and day of the lunar month.
Computation indicates that the true duration was only about 3.9 hours
(58 degrees), so that the reported duration was too long.

In medieval Islamic history, there seems to be no definite instance of

the Moon described as turning red during totality. The earliest reliable
observation of this kind is relatively recent. It is reported by the eighteenth
century historian al-'Umari of Mosul, who describes an event occurring
during his own time:
(1186 AH.) In this year, the Moon was totally eclipsed in the first half
of the night. All of its light disappeared and it became like a copper disc.
[Al-'Umari: al-Athar al-Jaliyyah fi 7 -Hawadith al-Ardiyyah, pp. 136-7.]
Only the year is given, corresponding to AD 1772-3. Either the total
lunar eclipse of 17/18 April or that of 11/12 October (both in AD 1772)
could be intended.

Although several medieval annals describe alleged eclipses in which

the Moon appeared red, all events of this kind can be explained by
atmospheric phenomena such as dust storms. For example, the Damascus
chronicler Abu Shama records the following observations, made in his
own time. These began on a date corresponding to AD 1256 Jul 11:
(654 AH.) On the night of the 16th of the month Jumada al-Ukhra, the
Moon was eclipsed at the beginning of the night and it was extremely red.
It then reappeared. The Sun was eclipsed the following day and it became
reddish at both the time of its rising and setting. It remained like this for
(several) days...
[Abu Shama: Tarajum Rijal al-Qarnain al-Sadis wa'l Sabi\
also known as al-Dhail (ala al-Rawdatain, pp. 189-191.]
No solar eclipse would be visible near Damascus around this time.
Although there was a lunar obscuration on the night of 8/9 July - i.e. two
days earlier than the date reported by Abu Shama - this was only partial
(magnitude 0.19).
454 12 Solar and lunar eclipses recorded in medieval Arab chronicles

2800 - Key
Total solar eclipse

2400 - /\ Lower limit from

horizon eclipse

2000 -

1600 -

1200

800 -

400 -

0 i i i i i i i i

+900 +1000 +1100 +1200 +1300

Year
Fig. 12.4 AT limits derived from eclipse observations recorded in Arab chroni-
cles.

12.9 Conclusion
The few AT limits obtained in this chapter are shown diagrammatically in
figure 12.4. Although the observations of total solar eclipses yield valuable
limits for AT, the number of results is disappointing.

12.10 Appendix: Arab chronicles consulted

Abu Shama: Kitab al-Rawdataimfi Akhbar al-Dawlatain (2 vols.). Cairo,
1870-71.
Abu Shama: Tarajum Rijal al-Qarnain al-Sadis wal Sabi\ also known
as al-Dhail 'ala al-Rawdatain. Cairo, 1947.
al-'Asqalani: Inba al-Ghumr bi 'Bna al-'Umr (8 vols.). Hyderabad-
Deccan, 1967-75.
al-Maqrizi: al-Sulukfi Ma'rifat Duwal al-Muluk (4 vols.). Cairo, 1934-
71.
al-Maqrizi: IttVaz al-Hunafa bi Akhbar al-A'imma al-Fatimiyin al-
Khulafa (3 vols.). Cairo, 1967-73.
al-Tabari: Tarikh al-Tabari: Tarikh al-Rusul wa'l Muluk (10 vols.).
Cairo, 1960-69.
al-'Umari: al-'Athar al-Jaliyyah fi 7 -Hawadith al-Ardiyyah. Al-Najaf,
1973.
 12.10 Appendix: Arab chronicles consulted 455

Ibn al-Athir: al-Kamilfi al-Tarikh (9 vols.). Cairo, 1929-38.

Ibn al-Jawzi: al-Muntazamfi Tarikh al-Muluk wa'l-Umam (10 vols.: only
vol. V, part 2 to vol. X accessible). Hyderabad-Deccan, 1938-42.
Ibn Hayyan: al-Muqtabis fi Tarikh al-Andalus (10 vols.: only vols. Ill
and V accessible). Vol. Ill: Paris, 1937; vol. V: Madrid, 1979.
Ibn Iyas: [BadaT al-Zuhur fi Waqa'V al-Duhur. New edition (5 vols.),
Wiesbaden, 1960-75. Older edition (3 vols.), Cairo, 1894-5.
Ibn Taghri Birdi: Hawadith al-Duhur fi Mada 7 -Ayyam wa'l Shuhur.
Only extracts from vol. 7, parts 2 and 4 have been published, by the
University of California, 1931 and 1942.
Ibn Tulun: Mufakahat al-Khullanfi Hawadith al-Zaman (2 parts). Cairo,
1962^.
 13
Observations of eclipses by
medieval Arab astronomers

13.1 Introduction
The eclipse observations made by medieval Arab astronomers are among
the most accurate and reliable data from the whole of the pre-telescopic
period. Careful records of both solar and lunar eclipses are contained in
a number of compendia - some known as zijes (astronomical handbooks
containing various tables along with explanatory text). These include mea-
surements of the times of occurrence and other details such as magnitude
estimates. Although the main emphasis in this chapter will be on timed
data, solar magnitude estimates and horizon observations of eclipses will
also be considered.
Many of the observations discussed below were investigated by New-
comb (1878) and Newton (1970). However, these authors relied on pub-
lished translations which sometimes contained significant errors, while
their own interpretations are occasionally suspect. Furthermore, in nei-
ther case was a direct solution made for AT.

13.2 Sources of data

Most of the accessible eclipse observations by medieval Arab astronomers
are contained in a single treatise - the zij compiled by the great Cairo
astronomer Ibn Yunus, who died in AD 1009 (his date of birth is un-
known). A few eclipses are also recorded in works by al-Battani (who
lived between AD 850 and 929) and al-Biruni (AD 973-1048).
Ibn Yunus cites reports of some thirty solar and lunar eclipses from
between AD 829 and 1004. His treatise, dedicated to Caliph al-Hakim,
is entitled al-Zij al-Kabir al-Hakimi. Not all of Ibn Yunus' text survives
today, but portions of it are extant in manuscripts at Leiden and Oxford; a
further manuscript at Paris contains abridged versions of certain sections
of the text (King, 1976). Only the manuscript which is preserved in
the library at Leiden University (cat. no. Or 143) contains the eclipse

456
 13.2 Sources of data 457

observations; this also notes some planetary conjunctions. This text was
published in its original Arabic, along with a translation into French, by
Caussin (1804). Although Caussin's translation is generally very sound,
it contains several errors. Recently Dr Said S. Said, in conjunction with
the author, has made a detailed study of a microfilm of the manuscript
supplied by Leiden University, and has re-translated all of the eclipse
records (Said and Stephenson, 1997). These translations, which preserve
the original astronomical terminology of the record as closely as possible,
form the basis of much of the present chapter.
The observations compiled by Ibn Yunus fall into three distinct groups:
AD 829-866; 923-933; and finally 977-1004. Most of the eclipses in the
first set were witnessed by al-Mahani of Baghdad, while the principal
observer in the second group was Ali ibn Amajur, also of Baghdad. Most
of the eclipses in the third set were observed by the Cairo astronomer
Ibn Yunus himself, as is clear from his use of the first person in the text.
Sometimes, the place where the eclipse was seen is directly specified in the
record; this is particularly true of the Cairo observations. However, more
often it may be inferred from the reference to the astronomer's name.
There can be little doubt that all of the observations from between AD
829 and 933 were made at Baghdad, while the remaining observations
were made in Cairo (al-Qahirah), which was founded in AD 969.
In assembling these various accounts, Ibn Yunus had two main motives.
Firstly, he wished to emphasise the need for improved planetary tables
by illustrating how poorly computations made with existing tables were
supported by observation and secondly he simply wanted to list these
observations for the benefit of future astronomers (see King, 1976).
Four eclipses witnessed by al-Battani and his colleagues are reported in
his treatise entitled al-Zij al-Sabi\ written c. AD 910. This work has been
published along with a translation into Latin by Nallino (1899). The places
of observation were al-Raqqah (in present-day Syria) and Antakyah (in
Turkey). About a century later, al-Biruni also recorded four eclipses, as
seen from Jurjan (in present-day Iran), Jurjaniyyah (in Turkmenistan) and
Ghaznah (Afghanistan). These observations are contained in al-Biruni's
Kitab Tahdid al-Amakin H-Tashih Masafat al-Masakin (dating from AD
1025) and his al-Qanun al-Mas'udi (composed in AD 1030). The former
work has been translated into English by Ali (1967). The main aim
of both al-Battani and al-Biruni in observing eclipses appears to have
been the determination of longitude difference between selected sites (see
section 13.3).
The eclipse reports cited by al-Battani and al-Biruni probably represent
only a small proportion of the number actually available to them. Sum-
maries of a few additional observations are in fact recorded by al-Biruni
in his Kitab Tahdid but unfortunately he does not give any useful timed
458 13 Observations of eclipses by medieval Arab astronomers

Table 13.1 Locations from which eclipses were reported by medieval

Muslim astronomers.
Original name Modern name N lat. (deg) E long, (deg)
al-Qahirah Cairo 30.05 31.25
al-Raqqah Raqqa 35.94 39.02
Antakyah Antakya 36.20 36.17
Baghdad Baghdad 33.34 44.40

Ghaznah Ghazni 33.55 68.43

Jurjan Gorgan 36.83 54.48
Jurjaniyyah Kunya-Urgench 42.32 59.16
Nishapur Neyshabur 36.21 58.83

information. Further observations by other medieval Arab astronomers

may well survive in unpublished texts. However, at present the range of
source material is rather restricted.
A careful account of a central annular eclipse seen in AD 873 at
Nishapur (in present-day Iran) is given by al-Biruni in his al-Qanun al-
Mas'udi, although no times are reported. This observation will also be
considered below.
Geographical co-ordinates (degrees and decimals) of the places of ob-
servation discussed in this chapter are listed in table 13.1. This table gives
both original names and their present-day equivalents. The locations of
the various sites are also shown in figure 13.1. Throughout this chapter
I have normally used the original names except for Cairo, for which it
would seem pedantic to use al-Qahirah.

13.3 Motives for observation

Some of the main reasons why medieval Arab astronomers made careful
observations of eclipses have already been briefly mentioned in the previ-
ous section. Further comments now seem desirable. Ibn Yunus compared
some of the observations made in Baghdad with tables produced around
AD 810 by Yahya ibn Abi Mansur. These were contained in the latter
writer's al-Zij al-Mumtahan (Tables verified by observation') and were
developed from Ptolemy's methods as laid down in his Almagest. Ibn
Yunus was able to demonstrate that the times of eclipses calculated from
the tables of Yahya frequently deviated from the observed times by half
an hour or more. As is apparent from the records of the eclipses of AD
923 to 933 cited by Ibn Yunus, the contemporary astronomer Ibn Amajur
had already shown that the tables of Habash al-Hasib, produced around
 13.3 Motives for observation 459

30° E 35° E 40° E 45° E 50° E 55° E 60° E 65° E 70° E

40° N

J / |
p <3
35° N

i M w -jurjan
r .Nishapur^3^7\
^ ^ x ^ ^ -
\ Ghaznah
35° N

h
_4-—-t^\ \ \ 30° N
1 \ \ \ ^r
Baghdad \ \ \ r ^ ^ \
30° N
25° N

25° N

35° N 40° E 45° E 50° E 55° E 60° E 65° E

Fig. 13.1 Map showing the locations of the various sites where eclipses were
observed by medieval Muslim astronomers.

AD 840, were in error by similar amounts. These tables, also entitled

al-Zij al-Mumtahan, were based on observations made by Habash. For
instance, the account of the solar eclipse of AD 923 Nov 11 - as quoted
by Ibn Yunus - contains the following details:

(This) solar eclipse was calculated and observed by Abu al-Hasan Ali
ibn Amajur, who used the al-Zij al-Arabi of Habash... We as a group
observed and clearly distinguished it... We observed this eclipse at several
sites on the Tarmah (an elevated platform on the outside of the building)...
According to calculation from the conjunction tables in the Habash Zij the
middle was at 0;31 h (i.e. 31 min) and its clearance at 0;44 hours (i.e. 44
min), calculation being in advance of observation.
[Trans. Said and Stephenson (1997).]
The Arabic text of the above record (from the Leiden manuscript of
Ibn Yunus) is shown in figure 13.2. This figure includes the report of the
lunar eclipse of AD 923 Jun 1/2.
A more extensive translation of the above record is given in section
13.7 below. It is apparent that eclipse predictions - although of limited
accuracy - were evidently sufficiently reliable to enable careful advance
planning of observations. This was particularly important when eclipses
were used to determine the difference in longitude between two separate
sites.
The technique of determination of the longitude difference between two
cities by using simultaneous timings of lunar eclipse - widely practised
460 13 Observations of eclipses by medieval Arab astronomers

Fig. 13.2 Arabic text of eclipse records from AD 923 Jun 1/2 (lunar) and 923
Nov 11 (solar) in the Ibn Yunus manuscript (Leiden University library: Or. 143).
 13.4 Calendrical remarks 461

in Europe during the eighteenth and early nineteenth centuries - was

pioneered by medieval Arab astronomers. Hipparchus in the second
century BC was already well aware of the value of lunar eclipses for
measuring longitude, as noted by Strabo:
In like manner, we cannot accurately fix points that lie at varying
distances from us, whether to the east or to the west, except by a
comparison of the eclipses of the Sun and Moon. That, then, is what
Hipparchus says on the subject.
[Strabo: Geography, I, 1, 13; trans. Jones (1917, pp. 23-25).]
However, 'the total absence in antiquity of any scientific organisation
deprived the whole method of its practical importance' (Neugebauer, 1975,
p. 667). Hipparchus would have been unable to obtain the observations
which he needed for such a goal. Similar remarks apply to Ptolemy in
the second century AD. Hence the technique remained no more than an
idealistic concept until it was put to use by the Arabs.
Al-Biruni, in his Kitab Tahdid, gives the following outline of the method
as applied in his time:
If we know beforehand of the formation of a lunar eclipse and we wish
to determine the longitudinal difference between two towns, we make
arrangements beforehand for someone in each town who can measure the
times accurately by instruments, to obtain as accurately as possible the
times of the beginning of an eclipse and its end and those of the beginning
of clearance and its end.
[Trans. Ali (1967, p. 130).]
Using lunar eclipse observations, al-Biruni determined the longitude
difference between Jurjan and Ghaznah as (2;21) minutes of day and
between Jurjaniyyah and Ghaznah as (1;42,12) minutes of day. Since one
minute of day corresponded to 1/60 day or 6 deg, and al-Biruni followed
the standard practice of using sexagesimal arithmetic, these results were
equivalent to 14.1 deg and 10.22 deg respectively. The true values are
respectively 14.0 and 9.4 deg, representing tolerable precision at this
epoch.
A diagram by al-Biruni explaining the formation of lunar eclipses is
illustrated in figure 13.3.

13.4 Calendrical remarks

Dates of the various observations cited by Ibn Yunus, al-Battani and al-
Biruni are usually expressed in terms of the Islamic lunar calendar. Years
are numbered from al-Hijrah, the migration of the Prophet Muhammad
to Medina in AD 622. The operation of this calendar and conversion of
462 13 Observations of eclipses by medieval Arab astronomers

Fig. 13.3 Diagram by al-Biruni explaining the formation of lunar eclipses (Nasr,
1976, Fig. 40).

dates to the Julian or Gregorian calendars have already been discussed

in chapter 12. However, some of the dates cited by Ibn Yunus are also
expressed in terms of other systems: these are the Persian, Syrian and
Coptic calendars. In giving translations of the various eclipse records
below, I have - for the sake of brevity - normally given only the date
according to the Muslim calendar in order to avoid cumbersome entries;
certain texts cite all four dating schemes mentioned above! Invariably,
 13.5 Observational techniques 463

when the Muslim date is converted to the Julian calendar there is either
exact agreement with the calculated date of an eclipse or the discrepancy
between them is unimportant. This is also true of dates given on the other
three calendars.
The Syrian and Coptic calendars have already been briefly discussed
(chapters 11 and 12). Some remarks on the Persian calendar are as follows.
The Persian calendar resembles to some extent the ancient Egyptian
calendar. Years consist of 12 months, each of 30 days, with five epagomenal
(additional) days placed after the end of the eighth month. Years are
counted from the era of Yazdijerd, the accession of King Yazdijerd III:
AD 632 Jun 16. Because of its simple rules, the Persian calendar - like
the ancient Egyptian scheme - is convenient for counting days between
two epochs.

13.5 Observational techniques

Medieval Arab astronomers were well aware of the hazards of observing
solar eclipses with the naked eye. As a result, they were in the habit of
viewing the Sun by reflection in water in order to reduce its glare. Thus
al-Biruni in his Kitab Tahdid makes the following remarks:
The faculty of sight cannot resist it (the Sun's rays), which can inflict a
painful injury. If one continues to look at it, one's sight becomes dazzled
and dimmed, so it is preferable to look at its image in water and avoid a
direct look at it, because the intensity of its rays is thereby reduced...
Indeed such observations of solar eclipses in my youth have weakened my
eyesight.
[Trans. Ali (1967, p. 131).]
Eclipse magnitudes were frequently estimated by the Muslim
astronomers. These were usually expressed as a fraction of the diam-
eter of the luminary obscured, but occasionally were quoted in terms of
surface area. Often these estimates were expressed to the nearest digit or
twelfth of the solar or lunar diameter, following ancient Babylonian and
Greek practice (see chapter 3). On a few occasions it is reported that the
Moon or Sun was eclipsed when it rose or set, sometimes the proportion
of the disk obscured at the time being used. However, the numerous timed
data are the most valuable for determining AT.
For partial solar and lunar eclipses, medieval Arab astronomers timed
three separate phases: 'beginning' (i.e. first contact), 'middle' and 'clear-
ance' (last contact). No total solar eclipse is known to have been recorded
by them (unlike in the case of chroniclers), but total lunar obscurations
were frequently reported. Here two extra phases were timed: 'beginning of
staying' (second contact) and 'end of staying' or 'beginning of clearance'
464 13 Observations of eclipses by medieval Arab astronomers

(third contact). In such cases, the time of mid-eclipse was merely the
average of the measurements for second and third contact.
Owing to the mediocre quality of clocks at this early period, eclipse
times were usually determined indirectly by the Arab astronomers - by
measuring the altitude of the Sun in the case of a solar eclipse and of
either the Moon or a selected bright star for a lunar obscuration. As noted
above (section 13.3), often several astronomers would make independent
observations at the same site.
The type of instrument used for determining altitudes is hardly ever
mentioned in the text, although it is most likely to have been a hand-held
astrolabe. Among astronomical instruments, the astrolabe was the most
widely used in both the Arab dominions and Europe during the Middle
Ages. (For a discussion of its construction and use, see North (1974);
a photograph of a tenth century example is shown in figure 13.4.) Of
ancient Greek origin, the astrolabe served two main purposes: (i) as an
observational device for the determination of the altitudes of celestial
bodies; and (ii) as an analogue computing device, particularly for the
determination of local time.
Most altitude measurements relating to lunar and solar eclipses were
made to the nearest degree or half degree, probably the best that could be
achieved with an astrolabe. On one occasion (AD 928) the altitude of the
Sun at an eclipse contact was recorded with significantly higher precision,
but this seems to be unique:
We (found) that it cleared and nothing of the eclipse remained and we
distinguished the (full) circle of the Sun's body in water; (that was) when
the altitude was 12 deg in the east, less | of a division of the al-halaqa (i.e.
the ring), which is graduated in thirds (of a degree), that is (less by) |
degree
[Trans. Said and Stephenson (1997).]
By comparison, certain altitude determinations, although quoted to the
nearest degree or even half degree, are said to be only approximate. Just
how inferior these are to normal measurements is not clear. Evidently, me-
dieval Arab astronomers were capable of making altitude measurements
of very high accuracy. A recent analysis of many determinations of the
meridian altitude of the Sun made by Muslim astronomers between AD
832 and 1018 revealed that several observers achieved consistent precision
using a fixed instrument of around 1 arcmin (Stephenson and Said, 1991).
Presumably owing to the large changes in azimuth of the Sun or Moon
(or clock star) during the course of an eclipse - and the mediocre defini-
tion of the contacts using the unaided eye - it was normally considered
adequate to use a hand-held instrument for determining altitudes under
these circumstances.
 13.5 Observational techniques 465

Fig. 13.4 Tenth century Arab astrolabe (Courtesy: Dar al-Athar al-Islamiyyah,
Ministry of Information, Kuwait).

Altitudes taken during an eclipse were afterwards reduced to local time

(LT) by the observers either with the aid of tables or an astrolabe. Detailed
examples of the use of tables by medieval astronomers to convert from
solar altitudes to LT are given by King (1973). The text of Ibn Yunus
notes application of an astrolabe for conversion of star altitudes on two
separate occasions, as in the following quotation from al-Mahani on a
date corresponding to AD 854 Aug 11/12:

... It was found by observation that the time of beginning of the eclipse
was when the altitude of (the star) al-dabaran (Aldebaran: a Tau) was
45;30 deg in the east... We determined the time of the beginning from the
466 13 Observations of eclipses by medieval Arab astronomers

altitude of al-dabaran by the astrolabe and found it to be 44 deg (of the

celestial sphere) after midnight
[Trans. Said and Stephenson (1997).]
In common with other altitude determinations, the above measurement
is effectively expressed in degrees and arcmin. Following the customary
convention, I have rendered this as 45 ;30 deg.
By computing the LT direct from the altitude determination, it is
possible to assess the accuracy with which the above reduction was made.
The estimate of 44 degrees (i.e. 2h 56m) after midnight is virtually identical
to the LT which may be calculated directly from the altitude measurement
(2h 55.6m). By chance this result is unusually precise. Among the
reductions reported by Ibn Yunus, the average error - whether using the
astrolabe or tables - can be shown to be about 4 minutes (Stephenson
and Said, 1991). This still represents tolerable precision.
On some dates, only contact timings are available, the original altitude
measurements being no longer preserved. Eclipse times were normally
expressed in terms of hours and minutes, both equal hours and seasonal
hours being used with comparable frequency. Usually it is clear from the
text which system was adopted. For instance, if a text quotes times in
'hours of daytime' or 'hours of night', these are evidently seasonal hours.
At the relatively low latitudes where ipost observations were made
(c. 30 to 35 deg) the altitude of a celestial body changes by one degree
in about 5 minutes (except when near the meridian). In principle, this is
the typical accuracy with which it is possible for us to fix the LT of an
eclipse contact when an altitude measurement is reported by a medieval
Arab astronomer. Such precision is significantly greater than that attained
by Chinese sky watchers using a clepsydra (no better than the nearest 15
minutes or so - see chapter 9) and was not significantly improved upon
in Europe until the sixteenth century.
When recording first contact of an eclipse, the Baghdad astronomers
simply noted the time of the 'beginning' of the event. However, at Cairo
Ibn Yunus on two occasions recorded both this moment and the time
when an eclipse was first 'perceived'. Evidently he was aware of the delay
in detecting the start of an eclipse owing to the limited resolution of the
unaided eye. His contemporary al-Biruni offered the following specific
remarks on this problem for the case of lunar eclipses:

A (lunar) eclipse does not become noticeable to the observer until the
segment removed from it (i.e. from the Moon) according to some authors
of zijes reaches a limit of one digit; I mean one part in twelve of its body
(i.e. disk). A limit is also set to its time, which is 1 ;49 azman (time degrees)
or 0;7,16 (equal) hours. By this amount of time the true beginning of the
eclipse is in advance of the apparent beginning, and the true completion of
 13.6 Large solar eclipses in astronomical treatises 467

clearance delays behind the apparent completion of clearance. I would

think that the amount of a digit in this respect is (too) large because even
though the first contact between the shadow (of the Earth) and the Moon
is not perceptible, a small indentation (on the Moon's edge) could be
seen
[Kitab Tahdid; trans. Said and Stephenson (1997).]
Obviously, the effect of visual acuity on the detection of eclipse contacts
is a very subjective issue. In practice, recorded observations only allude
to the delay in detection of first contact, never making reference to an
advance of last contact. When an observation contains estimates of both
the time when an eclipse was first 'perceived' and the time of 'beginning'
I have given it special consideration. However, if only one measurement
is reported for first contact, I have accepted this without comment.

13.6 References to large solar eclipses in astronomical treatises

Only two observations of this kind are known. A description of the
ring phase in AD 873 was quoted by al-Biruni, while in 1004 Ibn Yunus
witnessed a very large partial eclipse which was not quite annular. These
two observations will be considered before the other types of data.

(1) AD 873 Jul 28 [Tuesday] (annular: mag. = 0.94), Nishapur

This solar eclipse was observed by Abu al-Abbas-al-Iranshahri at
Nishapur early in the morning on Tuesday the 29th of the month of
Ramadan in the year 259 of al-Hijrah... (date on Persian calendar)... He
mentioned that the Moon's body (i.e. disk) was in the middle of the Sun's
body. The light from the remaining uneclipsed portion of the Sun
surrounded it (i.e. the Moon). It was clear from this that the Sun's
diameter exceeded in view that of the Moon.
[al-Qanun al-Mas'udi; trans Said and Stephenson (1997).]
Attention was drawn to this observation by Goldstein (1979). Although
the eclipse occurred long before al-Biruni's own time, he cited it as
evidence that the angular diameter of the Sun could exceed that of the
Moon. Ptolemy (Almagest, V, 14) had implied that annular eclipses could
not take place because he believed that the apparent lunar diameter at
apogee was equal to the apparent solar diameter (assumed to be fixed).
This issue was still being debated in seventeenth century Europe - see
chapter 11.
Goldstein notes that none of the works of al-Iranshahri survive, but
this author is mentioned by al-Biruni in several other places.

RESULTS
For annularity at Nishapur, 1820 < AT < 3750 sec.
(NB computed LT of maximum phase = 5.6 h; solar altitude = 7 deg.)
468 13 Observations of eclipses by medieval Arab astronomers

(2) AD 1004 Jan 24 [Monday] (annular, mag. = 0.98): Cairo

This solar eclipse was in the sign of Aquarius and was in the late
afternoon of Monday the 29th of the month of Rabi' al-Awwal in the year
394 of al-Hijrah... (date on Persian, Syrian and Coptic calendars)... The
Sun was eclipsed until what remained of it resembled the crescent Moon
on the first night of the month. I estimated the eclipsed portion of the Sun
to be 11 digits (i.e. of surface). The altitude of the Sun when the eclipse
became noticeable in it (i.e. on its disk) was 16^ deg in the west; thus I
estimated (the altitude) at the beginning to be 18 \ deg. About quarter of
the diameter was eclipsed when the altitude was 15 deg, and half of the
diameter was eclipsed when the altitude was 10 deg. The eclipse was
complete (i.e. maximum phase) when the altitude was 5 deg.
[Al-Zij al-Kabir al-Hakimi; trans. Said and Stephenson (1997).]
The Arabic text of the above record from the Ibn Yunus manuscript is
illustrated in figure 13.5. This figure also includes the records of the solar
eclipse of AD 993 Aug 20 and the lunar eclipses of AD 1001 Sep 5/6 and
1002 Mar 1/2.
The solar altitude measurements will be considered in section 13.7.
Although the eclipse was generally annular, it is clear from the text that
the observed phase at Cairo was only partial. The progress of the eclipse
was carefully followed, and it thus seems very unlikely that the observers
overlooked the annular phase.

RESULTS
For a partial eclipse at Cairo, either AT < 1770 or AT > 1940 sec.
Since the belt of annularity was rather narrow, only a restricted range
of AT is ruled out by the observation of a partial eclipse. It is unfortunate
that Ibn Yunus does not state whether the upper or lower part of the Sun
remained visible. If he had done so, one of the above AT ranges would
have been excluded. His remark that what remained of the eclipsed Sun
'resembled the crescent Moon on the first night of the month' may simply
be an indication of the narrowness of the solar crescent, not necessarily
the direction in which it was facing.
(NB computed LT of maximum phase = 16.7 h; solar altitude = 7 deg.)

13.7 Solar eclipse times recorded by Ibn Yunus

Although Ibn Yunus lists the various observations which he cites purely
in chronological order (i.e. without regard to type), he records so many
observations that I have decided to group the solar and lunar eclipses
separately. Lunar eclipse timings reported in his treatise will be discussed
in section 13.8. Except where otherwise stated, all translations in this and
the following sections are by Said and Stephenson (1997), with occasional
 13.7 Solar eclipse times recorded by Ibn Yunus 469

;;

^rt^ife^E^W ^ ^ % i ^ i . ^ .
j Jk>tt>r

Fig. 13.5 Arabic text of eclipse records of AD 993 Aug 20 (solar), 1001 Sep
5/6 (lunar), 1002 Mar 1/2 (lunar) and 1004 Jan 24 (solar) in the Ibn Yunus
manuscript (Leiden University library: Or. 143).
470 13 Observations of eclipses by medieval Arab astronomers

minor amendments. Predictions by the Arab astronomers will usually

be ignored.
Each entry below begins with the calculated date on the Julian calendar,
followed by the place of observation, including the name of the principal
observer, where known. The magnitude of the eclipse calculated using
equation (8.1) is given for reference only. In the case of altitude measure-
ments, allowance for refraction is unnecessary since no measured altitude
is less than about 5 deg. Under these circumstances, refraction corrections
amount to no more than about 0.1 deg (see the tables of Allen, 1976). It
should be noted that although the time of mid-eclipse is often specified,
this will be taken as the moment of maximal phase instead. In the case
of a solar eclipse - but not a lunar obscuration - these instants are not
identical owing to the irregular rate of advance of the lunar shadow across
the Earth's surface (see chapter 3).

(1) AD 829 Nov 30 [Tuesday] (computed mag. = 0.56): Baghdad

Ahmad b. 'Abd Allah known as Habash said: There was a lunar
eclipse... in the year 198 of Yazdijerd... As for the solar eclipse, which
(occurred) in this year at the end of the month of Ramadan, all
calculations were in error. The altitude of the Sun at the beginning of the
eclipse was 7 deg as they (the astronomers) claim. The eclipse ended when
the altitude of the Sun was about 24 deg, as though it was 3 hours of day
(after sunrise)'.

Here we have a combination of a Persian year and a Muslim month.

The year 198 of Yazdijerd (= 214/215 AH) covered the period from AD
829 Apr 28 to 830 Apr 27. A date at the very end of Ramadan in that
year closely corresponds (within a day or so) to AD 829 Nov 30 - the
calculated day of a solar eclipse visible in Baghdad. No observational
details are recorded for the lunar eclipse; for the present purpose, the
allusion to this event merely serves to establish the year.
The remark that the astronomers claimed that the altitude of the Sun
at first contact was 7 deg suggests that the measurement was regarded as
only approximate.

RESULTS
(i) Solar altitude at first contact ~ 7 deg (in the east). Equivalent LT ~
7.68 h, UT ~ 4.60 h. Computed TT = 4.85 h, thus AT - 900 sec.
(ii) Solar altitude at last contact ~ 24 deg (in the east). Equivalent LT
- 9.52 h, UT - 6.44 h. Computed TT = 7.06 h, thus AT - 2250 sec.
Presumably the recorded time of the end of the eclipse (i.e. three hours
after sunrise) was derived from the solar altitude measurement. The fact
that the time was expressed in 'hours of day' implies seasonal hours. Since
the LT of sunrise = 6.87 h, 1 seasonal hour = 0.855 h. Hence the estimated
 13.7 Solar eclipse times recorded by Ibn Yunus All

LT of last contact = 9.44 h. This compares fairly well with the above
result deduced from the altitude measurement of 9.52 h.

(2) AD 866 Jun 16 [SundayJ (mag. = 0.66): Baghdad

This solar eclipse was mentioned by al-Mahani. He said: The Sun is to
be eclipsed on Sunday the 28th of (the month of) Jumada al-Ula in the
year 252 of al-Hijrah... (date on Persian calendar)... It was found (by
observation) that this eclipse began (a little) more than a third of an hour
after Zawal; the middle of the eclipse, as we estimated, was at 7 hours and
3 and JQ (i.e. 7 ;26 h after sunrise); then the eclipse cleared at 8 hours \
(i.e. 8 ;30 h)... (calculated details)... The eclipsed part of the Sun's diameter,
as we estimated, was more than 7 digits and less than 8 digits'.

The month should be Jumada al-Ukhra, rather than the previous month
Jumada al-Ula; otherwise the recorded Islamic date (including the week-
day) is correct - as is the date on the Persian calendar. Elsewhere in the
above text, it is stated that times are expressed in seasonal hours. No
altitude measurements are preserved, only the reduced times. Zawal means
the moment when the shadow of a gnomon begins to noticeably increase
after the Sun transits the meridian and is thus a few minutes after noon.
It will be assumed that the estimated time of first contact ('more than a
third of an hour after Zawal') means roughly half an hour after midday.

RESULTS
(i) First contact ~ 0.52 h after midday. LT of contact ~ 12.50 h, UT ~
9.51 h. Computed TT = 10.12 h, thus AT - 2200 sec.
(ii) Maximum phase 7.43 seasonal hours after sunrise. LT of sunrise
on Jun 16 = 4.82 h, thus 1 seasonal hour = 1.20 equal hours. LT of
observation = 13.71 h, UT = 10.73 h. Computed TT = 11.43 h, thus AT
= 2500 sec.
(iii) Last contact 8.5 seasonal hours after sunrise. LT of contact = 14.99
h, UT = 12.01 h. Computed TT = 12.70 h, thus AT = 2450 sec.
These three AT results are fairly self-consistent.

(3) AD 923 Nov 11 [Tuesday] (mag. = 0.80): Baghdad

This solar eclipse was calculated and observed by Abu al-Hasan Ali ibn
Amajur, who used the al-Zij al-Arabi of Habash. This eclipse was at the
conjunction (i.e. new Moon) of (the month of) Sha'ban in the year 311
(AH). We as a group observed (this eclipse) and clearly distinguished it.
The estimate of all (observers) for the middle of the eclipse was that it
occurred when the altitude of the Sun was 8 deg in the east; its clearance
was at 2^ seasonal hours (after sunrise), when the altitude of the Sun was
20 deg. We observed this eclipse at several sites on the Tarmah (an elevated
platform on the outside of the building). The estimate of Abu al-Hasan for
472 13 Observations of eclipses by medieval Arab astronomers

the middle of the eclipse at his house was when the altitude of the Sun was
8 deg, as I estimated myself at my house before he arrived. The magnitude
of the eclipse was \ and \ (i.e. | ) of the Sun's diameter; the middle of the
eclipse, which we estimated when the Sun's altitude was 8 deg, would be
when the elapsed time (after sunrise) was 0;50 seasonal hours, and the
(celestial) sphere had revolved (through) 10;40 deg. (The interval) between
the middle of the eclipse and its clearance in this observation was 1 ;22
seasonal hours... (alternative times in equal hours)...'.
The statement that the eclipse 'was at the conjunction of (the month
of) Sha'ban' implies that it occurred at the new Moon of Sha'ban - i.e.
the very end of the previous lunar month Rajab.

RESULTS
(i) Solar altitude at maximum phase = 8 deg (in the east). Equivalent LT
= 7.56 h, UT = 4.37 h. Computed TT = 4.91, thus AT = 1950 sec.
(ii) Solar altitude at last contact = 20 deg (in the east). Equivalent LT
= 8.72 h, UT = 5.53 h. Computed TT = 5.98 h, thus AT = 1600 sec.
NB the LTs derived in (i) and (ii) above agree well with the times
deduced by the observers from their own altitude determinations (i.e. 7.51
h and 8.70 h respectively).

(4) AD 928 Aug 18 [Monday] (mag. = 0.25): Baghdad

This solar eclipse was calculated and observed by Ali ibn Amajur.
(According to calculation), the beginning was to be at... (calculated
details)... on Monday. He said: 4I observed this eclipse with my son Abu
al-Hasan and Muflih and (found) that the Sun rose (already) eclipsed by
less than one \ of its surface. The eclipse continued to increase by an
amount that we could perceive until | (of its surface) was eclipsed. We
observed the Sun distinctly (by reflection) in water. We (found) that it
cleared and nothing of the eclipse remained and we distinguished the (full)
circle of the Sun's body in water; (that was) when the altitude was 12 deg
in the east, less \ of a division of the al-halaqa (i.e. the ring), which is
graduated in thirds (of a degree), that is (less by) | degree.... (comparison
with calculation)...' .
Although the full date is not stated, it can fairly readily be established.
The observation is cited between the records of two lunar eclipses: those
of AD 927 Sep 13/14 and 929 Jan 27/28 (entries 6 and 7 of section
13.8 below). Only one solar obscuration would be visible at Baghdad
during this interval - that of AD 928 Aug 18, which indeed occurred on
a Monday. Hence the date is firmly established.

RESULTS
Solar altitude at last contact = 11.9 deg (in the east). Equivalent LT =
6.44 h, UT = 3.50 h. Computed TT = 4.01 h, thus AT = 1800 sec.
 13.7 Solar eclipse times recorded by Ibn Yunus 473

(5) AD 977 Dec 13 [Thursday] (mag. = 0.60): Cairo

This solar eclipse was in the early morning of Thursday the 28th of the
month of Rabi 'al-Akhir, in the year 367 of al-Hijrah... (date on Persian
calendar)... We, a group of scholars (ten names are given), attended at
al-Qarafah (a district of Cairo) in the Mosque of Abu Ja'far Ahmad ibn
Nasr al-Maghribi to watch this eclipse. Everyone waited for the beginning
of this eclipse. It began to be perceived when the altitude of the Sun was
more than 15 deg but less than 16 deg. (Those) present all agreed that
about 8 digits of the Sun's diameter were eclipsed, that is (a little) less than
7 digits of surface. The Sun completely cleared when its altitude was more
than 33 deg by about | of a degree, as estimated by me, and agreed by all
those present... (calculated details)...

RESULTS
(i) Solar altitude at first contact = 15.5 deg (in the east). Equivalent LT
= 8.38 h, UT = 6.29 h. Computed TT = 6.80 h, thus AT = 1800 sec.
(ii) Solar altitude at last contact = 33.3 deg (in the east). Equivalent LT
= 10.69 h, UT = 8.60 h. Computed TT = 9.15 h, thus AT = 2000 sec.

(6) AD 978 Jun 8 [Saturday] (mag. = 0.50): Cairo

This solar eclipse was on Saturday the 29th of (the month of) Shawwal
in the year 367 of al-Hijrah... (date on Persian, Syrian and Coptic
calendars)... A maximum of 5^ digits of the Sun's diameter were eclipsed,
according to estimation, that is 4 digits 10 minutes (i.e. 4 \ digits) of
surface. The altitude of the Sun when a portion of the eclipse began to be
perceived was 56 deg approximately. The completion of the clearance was
when the altitude of the Sun was 26 deg or about so... (calculated
details)...

RESULTS
(i) Solar altitude at first contact ~ 56 deg (in the west). Equivalent LT ~
14.49 h, UT - 12.37 h. Computed TT = 12.73 h, thus AT - 1300 sec.
(ii) Solar altitude at last contact ~ 26 deg (in the west). Equivalent LT
- 16.82 h, UT - 14.70 h. Computed TT = 15.26 h, thus AT - 2000 sec.
Although the direction in which the Sun was located is not stated, it
is obvious from the context that at last contact the solar altitude was
declining. For first contact, a solar altitude of 56 deg leads to a LT of
either 9.51 h or 14.49 h. The former would imply an impossible duration
of more than 7 hours, so that a westerly azimuth at this phase may also
be inferred.

(7) AD 979 May 28 [Wednesday] (mag. = 0.45): Cairo

This solar eclipse was in the late afternoon of Wednesday the 23rd
[read: 28th] of (the month of) Shawwal in the year 368 of al-Hijrah... (date
474 13 Observations of eclipses by medieval Arab astronomers

on Persian, Syrian and Coptic calendars)... The eclipse was perceptible

when the altitude of the Sun was 6^ deg. About 5^ digits of the Sun's
diameter were eclipsed, as I estimated, that is 4;10 digits of surface. The
Sun set eclipsed... (comparison with magnitude of previous solar eclipse)...

The 23rd of a lunar month is too early for a solar eclipse, implying a
minor scribal error. The day of the month - like the other numbers -
is expressed using numerals based on the letters of the Arabic alphabet
(similar to the practice of the ancient Greeks). The presence or absence
of diacrytical points (dots) can lead to widely differing numbers. In this
particular instance, the numerals for 23 and 28 are very similar.

RESULTS
Solar altitude at first contact = 6.5 deg (in the west). Equivalent LT =
18.37 h, UT = 16.21 h. Computed TT = 16.61 h, thus AT = 1450 sec.

(8) AD 985 Jul 20 [Monday] (mag. = 0.30): Cairo

This solar eclipse was in the late afternoon on Monday at the end of
(the month of) Safar in the year 375 of al-Hijrah. The altitude of the Sun
when I perceived its eclipse by eye was 23 deg approximately. The altitude
was 6 deg when nothing of its eclipse remained to be perceived by the eye.
A maximum of \ of the Sun's diameter was eclipsed.

As noted in the comment on the previous entry, the Arabic numeral for
23 closely resembles that for 28. Hence the altitude of the Moon at first
contact could well have been 28 deg. A mean of 25.5 deg will be assumed.

RESULTS
(i) Solar altitude at first contact ~ 25.5 deg (in the west). Equivalent LT
- 16.74 h, UT - 14.73 h. Computed TT = 15.15 h, thus AT - 1500 sec.
(ii) Solar altitude at last contact = 6 deg (in the west). Equivalent LT
= 18.30 h, UT = 16.29 h. Computed TT = 16.51 h, thus AT = 750 sec.

(9) AD 993 Aug 20 [Thursday] (mag. = 0.96): Cairo

This solar eclipse was in the forenoon of Sunday the 29th of (the month
of) Jumada al-Ukhra in the year 383 of al-Hijrah... (date on Persian,
Syrian and Coptic calendars)... The eclipse began when the altitude of the
Sun was 27 deg in the east and was complete (i.e. reached its maximum)
when the altitude was 45 deg in the east. The Sun cleared when its altitude
was 60 deg in the east. About | of it (i.e. the surface) was eclipsed.

Use of the provisional AT result obtained from equation (8.1) (1750

sec) leads to a calculated magnitude of 0.96. The observed magnitude
(equivalent to 0.59 in terms of the solar diameter) would thus appear to
represent a considerable underestimate (see also section 13.10). The eclipse
 13.7 Solar eclipse times recorded by Ibn Yunus 475

was independently reported to be total by the Cairo chronicler, al-Maqrizi

(see chapter 12). However, al-Maqrizi lived more than four centuries after
the event and his source is unknown.

RESULTS

(i) Solar altitude at first contact = 27 deg (in the east). Equivalent LT =
7.67 h, UT = 5.60 h. Computed TT = 6.16 h, thus AT = 2000 sec.
(ii) Solar altitude at maximum phase = 45 deg (in the east). Equivalent
LT = 9.08 h, UT = 7.01 h. Computed TT = 7.40 h, thus AT = 1400 sec.
(iii) Solar altitude at last contact = 60 deg (in the east). Equivalent LT
= 10.35 h, UT = 8.28. Computed TT = 8.72 h, thus AT = 1600 sec.
NB for this eclipse to be partial at Cairo, in accordance with the
observation recorded by Ibn Yunus, either AT < -2800 or AT > -750
sec. Neither of these limits are at all critical and will be disregarded in
further investigation.

(10) AD 1004 Jan 24 [Monday] (mag. = 0.98): Cairo

This solar eclipse was in the sign of Aquarius and was in the late
afternoon of Monday the 29th of the month of Rabi' al-Awwal in the year
394 of al-Hijrah... (date on Persian, Syrian and Coptic calendars)... The
Sun was eclipsed until what remained of it resembled the crescent Moon
on the first night of the month. I estimated the eclipsed portion of the Sun
to be 11 digits (i.e. of surface). The altitude of the Sun when the eclipse
became noticeable in it (i.e. on its disk) was 16\ deg in the west; thus I
estimated (the altitude) at the beginning to be 18^ deg. About quarter of
the diameter was eclipsed when the altitude was 15 deg and half of the
diameter was eclipsed when the altitude was 10 deg. The eclipse was
complete (i.e. maximum phase) when the altitude was 5 deg.

The observation that the eclipse fell only a little short of annularity has
already been discussed in section 13.6. Since the altitude at greatest phase
was only 5 deg, the Sun would set still partially obscured; this explains the
lack of any reference to last contact. Ibn Yunus' adjustment of as much as
2 deg to the measured altitude at the start suggests that the eclipse was not
noticed until it was already fairly well advanced. The altitude correction
represents a delay in sighting the eclipse by as much as 0.15 h; possibly
intermittent cloud was responsible. Presumably the observer made an
empirical correction based on the phase when the eclipse was first seen. It
will be assumed that the amended altitude (i.e 18.5 deg) is accurate.
The determinations of solar elevation when it was estimated that re-
spectively one quarter and half of the solar diameter was covered are
probably also worth considering; the phase was changing fairly rapidly
- by about 1 digit every 5 or 6 minutes. However, these observations
476 13 Observations of eclipses by medieval Arab astronomers

are likely to lead to less accurate AT results than those derived from the
measurements at first contact and maximum phase.
RESULTS
(i) Solar altitude at first contact = 18.5 deg (in the west). Equivalent LT
= 15.68 h? UT = 13.85 h. Computed TT = 14.25 h, thus AT = 1450 sec.
(ii) Solar altitude when ~ 0.25 of solar diameter covered = 1 5 deg (in
the west). Equivalent LT = 16.00 h, UT = 14.17 h. Computed TT = 14.53
h, thus AT - 1300 sec.
(iii) Solar altitude when ~ 0.50 of solar diameter covered = 10 deg (in
the west). Equivalent LT = 16.45 h, UT = 14.62 h. Computed TT = 14.81
h, thus AT - 700 sec.
(iv) Solar altitude at maximum phase = 5 deg (in the west). Equivalent
LT = 16.88 h, UT = 15.05 h. Computed TT = 15.37 h, thus AT = 1150
sec.
The various AT results derived in this section are assembled in table
13.2.

The AT values listed in table 13.2 are plotted in figure 13.6. The scale
of this diagram has been chosen to enable ready comparison with later
AT plots in this chapter (figures 13.8, 13.9 and 13.10).

13.8 Lunar eclipse times recorded by Ibn Yunus

For all lunar eclipses, calculated double dates (centred on local midnight)
are used systematically in this chapter - as elsewhere in this book. How-
ever, I have computed weekdays for the second of each pair of dates since
the Islamic day begins at sunset; as a result direct comparison can be
made with the recorded weekday. The form of the various entries follows
much the same pattern as in the previous section. It should be noted that
whereas the Baghdad observers regularly measured altitudes of selected
clock stars rather than the Moon itself, the Cairo astronomers generally
preferred to determine the lunar elevation.
Among the lunar records cited by Ibn Yunus from al-Mahani is an
account of an eclipse whose magnitude was so small that the observers
believed it to be penumbral. Calculations from tables had indicated a
likely magnitude of 1.5 digits. However, al-Mahani stated that although
the brightness at the northern edge of the Moon appeared diminished,
no part of the lunar disk was lost to view. He did not note the time of
occurrence so that the record cannot be utilised to determine AT. The
Islamic date of this event (252 AH, Dhu al-Qa'dah 15) corresponds to AD
866 Nov 25/26, on which day the computed magnitude was only 0.03. A
truly penumbral eclipse was seen by Babylonian astronomers in 188 BC
(see chapter 6).
 13.8 Lunar eclipse times recorded by Ibn Yunus All

Table 13.2 AT results from solar eclipse timings

recorded by Ibn Yunus.
Year" Cih ATC
+829 1 -900
+829 4 -2250
+866 1 -2200
+866 M 2500
+866 4 2450

+923 1 1950
+923 4 1600
+928 4 1800
+977 1 1800
+977 4 2000

+978 1 -1300
+978 4 -2000
+979 1 1450
+985 1 -1500
+985 4 750

+993 1 2000
+993 M 1400
+993 4 1600
1004 1 1450
1004 E -1300
1004 E -700
1004 M 1150
a
Year
b
Contact, etc. (M = maximal phase; E = estimated phase - in
AD 1004 only).
c
Derived AT value.

(1) AD 854 Feb 16/17 [Saturday] (mag. = 0.92): Baghdad

This lunar eclipse was mentioned by al-Mahani. 'There was an eclipse of
the Moon in the month of Ramadan in the year 239 of al-Hijrah on the
night of Saturday, the middle of the month. It was found by observation
that the beginning of this eclipse was at 10 hours and something like half
of one-tenth of an hour (i.e. 10;03 h) after midday of Friday. We did not
determine its times apart from the beginning. It was found that the
uneclipsed part of its body was (a little) more than 1/10...(calculated
details)...'.
The fact that only the time of first contact is reported is surprising; as an
478 13 Observations of eclipses by medieval Arab astronomers

4000 -

3500 -

3000 •

_ 2500
o
jT 2000 - o 00 o
< o 0
o o
1500 o° 0 °
o o
o
1000 -
o o
500

n
+ 800 +820 +840 +860 +880 +900 +920 +940 +960 +980+1000+1020+1040
Year
Fig. 13.6 AT values derived from solar eclipse timings recorded by Ibn Yunus.

evening eclipse, the entire course of events should have been visible unless
unfavourable weather intervened. Although the text does not specify
whether equal or unequal hours were used, since part of the measured
interval of 10;03 hours after midday was in daylight and part in darkness,
only equal hours would be meaningful.

RESULTS
LT of first contact - 22.05 h on Feb 16. UT 19.35 h. Computed TT =
20.22 h, thus AT - 3150 sec.

(2) AD 854 Aug 11/12 [Sunday] (mag. = 1.14): Baghdad

This lunar eclipse was mentioned by al-Mahani. The Moon was
eclipsed on the night of Sunday 13th of the month of Rabi' al-Awwal in
the year 240 of al-Hijrah. It was found by observation that the time of
beginning of the eclipse was when the altitude of (the star) al-dabaran
(Aldebaran: a Tau) was 45;30 deg in the east. We did not find its times
(accurately) except this time, which was exact and precise. We measured
the time of the completion of (the first phase of) the eclipse, which is the
time of the beginning of the staying (al-makth) (in totality) and found it (to
be) when the altitude of (the star) (al-shVra) al-shamiyyah (Procyon: a
CMi) was between 22 and 23 deg in the east. This (latter) measurement is
not exact but approximate. We determined the time of the beginning from
the altitude of al-dabaran by the astrolabe and found it to be 44 deg (of
 13.8 Lunar eclipse times recorded by Ibn Yunus 479

the celestial sphere) after midnight... (calculated details)... We (also)

determined the time (of the beginning of) the stay by the astrolabe, taking
the altitude of al-shamiyyah as 23 deg and found it to be 23^ parts (i.e.
degrees) of the celestial sphere after the beginning (of the eclipse)'.
RESULTS
(i) Altitude of a Tau at first contact = 45.5 deg in the east. Equivalent
LT (Aug 12) = 2.92 h, UT = 0.01 h. Computed TT = 0.71 h, thus AT =
2500 sec.
(ii) Altitude of a CMi at second contact ~ 22.5 deg in the east. Equiva-
lent LT - 4.45 h, UT - 1.53 h. Computed TT = 1.91 h, thus AT - 1350
sec.
The LTs deduced from the measured altitudes of both a Tau (2.92 h)
and a CMi (4.45 h) agree well with the astrolabe reductions (respectively
2.93 and 4.50 h).

(3) AD 856 Jun 21/22 [Monday] (mag. = 0.59): Baghdad

(This is) the third lunar eclipse mentioned by al-Mahani. There was an
eclipse of the Moon on the night of Monday the middle of (the month of)
Safar in the year 242 of al-Hijrah... (date on the Persian Calendar)... It
was found by observation that the beginning of the eclipse was when the
altitude of (the star) al-dabaran (Aldebaran: a Tau) was 9;30 deg in the
east; the amount of the revolution of the (celestial) sphere from midnight
to this time, as we determined (from this measurement) with the astrolabe
was 50 deg. We did not determine its times except for the beginning. It was
found (by observation) that the uneclipsed part of its body (i.e. disk) was
more than one-quarter and less than one-third (calculated details)...'.
RESULTS
Altitude of a Tau at first contact = 9.5 deg in the east. Equivalent LT
(Jun 22) = 3.39 h, UT = 0.43 h. Computed TT = 1.08 h, thus AT = 2350
sec.
Once again, the astrolabe reduction (3.33 h) is fairly precise.

(4) AD 923 Jun 1/2 [Monday] (mag. = 0.66): Baghdad

This lunar eclipse was calculated by Ali ibn Amajur al-Turki from the
al-Zij al-'Arabi of Habash and observed by him and his son Abu al-Hasan
and his freedman Muflih. (He said:) There was an eclipse of the Moon in
(the month of) Safar in the year 311 of al-Hijrah... (calculated
details)... The Moon rose at sunset (already) eclipsed by \ or (a little)
more of the digits of the diameter (i.e. 3 digits or a little more). The Moon
was eclipsed by (a little) more than 9 digits of diameter. The middle of the
eclipse was at 1 and | of equal hours of night (i.e. after sunset). The
clearance of the eclipse was at 3 equal hours (after sunset) and (that was)
when the altitude of (the star) al-ridf (Deneb: a Cyg) was 29;30 deg in the
east... (calculated details)...'.
480 13 Observations of eclipses by medieval Arab astronomers

The alternative Arabic name for al-ridf is Dhanab, from which the
modern name Deneb is derived.

RESULTS
(i) Mid-eclipse 1.67 h after sunset on Jun 1. LT of sunset = 19.14 h. Hence
LT of observation = 20.81 h, UT = 17.78 h. Computed TT = 18.12 h,
thus AT = 1200 sec.
(ii) Altitude of a Cyg at last contact = 29.5 deg in the east. Equivalent
LT = 21.97 h, UT = 19.94 h. Computed TT = 19.50 h, thus AT = 2000
sec.

(5) AD 925 Apr 11/12 [Tuesday] (mag. = 1.08): Baghdad

This lunar eclipse was calculated and observed by Abu al-Hasan ibn
Amajur. This eclipse was on the night of Tuesday, the 15th of (the month
of) Muharram, year 313 of al-Hijrah. He mentioned that the Moon was
totally eclipsed and reported its five times (as found by calculation but not
given in the text). Then he said: 'I observed this eclipse. The beginning was
when the altitude of (the star) (al-simak) al-ramih (Arcturus: a Boo) was 11
deg in the east. The end of clearance was when the altitude of (the star)
al-nasr al-waqi' (Vega: a Lyr) was 24 deg'. He then said: The beginning of
the eclipse would be when the time elapsed from the start of night (i.e.
from sunset) was 0;55 seasonal hours; the observation was later by 0;23
seasonal hours than that calculated from the al-Zij al-Mumtahan of
Habash. The end of clearance by observation would be at 4;36 seasonal
hours; observation was (again) later than calculation by 0;17 seasonal
hours'.
There must be an error in the preserved altitude of a Boo at first contact.
This star would be 11 deg above the eastern horizon at about 55 minutes
before sunset rather than after it but the text clearly implies an observation.
Presumably a scribe made an error in reporting the elevation of the star,
but there is no obvious alternative reading here. The reported time of 55
minutes after sunset will be adopted instead. When Ibn Amajur compared
this time with the result based on the tables of Habash, it showed much
the same discrepancy as for the reported time of last contact.
It is not stated whether the azimuth of a Lyr was east or west of the
meridian at last contact. However, it is clear from the reported time of
4 h 36 m after sunset (LT = 22.72 h) that the star was to the east of the
meridian; an elevation of 24 deg would correspond to a rather similar LT
of 22.77 h.

RESULTS
(i) First contact 0.92 h after sunset on Apr 11. LT of sunset = 18.52 h.
Hence LT of observation = 19.44 h, UT = 16.46 h. Computed TT = 17.13
h, thus AT = 2400 sec.
 13.8 Lunar eclipse times recorded by Ibn Yunus 481

(ii) Altitude of a Lyr at last contact = 24 deg in the east. Equivalent

LT (Apr 11) = 22.77 h, UT = 19.74 h. Computed TT = 20.44 h, thus AT
= 2350 sec.
These two results for AT are almost identical.
(6) AD 927 Sep 13/14 [Friday] (mag. = 0.22): Baghdad
This lunar eclipse was calculated and observed by AH ibn Amajur, who
used the al-Zij al-'Arabi of Habash. This eclipse was on the night of
Friday, in the year 315 of al-Hijrah...(calculated details)...' He said: 'This
eclipse was observed by my son Abu al-Hasan. The beginning of the
eclipse was when the altitude of (the star) al-shi'ra al-yamaniyyah (Sirius: a
CMa) was 31 deg in the east; the part of the celestial sphere which has
revolved between sunset and the beginning of the eclipse is 148 deg plus a
third of a degree and this is (equivalent to) 9;52 equal hours, which is 10 ;0
seasonal hours. The estimated digits of the eclipse was more than \ but
less than |, as though it was 3 | digits...(calculated details...)'.
Although the month (and day of the month) is not cited, no other
lunar eclipse apart from that identified above occurred in 315 AH (AD
927 Mar - 928 Feb). Confirmation of the calculated date is provided by
the recorded weekday (Friday). In addition, the estimate of magnitude is
in tolerable accord with calculation for this eclipse.

RESULTS
Altitude of a CMa at first contact = 31 deg in the east. Equivalent LT
(Sep 14) = 3.98 h, UT = 0.91 h. Computed TT = 1.73 h, thus AT = 2950
sec.
(7) AD 929 Jan 27/28 [Wednesday] (mag. = 1.19): Baghdad
This lunar eclipse was calculated and observed by Abu al-Hasan AH ibn
Amajur. This eclipse was at the opposition (i.e. full Moon) of (the month
of) Dhu al-Hijjah, year 316 of al-Hijrah... (calculated details)... He said: 'I
observed this eclipse at its beginning when the altitude of (the star)
(al-simak) al-ramih (Arcturus: a Boo) was 18 deg in the east; the time
elapsed from the start of the night (i.e after sunset) (to the beginning) was
5 hours seasonal, as required by calculation from the (al-Zij) al-Mumtahan\
RESULTS
Altitude of a Boo at first contact = 1 8 deg in the east. Equivalent LT (Jan
27) = 22.79 h, UT = 20.10 h. Computed TT = 21.77 h, thus AT = 6000
sec.
Although the LT derived from the altitude measurement is in fair accord
with the stated time of 5 unequal hours after sunset (equivalent to a LT of
22.88 h), and was said to confirm calculation from tables, the resulting AT
value is very discordant compared with roughly contemporaneous results.
Hence the observation will be rejected.
482 13 Observations of eclipses by medieval Arab astronomers

(8) AD 933 Nov 4/5 [Tuesday] (mag. = 1.42): Baghdad

This lunar eclipse was calculated and observed by Ali ibn Amajur
al-Turki. He said: '(According to calculation) from the al-Zij al-Mumtahan
al-Arabi of Habash, the eclipse opposition was on the night of Tuesday the
13th of (the month of) Dhu al-Qa'dah, year 321 of al-Hijrah... (calculated
details)...' He said 'I observed this eclipse when (the Moon) entered (the
shadow) and (that was) when the altitude of (the star) (al-simak) al-ramih
(Arcturus: a Boo) was 15 deg in the east and when the time elapsed from
the start of the night was 9;56 seasonal hours... (calculated details)...'.

RESULTS

Altitude of a Boo at first contact = 1 5 deg in the east. Equivalent LT

(Nov 5) = 4.56 h, UT = 1.36 h. Computed TT = 1.99 h, thus AT = 2250
sec.
The LT deduced from the altitude of a Boo agrees reasonably well with
the stated time after sunset (corresponding to a LT of 4.41 h).

(9) AD 979 May 14/15 [ThursdayJ (mag. = 0.70): Cairo

This lunar eclipse was in (the month of) Shawwal in the year 368 of
al-Hijrah. The Moon rose eclipsed on the night whose morning was
Thursday... (date on Persian, Syrian and Coptic calendars)... More than 8
digits but less than 9 of the Moon's diameter were eclipsed... (calculated
details)... The eclipse cleared when about one equal hour and a fifth of
night had elapsed, as I estimated... (calculated details)....

RESULTS

Last contact ~ 1.20 h after moonrise on May 14. LT of moonrise = 18.87

h. Hence LT of contact - 20.07 h, UT - 17.90. Computed TT - 18.31 h,
thus AT ~ 1650 sec.

(10) AD 979 Nov 6/7 [Friday] (mag. = 0.84): Cairo

This lunar eclipse was in the month of Rabi 'al-Akhir in the year 369
(of al-Hijrah), on the night whose morning was Friday the 13th of the
month... (date on Persian, Syrian and Coptic calendars)... A group of
scholars gathered to observe this eclipse. They estimated that what was
eclipsed of the Moon's circular surface was 10 digits. The altitude of the
Moon when they perceived the eclipse was 64^ deg in the east. The altitude
when its clearance completed was 65 deg in the west... (calculated details).

Although the Moon was high in the sky at both contacts, each altitude
was sufficiently below the meridian elevation (77 deg) for the rate of
change with time to be significant.
 13.8 Lunar eclipse times recorded by Ibn Yunus 483

RESULTS
(i) Altitude of Moon at first contact = 64.5 deg in the east. Equivalent LT
(Nov 6) = 22.40 h, UT = 20.08 h. Computed TT = 20.61 h, thus AT =
1900 sec.
(ii) Altitude of Moon at last contact = 65 deg in the west. Equivalent
LT = 1.62 h, UT = 23.30 h. Computed TT = 23.67 h, thus AT = 1300
sec.
(11) AD 980 May 2/3 [Monday] (mag. = 1.61): Cairo
The Moon was totally eclipsed in (the month of) Shawwal in the year
369 of al-Hijrah on the night whose morning was Tuesday... (date on
Persian calendar)... A group of scholars gathered to observe this eclipse.
They perceived the trace of the eclipse when the altitude of the Moon was
471 deg. The eclipse cleared when about | of an equal hour remained to
the end of the night (i.e. before sunrise). Our gathering to observe this
eclipse was in the Mosque of Ibn Nasr (al-Maghribi) at al-Qarafa.
The day of the week is in error; it should be Monday rather than
Tuesday.
There is an obvious error in the recorded lunar altitude when the eclipse
began; this significantly exceeds the meridian altitude of the Moon (42.5
deg). A possible alternative reading in place of 47 is 40 (the figures are
similar), leading to an altitude at first contact of 4 0 | deg. However, even
if this restoration is correct, the Moon would be too close to the meridian
for the rate of change of altitude to be significant. Hence only the timing
of fourth contact is usable.

RESULTS
Last contact ~ 0.60 h before sunrise on May 3. LT of sunrise = 5.25 h.
Hence LT of contact ~ 4.65 h, UT ~ 2.48 h. Computed TT = 3.07 h, thus
AT - 2100 sec.

(12) AD 981 Apr 21/22 [Friday] (mag. = 0.18): Cairo

This lunar eclipse was in the month of Shawwal in the year 370 of
al-Hijrah on the night whose morning was Friday... (date on Persian,
Syrian and Coptic calendars)... We gathered to observe this eclipse at
al-Qarafa in the Mosque of Ibn Nasr al-Maghribi. We perceived the
beginning of this eclipse when the altitude of the Moon was approximately
21 deg. About one-quarter of the Moon's diameter was eclipsed. The
Moon cleared completely when about | of an hour remained to sunrise.
As the eclipse ended shortly before sunrise, the Moon must have been
to the west of the meridian when its altitude was measured. Although it
is not stated whether the time interval of one-quarter of an hour before
sunrise for last contact was in equal or seasonal hours, the latter would
only be 0.02 h less. A mean of 0.24 h will be assumed.
484 13 Observations of eclipses by medieval Arab astronomers

RESULTS
(i) Altitude of Moon at first contact ~ 21 deg in the west. Equivalent LT
(Apr 22) - 3.53 h, UT - 1.39 h. Computed TT = 1.97 h, thus AT - 2000
sec.
(ii) Last contact ~ 0.24 h before sunrise on Apr 22. LT of sunrise =
5.39 h. Hence LT of contact ~ 5.15 h, UT ~ 3.01 h. Computed TT = 3.67
h, thus AT ~ 2350 sec.

(13) AD 981 Oct 15/16 [Sunday] (mag. = 0.36): Cairo

The Moon was eclipsed in the month of Rabi' al-Akhir, in the year 371
of al-Hijrah on the night whose morning was Sunday. About 5 digits of
diameter were eclipsed. The altitude of the Moon at the contact from
outside (external or first contact) was about 24 deg, as I estimated...
(calculated details)...
It is not specified whether the Moon was to the east or west of the
meridian at first contact. The stated altitude of 24 deg corresponds to
a LT of either about 19.56 h (UT = 17.66 h) or 4.44 h (UT = 2.11 h)
depending on whether the Moon was in an easterly or westerly azimuth.
Since the computed TT is 2.54 h, only the latter alternative would lead
to a meaningful result for AT (i.e. ~ 1500 sec rather than ~ 32000 sec).
Hence it will be assumed that the Moon was to the west of the meridian.

RESULTS
Altitude of Moon at first contact ~ 24 deg in the west. Equivalent LT
(Oct 16) - 4.44 h, UT - 2.11 h. Computed TT = 2.54 h, thus AT - 1500
sec.

(14) AD 983 Mar 1/2 [Friday] (mag. = 1.07): Cairo

This lunar eclipse was in the month of Ramadan in the year 372 of
al-Hijrah on the night whose morning was Friday, the 15th of the
month... (date on Persian calendar)... The Moon was totally eclipsed. The
altitude of the Moon when the eclipse became perceivable was 66 deg. The
altitude when the Moon cleared completely was 35 plus ^ and | (i.e. 35;5O
deg); the Moon was totally dark for about an hour... (calculated details)...
The recorded altitude of the Moon at first contact slightly exceeds the
meridian altitude of 65.8 deg. However, if the Moon was near the meridian
at first contact, this could still represent a good measurement, although
of course a value of AT cannot be derived.

RESULTS
Altitude of Moon at last contact = 35.8 deg in the east. Equivalent LT
(Mar 2) - 3.43 h, UT = 1.55 h. Computed TT = 1.87 h, thus AT = 1200
sec.
 13.8 Lunar eclipse times recorded by Ibn Yunus 485

(15) AD 986 Dec 18/19 [Sunday] (mag. = 0.91): Cairo

This lunar eclipse was on the night whose morning was Sunday, the
15th of (the month of) Sha'ban in the year 376 of al-Hijrah. The eclipse
became noticeable when the altitude of the Moon was 24 deg in the west. I
estimated the (first) contact was when the altitude was 50^ deg. About 10
digits of the Moon's diameter were eclipsed. The observation was in the
Mosque of Abu Ja'far Ahmad ibn Nasr al-Maghribi at al-Qarafa in the
presence of Abu Ahmad ibn 'Asim and *Abd al-Rahman ibn Isa ibn
Tabyan. The Moon set eclipsed.

In using the Arabic alphabet for numerals to express altitude in degrees,

the 50 of the text could possibly be read as 30, which is much closer to
the altitude when the eclipse first 'became noticeable'. If so, there would
still be a delay of about half an hour between true first contact and
the observed start. Although this could be explained by cloud, there is
no mention of unfavourable weather in the record. It thus seems best
to consider only the original measurement of 24 deg. Since Ibn Yunus
judged that he was late in detecting first contact, it will be assumed that
this latter value is only approximate.

RESULTS
Altitude of Moon at first contact ~ 24 deg in the west. Equivalent LT
(Dec 19) - 4.89 h, UT - 2.84 h. Computed TT = 3.07 h, thus AT - 800
sec.
NB an earlier moment for the detection of the eclipse would, of course,
lead to a higher result for AT.

(16) AD 990 Apr 12/13 [Sunday] (mag. = 0.74): Cairo

This lunar eclipse was on the night whose morning was Sunday the 16th
of (the month of) Muharram in the year 380 of al-Hijrah. l\ digits of the
Moon's diameter were eclipsed, as I guessed. The Moon cleared when the
ascendant was the beginning of Aquarius. The altitude of the Moon when
the eclipse began, I mean at the time of contact, was 38 deg.

It is not mentioned whether the Moon was east or west of the meridian
at first contact. For an altitude of 38 deg, the corresponding LTs are
respectively 21.80 h and 2.20 h. Since the latter time is later than the time
of last contact as inferred from the rising of the beginning of Aquarius
(approximately 1.3 h) the Moon must have been in an easterly azimuth
when the eclipse began.
The way in which the time of end is expressed is without parallel
among medieval Arabic eclipse records and presumably was derived from
a separate measurement. Since the nature of this measurement cannot
be established, and the phrase 'the beginning of Aquarius' may represent
486 13 Observations of eclipses by medieval Arab astronomers

only an approximate celestial longitude, this result will not be used to

determine AT. (NB Aquarius extends from long. 300 to 330 deg.)

RESULTS
Altitude of Moon at first contact = 38 deg in the east. Equivalent LT
(Apr 12) = 21.80 h, UT = 19.70 h. Computed TT = 20.69 h, thus AT =
3500 sec.
(17) AD 1001 Sep 5/6 [Saturday] (mag. = 0.86): Cairo
This lunar eclipse was in (the month of) Shawwal in the year 391 of
al-Hijrah at the start of the night of Saturday, the 14th of the
month... (date on Persian calendar)... The Moon cleared when about 2
seasonal hours of night had elapsed (i.e. after sunset). I saw the Moon
before its clearance and it was like a crescent.
It is not clear whether the estimate of 2 hours is only approximate, but
this will be assumed.

RESULTS
(i) Last contact ~ 2.00 h after sunset on Sep 5. LT of sunset = 18.24 h.
2 h unequal = 1.92 h. Hence LT of contact - 20.12 h, UT ~ 18.01 h.
Computed TT = 18.17 h, thus AT - 600 sec.
(18) AD 1002 Mar 1/2 [Monday] (mag. = 1.44): Cairo
This lunar eclipse was on the night whose morning was Monday the
15th of the month of Rabi' al-Akhir in the year 392 of al-Hijrah... (date on
Persian calendar)... The Moon was totally eclipsed and had a staying
(al-makth: i.e. in totality or in darkness). The eclipse began when the
altitude of (the star) (al-simak) al-ramih (Arcturus: a Boo) was 12 (or 52)
deg east and when the altitude of (the star) al-hadi (Capella: a Aur) was
14 deg in the west. The altitude of (the star) (al-simak) al-ramih at the
complete clearance was 35 deg.
There is evidence of scribal errors in recording both altitudes of a Boo.
The elevation of this star at first contact (implying an LT of 20.25 h)
is incompatible with that of a Aur at the same moment (LT = 23.55
h). However, the symbols for the numbers 12 and 52 are so similar that
confusion is not uncommon - as noted by Caussin (1804). Reading 52
deg for the altitude of a Boo leads to a LT of 23.48 h, which is in fairly
good accord with that derived from the a Aur measurement. This altitude
will thus be adopted.
Assumption of an altitude for a Boo of 35 deg in the west at the end of
the eclipse implies an LT of 6.41 h. This is about 40 minutes after sunrise,
so that the star would be invisible; further, an excessive duration for the
eclipse of nearly 7 h would be implied. In this case, it is not possible to
satisfactorily restore the altitude measurement at last contact.
 13.8 Lunar eclipse times recorded by Ibn Yunus 487

Table 13.3 AT results from lunar eclipse timings

recorded by Ibn Yunus.

Year Ct AT
+854a 1 -3150
+854b 1 2500
+854b 2 -1350
+856 1 2350
+923 M 1200

+923 4 2000
+925 1 2400
+925 4 2350
+927 1 2950
+933 1 2250

+979a 4 -1650
+979b 1 1900
+979b 4 1300
+980 4 -2100
+981a 1 -2000

+981a 4 -2350
+981b 1 -1500
+983 4 1200
+986 1 -800
+990 1 3500

1001 4 -600
1002 1 1750
1002 1 1950

RESULTS
(i) Altitude of a Aur at first contact = 14 deg in the east. Equivalent LT
(Mar 1) = 23.55 h, UT = 21.67 h. Computed TT = 22.15 h, thus AT =
1750 sec.
(ii) Altitude of a Boo at first contact = 52 deg in the east. Equivalent
LT = 23.48 h, UT = 21.61 h. Computed TT = 22.15 h, thus AT = 1950
sec.
The AT results obtained in this section are listed in table 13.3, and are
plotted in figure 13.7.
Comparison between figure 13.7 and figure 13.6 shows that the scatter
is considerably greater for the lunar data. A partial explanation is poorer
488 13 Observations of eclipses by medieval Arab astronomers

4000 -

3500 -

3000 -

_ 2500
o
3 •
^ 2000

1500

1000

500 -

0
+800 +820 +840 +860 +880 +900 +920 +940 +960 +980+1000+1020+1040
Year
Fig. 13.7 AT values derived from lunar eclipse timings recorded by Ibn Yunus.

resolution of the contacts, perhaps combined with the greater difficulty of

reading an instrument at night.

13.9 Solar and lunar eclipse timings recorded by al-Battani and al-Biruni
The following translations are from Said and Stephenson (1997) except
where otherwise stated. Observations by al-Battani (AD 883-901) are
taken from his al-Zij al-Sabi; those by al-Biruni are extracted from both
his al-Qanun al-Mas'udi (AD 1003-4) and his Kitab Tahdid al-Amakin
H-Tashih Masafat al-Masakan (AD 1019 only). Apart from the very last
observation, only the times of maximal phase are preserved. Dates cited
by al-Battani are given in terms of the Syrian calendar, each year of which
began on 1 October of the Julian calendar.

(1) AD 883 Jul 23/24 [Wednesday] (lunar, mag. = 0.95): al-Raqqah

This lunar eclipse was observed by us at the city of al-Raqqah on the
23rd of (the month of) Tammuz in the year 1194 of Dhu al-Qarnayn ('the
Two-horned', i.e. Alexander IV of Macedonia), which is the year 1206 of
al-Iskandar (i.e. Alexander the Great). The middle of the eclipse was at a
little more than 8 equal hours after midday. A little more than \ plus | (i.e.
| or 10 digits) of the Moon's diameter was eclipsed... (calculated details)...
[Al-Battani.]
 13.9 Eclipse timings recorded by al-Battani and al-Biruni 489

The reference to Alexander IV of Macedonia is evidently an indirect

allusion to the Seleucid era. Both the year 1194 from the Seleucid
era (adopted as 1 Oct in BC 312) and the year 1206 from the death
of Alexander the Great (323 BC) correspond to AD 883/4. Tammuz
(equivalent to July on the Julian calendar) was the tenth month of the
year.
The 'little more than 8 equal hours' will be assumed to be 8.1 h.
Al-Battani expresses most times to the nearest \ or \ of an hour.

RESULTS

LT of mid-eclipse - 20.10 h, UT - 17.57 h. Computed TT = 17.82 h,

thus AT - 900 sec.

(2) AD 891 Aug 8 [Sunday] (solar, mag. = 0.89): al-Raqqah

This solar eclipse was observed by us at the city of al-Raqqah on the
8th of (the month of) Ab in the year 1202 of Dhu al-Qarnayn, which is the
year 1214 after the death of al-Iskandar. The middle of the eclipse was at
one seasonal hour after midday. (A little) more than \ of the Sun (i.e of
surface) was eclipsed in view... (calculated details)...
[Al-Battani.]
Ab (corresponding to August) was the 11th month of the year.

RESULTS

1 unequal hour = 1.14 h. Hence LT of mid-eclipse ~ 13.14 h, UT ~ 10.59

h. Computed TT = 11.06 h, thus AT = 1700 sec.

(3) AD 901 Jan 23 [Friday] (solar, mag. = 0.67): Antakyah

This solar eclipse was observed by us at the city of Antakyah on the
23rd of (the month of) Kanun al-thani in the year 1212 of Dhu al-Qarnayn
which is the year 1224 after the death of al-Iskandar. The middle of the
eclipse was about 3 | equal hours before midday. (A little) more than \ of
the Sun (i.e Sun's surface) in sight was eclipsed (calculated details)...
[Al-Battani.]
Kanun al-thani (corresponding to January) was the fourth month of
the year.

RESULTS

LT of mid-eclipse ~ 8.33 h, UT - 6.17 h. Computed TT = 6.63 h, thus

AT - 1650 sec.
490 13 Observations of eclipses by medieval Arab astronomers

(4) AD 901 Jan 23 [Friday] (solar, mag. = 0.68): al-Raqqah

... This eclipse was observed by someone on our behalf at the city of
al-Raqqah (on the same date as above). The middle of the eclipse was (a
little) less than 3 | equal hours before midday. (A little) less than | of the
Sun (i.e. Sun's surface) in view was eclipsed... (calculated details)...
[Al-Battani.]

This is a continuation of the previous text. The interval of '3 hours and
less than half an hour equal' will be assumed to be 3.4 h.

RESULTS

LT of mid-eclipse ~ 8.60 h, UT ~ 6.26 h. Computed TT = 6.69 h, thus

AT - 1550 sec.

(5) AD 901 Aug 2/3 [Sunday] (lunar, mag. = 1.06): Antakyah

This lunar eclipse was observed by us at the city of Antakyah on the
2nd of (the month of) Ab in the year 1212 of Dhu al-Qarnayn, which is the
year 1224 from the death of al-Iskandar. The middle of the eclipse was at
approximately 15 plus | (i.e 15;20 h) equal hours after midday. The Moon
was eclipsed by less than its diameter by a small amount... (calculated
details)...
[Al-Battani.]

Although this eclipse was recorded as marginally partial, it was actually

total according to computation.

RESULTS

LT of mid-eclipse ~ 3.33 h on Aug 3, UT ~ 0.98 h. Computed TT = 1.16,

thus AT - 650 sec.

(6) AD 901 Aug 2/3 [Sunday] (lunar, mag. = 1.06): al-Raqqah

This eclipse was observed (by someone on our behalf) at the city of
al-Raqqah (at the same date as above). The middle of the eclipse was
approximately 15 plus | and ^ (i.e. 15;35 equal hours) after
midday... (calculated details)...
[Al-Battani.]

This is a continuation of the preceding account.

RESULTS

LT of mid-eclipse ~ 3.58 h on Aug 3, UT ~ 1.04 h. Computed TT = 1.16,

thus AT - 450 sec.
 13.9 Eclipse timings recorded by al-Battani and al-Biruni 491

(7) AD 1003 Feb 19/20 [Saturday] (lunar, mag. = 0.14): Jurjan

This lunar eclipse was on the night of Saturday the 14th of the month
of Rabi' al-Akhir in the year 393 (of al-Hijrah). I observed the beginning
and clearance at Jurjan by the altitude of the (two stars) al-ShVrayan (i.e.
al-shira al-yamaniyyah - Sirius: a CMa - and al-shi'ra al-shamiyyah -
Procyon: a CMi). The Moon was eclipsed by \ of its diameter by estimate.
The longitude difference between Jurjan and Ghaznah is 2;21 minutes of
day. The middle of the eclipse at it (presumably at Ghaznah) was 19;11
(minutes of day) after midday of Friday... in the year 1751 of Buktinassar
(i.e. Nabonassar).
[Al-Biruni: al-Qanun]
In the following observation (AD 1003 Aug 14/15), al-Biruni quoted
only the measured time of mid-eclipse at Ghaznah and it will be presumed
that he did the same in this case.
A minute of day (daqa'iq al-Ayyar) - i.e. 1/60 of a day - was equivalent
to 0.4 of an equal hour. Hence mid-eclipse was 7.67 h after noon at
Ghaznah.

RESULTS
LT of mid-eclipse at Ghaznah = 19.67 h on Feb 19, UT = 15.35 h.
Computed TT = 15.75 h, thus AT = 1450 sec.
NB If the LT of mid-eclipse (7.67 h after noon) related to Jurjan, the
corresponding value for AT would be —1900 sec, an obviously errant
result.

(8) AD 1003 Aug 14/15 [SundayJ (lunar, mag. = 0.14): Jurjan

This lunar eclipse was on the night of Sunday the 13th of (the month
of) Shawwal in the year 393 (of al-Hijrah). I observed it at Jurjan by the
altitudes of (the two stars) al-nasran (i.e. 'the two eagles': al-nasr al-tair -
Altair: a Aql - and al-nasr al-waqi' - Vega: a Lyr) and (the star) al-'ayyuq
(i.e. Capella: a Aur). The middle of the eclipse at Ghaznah occurred when
more than | of its diameter was eclipsed and that was 31 ;21 (minutes of
day) after midday of Saturday... in the year 1751 (of Nabonassar).
[Al-Biruni: al-Qanun]
It seems curious that despite al-Biruni's mention of the star altitudes
which he made at Ghaznah, he does not quote any LTs for that city in
both this and the previous entry.
Mid-eclipse was 12.54 h after noon at Ghaznah.

RESULTS
LT of mid-eclipse at Ghaznah = 0.54 h on Aug 15, UT = 20.01 h.
Computed TT = 20.20 h, thus AT = 700 sec.
492 13 Observations of eclipses by medieval Arab astronomers

(9) AD 1004 Jul 4/5 [Wednesday] (lunar, mag. = 0.17): Jurjaniyyah

This lunar eclipse was on the night of Wednesday the 14th of the month
of Ramadan in the year 394 (of al-Hijrah). I observed its middle at
Jurjaniyyah of Khwarizm and found it to be 36;32 (minutes of day) after
midday of Tuesday... in the year 1752 (of Nabonassar). Ghaznah is east of
Jurjaniyyah by 1;42,12 (minutes of day).
[Al-Biruni: al-Qanun]

In this instance, al-Biruni did not mention the clock stars which he used
to determine the LT of observation.
It seems most likely that in this instance the place to which the LT refers
is Jurjaniyyah; the allusion to Ghaznah would appear to be incidental.
(The longitude difference between Ghaznah and Jurjaniyyah according to
al-Biruni is equivalent to 0.68 h.) Mid-eclipse was observed 14.61 h after
noon at Jurjaniyyah.

RESULTS

LT of mid-eclipse at Jurjaniyyah = 2.61 h on Jul 5, UT = 22.72 h.

Computed TT = 23.45 h, thus AT = 2600 sec.

(10) AD 1019 Sep 16/17 [Thursday] (lunar, mag. = 0.76): Ghaznah

This lunar eclipse was (seen) at Ghaznah in (the month of) Jumada
al-Ula, in the year 410 (of al-Hijrah). I observed it and (found that) at the
moment when the indentation at (the edge of) the full Moon became
noticeable the altitude of (the star) al-'ayyuq (Capella: a Aur) from the
east was slightly less than 66 deg, that of (the star) al-shVra al-yamaniyyah
(Sirius: a CMa) was 17 deg, that of (the star) al-shamiyyah (Procyon: a
CMi) was 22 deg, and that of (the star) aldabaran (Aldebaran: a Tau) was
63 deg; all in the east. All these (altitude measurements) necessitate that
the beginning of the eclipse would be when approximately 8 hours of night
had elapsed (i.e. after sunset)... (calculated details)... It was clear to the
sight that the world was lit up, the stars had disappeared, the Sun was
about to rise, and the Moon was about to set behind the mountains which
screened it. A small portion of the eclipse (still) remained in its body (i.e.
disk) and I was unable to observe it (the time of completion of clearance)
exactly.

[Al-Biruni: Kitab]

All four star altitudes refer to first contact. It will be assumed that the
altitude of Capella (a Aur) was 65.7 deg. Since visibility of the eclipse was
eventually interrupted by mountains, last contact could not be observed
from Ghaznah.
 13.9 Eclipse timings recorded by al-Battani and al-Biruni 493

Table 13.4 AT results from solar and lunar eclipse timings

recorded by al-Battani and al-Biruni.

Year Type Ct AT
+883 Moon M -900
+891 Sun M 1700
+901 Sun M -1650
+901 Sun M -1550
+901 Moon M -650
+901 Moon M -450

1003a Moon M 1450

1003b Moon M 700
1004 Moon M 2600
1019 Moon 1 1900
1019 Moon 1 1700
1019 Moon 1 1800
1019 Moon 1 1600

RESULTS

(i) Altitude of a Aur at first contact = 65.7 deg in the east. Equivalent LT
= 2.29 h, UT = 21.60 h. Computed TT = 22.13 h, thus AT = 1900 sec.
(ii) Altitude of a CMa at first contact = 1 7 deg in the east. Equivalent
LT = 2.34 h, UT = 21.65 h. Computed TT = 22.13 h, thus AT = 1700
sec.
(iii) Altitude of a CMi at first contact = 22 deg in the east. Equivalent
LT = 2.32 h, UT = 21.63 h. Computed TT = 22.13 h, thus AT = 1800
sec.
(iv) Altitude of a Tau at first contact = 17 deg in the east. Equivalent
LT = 2.38 h, UT = 21.69 h. Computed TT = 22.13 h, thus AT = 1600
sec.
These four results are fairly self-consistent.

The various AT values from this section are assembled in table 13.4.
These values are plotted in figure 13.8, which also shows the limits obtained
later in this chapter from horizon observations - see section 13.11. It is
evident that apart from the individual measurements made at the eclipse
of AD 1019, the scatter in these results is significantly greater than for the
data from Ibn Yunus.
494 13 Observations of eclipses by medieval Arab astronomers

Key
o Solar results
4000 - • Lunar results
y\ Lower limit
- v Upper limit

3000 - -

2000
0
o
0

1000
•

0
+800 +820 +840 +860 +880 +900 +920 +940 +960 +980+1000+1020+1040
Year

Fig. 13.8 AT values derived from eclipse timings recorded by al-Battani and
al-Biruni along with AT limits derived from horizon observations.

13.10 Solar magnitude estimates

The solar eclipse records cited in sections 13.7 and 13.9 above contain
a number of estimates of magnitude. Most of these are expressed as a
fraction of the apparent diameter of the Sun, sometimes accompanied by
the surface equivalent, but in one case (AD 928) only a surface magnitude
appears to be given. In this year it is stated that:

... The Sun rose (already) eclipsed by less than one \ of its surface. The
eclipse continued to increase by an amount that we could perceive until \
(of its surface) was eclipsed...

It seems a reasonable presumption that the peak magnitude was also

expressed in terms of the visible surface of the Sun. From table 3.1
(chapter 3), it may be seen that the equivalent linear magnitude would
be 0.36.
Where both linear and surface magnitudes are quoted, I have adopted
only the former result. At the eclipses of AD 891 and 901 (two separate
observations were made in the latter year), the magnitude was said to
be (i) > | ; (ii) > \\ and (iii) < \ (see section 13.9). As the observers
were familiar with the use of digits, I have taken these estimates to be
 13.10 Solar magnitude estimates 495

Table 13.5 AT results from medieval Arab estimates of solar eclipse

magnitude.

Date AD Place Mag. est. AT

891 Aug 8 al-Raqqah >§ 50
901 Jan 23 Antakyah >\ 5500
901 Jan 23 al-Raqqah <\ 5850
3
923 Nov 11 Baghdad 4 2700
928 Aug 18 Baghdad \ surf. —

8
977 Dec 13 Cairo 12 2900
5.5
978 Jun 8 Cairo 12 500
5.5
979 May 28 Cairo 12 —
1
985 Jul 20 Cairo 4
850
993 Aug 20 Cairo f surf. -8050

as follows: (i) between 8 and 9 twelfths (i.e. between \ and | ) ; (ii) and
(iii) j2 (i-e* between \ and | ) . I have not made any amendments to
the observed magnitudes to allow for possible systematic errors. There
is sound evidence for such errors in the case of lunar obscurations - see
chapter 7 - but not for solar eclipses.
As discussed in chapter 3, unless it is clear whether the upper or lower
limb of the Sun was covered, in principle an observation of magnitude can
lead to two separate values for AT - typically some 20000 sec apart. Since
numerous contemporary measurements (eclipse timings, etc.) indicate that
AT at this period was of the order of 2000 or 3000 sec, usually only one
solution need be considered.
The observations and AT results are summarised in table 13.5. Full
(AD) dates are cited in each case since translations of individual records
are not given in this section.
No value of AT can satisfy the estimates of magnitude in AD 928 (0.37)
or 979 (0.46). The geographical circumstances are such that the actual
magnitudes cannot have exceeded 0.33 and 0.42 respectively. The AT
value of —8050 sec obtained in AD 993 (the only possible result on this
date for a magnitude of 0.67) is clearly anomalous. Not only is the AT
figure itself highly discordant, it also leads to a computed solar altitude at
greatest phase of 60 deg rather than the observed figure of 45 deg. Clearly
the scatter of the remaining AT values in table 13.5 is considerably larger
than that obtained from timed data (tables 13.2 to 13.4). It is apparent
496 13 Observations of eclipses by medieval Arab astronomers

that the use of magnitude estimates to determine AT is scarcely justifiable

at this relatively late period (see also chapters 5 and 9). Accordingly, the
results in table 13.5 will not be considered in subsequent analysis.

13.11 Rising and setting phenomena

Among the observations discussed in sections 13.7 to 13.9 are six instances
where it is stated that the Sun or Moon rose or set whilst eclipsed. Five
of these observations, ranging in date from AD 923 to 986, were made
at either Baghdad or Cairo. In Baghdad, the terrain is flat and the
horizon is extremely level. This is true also of Cairo except towards the
south-east, where the horizon is screened by hills. A remarkably vivid
description of a moonset eclipse in AD 1019 was reported from Ghaznah
by al-Biruni (see section 13.9). Unfortunately, this city is located in a
very mountainous region so that an accurate horizon profile would be
needed to make proper use of the observation. Hence only the five earlier
records from Baghdad and Cairo will be considered here, along with a
sixth observation from Cairo in AD 981 which states that a lunar eclipse
ended not long before sunrise.
In two instances (AD 923 and 928), an estimate of the degree of
obscuration of the Sun or Moon when on the horizon is clearly given. On
the former occasion, it is stated that the Moon rose already rather more
than \ covered, the magnitude eventually reaching 9 digits. Hence this
observation can be utilised to obtain an independent value for AT. In
AD 928, the fraction of the solar disk said to be covered at sunrise (less
than \) is very close to the observed peak magnitude of \. Hence on this
occasion a useful result for AT cannot be obtained.
Extracts from the relevant texts - all quoted in full in sections 13.7 and
13.8 - are given below along with comments and the derived AT limits.
All six records are taken from the zij of Ibn Yunus.

(1) AD 923 Jun 1/2 (lunar, mag. = 0.66): Baghdad

... The Moon rose at sunset (already) eclipsed by \ or (a little) more of

the digits of the diameter (i.e. 3 digits or a little more). The Moon was
eclipsed by (a little) more than 9 digits of diameter...

In order to determine limits to AT, it will be assumed that the Moon

rose between first contact and mid-eclipse. However, the estimate of the
fraction of the Moon obscured as it reached the horizon leads to a separate
determination of AT.
 13.11 Rising and setting phenomena 497

RESULTS

(1) First contact before moonrise on Jun 1. LT of moonrise = 19.07 h.

Hence LT of contact < 19.07 h, U T < 16.04 h. Computed TT = 16.74 h,
thus A T > 2500 sec.
(ii) Mid-eclipse after moonrise. LT of moonrise = 19.13 h. Hence LT
of contact > 19.13 h, U T > 16.10 h. Computed TT = 18.12 h, thus A T
< 7250 sec.
Combining these limits yields 2500 < A T < 7250 sec.
(iii) If one assumes between 3 and 4 twelfths - a mean of 0.29 - and
adjusts for the likely systematic error made by the unaided eye in assessing
the magnitude of a lunar eclipse (see chapter 7), the estimated proportion
of the Moon's diameter actually in shadow at moonrise is 0.22. For 0.22
of the lunar diameter to be obscured at moonrise (phase increasing), A T
= 3450 sec.

(2) AD 928 Aug 18 (solar, mag. = 0.25): Baghdad

... The Sun rose (already) eclipsed by less than one \ of its surface. The
eclipse continued to increase by an amount that we could perceive until \
(of its surface) was eclipsed...

It may be inferred that the Sun rose at some time between first contact
and maximal phase.

RESULTS

(i) First contact before sunrise on Aug 18. LT of sunrise = 5.44 h. Hence
LT of contact < 5.44 h, UT < 2.50 h. Computed TT = 2.57 h, thus AT
> 250 sec.
(ii) Maximal phase after sunrise. Hence LT of contact > 5.44 h, UT >
2.50 h. Computed TT = 3.31 h, thus AT < 2900 sec.
Combining these limits yields 250 < AT < 2900 sec.

(3) AD 979 May 14/15 (lunar, mag. = 0.70): Cairo

... The Moon rose eclipsed... More than 8 digits but less than 9 of the
Moon's diameter were eclipsed... The eclipse cleared when about one equal
hour and a fifth of night had elapsed, as I estimated....

As the horizon at Cairo is somewhat hilly towards the south-east,

accurate limits to AT cannot be derived without considering the horizon
profile in detail. However, making no allowance for this effect leads to
such wide limits for AT (-4100 < AT < 5750 sec) that attempts to refine
this solution would not be worthwhile. The observation will thus not be
considered further.
498 13 Observations of eclipses by medieval Arab astronomers

(4) AD 979 May 28 (solar, mag. = 0.45): Cairo

...The eclipse was perceptible when the altitude of the Sun was 6| deg.
About 51 digits of the Sun's diameter were eclipsed, as I estimated, that is
4; 10 digits of surface. The Sun set eclipsed...

The computed magnitude is very close to the observed value (0.46). The
fact that the solar altitude was estimated as only 6.5 deg (in the west) at
first contact suggests that this eclipse had not reached maximum phase
when the Sun set. This altitude measurement leads to a very average figure
for AT of 1450 sec (as deduced in section 13.7) and on this assumption
(or the AT value of 1720 sec indicated by equation (8.1)) the computed
depression of the Sun below the horizon at last contact would be about
12 deg. Hence it seems reasonable to assume that the Sun reached the
western horizon at some time between first contact and maximal phase.

RESULTS
(i) First contact before sunset on May 28. LT of sunset = 18.95 h. Hence
LT of contact < 18.95 h, UT < 16.80 h. Computed TT = 16.52 h, thus
AT > -1000 sec.
(ii) Maximum phase after sunset. LT of maximum > 18.95 h, UT >
16.80 h. Computed TT = 17.48 h, thus AT < 2450 sec.
Combining these limits yields -1000 < AT < 2450 sec.
NB the latter figure can be only approximate.

(5) AD 981 Apr 21/22 (lunar, mag. = 0.18): Cairo

... About one-quarter of the Moon's diameter was eclipsed. The Moon
cleared completely when about \ of an hour remained to sunrise.

Sunrise (LT = 5.40 h) would be marginally later than moonset. Hence

the observation that last contact occurred before the Moon reached the
horizon (implicit in the text) yields a slightly more critical limit for AT.

RESULTS
Last contact before moonset on Apr 22. LT of moonset = 5.39 h. Hence
LT of contact < 5.39 h, UT < 3.25 h. Computed TT = 3.67 h, thus AT
> 1500 sec.
Only a single (lower) limit to AT may be derived on this occasion.

(6) AD 986 Dec 18/19 (lunar, mag. = 0.91): Cairo

... The eclipse became noticeable when the altitude of the Moon was 24
deg in the west.. .About 10 digits of the Moon's diameter were
eclipsed... The Moon set eclipsed.
 13.12 Conclusion 499

Table 13.6 AT limits from eclipses observed by Arab astronomers.

AT Range (sec)
Year Type Observation LL UL
+837 Sun annular 1820 3750
+923 Moon moonrise 2500 7250
+928 Sun sunrise 250 2900
+979 Sun sunset -1000 2450

+981 Moon moonset 1500

+986 Moon moonset -6450 5450
1004 Sun partial 1940 1770

The text does not state clearly whether the eclipse reached a maximum
before moonset or if the phase was still growing when the Moon reached
the horizon. Moonset at some time between first and last contact will be
assumed.

RESULTS
(i) First contact before moonset on Dec 19. LT of moonset = 6.92 h.
Hence LT of contact < 6.92 h, UT < 4.87 h. Computed TT = 3.07 h, thus
AT > -6450 sec.
(ii) Last contact after moonset. LT of moonset = 7.01 h. Hence LT of
contact > 7.01 h, UT > 4.97 h. Computed TT = 6.48 h, thus AT < 5450
sec.
Combining these limits yields —6450 < AT < 5450 sec.

The AT limits (lower = LL, upper = UL) obtained from the five
selected eclipses are listed in table 13.6. This table also includes the AT
ranges derived earlier (section 13.6) from the observation of an annular
solar eclipse in AD 873 and a large partial eclipse in AD 1004.
The AT limits in table 13.6 lying between 0 and +4000 sec, along with
the single AT value obtained from the estimate of phase at moonrise in
AD 923 (AT = 3450 sec), are plotted in figure 13.8. This diagram also
displays the few AT results obtained from solar and lunar eclipse timings
recorded by al-Battani and al-Biruni (section 13.9).

13.12 Conclusion
With the AT results obtained in this chapter from observations by me-
dieval Arab astronomers, we come to the end of our compilation of data.
Apart from solar eclipse magnitudes - discussed in section 13.10 - the sets
500 13 Observations of eclipses by medieval Arab astronomers

4500

4000

3500

3000

'o 2500

< 2000

1500

1000

500

+800 +820 +840 +860 +880 +900 +920 +940 +960 +980+1000+1020+1040
Year

Fig. 13.9 AT values and limits derived from medieval Arab eclipse observations.

of both individual AT values and limits derived from this material are
unusually important for the study of the Earth's past rotation. The various
results in tables 13.2, 13.3, 13.4 and 13.6 - along with an isolated value
for AT derived in section 13.11 from an estimate of phase at moonrise
(i.e. 3450 sec at AD 923) are plotted in figure 13.9.
In this diagram, as previously, open circles represent solar eclipse tim-
ings and shaded circles lunar eclipse timings. Arrow heads denote upper
and lower limits to AT derived from rising and setting events. The full
vertical line indicates the range in AT determined from the annular eclipse
of AD 873 and the broken line the two possible ranges obtained from the
partial eclipse of AD 1004.
 14
Determination of changes
in the length of the day

14.1 Introduction
In chapters 4-13, more than 400 timed and untimed observations of both
solar and lunar eclipses from the pre-telescopic period (date range —700
to +1600) have been investigated in detail. AT values or limits (depending
on whether the observations were timed or untimed) have been derived in
almost every case. The fundamental objective of the present chapter is to
use these results to obtain the best-fitting AT curve to the data and hence
to determine changes in the length of the day (LOD) over the historical
period. In addition, the various geophysical mechanisms responsible for
the observed variations will also be discussed. I am grateful to Dr L. V.
Morrison of the Royal Greenwich Observatory for undertaking the data
analysis which forms the basis of much of this chapter and producing
several of the diagrams. This chapter is essentially an enlargement of
section 5 of the paper by Stephenson and Morrison (1995). Although
several tens of further historical data have been added (notably medieval
Chinese timing - see chapter 9), the basic conclusions obtained in that
paper remain unchanged.
All of the AT results derived in chapters 4-13 are summarised in
tabular form in the Appendices for ready reference. For each individual
observation, only the year (— or +) and appropriate AT value or limits
are tabulated. Appendix A contains timed data and Appendix B untimed
material, each being divided into sub-groups depending on the source
and type of observation it contains. An outline of the contents of each
sub-group is as follows:

Al. Babylonian timings of solar and lunar eclipses (intervals < 25 deg).

A2. Babylonian timings of solar eclipses (intervals > 25 deg).

A3. Babylonian timings of lunar eclipses (intervals > 25 deg or approxi-

mate measurements < 25 deg + estimates of phase on horizon).

501
502 14 Determination of changes in the length of the day

A4. Chinese measurements of solar eclipse times to ^ day (+ estimates

of phase on horizon).
A5. Chinese measurements of lunar eclipse times to | night-watch (+
estimates of phase on horizon).
A6. Chinese measurements of lunar eclipse times to ^ day.
A7. Ancient Greek timings of lunar and solar eclipses.
A8. Arab timings of solar eclipses.
A9. Arab timings of lunar eclipses.
Bl. Untimed total and annular solar eclipses.
B2. Untimed partial solar eclipses (AT limits reversed).
B3. Sunrise/sunset or moonrise/moonset observations.
Estimates of the degree of obscuration of the Sun or Moon when on
the horizon are ranked with timed observations since they yield results of
similar form). The total numbers of timed and untimed data listed are
respectively about 305 and 105.

14.2 Timed and untimed observations

As these two groups form independent data sets, the AT values determined
from them can be considered separately. The results in Appendix A are
largely derived from individual timings by astronomers of the various
stages of solar or lunar eclipses. Since each measurement leads to a
specific value for AT (rather than a limit), the results can be treated
statistically. In the analysis of this material, the AT values in each group
will be assigned unit weight except for data in categories Al and A8.
The results derived from Babylonian records in which the measured time-
interval was less than 25 deg (group Al) and also Arab timings of solar
eclipses (A8) show an unusually high degree of self-consistency even when
the observations were said to be only approximate. Hence they will be
allotted double weight. Although the Arab observations originated from
several independent sources, it seems best to regard them as a unit.
None of the data in Appendix B are based on measurement. Here
the observers, many of whom were not astronomers, simply described
what they saw - sometimes in graphic terms. Untimed observations set
limits on AT. Unless the observations are questionable in any way, a
viable AT curve should not infringe these limits. Nevertheless, the form
of the results in each of the three categories Bl, B2 and B3 have certain
 14.2 Timed and untimed observations 503

mutual differences. Data in group Bl define ranges (solution space) of AT

anywhere within which the true value of this parameter is equally likely to
lie. Most AT ranges in this class are fairly narrow, seldom exceeding 2000
sec in width. Furthermore, the limits are extremely sharp; a total solar
eclipse, in particular, is a remarkably well-defined event. Observations in
category B2 define the inverse of the solution space in category Bl; in
each case an extensive range of AT is indicated (several tens of thousands
of arcsec in width) apart from a fairly narrow gap - up to about 2000 sec
wide - in which the solution cannot lie; otherwise the eclipse would be
fully total or annular instead of partial.
AT ranges in group B3 have similar properties to those in Bl, but are
much broader - typically between about 5000 and 10000 sec in width.
Because of this feature, many data in this category prove to be redundant.
Additionally, some limits are rather poorly defined - especially when it is
inferred that the Moon or Sun reached the horizon at mid-eclipse.
In figure 14.1 are plotted the AT ranges obtained from the untimed
data, as listed in Appendix B. A different symbolism from that adopted in
earlier chapters is used, as explained in the key to the diagram. Solution
space derived from a total or annular solar eclipse (Bl) is denoted by a
heavy vertical line. In the case of a partial solar obscuration (B2), the
solution space at that date is indicated by a pair of lines of medium width
- separated by a gap within which the value of AT must not lie. Extending
these lines to the upper and lower edges of the diagram tends to give them
undue prominence. Hence beyond the excluded zone only short lines
terminated by arrow heads are used; beyond the arrow heads any value
of AT is still acceptable. Finally, solution space deduced from a rising
or setting event (B3) is represented by a fine vertical line, with an arrow
head where appropriate. These various symbols replace respectively the
full lines, broken lines and isolated arrow heads of earlier diagrams. For
most of the partial solar obscurations, the solution space on one side of
the belt of totality or annularity proves to be redundant; it is obviated by
other roughly contemporaneous observations. These redundant solutions
are not shown in figure 14.1 in order to simplify the diagram. Certain
solar observations which lead to extremely narrow forbidden AT ranges
(in the years —187, —79, —1, +1333 and +1354) are entirely omitted, as
are limits obtained from horizon observations which do not contribute
useful constraints to the solution space for AT.
Figure 14.2 is similar to figure 14.1 but here the individual AT values
obtained from the timed data (Appendix A) are shown. Once again a
revised set of symbols is used, as defined in the key to the diagram;
this distinguishes observations in the various categories. In particular,
double-weighted results (from sub-groups Al and A8 of Appendix A) are
represented by heavy black symbols.
504 14 Determination of changes in the length of the day

20000 Untimed total, annular and partial solar eclipses

solar and lunar eclipses observed on the horizon
and modern occulations (inset)

Expected on the basis

15000 of tidal friction I Solar; total, annular
la limits

Solar and lunar horizon

observations
10000 -

5000- 200 -^

-500 2000

Fig. 14.1 AT ranges obtained from untimed total, annular and partial solar
eclipses and also solar and lunar eclipses on the horizon: —800 to +1600. Also
shown (inset) is the AT curve obtained from modern occultations. (Courtesy: Dr
L. V. Morrison.)

20000 -

-500 2000

Fig. 14.2 AT values obtained from timed solar and lunar eclipses. Also shown
(inset) is the AT curve obtained from modern occultations. (Courtesy: Dr L.V.
Morrison.)
 14.2 Timed and untimed observations 505

25-

20

15

10

-750 -500 -250 0 250 500 750 1000 1250 1500

Year
Fig. 14.3 AT ranges and individual values obtained from untimed and timed
solar and lunar eclipses: —800 to +1600. In this diagram, each allowed AT range
is indicated by a gap between two vertical lines; this convention in the opposite
to that adopted in figures 14.1, 14.4 and 14.5. (Courtesy: Dr R. S. Roberts).

Also shown in both figure 14.1 and figure 14.2 (inset, on a much
larger scale) is the AT curve since the year +1600, obtained mainly from
telescopic observations of occultations (Stephenson and Morrison, 1984).
This is depicted as an irregular full line. The nature of the four smooth
curves will be discussed below.
It is obvious from comparison of figures 14.1 and 14.2 that the untimed
and timed data define similar trends in AT over the past 2700 years. This
is particularly clear from figure 14.3, which displays both the timed and
untimed data. In this diagram, each individual AT value is represented
by a dot of equal size. Each allowed AT range as indicated by a total
or annular solar eclipse or a horizon observation is denoted by a gap
between two vertical lines. In the case of a partial solar eclipse, almost
the entire solution space is permitted except for that indicated by a short
vertical line. This method of depicting AT limits was suggested to me by
Dr R. S. Roberts of the University of Durham, to whom I am grateful
for producing figure 14.3. The convention adopted is the reverse of that
followed in figure 14.1.
In the following sections, the timed and untimed data will be used to
assess the viability of the following assumptions: (i) lunar and solar tidal
506 14 Determination of changes in the length of the day

friction only (section 14.3); (ii) tidal friction plus an additional non-tidal
component - assumed constant (section 14.4); (iii) tidal friction on which
a variable non-tidal component is superposed (section 14.5).

14.3 Lunar and solar tidal friction

As discussed in chapter 2, lunar and solar tides should produce an es-
sentially constant deceleration of the Earth's rotation over the last few
millennia. Hence if non-tidal mechanisms were insignificant, the resulting
AT curve would be parabolic. The analysis of perturbations of the orbits
of near-Earth artificial satellites by Christodoulidis et al. was shown in
chapter 2 to be commensurate with a quadratic term in AT of +44 + 2
sec/cy2. The pair of dotted curves in figures 14.1 and 14.2 represent this
result and its associated standard error. They have equations:
= 46£ 2 -20sec, (14.1)

= 42; 2 -20sec, (14.2)

where t is measured in Julian centuries from the epoch +1820 (rather
than +1800, as defined earlier). This is the approximate date at which the
average LOD is equal to the standard LOD on the TT scale (i.e. 86400
SI sec), as discussed in chapter 1. The constant (—20 sec) is included
to obtain a satisfactory fit to the telescopic curve shown in the insets to
figures 14.1 and 14.2; over the pre-telescopic period, this term is, of course,
negligible.
It is clear from figures 14.1 and 14.2 that the curve of best fit to the
pre-telescopic data (whether timed or untimed) lies far below the tidal
range; the discrepancy in AT amounts to about one hour by the epoch
+150 and two hours by —550. There is thus evidence of marked non-tidal
behaviour. The existence of this was first clearly shown by Stephenson
and Morrison (1984) from the analysis of smaller data sets than those
displayed in figures 14.1 and 14.2.

14.4 Constant non-tidal component

A much better approximation to the observational data in the pre-
telescopic period is provided by a solution which incorporates a quadratic
non-tidal term in AT in addition to the tidal parabola. On the (pro-
visional) assumption that the rate of increase in the LOD - both tidal
and non-tidal - has been constant over the past 2700 years, the parabola
represented by a broken line in figures 14.1 and 14.2 has been chosen
to satisfy the greatest number of boundary conditions set by the eclipse
 14.4 Constant non-tidal component 507

limits. This parabola has the equation:

= 31t 2 -20sec. (14.3)
Although the parabola is a remarkably good fit to most of the data
(including the telescopic observations) there are important discrepancies
- especially in the period from +700 to +1400. During this interval, the
eclipses of +761, +1133, +1221 and +1267 were all clearly described as
total and that of +1147 was recorded as annular. Yet in each case the AT
values yielded by equation (14.3) render the phase as only partial at the
appropriate place of observation. Indeed the observational limits to AT
all lie substantially below the parabola, as shown in figure 14.1. The dis-
crepancies (in chronological order) are —170, —290, —190,-110 and —110
sec. There are also three marked discrepancies among the results obtained
from partial eclipses: in +1178, +1330 and +1361. In each case, use of
equation (14.3) would indicate a total eclipse but the record specifically
denies this. Furthermore, the AT discrepancies (respectively —120, —170
and —150 sec) are of the same sign as for the central eclipses listed above.
Strong evidence for a systematic discrepancy between a parabolic rep-
resentation and the results of observation during the same interval is
provided by two independent sets of timed data. These are derived from
Arab and Chinese measurements in the approximate date range from
+800 to +1300. In this interval more than 80 per cent of all data points
lie below the parabola (see figure 14.2). The situation is more clearly
shown in figure 14.4, which depicts on a larger scale than in figure 14.2
both the timed and untimed data between the years +400 and +1600.
Around the epoch +500, there is a discrepancy between observation
and the mean parabolic fit (equation (14.3)) of opposite sign to that noted
above. Amounting to some +350 sec, this is mainly demonstrated by the
total solar eclipse of +454; a value for AT of at least 6130 sec would be
required to render this eclipse total at the Chinese capital of Chien-k'ang,
where, on account of the darkness 'all the constellations were brightly lit'
(see chapter 8). A figure for AT of only 5780 sec is yielded by equation
(14.3) at this date.
Near this same date, the Greek and Chinese timed observations (cover-
ing the period from +364 to +596 indicate a solution above the parabola,
most AT values lying above +5500 sec. Since there are only a few obser-
vations and the scatter in AT is rather large, the evidence is somewhat
weak. However, it should be mentioned that the two low data points
(AT = +1750 and +2800 sec) are obtained from two phases of the same
lunar eclipse - that of +434 - and are thus are not independent. Both
results are about an hour less than the mean for that date derived from
the remainder of the observations in this group. Possibly on this occasion
the clepsydra was malfunctioning or careless readings were taken.
508 14 Determination of changes in the length of the day

8000 P-TT
Timed and untimed o Chinese lunar eclipses (-night-watch); also
D
data:+400 to+1600 phase on horizon
x Chinese solar eclipses ( ^ day)

» Chinese lunar eclipses ( ^ day)

6000 • Arab solar eclipses
Arab lunar eclipses

Expected on the basis

of tidal friction
n
1<7 limits

2000

| Solar; total, annular

Solar; partial

I Solar and lunar horizon observations

2000, i i < I J i i I i i i _L
00 600 800 1000 1200 1400 1600
Year

Fig. 14.4 Combined timed and untimed data for the period from +400 to +1600.
(Courtesy: Dr L. V. Morrison.)

The accord between the parabola represented by equation (14.3) and

the observational results is closer in ancient times, but fewer AT limits
are critical. Figure 14.5 displays the limits set by the untimed data over
the period from —800 to +500 on an enhanced scale.
By choosing a different origin for the cusp of the parabola and varying
the quadratic term, it is possible to obtain a better fit to the set of pre-
telescopic data. However, this is at the expense of the modern observations.
Stephenson and Morrison (1995) found that the following parabola fitted
the ancient and medieval observations more closely than equation (14.3):
AT = 35T 2 -20sec,
with T measured in centuries from a cusp at the epoch +1735. However,
they remarked that this parabola does not 'remotely resemble the AT
curve between +1600 and the present'. In addition, it does not meet the
condition that the cusp lie near the beginning of the nineteenth century, in
keeping with the requirement that the average LOD at that epoch should
be equal to the standard LOD on the TT scale - see chapter 1.
It is thus apparent that no single parabola can satisfactorily represent
both the pre-telescopic and telescopic data, implying that the non-tidal
component of the Earth's spin is variable with time. The degree of
variability will be investigated in the following section by fitting a more
flexible curve than a parabola to the data.
 14.5 Variable non-tidal component 509

Untimed total, annular and partial solar

20000 K eclipses; solar and lunar eclipses on
the horizon -800 to +500
18000

16000
Expected on the basis
^ of tidal fraction
14000 1(7 limits

12000

10000
I Solar; total, annular

! Solar; partial
8000
| Solar and lunar horizon observations

6000
i i 1 i i i 1 i i i I
-800 -600 -400 -200 200 400
Year
Fig. 14.5 AT limits obtained from untimed total, annular and partial so-
lar eclipses and also solar and lunar eclipses on the horizon: —800 to +700.
(Courtesy: Dr L. V. Morrison.)

14.5 Variable non-tidal component

Before attempting to obtain a better fit to both the pre-telescopic and tele-
scopic results, the following limitations should be noted. In particular, it is
clear from figure 14.3 that the temporal distribution of the pre-telescopic
data is far from uniform. In summary, there are many observations (both
timed and untimed) between —600 and —50, scarcely any between —50
and +400, a small but useful number of data from +400 to +800, and
many from +800 to +1300. Between +1300 and +1600 - a period when
good data would have been especially helpful in extending the well-defined
telescopic curve backwards in time - there is only a mediocre selection
of observations (all untimed). Further, in the whole of the period back
to —500, the parabola represented by equation (14.3) does not deviate by
more than about 500 sec from observation. As a result of these circum-
stances, it is only feasible to use the pre-telescopic material to investigate
minor departures from a parabola on a time-scale of several centuries
to millennia. Hence in fitting a more flexible curve than a parabola to
the data, it should be tightly constrained to avoid introducing spurious
short-term fluctuations.
Under these circumstances, cubic splines provide the most effective
method of curve-fitting. Cubic splines consist of a number of cubic
510 14 Determination of changes in the length of the day

polynomial segments (each fitted by least-squares), joined end to end and

having continuity in both first and second derivatives at the joins (knots).
Continuity in derivatives provides a realistic model for changes in the
LOD. Further, by restricting the number of knots, it is readily possible
to constrain the degrees of freedom. On this basis, cubic splines with five
knots (at the epochs -200, +300, +1100, +1700 and +1990) were fitted to
the data plotted in figures 14.1 and 14.2 as follows. The AT results derived
from pre-telescopic eclipse timings in Appendix A were first weighted as
discussed in section 14.2. They were then combined with the modern
occultation data which were represented by normal points of high weight
at 50-year intervals between +1650 to +1900, and at 25-year intervals
in the present century. Finally, critical limits imposed by observations of
six solar eclipses were added with high weight. Four of the solar eclipses
were total: in —135 (see chapter 5), +454 (chapter 8), +761 (chapter 8)
and +1241 (chapter 11) - totality being vividly described in each case. A
further eclipse in —393 (see chapter 10) was partial, and one in +1567 (see
chapter 11) was annular-total. The resulting cubic spline curve is plotted
in figures 14.1, 14.2, 14.4 and 14.5 as a continuous curve, extending back
to the epoch —500.
The total eclipse in +454 and an observation of a partial eclipse in
+360 set close limits to the form of the AT curve at this epoch. The spline
cannot pass much above the lower boundary of +454 without infringing
the upper limit in +360. The account of the eclipse of +360 asserts that
at maximum phase the Sun had the appearance of a hook (see chapter 8).
Hence it may be concluded that the observers did not witness the annular
phase but only saw a large partial eclipse.
Despite the small number of critical limits, the spline curve since —500
satisfies the numerous constraints imposed by all the untimed solar and
lunar eclipses with only three exceptions: horizon observations of lunar
eclipses in —382, —239 and +1067. In —382, only a Greek transcription of
a Babylonian record is available (see chapter 4). This may have become
corrupted. In —239, Babylonian astronomers recorded a lunar eclipse
which began 'at 3 deg (i.e. 12 min) before sunrise,' adding that 'it set
eclipsed' (see chapter 7). For any value of AT less than 13 650 sec, the
Moon would set at Babylon before the eclipse began, but the spline fit
passes some 300 sec below this limit. Possible explanations are confusion
of the deep penumbral shadow with the umbra or anomalous refraction.
In +1067 Chinese astronomers recorded a lunar eclipse of 'more than 8
divisions. It set eclipsed...' (see chapter 9). As this eclipse was, in fact,
total, it would appear that as seen from the capital (Pien) the Moon set
before totality. The indicated range of AT < +750 sec corresponds to the
Moon setting just before the total phase. This result is in dispute with the
roughly contemporaneous lower limits set by the well-observed total solar
 14.6 Further discussion of timed observations 511

eclipses of +1061 (see chapter 12) and +1124 (chapter 11), and would
severely distort the form of the AT curve around this date. Possibly cloud
interfered with observation, although this is not mentioned in the record.
It should be stressed that the limits set by horizon observations are not
as reliable as for central and partial solar eclipses.

14.6 Further discussion of timed observations

In figures 14.6a and b are plotted histograms showing the residuals from
the spline curve for the Arab timings (date range +829 to +1019) and the
Chinese timings (+1040 to +1280). In figure 14.6b, five highly discordant
Chinese results are not represented. A bin of width 100 sec is used in each
diagram. The arrows denote the mean position of the parabola 3It2 — 20
sec (equation (14.3)). Clearly, the distribution of the Arab results (standard
deviation a = 550 sec) is significantly narrower than for the Chinese data
(a = 700 sec). However, the parabola is at variance with both datasets by
almost identical amounts (—500 sec: in the sense observed-computed).
In principle, this discrepancy of 500 sec could be explained by a system-
atic delay by this amount in the Arab and Chinese timings of both solar
and lunar eclipses. Nevertheless it seems most unlikely that it arises from
a data artefact. The Arab and Chinese astronomers employed quite dif-
ferent techniques for measuring time (respectively altitude determinations
and water clocks). Some delay in detecting the start of an eclipse might
be expected owing to the limited resolution of the unaided eye, although
since both sets of astronomers made approximate eclipse predictions, they
would be able to keep a watch for these events (see chapters 9 and 13). In
500 sec, the depth of the indentation at the solar or lunar limb increases
by up to 4 arcmin - far greater than the unaided eye limit (typically 1
arcmin). In addition, at last contact the opposite effect would be expected.
Since the observations from both Arab and Chinese sources are roughly
equally distributed between first contact, last contact and other phases
(principally mid-eclipse), any residual bias should be minimal. A real
variation in the Earth's rate of rotation is thus indicated.
Most of the Arab altitude measurements (whether of the Sun, Moon
or a reference star) were made to the nearest degree. Except when a
celestial body is near the meridian, at mid-latitudes its altitude changes
by this amount in about 6 min. For the solar eclipses in figure 14.6a, the
standard deviation is 6 min, so that these measurements are fairly accurate.
However, in the case of lunar eclipses a is as much as 11 min. In addition
to the poorer definition of the contacts for a lunar eclipse, the fact that
scales would then have to be read at night may have contributed to this
significant discrepancy. In the case of the Chinese measurements (figure
14.6b), a = 12 min (ignoring the five severely deviant results mentioned
512 14 Determination of changes in the length of the day

(a)

-2000 -1000 1000 2000

Residual (sec)

(b)

E 10 -

5 -

-2000 -1000 1000 2000

Residual (sec)

Fig. 14.6 Histograms of the residuals of the medieval timed data from the spline
curve: (a) Arab observations from +829 to +1019; (b) Chinese observations
from +1046 to +1280. The arrow indicates the mean position of the best-fitting
parabola 31t2. (Courtesy: Dr L. V. Morrison.)
 14.7 Changes in the LOD 513

above). This represents tolerable precision since times were only estimated
to the nearest 15 to 30 min.
From the frequency distribution of the residuals for all the Babylonian
timings of eclipses (not shown diagrammatically), the estimate for the
standard deviation is about 30 min. However, in the case of the restricted
set of lunar eclipses which occurred within 25 deg (1 h 40 min) of
sunset/sunrise (shown by black dots in figure 14.2), the standard deviation
is much smaller: 13 min, after the rejection of one outlier. These various
measurements were probably made made with a clepsydra, which although
presumably showing a considerable drift over several hours might be
expected to be more precise in measuring shorter intervals of time. All but
the very earliest lunar eclipses were timed to the nearest degree (4 min) -
see chapter 4 - so that the real accuracy of even the best measurements
was relatively poor. Nevertheless, the precision attained by the Babylonian
astronomers was not much inferior to that achieved more than 1000 years
later by the Arabs and Chinese.

14.7 Changes in the LOD

By taking the first derivative along the cubic spline in figures 14.1 and 14.2,
changes in the LOD (ALOD) relative to the standard reference day of
86 400 SI sec) may be derived. The resulting LOD variation is represented
by a continuous curve in figure 14.7. The irregular decade fluctuations in
the period from 1830 to the present are also shown; these are taken from
Jordi et al. (1994). Similar variations have been detected using earlier
telescopic observations (Stephenson and Morrison, 1984) but because of
the lower precision of measurement these are less well defined. It may be
presumed that decade fluctuations on a similar scale occurred throughout
the entire period covered by the historical data but the integrated effect of
these variations (i.e. their contribution to AT) is too small to be detectable
using observations made with the unaided eye.
In figure 14.7 are also shown the straight lines corresponding to the
effect of tidal friction alone (represented by a dotted line) and the mean
observed long-term trend in the LOD (broken line). The rate of increase
in the LOD owing to lunar and solar tides is +2.3 + 0.1 milliseconds per
century (ms/cy). As noted in chapter 1, there is no evidence to suggest
that tidal dissipation has altered significantly over the past 2700 years.
The observed mean rate of increase in the LOD (corresponding to the
parabolic fit to AT in figures 14.1 and 14.2) is substantially less: +1.7
ms/cy. In order to obtain an indication of the uncertainty in this latter
result, comparison may be made between the spline curve and the mean
parabola in figures 14.1 and 14.2. In the whole of the period covered
by reliable data (since —500), the two curves do not deviate from one
514 14 Determination of changes in the length of the day

1 i I I I i I I i i i I i i i r

Reference LOD 86400 SI seconds exactly

Observed

g
c -20 — Average +1.7 ms/c
CD
c
(0

-40 Tidal friction +2.3 ms/cy

-500 500 1000 1500 2000

Year
Fig. 14.7 Changes in the length of the mean solar day (LOD) from —500 to
+1990 obtained by taking the first derivative along the spline curves shown in
figures 14.1, 14.2, 14.4 and 14.5. (Courtesy: Dr L. V. Morrison). The decade
fluctuations in the LOD from +1830 to +1990 are taken from Jordi et al.
(1994).

another by more than about 500 sec, the mean deviation being about half
of this amount. Taking the average value of t2 over this date range of
250 cy2, the estimated error in the coefficient of t2 is 0.10 ms/cy 2 . The
corresponding uncertainty in the rate of change in the LOD is 0.05 ms/cy,
a result of +1.70 + 0.05 (ms/cy).
Comparison of the tidal and observed mean rates of increase in the LOD
indicates that in opposition to the tides there is an accelerative component
producing an average change in the LOD of —0.6 + 0.1 ms/cy. On this
residual non-tidal term are superposed oscillations of amplitude ~4 ms
and periodicity ~1500 years. These oscillations are of similar amplitude to
the decade fluctuations in the LOD observed in recent centuries, although
of much longer period. The scale of the long-term variations is such that
for several centuries around the epoch +300 there was negligible mean
change in the LOD (the tidal term was effectively annulled) while at other
periods (around —500 and +1000) the average rate of increase in the
LOD was roughly twice the tidal figure.
In table 14.1 are shown side by side the AT and ALOD (relative to the
reference LOD) values at 50-year intervals from the epoch —500 to +1600.
(A lunar acceleration of —26 arcsec/cy2 is assumed.) Before —500 the data
are too sparse to derive reliable results for both parameters. More precise
values since +1600 are tabulated by Stephenson and Morrison (1984). It
is hoped that the AT values in this table will be of interest to historians
of astronomy as well as geophysicists. An accurate knowledge of AT is,
of course, essential in the calculation of historical eclipses.
 14.7 Changes in the LOD 515

Table 14.1 AT and ALOD values from the spline curve in

figures 14.1, 14.2, etc. (NB lunar accel. = -267cy 2 .)

Year AT (s) ALOD (ms)

-500 16 800 -42
-450 16000 -40
-400 15 300 -38
-350 14600 -36
-300 14000 -35

-250 13 400 -33

-200 12 800 -32
-150 12200 -30
-100 11600 -29
-50 11100 -29

0 10600 -28
+50 10100 -27
+100 9600 -27
+150 9100 -27
+200 8 600 -26

+250 8200 -27

+300 7 700 -27
+350 7 200 -27
+400 6700 -27
+450 6200 -27

+500 5 700 -27

+550 5 200 -27
+600 4700 -26
+650 4300 -25
+700 3 800 -24

+750 3400 -23

+800 3000 -22
+850 2600 -20
+900 2200 -19
+950 1900 -17

1000 1600 -15

1050 1350 -14
1100 1100 -12
1150 900 -10
1200 750 -9
516 14 Determination of changes in the length of the day

Table 14.1 continued.

Year (s) ALOD (ms)
1250 600 -7
1300 470 -6
1350 380 -5
1400 300 —4
1450 230 -3

1500 180 -3
1550 140 -2
1600 110 -2

14.8 Geophysical discussion

It has been shown above that over the past 2500 years there is a variable
non-tidal decrease in the LOD at an average rate of —0.6 ms/cy in
opposition to the steady tidal increase of +2.3 ms/cy. The non-tidal
acceleration and the fluctuations which occur on a time-scale of 1500
years may have separate geophysical origins. The most plausible cause
of the former would appear to be post-glacial uplift. Following the
deglaciation after the last ice-age, there has been a gradual decrease in
the terrestrial oblateness owing to viscous rebound of the solid Earth.
Artificial satellite measurements of the present-day rate of change in the
Earth's zonal harmonic J2 lead to similar results to that obtained above
for the mean non-tidal rate of decrease in the LOD. By analysing the
acceleration of the node of the orbit of the Starlette satellite, Cheng et al.
(1989) obtained a value for J2 of -(2.5 + 0.3) x 10~ n /y, corresponding
to a rate of decrease in the LOD of —0.44 + 0.05 ms/cy. This result was
fairly close to that obtained previously by Yoder et al. (1983) based on
orbital studies of the Lageos satellite: equivalent to —0.6 ms/cy.
If reliable historical data had extended much further back in time, it
might have proved possible to detect a gradual decrease in J2 - as expected
from the theory of post-glacial rebound. However, the time-scale of 2700
years is too short to achieve this objective, bearing in mind the mediocre
quality of the early observational data.
With regard to the quasi-periodic fluctuations on a time-scale of some
1500 years, a possible mechanism would appear to be electromagnetic
coupling between the core and mantle of the Earth. This is the most
likely cause of the decade fluctuations (Lambeck, 1980, p. 247). Since the
observed peak angular acceleration - relative to the mean non-tidal term
- is ~ + 2 ms/cy (~ + 5 X 10~22 rad/sec 2 ), it is readily calculated that
 14.9 Concluding remarks 517

the required torques operating on the mantle are ~ + 4 x 1016 Nm. The
decade fluctuations observed over the last few centuries require torques
an order of magnitude greater - but, of course, for a much shorter period
of time. The possibility of appreciable changes in the moment of inertia
of the Earth caused by global sea-level variations cannot be ruled out.
Significant long-term alterations in climate have been detected in the last
few millennia (Lamb, 1982). However, at present the associated changes
in the LOD cannot be quantified owing to the lack of reliable sea-level
measurements on a global scale (Pirazolli, 1991).
The investigation of long-term changes in the rate of rotation of the
Earth using historical observations thus raises a number of geophysical
issues which await further development.

14.9 Concluding remarks

The study of Earth's past rotation is just one of several modern scientific
disciplines in which ancient and medieval astronomical observations play
a major role. It is particularly satisfying that observations from a wide
variety of early cultures made using different techniques can be integrated
into a viable entity.
 Appendix A Timed data

Table Al Babylonian timings of Table A2 Babylonian timings of

solar and lunar eclipses (intervals solar eclipses (intervals > 25 deg).
< 25 deg).
Year A T (sec)
Year A T (sec)
-253 11450
-665 21050 -253 11500
-536 18 800 -253 11650
-482 17 300 -248 13 850
-420 15 500 -189 12900
-407 15 250
-189 13400
-406 16200 -189 14150
-405 16 500 -169 12200
-352 14 550 -135 12100
-352 15 250 -135 12250
-321 14150

-316 15 550
-307 14100
-280 12950
-239 14200
-214 17 750

-211 11800
-193 13 650
-188 11250
-169 12 300
-142 12 550

-135 12 600
-133 10950
-128 12600
-108 12050
-66 10150

518
 Appendix A Timed data 519

Table A3 Babylonian timings of Table A3 -continued

lunar eclipses (intervals > 25 deg Year A T (sec)
or approximate measurements
< 25 deg + estimates of phase -381 16000
on horizon). -377 16000
Year A T (sec) —377 16200
-377 15 750
-720 21650 -377 14950
-719 19000
-694 16750 -370 12 750
-685 22 500 -366 19 550
-684 19000 -352 15 950
-352 14450
-666 17250 -316 16 550
-620 17 800
-600 18 000 -316 16100
-598 14400 -316 19 950
-586 18 850 -316 20200
-238 8 250
-579 19050 -225 17 550
-572 18050
-561 15 950 -225 17 550
-536 18 200 -225 18 950
-522 19250 -225 19 350
-225 15 450
-501 18 950 -214 16 500
-500 14950
-500 15 300 -214 17 650
-500 14950 -214 16 850
-500 13000 -214 14150
-211 21350
-490 13 800 -193 14000
-423 16700
-423 17000 -189 12 700
-423 17050 -188 10750
-407 14 300 -162 9 700
-159 14050
-406 14950 -153 13150
—406 13 850
-396 14150
-382 17 750
-381 16400
520 Appendix A Timed data

Table A3 -continued Table A 4 Chinese measure-

Year AT (sec) ments of solar eclipse times to
J^Q day + estimates of phase on
-142 12900 horizon.
-142 12750 Year A T (sec)
-135 11800
-128 11850 +586 3 650
-128 12050 +680 2 500
+691 1000
-119 12 500 +702 2 350
-119 11700 +937 2100
-119 10900
-119 11950 1040 2 850
-98 13 050 1042 1200
1052 2900
-95 13150 1054 1900
-79 10350 1059 2400
-79 11450
-79 10800 1066 1700
-79 11300 1068 1600
1068 1200
-79 12050 1068 850
-66 11400 1069 500
-65 9900
-65 10450 1080 0

1094 2 650
1100 -1000
1107 1050

1107 1400
1107 1150
1173 900
1173 2000
1173 1400

1183 800
1195 850
1202 700
1202 1700
1216 400

1243 350
1245 1350
1277 300
1277 500
1277 600
 Appendix A Timed data 521

Table A5 Chinese measure- Table A6 Chinese measure-

ments of lunar eclipse times to ments of lunar eclipse times to
^ night-watch (+ estimates of 100
day.
pnase on normm). Year A T (sec)
Year AT (sec)
1052 2100
+434 1750 1069 2650
+434 2 800 1069 1750
+437 7450 1069 800
+437 6000 1071 500
+437 6 550
1071 300
+438 5 800 1071 950
+440 6200 1073 900
+440 8100 1073 1400
+503 6000 1073 800
+513 5 750
1078 300
+543 4600 1078 1200
+585 6000 1081 1350
+585 5 700 1082 1650
+585 7 200 1082 1150
+592 5 900
1082 1200
+593 4650 1085 150
+595 5450 1085 50
+595 4050 1085 200
+595 3 500 1088 650
+596 4200
1088 -900
+596 4 800 1088 150
+948 4450 1089 1100
1042 700 1092 700
1067 100 1092 -1100
1068 950
1092 1050
1071 1300 1097 1050
1073 1850 1099 5100
1074 1600 1099 3 850
1074 500 1099 650
1081
1 V O -L 100
-L \J\J

1168 -2 300
1185 100
522 Appendix A Timed data

Table A6 -continued
Year ings of lunar and solar eclipses.
A T (sec)
1099 1150 Year A T (sec)
1099 700 -200 12400
1099 250 -199 11800
1106 200 -199 12 650
1106 950 -199 13 550
-173 11000
1168 0
1168 100 -173 10750
1270 -1300 -140 8 550
1270 -450 +125 10 350
1270 -200 +364 8100
+364 8 300
1272 -200
1272 -750 +364 8 400
1272 750
1277 650
1277 -50

1277 900
1277 200
1279 450
1279 50
1279 -400

1279 700
1279 900
1279 550
1280 1050
 Appendix A Timed data 523

Table A8 Arab timings of solar Table A9 Arab timings of lunar

eclipses. eclipses.

Year AT (sec) Year AT (sec)

+829 900 +854 3150
+829 2 250 +854 2 500
+866 2200 +854 1350
+866 2 500 +856 2 350
+866 2450 +883 900

+891 1700 +901 650

+901 1650 +901 450
+901 1550 +923 1200
+923 1950 +923 2000
+923 1600 +923 3 450

+928 1800 +925 2400

+977 1800 +925 2 350
+977 2000 +927 2950
+978 1300 +933 2 250
+978 2000 +979 1650

+979 1450 +979 1900

+985 1500 +979 1300
+985 750 +980 2100
+993 2000 +981 2000
+993 1400 +981 2 350

+993 1600 +981 1500

1004 1450 +983 1200
1004 1300 +986 800
1004 700 +990 3 500
1004 1150 1001 600

1002 1750
1002 1950
1003 1450
1003 700
1004 2 600

1019 1900
1019 1700
1019 1800
1019 1600
 Appendix B Untimed data

Table Bl Untimed total and annu- Table Bl -continued

lar solar eclipses.
AT range (sec)
AT range (sec) Year LL UL
Year LL UL
1431 10 700
-708 20230 21170 1485 - 5 500 780
-309 13 300 17 200 1560 -475 205
-180 11800 12 720 1567 145 165
-135 11210 12140 1575 -2150 1390
+454 6130 7900
1601 -1460 3170
+761 1720 3 290
+840 1610 6 800
+873 1820 3 750
+912 880 2 600
+968 1580 2 600

+975 1230 4480

1061 800 2140
1124 960 2 700
1133 540 1150
1147 300 1190

1176 600 1600

1185 -2200 10 500
1221 — 980
1239 -500 1450
1241 630 1400

1267 -850 820

1275 -700 1300
1292 -50 1830
1406 170 740
1415 -650 670

524
 Appendix B Untimed data 525

Table B2 Untimed partial solar Table B3 Sunrise/sunset or moon-

eclipses (AT limits reversed). rise/moonset observations.
AT range (sec) AT range (sec)
Year LL UL Year LL UL
-430 12 650 11800 -719 — 21400
-393 14250 12710 -701 18 350 25 000
-187 14140 13 830 -694 9 700 19 850
-79 8 520 8 200 -666 11400 23 500
-27 9 530 8090 -665 — 21750

-1 2150 2050 -590 12 550 24700

+120 8 970 8150 -587 11000 20 850
+360 9400 7120 -576 8 250 20 300
+494 6 600 5 980 -562 13 050 20600
+702 2760 1460 -536 15450 19 300

+729 1190 420 -482 14900 18 700

+822 — 4020 -441 13 350 24650
1004 1940 1770 -382 15 700 21450
1135 3 810 1840 -381 — 17 850
1147 2 870 1680 -352 13450 17 350

1178 2 320 1140 -330 13000 25 050

1330 1210 890 -321 13400 18 000
1333 4250 4150 -280 12 350 14450
1354 4 320 4200 -247 11450 23100
1361 1760 500 -240 12 800 —

1605 — 1060 -239 13 650 17 300

-211 6 800 16 550
-200 10 500 15 900
-194 — 15 350
-189 10700 13100

-188 6900 15 800

-149 8 350 16050
-98 10100 13 800
-88 7 250 —
-79 6850 11700
526 Appendix B Untimed data

Table B3 -continued
AT range (sec)
Year LL UL
-66 4150 15 350
-65 9 500 13 350
-34 10000 12000
+438 4 800 8 850

+503 5 000 9 800

+513 2250 6 300
+514 -250 5 500
+532 2400 6 850
+882 1250 7000

+937 500 3 500

+923 2 500 7 250
+928 250 2900
+979 -1000 2450
+981 1500 —

+986 -6450 5 450

1042 -850 2050
1067 -3150 750
1068 -200 3 400

1071 -1150 2600

1073 -1850 3100
1081 -850 2750
1178 -5650 3450
 References

Aaboe, A., Britton, J. P., Henderson, J. A., Neugebauer, O. and Sachs, A. J.,
1991. Trans. Amer. Phil. Soc, 81, No. 6, 1-75.
Adams, J. C, 1853. Phil. Trans. R. Soc Lond., 143, 397-406.
Adams, J. C, 1859. Comptes Rendus des Seances de VAcademie des Sciences, 48,
247-8.
Airy, G., 1853. Phil. Trans. R. Soc. Lond., 143, 179-200.
Airy, G , 1857. Mem. R. Astr. Soc, 26, 131-49.
Ali, J., 1967, The Determination of the Co-ordinates of Positions for the
Correction of Distances between Cities. Beirut.
Allen, C. W., 1976. Astrophysical Quantities. London.
Arnold, T., 1885. Symeonis Monachi Opera Omnia. Longman and Co., London.
Aston, W. G., 1972. Nihongi: Chronicles of Japan from the Earliest Times to AD
697. Tokyo.
Barron, J. P. and Easterling, P. E., 1985. In The Cambridge History of Classical
Literature, vol. I: Greek Literature (ed. by P. E. Easterling and B. M. W.
Knox). Cambridge University Press, pp. 117-28.
Beaujouan, G. 1974. In Actes du XVIe Congres International d'Histoire des
Sciences, vol. 2, pp. 27-30. Tokyo-Kyoto.
Beijing Observatory, 1988. Zhongguo Gudai Tianxiang Jilu Zongji (A Union
Table of Ancient Chinese Records of Celestial Phenomena). Kexue Jishi
Chubanshe, Kiangxu.
Berger, A. L., 1977. Celestial Mechanics, 15, 53-74.
Bey, M., 1860. Comptes Rendus, 51, 684.
Bickerman, E. J., 1980. Chronology of the Ancient World. Cornell University
Press, Ithaca, N. Y.
Bielenstein, H., 1979. Bull. Museum Far Eastern Antiquities, Stockholm, 51,
3-300.
Blair, P. H., 1963. In Studies on the Early British Border (ed. by N. K.
Chadwick). Cambridge, pp. 63-118.
Bouquet, M. (ed.), 1781. Recueil des Historiens des Gaules et de la France, vol 10.
Paris.
Bowra, C. M., 1964. Pindar. Clarendon Press, Oxford.

527
528 References

Bretagnon, P, 1982. Astron. Astrophys., 114, 278-88.

Brinkman, J. A., 1968. A Political History of Post-Kassite Babylonia (Analecta
Orientalia, 43). Pontifical Biblical Institute, Rome.
Britton, J. P., 1985. The Quality of Ptolemy's Solar and Lunar Observations and
Parameters. (Unpublished manuscript.)
Brown, E. W., 1915a. Mon. Not. R. Astr. Soc, 75, 508-16.
Brown, E. W., 1915b. Report of the 84th Meeting of the British Association for
the Advancement of Science, 311-21.
Brown, E. W., 1919. Tables of the Motion of the Moon, Yale University Press,
New Haven, Conn., vol. I.
Brown, E. W, 1926. Trans. Astr. Obs. Yale Univ., 3, 205-35.
Budge, W. A., 1925. The Rise and Progress of Assyriology. Hopkinson and Co.,
London.
Calame, O. and Mulholland, D , 1978. Science, 199, 977-78.
Carter, T. F., 1925. The Invention of Printing in China and its Spread Westwards.
New York.
Caussin C, 1804. Le Livre de la Grande Table Hakemite par ebn Iounis. Paris.
Celoria, G., 1877a. Memorie del Reale Istituto Lombardo di Scienze e Letteri,
Classe di Scienze, Mathematiche e Naturali, 13, 275-300.
Celoria, G., 1877b. Memorie del Reale Istituto Lombardo di Scienze e Letteri,
Classe di Scienze, Mathematiche e Naturali, 13, 367-82.
Chabot, J. B., 1905. Chronique de Michel le Syrien. 4 vols. Leroux, Paris.
Chapront-Touze, M. and Chapront, J., 1983. Astron. Astrophys., 124, 50-62.
Chen Jiujin, 1983a. Zhi-ran Ke-xue Shi-yan-jiu (Studies in the History of
Natural Sciences), 2, No. 2, 118-32.
Chen Jiujin, 1983b. Zhi-ran Ke-xue Shi-yan-jiu (Studies in the History of
Natural Sciences), 2, No. 4, 303-15.
Cheng, M. K., Eanes, R. J., Shum, C. K., Schutz, B. E. and Tapley, B. D., 1989.
Geophys. Res. Lett., 16, 393-6.
Cherniss, H. and Helmbold, W.C., 1957. Plutarch's Moralia, vol. XII.
Heinemann, London.
Chou Hung-hsiang, 1979. Scientific American, 240, No. 4, 134-49.
Christodoulidis, D. C, Smith, D. E., Williamson, R. G. and Klosko, S. M , 1988.
J. Geophys. Res., 93, 6216-36.
Clark, D. H. and Stephenson, F. R., 1977. The Historical Supernovae. Pergamon
Press, Oxford.
Clavius, C, 1593. In sphaeram Ioannis de Sacrobosco, Commentarius. Sumptibus
Fratrum de Gabiano, Lugduni.
Clemence, G. M., 1948. Astron. J , 53, 169-79.
 References 529

Clemence, G. M , 1971. J. Hist. Astr., 2, 73-9.

Cohen A. P. and Newton, R. R., 1981-83. Monumenta Serica, 35, 347-430.
Corcoran, T. H., 1971. Seneca: Naturales Quaestiones. Heinemann, London.
Cowell, P. H , 1905. Mon. Not. R. Astr. Soc, 66, 3-5.
Cowell, P. H , 1906a. Mon. Not. R. Astr. Soc, 66, 352-5.
Cowell, P. H , 1906b. Mon. Not. R. Astr. Soc, 66, 523-41.
Cowell, P. H., 1907. Mon. Not. R. Astr. Soc, 68, 18-9.
D'Elia, P. M , 1960. Galileo in China. Cambridge, Mass.
Dakyns, H. G., 1892. The Works of Xenophon, vol II. Macmillan and Co.,
London.
Danjon, A., 1929. VAstronomie, 43, 13-22.
Darwin, G. H , 1880. Phil. Trans. R. Soc. Lond., Ill, 713-891.
Dasent, G. W., 1894. The Saga of Haeon and a Fragment of the Saga of Magnus.
H.M. Stationery Office, London.
de Sitter, W., 1927. Bull. Astr. Inst. Netk, 4, 21-38.
Delaporte, L. J., 1910. Chronographie de Mar Elie bar Sinaya, Metropolitaine de
Nisibe. Paris.
Delaunay, C.-E., 1859. Comptes Rendus des Seances de VAcademie des Sciences,
48, 817-27.
Delaunay, C.-E., 1860a. Comptes Rendus des Seances de VAcademie des Sciences,
51, 695-702.
Delaunay, C.-E., 1860b. Connaissance des Temps, Additions pour VAnnee 1862,
1-58.
Delaunay, C.-E., 1865. Comptes Rendus des Seances de VAcademie des Sciences,
61, 1023-32.
Dicke, R. H., 1966. In The Earth-Moon System (ed. by B. G. Marsden and A. G.
W. Cameron). Plenum Press, New York, pp. 98-163.
Dickey, J. O , Williams, J. G. and Yoder, C. F, 1982. IAU Colloquium No. 63 (ed.
by O. Calame). D. Reidel Publishing Company, Dordrecht, p. 217.
Dickey, J. O , Bender, P. L , Faller, J. E. et al. 199>4. Science, 265, 482-490.
Dicks, D. R., 1960. The Geographical Fragments of Hipparchus. University of
London Press.
Dubs, H. H., 1938, 1944, 1955. History of the Former Han Dynasty. Harvard
University Press, 3 vols.
Dubs, H. H., 1947. Harvard Journal of Asiatic Studies, 10, 162.
Duncombe, J. S., 1973. U.S. Naval Observatory Circular, No. 141. Washington,
DC.
Dunthorne, R., 1749. Phil. Trans. R. Soc. Lond., 46, 162-71.
530 References

Essen, L. and Parry, J.V.L., 1957. Phil Trans. R. Soc Lond., A, 250, 45-69.
Ferrari, A. J., Sinclair, W. S., Sjogren, W. L., Williams, J. G., and Yoder, C. K,
1980. J. Geophys. Res., 85, 3939-51.
Ferrel, W., 1865. Proc. Amer. Acad. Arts and Sciences, 6, 379-83.
Fitzgerald, C. P., 1976. China, a Short Cultural History. Cresset Press, London.
Florez, H., 1747. Espana Sagrada. Marin, Madrid.
Foley, N. B. F. and Stephenson, F. R., 1997. In preparation.
Forget, I., 1963. Synaxarium Alexandrinum. Louvain.
Fotheringham, J. K , 1908. Mon. Not. R. Astr. Soc, 69, 26-30.
Fotheringham, J. K , 1909. Mon. Not. R. Astr. Soc, 69, 204-10.
Fotheringham, J. K., 1915. Mon. Not. R. Astr. Soc, 75, 377-94.
Fotheringham, J. K., 1920a. Mon. Not. R. Astr. Soc, 80, 578-81.
Fotheringham, J. K., 1920b. Mon. Not. R. Astr. Soc, 81, 104-26.
Fotheringham, J. K , 1920c. Observatory, 43, 189-91.
Fotheringham, J. K., 1935. Mon. Not. R. Astr. Soc, 95, 719-23.
Freeman-Grenville, G. S. P., 1977. The Muslim and Christian Calendars. 2nd ed.
Oxford University Press.
Geer, R. M., 1954. The Library of History of Diodorus Siculus, Books XIX-XX.
Heinemann, London.
Gingerich, O., 1986. Scientific American, 251, No. 4, cover and pp. 74-83.
Ginzel, F. K., 1884a. Sitzungsberichte der Kaiserlichen Akademie des
Wissenschaften in Wien, Math.-naturwiss. Kl, Abt. II, 88, 629-755.
Ginzel, F K., 1884b. Sitzungsberichte der Kaiserlichen Akademie des
Wissenschaften in Wien, Math.-naturwiss. Kl, Abt. II, 89, 491-559.
Ginzel, F K., 1899. Spezieller Kanon der Sonnen-und Mondfinsternisse, Mayer
und Muller, Berlin.
Ginzel, F. K., 1918. Abhandlungen der Koniglichen Akademie der Wissenschaften,
Berlin, Phys.-Math. Kl, 4, 3-46.
Glauert, H , 1915. Mon. Not. R. Astr. Soc, 75, 489-95.
Goad, C. and Douglas, B., 1978. J. Geophys. Res., 83, 2306-10.
Godley, A. D., 1971. Herodotus, Books V-VII. Heinemann, London.
Godley, A. D., 1975. Herodotus, Books I and II. Heinemann, London.
Goldstein, B. R., 1979. Archives Internationales d'Histoire des Sciences, 29,
101-56.
Grayson, A. K., 1975. Assyrian and Babylonian Chronicles. J. J. Augustin, New
York.
Halley, E., 1693. Phil Trans. R. Soc Lond., 17, 913-21.
 References 531

Halley, E., 1695. Phil Trans. R. Soc Lond., 19, 106-75.

Halley, E , 1715. Phil Trans. R. Soc, 29, 245-262 and 314-16.
Han, Yu-shan, 1955. Elements of Chinese Historiography. Hawley, Hollywood.
Hansen, P. A., 1857. Tables de la Lune construites d'apres le Principe Newtonien
de la Gravitation Universelle. H. M. Stationery Office, London.
Heath, T. L., 1932. Greek Astronomy. Dent and Sons, London. Reprinted 1991
by Dover Publications Inc., Mineola, NY.
Heiskanen, W., 1921. Ann. Acad. Scient. Fennicae, A, 18, 1-84.
Herschel, W, 1781. Phil. Trans. R. Soc. Lond., 71, 115-21.
Hewlett, H. G. 1887. The Flowers of History by Roger of Wendover. H. M.
Stationery Office, London.
Hide, R , Birch, N. T, Morrison, L. V., Shea, D. J. and White, A. A , 1980.
Nature, 286, 114-17.
Hilton, J. L., Seidelmann, P. K. and Liu Ciyuan, 1992. Astron. J., 104, 2250-2.
Ho Peng Yoke, 1966. The Astronomical Chapters of the Chin Shu. Mouton, Paris.
Holder-Egger, O. (ed.), 1899. Scriptores Rerum Germanicarum, XLVII (1899), pp.
19, 56, 57 and 178.
Hsueh Chung-san and Ou-yang I, 1956. A Sino-Western Calendar for Two
Thousand Years. San-lien Shu-tien, Beijing.
Huber, P. J., 1973. Babylonian Eclipse Observations from 750 BC to 0.
(Unpublished manuscript.)
Hughes, D. W., Yallop, B. D. and Hohenkerk, C. Y, 1989. Mon. Not R. Astr.
Soc, 238, 1529-35.
Hunger, H., 1992. Astrological Reports to Assyrian Kings (State Archives of
Assyria, vol. VIII). Helsinki University Press.
Hunger H , 1995. Pre-print.
Hunger, H. and Pingree, D., 1989. MUL.APIN: an Astronomical Compendium in
Cuneiform. Ferdinand Berger, Horn (Austria). (Archiv fur Orientforschung, 24).
IAU, 1954. Trans. Int. Astr. Un., 8, 80.
IAU, 1968. Trans. Int. Astr. Un., XIIIB, 48.
ILE, 1954. Improved Lunar Ephemeris: 1952-1959, US Naval Observatory,
Washington, DC.
Innes, R. T. A., 1925. Astr. Nachr., 225, 109-10.
Jeffreys, H , 1920. Phil. Trans. R. Soc, A, 111, 239-64.
Jeffreys, H., 1976. The Earth: Its Origin, History and Physical Constitution (sixth
edition). Cambridge University Press.
Jeffreys, H., 1982. Geophys. J., 71, 555-66.
532 References

Jones, H. L. (e<±), 1917. The Geography of Strabo. Heinemann, London, vol. I,

pp. 23-5.
Jones, H. S., 1926. Mon. Not. R. Astr. Soc, 87, 4-31.
Jones, H. S., 1932. Ann. Cape Obs., 13, part 3.
Jones, H. S., 1939. Mon. Not. R. Astr. Soc, 99, 541-58.
Jordi, C, Morrison, L. V., Rosen, R. D., Salstein, D. A. and Rossello, G., 1994.
Geophys. J. Int., Ill, 811-8.
Kanda Shigeru, 1934. Nihon Temmon Shiryo. Tokyo.
Kanda Shigeru, 1935. Nihon Temmon Shiryo Soran. Tokyo.
Kant, I., 1754. Wochentliche Frag- und Anzeigungs-Nachrichten Nos. 23 and 24.
(See Ley, 1968.)
Karlgren, B., 1950. The Book of Odes. Stockholm, 1950.
Keightley, D. N. 1978a, Sources of Shang History: The Oracle-bone Inscriptions
of Bronze-age China. Berkeley and Los Angeles.
Keightley, D. N., 1978b. Harvard Journal of Asiatic Studies, 38, 423-38.
Kelvin, Lord, 1894. Popular Lectures and Addresses. Macmillan and Co. Ltd,
London, vol. II.
Kelvin, Lord and Tait, P. G., 1883. Treatise on Natural Philosophy. Cambridge
University Press.
Kepler, J., 1605. De Stella Nova in Pede Serpentarii, chap. XXIII, 4. Reprinted in
Joannis Kepleri Astronomi Opera Omnia (ed. by C. Frisch), vol. II, p. 696.
Frankfurt, 1859.
King, D. A., 1973. Archive for History of Exact Science, 10, 342-94.
King, D. A., 1976. Contribution to Dictionary of Scientific Biography, vol. 14, pp.
574-80. New York.
Kugler, F. X , 1907, 1909-24, 1913-14. Sternkunde und Sterndienst in Babel, vols.
I, II, Erganzungen. Munster.
Lagrange, J.-L., 1792. Nouveaux Memoir es de VAcademie Roy ale des Sciences et
Belles Lettres de Berlin, Oeuvres, 5, 687-98.
Lalande, J.-J. 1757. Memoire de VAcademie pour Vannee 1757.
Lamb, H. H., 1982. Climate History and the Modern World. Methuen, London.
Lambeck, K , 1977. Phil. Trans. R. Soc. Lond., A, 287, 545-94.
Lambeck, K., 1980. The Earth's Variable Rotation. Cambridge University Press.
Lane-Poole, S., 1926. Saladin and the Fall of the Kingdom of Jerusalem. London.
Laplace, P.-S., 1787a. Memoire de VAcademie pour Vannee 1786.
Laplace, P.-S., 1787b. Connaissance des Temps, Additions pour VAnnee 1790,
291-5.
Laskar, J., 1986. Astron. Astrophys., 157, 59-70.
 References 533

Legge, J., 1872. The Chinese Classics, vol. 5. Hong Kong. Reprinted Hong Kong,
1960.
Levi-Donati, G. R., 1987. Bolletino della Deputazione di Storia Patri per
I'Umbria, 84, 89-107.
Levi-Donati, G. R, 1989. I'astronomia, no. 89, 24-9.
Ley, W. 1968. Kant's Cosmogony. Greenwood Publishing Company, New York.
Liais, E., 1866. Memoires de I'Academie, 28, 1119-21.
Liu Baolin and Fiala, A. D., 1992. Canon of Lunar Eclipses: 1500 BC - AD
3000. Willmann-Bell, Inc., Richmond, Virginia.
Liu Chao-yang, 1944. Chung-kuo Wen-hua Yen-ch'iu Hui-k'an (China Cultural
Research Quarterly), 4 (2), 39.
Liu Chao-yang, 1945. Yu Chou (The Universe), 15, 15-16.
Livingstone, R. W., 1949. Thucydides: the History of the Peloponnesian War.
Oxford University Press, London.
Longomontanus, C. S., 1622. Astronomica Danica. Amsterdam.
Macey, S. L. (ed.), 1994. Encyclopedia of Time. Garland Publishing, Inc., New
York, p. 443.
Markowitz, W., Hall, R. G., Essen, L. and Parry, J. V. L., 1958. Phys. Rev. Lett.,
1, 105-6.
Maspero, H., 1939. Melanges Chinois et Bouddhiques, 6, 183-382 (esp. 212-3).
Maxwell, H., 1913. The Chronicle of Lonercost. James Maclehose and Sons,
Glasgow.
Mayer, T., 1753. Novae Tabulae Motuum Solis et Lunae. Gottingen.
Meeus, J., 1982. J. Brit. Astr. Assn., 92, 124-6.
Migne, J. P. (ed.), 1865. Patrologiae Graecae, vol. 148. Paris.
Millard, A. R., 1994. The Eponyms of the Assyrian Empire: 910-612 BC
University of Helsinki. (State Archives of Assyria Studies, vol. II).
Morrison, L. V, 1979, Mon. Not. R. Astr. Soc, 187, 41-82.
Morrison, L. V. and Stephenson, F. R., 1981. In Reference Coordinate Systems
for Earth Dynamics (ed. by E. M. Gaposchkin and B. Kolaczek). D. Reidel
Publishing Company, Dordrecht, pp. 181-5.
Morrison, L. V. and Ward, C. G., 1975. Mon. Not. R. Astr. Soc, 173, 183-206.
Muller, P. M., 1975. PhD Thesis, University of Newcastle upon Tyne.
Muller, P. M., 1976. Report SP 43-26. Jet Propulsion Laboratory, Pasadena,
California.
Muller, P. M. and Stephenson, F. R., 1975. In Growth Rhythms and the History
of the Earth's Rotation (ed. by G. D. Rosenberg and S. K. Runcorn). John
Wiley and Sons, London, pp. 459-534.
534 References

Munk, W. H. and MacDonald, G. J. R, 1960. The Rotation of the Earth.

Cambridge University Press.
Muratori, L. A. (ed.), 1723- . Scriptores Rerum Italicarum (25 vols.), Milan.
Nakayama Shigeru, 1969. A History of Japanese Astronomy. Harvard University
Press, Cambridge, Mass.
Nakiboglu, S. M. and Lambeck K., 1991. In Glacial Isostacy, Sea-level and
Mantle Rheology (ed. by R. Sabadini, K. Lambeck and E. Boschi). Kluwer,
Dordrecht, pp. 237-58.
Nallino, C. A., 1899. Al-Battani sive Albatenii Opus Astronomicum. Milan.
Nasr, S. H., 1976. Islamic Science: an Illustrated Study. World of Islam
Publishing Co. Ltd., London.
Needham, J., Lu Gwei-djen, Combridge, J. H. and Major, J. S., 1986. The Hall
of Heavenly Records. Cambridge University Press.
Needham, J., Wang Ling and Price, D deS., 1986. Heavenly Clockwork.
Cambridge University Press.
Neugebauer, O , 1947. Isis, 37, 37-43.
Neugebauer, O., 1952. The Exact Sciences in Antiquity. Princeton, New Jersey.
Neugebauer, O., 1955. Astronomical Cuneiform Texts, I—III. London.
Neugebauer, O., 1975. A History of Ancient Mathematical Astronomy.
Springer-Verlag, Berlin.
Nevill, E., 1906a. Mon. Not. R. Astr. Soc, 66, 404-20.
Nevill, E , 1906b. Mon. Not. R. Astr. Soc, 67, 2-13.
Newcomb, S, 1870. Amer. J. ScL, Ser. II, 50, 183-94.
Newcomb, S., 1878. Researches on the Motion of the Moon, Washington
Observations for 1875, Appendix II. US Naval Observatory, Washington, DC.
Newcomb, S., 1895a. Astr. Pap. Amer. Eph., 6, Washington, DC.
Newcomb, S. 1895b. The Elements of the Four Inner Planets and the
Fundamental Constants of Astronomy. Washington, DC.
Newcomb, S., 1898. Popular Astronomy. Macmillan and Co. Ltd, London.
Newcomb, S, 1909. Mon. Not. R. Astr. Soc, 69, 164-9.
Newcomb, S., 1912. Researches on the Motion of the Moon and Related
Astronomical Elements, Astr. Pap. Amer. Eph., 9, part I. Washington DC.
Newhall, X X, Standish, E. M. and Williams, J. C, 1983. Astron. Astrophys., 125,
150-67.
Newton, R. R., 1968. Geophys. J. R. Astr. Soc, 14, 505-39.
Newton, R. R., 1970. Ancient Astronomical Observations and the Accelerations of
the Earth and Moon. Johns Hopkins University Press, Baltimore.
Newton, R. R., 1972a. Mem. R. Astr. Soc, 76, 99-128.
 References 535

Newton, R. R., 1972b. Medieval Chronicles and the Rotation of the Earth. Johns
Hopkins University Press, Baltimore.
Niebuhr, B. G. (ed.), 1828. Corpus Scriptorum Historiae Byzantinae, vol. 33.
Weber, Bonn.
North, J. D., 1974. Scientific American, 230, No. 1, 96-106.
O'Keefe, J. A , 1970. J. Geophys. Res., 75, 6565-74.
Oates, J., 1986. Babylon (revised edition). Thames and Hudson, London.
Panchenko, D., 1994. J. Hist. Astr., 25, 275-88.
Pang, K. D , Yau, K. K. C, Chou Hung-hsiang and Wolf, R., 1988a. Vistas in
Astronomy, 31, 833-47.
Pang, K. D., Chou, Hang-hsiang and Yau, K. K. C, 1988b. Bull. Amer. Astr.
Soc, 21, 753.
Park, Seong-rae, 1979. J. Social Sciences and Humanities, 49, 53-117.
Parker, R. A., 1974. In The Place of Astronomy in the Ancient World (ed. by D.
G. Kendall and S. Piggott). Oxford University Press, pp. 51-65. Reprinted in
Phil. Trans. R. Soc. Lond. A, 276, 51-65.
Parker, R. A. and Dubberstein, W. H , 1956. Babylonian Chronology: 626 BC -
AD 75. Brown University Press, Providence, R.I.
Parkinson, J. H., Stephenson, F. R. and Morrison, L. V., 1988. Nature, 331, 421-3.
Parpola, S., 1983. Letters from Assyrian Scholars to the Kings Esarhaddon and
Assurbanipal. Part 2: Commentary and Appendices. Alter Orient und Altes
Testament, vol 5/1. Butzon und Bercker Kevelaer, Neukirchen-Vluyn.
Perrin. B., 1961. Plutarch's Lives, vol. V. Heinemann, London.
Pertz, G. H. (ed.), 1826- . Monumenta Germaniae Historica, Scriptores (32 vols.).
Hahn, Hannover.
Pingre, A.-G., 1901. Annales Celestes du Dix-Septieme Siecle (ed. by M. G.
Bigourdan). Paris.
Pirazzoli, P. A., 1991. In Glacial Isostacy, Sea-level and Mantle Rheology (ed. by
R. Sabadini, K. Lambeck and E. Boschi). Kluwer, Dordrecht, pp. 259-70.
Plechl, H. (ed.), 1972. Orbis Latinus. Kinkhardt and Biermann, Braunschweig.
Plummer, C, 1896. Venerabilis Bedae Opera Historica. Clarendon Press, Oxford.
Potter, K. R., 1955. The Historia Novella by William of Malmesbury. Nelson and
Sons, London.
Rabi, I. I., 1945. Richtmeyer Lecture to the American Physical Society, New
York. (Unpublished.)
Race, W. H., 1986. Pindar. Twayne, Boston, Mass.
Rackham, H., 1938. Pliny: Natural History, Books I—II. Heinemann, London.
Rankin, H. D., 1977. Archilochus of Paros. Noyes, Park Ridge (New Jersey).
536 References

Ranyard, A. C, 1879. Mem. R. Astr. Soc, 41.

Rassam, H., 1897. Asshur and the Land of Nimrod. Eaton and Mains, New York.
Ravn, O. E., 1942. Herodotus' Description of Babylon. Copenhagen.
Reade, J. E., 1986. In E. Leichty, Catalogue of the Babylonian Tablets in the
British Museum, vol VI, pp. xiii-xxxvi. British Museum Press, London.
Riccioli, G. B., 1665. Astronomiae Reformatae tomi duo. Bologna.
Rolfe, J. C, 1946. Quintus Curtius. Heinemann, London.
Rosan, L. J., 1949. The Philosophy of Proclus. Cosmos, New York.
Rowton, M. B , 1946. Iraq, 8, 94-110.
Sachs, A. J, 1948. J. Cuneiform Stud., 2, 271-90.
Sachs, A. J., 1967. Archaeologia, 15, Mars-Avril, 12-19.
Sachs, A. J., 1974. In The Place of Astronomy in the Ancient World (ed. by D. G.
Kendall and S. Piggott), Oxford University Press, pp. 43-50. Reprinted in Phil.
Trans. R. Soc. Lond. A, 276, 43-50.
Sachs, A. J., 1976. In Cuneiform Studies in Honor of Samuel Noah Kramer (ed.
by B. L. Eichler). Alter Orient und Altes Testament Sonderreihe, vol 25. Butzon
und Bercker Kevelaer, Neukirchen-Vluyn.
Sachs, A. J. and Hunger, H., 1988, 1989, 1996. Astronomical Diaries and Related
Texts from Babylonia, vols. I, II and III. Osterreichische Akademie der
Wissenschaften, Wien.
Sachs, A. J. and Schaumberger, J., 1955. Late Babylonian Astronomical and
Related Texts. Brown University Press, Providence, R.I.
Sachs, A. J. and Wiseman, D. J., 1954. Iraq, 16, 202-11.
Sage, E. T, 1936. Livy, vol. XL Heinemann, London.
Said, S. S. and Stephenson, F. R., 1991. J. Hist. Astr., 22, 297-310.
Said, S. S. and Stephenson, F. R., 1997. In preparation.
Said. S. S., Stephenson, F. R., and Rada, W. S., 1989. Bulletin of the School of
Oriental and African Studies, London Univ., 52, 38-64.
Sandys, J., 1978. The Odes of Pindar. Heinemann, London.
Sawyer, J. F. A., 1972. J. Theological Studies, 23, 124-8.
Schaumberger, J., 1952. Zeitschrift fur Assyriologie und Vorderasiatische
Archaologie, 50, 214-29.
Schroeter, J. Fr., 1923. Spezieller Kanon der Zentralen Sonnen- und
Mondfinsternisse, welche innerhalb des Zeitraums von 600 bis 1800 n. Chr. in
Europa sichtbar waren. Jacob Dybwad, Kristiania.
Seidelmann, P. K. (ed.), 1992. Explanatory Supplement to the Astronomical
Almanac. University Science Books, Mill Valley, California.
Standish, E. M., 1982. Astron. Astrophys., 114, 297-302.
 References 537

Stephenson, F. R., 1974. In The Place of Astronomy in the Ancient World (ed. by
D. G. Kendall and S. Piggott), Oxford University Press, pp. 118-21. Reprinted
in Phil. Trans. R. Soc. Lond. A, 276, 118-21.
Stephenson, F. R., 1992. Quart. J. Roy. Astr. Soc, 33, 91-8.
Stephenson, F. R., 1993. Encyclopedia Britannica, 17, 872b-7b.
Stephenson. F. R., 1997. In preparation.
Stephenson, F. R. and Fatoohi, L. J, 1993. J. Hist. Astr., 24, 255-7.
Stephenson, F. R. and Fatoohi, L. J., 1994a. Quart. J. R. Astr. Soc, 35, 81-94.
Stephenson, F. R. and Fatoohi, L. J, 1994b. J. Hist. Astr., 25, 99-110.
Stephenson, F. R. and Fatoohi, L. J., 1995. J. Hist. Astr., 26, 227-36.
Stephenson, F. R. and Houlden, M. A., 1986. Atlas of Historical Eclipse Maps:
East Asia, 1500 BC to AD 1900. Cambridge University Press.
Stephenson, F. R., Jones, J. E. and Morrison, L. V., 1997. Astr on. Astrophys. In
press.
Stephenson, F. R , Hunger, H. and Yau, K. K. C, 1985. Nature, 314, 587-92.
Stephenson, F. R. and Morrison, L. V., 1984. In Rotation in the Solar System (ed.
by R. Hide). The Royal Society, London. Reprinted in Phil. Trans. R. Soc
Lond., A, 313, 47-70.
Stephenson, F. R. and Morrison, L. V., 1995. Phil. Trans. R. Soc. Lond., A, 351,
165-202.
Stephenson, F. R. and Said, S. S, 1991. J. Hist. Astr., 22, 195-207.
Stephenson, F. R. and Said, S. S., 1997. In preparation.
Stephenson, F. R. and Walker, C. B. F. (eds.), 1985. Halleys Comet in History.
British Museum Publications, London.
Stephenson, F. R. and Yau, K. K. C, 1985. J. British Interplanetary Society, 38,
195-216.
Stephenson, F. R. and Yau, K. K. C, 1992a. Quart. J. R. Astr. Soc, 33, 91-8.
Stephenson, F. R. and Yau, K. K. G, 1992b. Astron. Astrophys., 260, 485-8.
Stier, H. E. and Kirsten, E., 1956. Westermanns Atlas zur Weltgeschichte.
Westermann, Braunschweig.
Stubbs, W., 1867. The Chronicle of the Reigns of Henry II and Richard I: AD
1169-1192. H. M. Stationery Office, London.
Taylor, G. I., 1919. Phil. Trans. R. Soc, A, 220, 1-33.
Thompson, J. E. S., 1974. In The Place of Astronomy in the Ancient World (ed.
by D. G. Kendall and S. Piggott). Oxford University Press, pp. 83-98.
Reprinted in Phil. Trans. R. Soc. Lond. A, 276, 83-98.
Thompson, R. C, 1900. The Reports of the Magicians and Astrologers of
Nineveh and Babylon. London, 2 vols. (reprinted New York, 1974).
538 References

Thorkelson, T., 1933. Visindafelag Islendinga (Reykjavik), 15.

Tihon, A., 1976-77. Bulletin de I'Institut Historique Beige de Rome, 46-47, 35-79.
Toomer, G. J., 1974. Archive for History of Exact Sciences, 14, 126-42.
Toomer, G. J., 1984. Ptolemy's Almagest: a Translation and Annotation.
Duckworth, London.
Toomer, G. J., 1988. In A Scientific Humanist: Studies in Memory of Abraham
Sachs (ed. by E. Leichty, M. deJ. Ellis and P. Gerardi). Samuel Noel Kramer
Fund, Philadelphia, pp. 353-62.
Tung Tso-Pin, 1945. Yin Li P'u (On the Calendar of the Yin Period). Lichuang.
Unger, E. 1931. Babylon. W. de Gruyter, Berlin.
Volland, H., 1990. In Earth's Rotation from Eons to Days (ed. by P. Brosche and
J. Sundermann). Springer-Verlag, Berlin, pp. 127-40.
von Oppolzer, T. R., 1887. Canon der Finsternisse. Kaiserl. Akad. der Wiss.,
Wien. Reprinted 1962 as Canon of Eclipses, Dover, New York.
Vyssotsky, A. N., 1949. Medd. fran. Lunds Astr. Obs. Historical Papers, No. 22.
Waley, A., 1934. Travels of an Alchemist. Routledge, London.
Williams, J. G., Sinclair, W. S. and Yoder, C. E, 1978. Geophys. Res. Lett., 5,
943-6.
Wise, D. V, 1969. J. Geophys. Res., 74, 6034-45.
Wiseman, D. J., 1965. Contribution to The Cambridge Ancient History.
Cambridge University Press, vol. II, chap. 31.
Xu Zhentao, Stephenson, F. R. and Jiang Yaotiao, 1995. Quart. J. Roy. Astr.
Soc, 36, 397-406.
Xu Zhentao, Yau, K. K. C. and Stephenson, F. R., 1989. Archaeoastronomy
(Supplement to J. History of Astronomy), no. 14, S61-72.
Yabuuchi Kiyoshi, 1979. Acta Asiatica, 36, 7-48.
Yoder, C. F, Williams, J. G., Dickey, J. O., Schutz, B. E , Eanes, R. J. and
Tapley, B. D., 1983. Nature, 303, 757-62.
Zawilski, M., 1994. The Collection of Texts Containing the Historical
Observations for the Solar Eclipses for Europe and the Near East.
(Unpublished manuscript.)
Zinner, E., 1990. Regiomontanus: his Life and Work. Amsterdam, p. 142.
 Acknowledgements

I am grateful to Dr Leslie V. Morrison of the Royal Greenwich Obser-

vatory for his continued help and advice. My thanks are also due to
Professor Peter J. Huber (Bayreuth University), and Professor Hermann
Hunger (University of Vienna) for permission to publish their translations
of Babylonian texts. I wish to thank Mrs Pauline Russell of the Depart-
ment of Physics, University of Durham for skilfully executing numerous
diagrams. I would also like to acknowledge with gratitude my indebted-
ness to the late Professor S. Keith Runcorn, who introduced me to this
subject and gave me much encouragement.
Much of the research for this book was undertaken at Jet Propulsion
Laboratory, Pasadena during my tenure of a Senior Resident Research
Associateship funded by the National Research Council, Washington DC.
I am grateful for the opportunities provided by this award.

539
 Index of eclipse records

Note: The letters in the 'Region' column refer to the following broad re-
gions or cultures: A Arab; B Babylonian (including Assyrian); C Chinese
(including Japanese and Korean); E European (including Alexandrian).

Principal solar eclipse records

Julian date Region Page(s)

BC 763 Jun 15 (B) 125-7
709 Jul 17 (Q 221, 226
669 May 27 (B) 125
657 Apr 15 (B) 125
648 Apr 6 ? (E) 338^2

BC 601 Sep 20 (C) 226

585 May 28 ? (E) 342^1
549 Jun 19 (C) 226-7
463 Apr 30 ? (E) 344-6
444 Oct 24 (Q 227

BC 431 Aug 30 (E) 346-8

394 Aug 14 (E) 366
369 Apr 11 (B) 138-9
322 Sep 26 (B) 79, 131-2, 141-2
310 Aug 15 (E) 348-51

BC 281 Jan 30 (B) 133, 142

254 Jan 31 (B) 72, 133-4
249 May 4 (B) 134, 139
247 Sep 7 (B) 122
241 Nov 28 (B) 142

BC 198 Aug 7 (C) 238

195 Jun 6 (B) 134-5, 139
190 Mar 14 (B) 121, 135-6, 139
188 Jul 17 (Q 234
188 Jul 17 (E) 367

540
 Index of eclipse records 541

Julian date Region Page(s)

BC 181 Mar 4 (C) 234-5
170 Jul 28 (B) 136, 140
147 Nov 10 (C) 235
136 Apr 15 (B) 64-8, 129-31, 136-7
134 Aug 19 (C) 312

BC 129 Nov 20 ? (E) 351-9

89 Sep 29 (Q 235, 315
80 Sep 20 (C) 235-6
35 Nov 1 ? (C) 236, 316
28 Jun 19 (C) 236-7

BC 16 Nov 1 (C) 230

15 Mar 29 (Q 230-1
2 Feb 5 (C) 237

AD 2 Nov 23 (C) 238

29 Nov 24 ? (E) 359-60
30 Nov 14 (C) 231
31 May 10 (C) 231
65 Dec 16 (C) 231-2

AD 71 Mar 20 ? (E) 360-4

120 Jan 18 (Q 237-9
360 Aug 28 (C) 232, 241-2
364 Jun 16 (E) 364-5
429 Dec 12 (C) 242

AD 454 Aug 10 (Q 67-8, 242-3

484 Jan 14 (E) 367-8
489 Mar 18 (C) 75, 312
493 Jan 4 (Q 312
494 Jun 19 (Q 243-4, 313

AD 512 Jun 29 (C) 313

523. Nov 23 (C) 241
529 Nov 17 ?1 (C) 316
532 Nov 13 (C) 317
562 Oct 14 (C) 244
542 Index of eclipse records

Julian date Region Page(s)

AD 585 Jul 31 (C) 313
586 Dec 16 (Q 288, 314
628 Apr 10 (C) 267
680 Nov 27 (C) 294
702 Sep 26 (C) 246-7

AD 729 Oct 27 (C) 247

733 Aug 14 (E) 422-3
754 Jun 25 (C) 247
756 Oct 28 (Q 245-6
761 Aug 5 (Q 247-9

AD 812 May 14 (E) 423

822 Apr 25 (C) 248
829 Nov 30 (A) 470-1
840 May 5 (E) 387, 389-90
866 Jun 16 (A) 471

AD 873 Jul 28 (A) 467

879 Apr 25 ? (C) 246
882 Aug 17 (A) 436-7, 449
891 Aug 8 (A) 489
901 Jan 23 (A) 489-90

AD 912 Jun 17 (A) 438

923 Nov 11 (A) 73, 459-60, 471-2
928 Aug 18 (A) 76, 464, 472, 494, 497
937 Feb 14 (Q 317, 327
939 Jul 17 (A) 443^

AD 968 Dec 22 (E) 52, 390-1

975 Aug 9 (C) 267-8
977 Dec 13 (Q 251
977 Dec 13 (A) 473
978 Jun 8 (A) 473

AD 979 May 28 (A) 473-4, 498

985 Jul 20 (A) 474
993 Aug 20 (A) 444, 474-5
1004 Jan 24 (A) 71, 75-76, 468-9, 475-6
1022 Aug 1 (Q 251-2
 Index of eclipse records 543

Julian date Region Page(s)

AD 1033 Jun 29 (E) 382-3, 416
1040 Feb 15 (C) 288
1042 Jun 20 (C) 318, 327
1046 Apr 9 (Q 289
1052 Nov 24 (Q 277, 289

1053 Nov 13 (C) 289

1054 May 10 (C) 252-3, 290
1059 Feb 15 (C) 290
1061 Jun 20 (A) 439^0
1068 Feb 6 (C) 72-3, 290-1

1094 Mar 19 (C) 291

1100 May 11 (C) 291
1124 Aug 11 (E) 391-2
1133 Aug 2 (E) 392-4, 416-17, 4 2 3 ^
1135 Jan 16 (C) 253

1140 Mar 20 (E) 424-5

1147 Oct 26 (E) 394, 417-18
1153 Jan 26 (E) 418-19
1173 Jun 12 (C) 292
1176 Apr 11 (E) 394-5

1176 Apr 11 (A) 439-42, 444-5

1178 Sep 13 (E) 80-1, 419
1178 Sep 13 (A) 450
1185 May 1 (E) 68-9, 395-7
1187 Sep 4 (E) 80-2

1191 Jun 23 (E) 381-2

1202 May 23 (C) 292
1214 Oct 5 (Q 252
1221 May 23 (C) 253-7
1230 May 14 (E) 425

1239 Jun 3 (E) 385, 397-403

1241 Oct 6 (E) 62, 402-4
1241 Oct 6 (A) 445-6
1245 Jul 25 (Q 263, 293
1260 Apr 12 (C) 295
544 Index of eclipse records

Julian date Region Page(s)

AD 1263 Aug 5 (E) 404
1267 May 25 (E) 404-5
1275 Jun 25 (C) 257-8
1277 Oct 28 (C) 294-5
1292 Jan 21 (C) 62, 258

1310 Jan 31 (E) 382

1312 Jul 5 (E) 419-20, 425
1330 Jul 16 (E) 420
1333 May 14 (E) 420-1, 427
1337 Mar 3 (E) 427-8

1354 Sep 17 (E) 421

1361 May 5 (Q 259-60, 263-4
1367 Jun 28 (Q 252
1406 Jun 16 (E) 405-6
1415 Jun 7 (E) 406-7, 425-6

1431 Feb 12 (E) 407-8

1431 Feb 12 (A) 446
1433 Jun 17 (E) 413-14
1433 Jun 17 (A) 442-3, 446
1460 Jul 18 (Q 268

1463 May 18 (A) 447-8

1476 Feb 25 (E) 426
1485 Mar 16 (E) 385, 408-9, 415
1491 May 8 (A) 448
1513 Mar 7 (A) 449

1514 Aug 20 (C) 261

1544 Jan 24 (E) 426
1560 Aug 21 (E) 409
1567 Apr 9 (E) 410-11
1575 May 10 (C) 261-2

1601 Dec 24 (E) 411

1605 Oct 12 (E) 411-13,415-16,421-2
1715 May 3 (E) 62
 Index of eclipse records 545

Principal lunar eclipse records

Julian date Region Page(s)

BC 731 Apr 9 (B) 122
721 Mar 19/20 (B) 99
720 Mar 8/9 (B) 100
720 Sep 1/2 (B) 100
702 Mar 19/20 (B) 194-5

BC 695 Apr 30/May 1 (B) 162, 195

686 Apr 21/22 (B) 162
685 Oct 3/4 (B) 163^
667 Oct 14/15 (B) 88, 164, 195-6
666 Apr 10/11 (B) 151, 206

BC 621 Apr 21/22 (B) 100

601 Apr 10/11 (B) 164
599 Feb 19/20 (B) 165
591 Mar 22/23 (B) 196
588 Jan 18/19 (B) 196-7

BC 587 Jan 7/8 (B) 165

580 Aug 14/15 (B) 165
577 Jun 13/14 (B) 197
573 Apr 1/2 (B) 166
563 Sep 5/6 (B) 197

BC 562 Mar 2/3 (B) 152

555 Oct 6/7 (B) 173-4
537 Oct 16/17 (B) 152, 198, 208
523 Jul 16/17 (B) 101, 166
522 Jan 9/10 (B) 167

BC 502 Nov 19/20 (B) 101-2

501 Nov 7/8 (B) 174
491 Apr 25/26 (B) 102
483 Nov 18/19 (B) 152-3, 198
442 Mar 23/24 (B) 198-9

BC 424 Sep 28/29 (B) 168-9, 181

421 Feb 2/3 (B) 153
408 Oct 31 /Nov 1 (B) 169
407 Oct 20/21 (B) 174-5
406 Oct 9/10 (B) 153
546 Index of eclipse records

Julian date Region Page(s)

BC 397 Apr 5/6 (B) 169-70
383 Dec 22/23 (B) 102-3
382 Jun 18/19 (B) 103^
382 Dec 12/13 (B) 104
378 Apr 6/7 (B) 175

BC 371 May 17/18 (B) 153^

367 Aug 29/30 (B) 154
353 Nov 21/22 (B) 176, 199, 208
331 Sep 20/21 (E) 372-5
317 Jun 18/19 (B) 154

BC 317 Dec 13/14 (B) 176-7

308 Jul 8/9 (B) 154-5
248 Oct 2/3 (B) 199-200
240 Nov 2/3 (B) 155, 200
239 Apr 28/29 (B) 181

BC 226 Aug 1/2 (B) 84, 177, 186

215 Dec 25/26 (B) 178, 186
212 Apr 29/30 (B) 155, 200-1
212 Oct 24/25 (B) 156
201 Sep 22/23 (E) 369-70

BC 200 Mar 19/20 (E) 370

200 Sep 11/12 (E) 370
194 Nov 4/5 (B) 156-57, 186-7
190 Feb 27/28 (B) 201
190 Aug 23/24 (B) 201-2, 209

BC 189 Feb 16/17 (B) 170, 202

188 Aug 1/2 (B) 55, 185
174 Apr 30/May 1 (E) 371
171 Aug 23/24 (B) 194
163 Mar 30/31 (B) 170-1

BC 160 Jan 26/27 (B) 157

154 Mar 21/22 (B) 182
150 Jul 2/3 (B) 187, 203
143 Feb 17/18 (B) 157, 182, 187
141 Jan 27/28 (E) 371-2
 Index of eclipse records 547

Julian date Region Page(s)

BC 136 Mar 31/Apr 1 (B) 188
134 Mar 9/10 (B) 157-8
129 Nov 4/5 (B) 171-2, 182
120 Jun 1/2 (B) 172, 183, 189
109 May 1/2 (B) 158-9

BC 99 Apr 11/12 (B) 203^, 209

96 Aug 3/4 (B) 159
80 Apr 10/11 (B) 160, 183, 189-90, 204, 209-10
80 Oct 5/6 (B) 160-1
67 Jan 19/20 (B) 161-2, 183, 204-5, 211
BC 66 Dec 28/29 (B) 88, 183, 205, 211

AD 125 Apr 5/6 (E) 372

328 Aug 15/16 (Q 328
434 Sep 4/5 (Q 298-9
437 Jan 8/9 (C) 284
437 Dec 28/29 (C) 281

AD 438 Jun 23/24 (C) 90, 319, 326

503 Jun 25/26 (C) 319, 326
513 Jun 4/5 (C) 320, 326
514 May 24/25 (Q 320-1, 326-7
590 Apr 25/26 (Q 283, 308-9

AD 590 Oct 18/19 (Q 309

595 Dec 22/23 (C) 309
596 Dec 10/11 (C) 299-300, 310
712 Sep 19/20 (C) 331-3
755 Nov 23/24 (E) 428-30

AD 854 Feb 16/17 (A) 84-6, 477-8

854 Aug 11/12 (A) 466, 478-9
856 Jun 21/22 (A) 479
866 Nov 25/26 (A) 476
882 Aug 2/3 (A) 436-7, 451

AD 883 Jul 23/24 (A) 488-9

901 Aug 2/3 (A) 490
923 Jun 1/2 (A) 479-80, 496
925 Apr 11/12 (A) 480-1
927 Sep 13/14 (A) 481
548 Index of eclipse records

Julian date Region Page(s)

AD 929 Jun 27/28 (A) 481
933 Nov 4/5 (A) 482
979 May 14/15 (A) 482, 497
979 Nov 6/7 (A) 482-3
980 May 2/3 (A) 483

AD 981 Apr 21/22 (A) 483-4, 498

981 Oct 15/16 (A) 484
983 Mar 1/2 (A) 484
986 Dec 18/19 (A) 485, 498-9
990 Apr 12/13 (A) 485-6

1001 Sep 5/6 (A) 486

1002 Mar 1/2 (A) 486-8
1003 Feb 19/20 (A) 491
1003 Aug 14/15 (A) 491
1004 Jul 4/5 (A) 492

1019 Sep 16/17 (A) 492-3

1042 Jul 5/6 (C) 321, 327
1063 Nov 8/9 (C) 303
1067 Mar 3/4 (C) 321, 327-8
1068 Aug 15/16 (C) 322

1069 Dec 30/31 (C) 302

1071 Jun 15/16 (Q 322, 328
1071 Dec 9/10 (C) 303, 323, 328
1073 Oct 18/19 (C) 323, 329
1081 Nov 19/20 (C) 324, 329

1106 Jan 21/22 (C) 276-7

1168 Mar 25/26 (C) 87-8, 300-2
1178 Aug 29/30 (A) 452
1277 May 18/19 (C) 303
1461 Jun 22/23 (A) 453

1487 Feb 7/8 (A) 432-3

1772 Apr 17/18 ? (A) 453
 Index of places of observation

Al-Raqqah (see also Raqqa) 457, 488-90 Erfurt 418-19

Aleppo 8, 446-7 Evreux 427
Alexandria 8-9, 58, 101-2, 135, 351-9, 362, 364-
5, 369-72, 375 Farfa 391
Antakyah (see also Antioch) 8, 457, 489-90 Florence 399-400
Antioch (see also Antakyah) 440, 442-3 Foligno 407
An-yang (see also Yin) 215, 219 Forcalquier 421
Arbela 357-8, 372-3, 375
Arezzo 397-8 Ghaznah 457, 461, 491-2, 496
Asshur 127
Athens 346-8, 367-8 Hamburg 405
Augsburg 392, 408-9 Hanyang (see also Seoul) 264
Avignon 382 Heian-kyo (see also Kyoto) 266-8
Heilsbronn 392-3
Babylon 2, 15, 48, 55, 58-9, 64, 66-8, 71-2, 78-9, Hellespont 135, 351-9
88, 93ff, 128ff, 344, 510 Hirschau 63
Baghdad 8, 15, 84, 436, 439-40, 449-52, 457-8,
466, 470-2, 476-82, 496-7 Jazirat Ibn 'Umar (see also Cizre) 439^3
Beijing (see also Ta-tu) 62, 78, 214, 250, 258, 260, Jurjan 457, 461, 491
287 Jurjaniyyah 457, 461, 492
Belvoir 425
Bergamo 387, 389-90 Kaifeng (see also Pien) 280
Bergen 411 Kaesong (see also Songdo) 262
Beze 382-3 Kangwha 262-3
Braunschweig 405, 417 Kerkrade 4 2 3 ^
Brauweiler 394 Kerulen River 253-7
Khalifah (see also Cueva de la Mora) 443-4
Cairo 9, 15, 71, 75-6, 442, 4 4 ^ 8 , 456-8, 466, Kobrin 425-6
468, 473-6, 482-6, 496-8 Kuang-ling 232
Cesena 398 Kyoto (see also Heian-kyo) 266, 268
Chaeroneia 360-4, 366
Ch'ang-an (see also Ta-hsing Ch'eng, Xi'an) 71, Lanercost 419-20
220, 228, 230-1, 234-8, 240, 244-8, 270, 278, Liege 405
294-5, 315-16 Lin-an 87, 250, 253, 257-8, 288, 292-3, 300, 303
Cheng 219-20 Lo-yang 74-5, 228, 230, 232, 234, 237-8, 240,
Chien-k'ang (see also Nanjing) 67, 90, 240-4, 244, 278, 312-13, 315-17, 319-20, 326
281, 300, 312-13, 315, 319, 326-9, 507
Ch'u-fu 223-7 Magdeburg 406, 418
Cizre (see also Jazirat Ibn 'Umar) 439-43 Malmesbury 424-5
Cluny 416 Marola 401-3
Coimbra 399, 402, 409, 420-1 Montpellier 400
Cologne 63 Mosul 451, 453
Constantinople 361, 390, 404-5 Mytho 425-6
Cordoba 438, 449
Cueva de la Mora (see also Khalifah) 4 4 3 ^ Nanjing (see also Chien-k'ang) 67-8, 90, 232,
240, 281
Damascus 448-9, 453 Naples 411-12
Delphi 361-2 Neider Alteich 406, 426
Durham 419 Nicaea 359-60

549
550 Index of places of observation

Nile Delta 445-6 Songdo (see also Kaesong) 2 6 2 ^

Nineveh 105-6, 124 Split 54, 401
Nishapur 458, 467 Stade 62, 402-4
Nisibis 451 Sung-chiang 252, 259-64
Novgorod 68-9, 391-2, 395, 397 Syracuse 348-50

Olmos (see also Welmish) 443-4 Ta-hsing Ch'eng (see also Ch'ang-an, Xi'an) 240,
Orkney 404 244-^5, 288, 300, 308-10, 313-15, 318
Orontes River 444-5 Ta-tu (see also Beijing) 62, 250, 258, 294-5, 303,
305
Paros 338^2 Thasos 338^2
Perugia 408, 421 Thebes 3 4 ^ 6
Pien (see also Kaifeng) 73, 250, 252-3, 280, 288- Toledo 385
91, 295, 300, 303, 315, 317-18, 321-4, 327,
510 Vigeois (see also Vosium) 419
P'ing-ch'eng 240, 312 Vosium (see also Vigeois) 419
Prague 406, 416-17
Pskov 426 Welmish (see also Olmos) 443-4
Wroclaw 407
Raqqa (see also al-Raqqah) 8
Reichersberg 393, 402 Xi'an (see also Ch'ang-an, Ta-hsing Ch'eng) 220-
Rhodes 3 5 2 ^ , 369, 371-2, 375 1, 227-8, 230, 245
Rome 362-4, 367, 410
Yamato Plain 266-7
St Germain des Pres 427 Yin (see also An-yang) 215
Salzburg 3 9 3 ^ York 429
Samarkand 2 5 3 ^ , 257
Sardis 343 Zbraslav 420
Seoul (see also Hanyang) 264 Zwiefalten 424
Siena 400
 Name index

Aaboe, A. 123 Cheng, M. K. 516

Abu Shama 444, 453 Ch'in-shih Huang-ti 95, 221, 228
Adams, J. C. 11 Chou Hung-hsiang 216
Agathocles 348-50, 352 Christodoulidis, D. C. 34, 37-8, 506
Agesilaus, King 366 Cicero, 346-7
Airy, G. B. 15 Clark, D. H. 130
Al-'Asqalani 446-7 Clavius, Christopher 31, 378, 409-10
Al-Battani 8, 456-7, 461, 488-90, 494, 499 Clemence, G. M. 23-5
Al-Biruni 45, 456-63, 466-7, 488, 491-2, 494, Cleomedes 351-3, 357
499 Cohen, A. P. 246
Al-Iranshahri 467 Combridge, J. H. 275
Al-Katib 444 Confucius 220-1
Al-Mahani 457, 465, 471, 476-9 Cosmas (of Prague) 417
Al-Maqrizi 442, 444, 446, 475 Cowell, P. H. 15, 16, 334
Al-Tabari 436, 449, 451 Curtius, 375
Al-'Umari 453
Albertus (of Stade) 403 Danjon, A.-L. 23
Alexander the Great 97, 105, 132, 372, 375, 488- Darwin, G. H. 13
9 D'Elia 261, 284
Anaxagoras 59, 343, 346-7 de Sitter, W. 17,21, 34, 111
Andreas (of Bergamo) 387, 389 Delaunay, C.-E. 11, 12
Archilochus 338-41, 360-1 Dicke, R. H. 336
Dickey, J. O. 35
Bartosek (of Drahonice) 414 Diocletian, Emperor 438
Bede, the Venerable 30, 378, 423 Diodorus Siculus 58, 72, 93, 97, 106, 348-9,
Berger, A. L. 10 366
Bertholdus (of Zwiefalten) 424 Dionysius Exiguus 30
Bey, M. 385 Dlugosz, Jan 426
Bickerman, E. J. 103, 347, 349, 366, 373 Douglas, B. 34
Bielenstein, H. 231 Dubberstein, W. H. 116, 155, 176
Botho (of Braunschweig) 405, 417 Dubs, H. H. 228, 230, 236, 330
Brahe, Tycho 9, 10 Duncombe, J. S. 410
Bretagnon, P. 40 Dunthorne, R. 9, 10
Brinkman, J. A. 95, 144
Britton, J. P. 99 Elias (of Nisibis) 451
Brown, E. W. 17, 20, 24, 41 Epping, J. 110
Budge, W. A. 106, 110 Eratosthenes 354
Bullock, Capt, 397 Essen, L. 26
Bunting, Heinrich, 413 Eusebius 359

Caesar, Julius 29, 30 Fatoohi, L. J. 74, 83, 89-91, 118-19, 121, 124,
Calame, O. 35 148, 180, 287
Carter, T. F. 220 Ferrari, A. J. 35
Caussin, C. 45, 61, 457, 486 Ferrel, W. 11, 12
Celoria, G. 15, 60, 377-8, 397, 399-400 Fiala, A. D. 55
Chabot, J.-B. 395-6, 423 Foley, N. B. 244, 246
Ch'ang Ch'un 253-7 Fosse, Andreas 411
Chapront, J. 35, 41 Fotheringham, J. K. 16, 17, 99, 127, 132, 144,
Chapront-Touze, M. 35, 41 146, 334ff, 369
Chen Chiujin 277, 279 Freeman-Grenville, G. S. P. 435

551
552 Name index

Gassarus, Achilles Pirminius 408 Laplace, P-S. 10, 11

Gaufredus (of Vosium) 419 Laskar, J. 15, 41
Ginzel, F. K. 15, 60, 80, 334, 357, 362, 365, Laurentius de Brezina 407
377-8, 390, 405, 407, 416-17, 420, 445 Leo Diaconus 390
Glaber, Rudolphus 416 Levi ben Gerson 427
Glauert, H. 20, 21 Levi Donati, G. R. 377, 401, 403, 407-9, 421
Goad, C. 34 Li Chih-ch'ang 253, 256
Goldstein, B. R. 467 Liais, E. 12
Grayson, A. K. 144 Liu Baolin 55
Gregoras, Nicephorus 404-5 Liu Chao-yang 218, 220
Gregory XIII, Pope 31 Livy 358, 367
Longomontanus 411
Habash al-Hasib, 458-9, 470-1, 479-82 Lu Gwei-djen 275
Halley, E. 7-8, 62 Luke, Saint 385
Han Yu-shan 228
Hansen, P A . 17 Ma Tuan-lin 274
Heiskanen, W. 39 MacDonald, G. J. F. 336
Herodotus 342-4 Macey, S. L. 4
Herschel, W. 7 Magnus (of Reichersberg) 393, 402
Hide, R. 27 Major, J. S. 275
Hilton, J. L. 47 Marinus Neapolitanus 367-8
Hipparchus 8, 44, 94^7, 101-4, 106, 351-9, 369- Maspero, H. 278
72, 375, 461 Mayer, T. 10
Honorius (of Augsburg) 392 Meeus, J. 54
Houlden, M. A. 220, 236 Michael the Syrian 378-9, 395-6, 423
Hsueh Chung-san 230 Migne, J.-P 404
Huber, P. J. 112-15, 122-3, 129ff, 149, 151ff, Millard, A. R. 125-6
181ff, 194ff Morrison, L. V. 1, 27-9, 34-5, 41-6, 123, 501,
Hughes, D. W. 4 505-6, 508, 513-14
Hunger, H. 59, lllff, 129ff, 150ff, 177-8, 181, Muhammad, Prophet 434, 461
185-7, 193-4, 199, 203, 209 Mulholland, D. 35
Muller, P. M. 32, 34, 45, 62, 81, 389, 394, 398
Ibn al-Athir 439, 441, 450 Munk, W. H. 336
Ibn al-Jawzi 439-40, 450, 452 Muratori, L. A. 377
Ibn Amajur 457-9, 471-2, 479-82
Ibn Hayyan 438, 443 Nabonassar, King 95-6, 101-5, 112, 369
Ibn Iyas 432-3, 447-8 Nakiboglu, S. M. 36
Ibn Taghri Birdi 453 Nebuchadrezzar II, King 93-4, 105, 147, 149,
Ibn Tulun 448 152, 164-6, 193, 196-7
Ibn Yunus 45-6, 61, 71, 444, 456-62, 465ff, 493, Needham, J. 230, 275-6, 280-1
496 Neugebauer, O. 97, 110, 118, 343, 358, 461
Isaiah, Prophet 93 Nevill, E. 99, 334
Newcomb, S. 15, 17-20, 24, 25, 40, 456
Jean de Stavelot 405 Newhall, X X 35
Jeffreys, H. 13, 39 Newton, R. R. 34, 39, 44-6, 60, 72, 80, 118, 338,
Johannes de Muris 378, 427-8 352, 354, 361, 369, 377, 394, 456
Jones, H. S. 21-3, 34
Jordi, C. 27, 513 O'Keefe, J. A. 13
Oates, J. 95, 105-6, 162
Kanda Shigeru 266, 268, 273 Ouyang, I. 230
Kant, I. 6, 7
Keightley, D. N. 216, 218-19 Panchenko, D. 344
Kelvin, Lord 12, 13, 18 Pang, K. 218-20, 330
Kepler, J. 360, 378, 412 Pappus 351-4, 357
King, D. A. 464-5 Parker, R. A. 58, 116, 155, 176
Kugler, F. X. 110 Parkinson, J. H. 62
Kuo Shou-ching 8, 285, 295 Parpola, S. 125
Parry, J. V. L. 26
Lamb, H. H. 517 Pericles 346
Lambeck, K. 34, 36-7, 39, 81, 516 Pertz, G. H. 377
Lane-Poole, S. 445 Phlegon 359-60
 Name index 553

Pietruccio di Giacomo 407 Strabo 93, 97, 106, 353-4, 358, 461
Pinches, T. G. 109-10 Strassmaier, J. N. 110
Pindar 344-6 Su Sung 280
Pingre, A.-G. 83 Symeon (of Durham) 428-9
Pingree, D. J. 185 Szu-ma Ch'ien 223, 234
Pirazolli, P. A. 36, 517
Pliny 335, 342-4, 357-8, 360-1, 372-3 Taylor, G. I. 39
Plutarch 346-7, 360-6, 373, 411 Thales 342-4
Price, D. deS. 276, 280-1 Theon, 364-5, 375
Proclus, 367-8 Thomas (of Split) 401
Ptolemy, Claudius 3, 8, 44, 76, 93-104, 106, 112, Thompson, J. E. S. 59
127, 335, 351, 353-4, 357-8, 368-9, 373, 458, Thompson, R. C. 124
461, 467 Thucydides 227, 346-8
Ptolemy (of Lucca) 382 Tichtel, Johannes 415
Toomer, G. J. 3, 93-104, 127
Rabi, I. I. 26 Traziger, Adam 405
Rada, W. S. 432, 435-6, 438ff Tung Tso-pin 330
Rassam, H. 107, 111, 121, 142, 177
Ravn, E. 72, 97
Reade, J. 107 Veghi, Antonio dei 408-9
Regiomontanus 9, 10, 428 von Oppolzer, T R. 50, 121, 127, 339, 343, 345,
349, 354, 362, 364
Riccioli, P. J. B. 415, 421
Ristoro d'Arezzo 397-8
Roger (of Wendover) 425 Waley, A. 253
Rositz, Sigmund 407 Walker, C. B. F. 113, 115, 117-18, 135, 150-1,
Rowton, M. B. 145 156, 164, 169, 173-5, 188-9
Walther, B. 9, 10, 428
Sachs, A. J. 59, HOff, 129ff, 150ff, 177-8, 193-4, Wang Ling 276, 280-1
199, 209 Ward, C. G. 34
Said, S. S. 57, 432-3, 435-6, 438ff, 452, 457, 464, Wendelin 83, 415-16, 421
466ff, 488ff William (of Malmesbury) 424
Saladin 444 Wise, D. V. 13
Scaliger, J. J. 31 Wiseman, D. J. 136, 145
Schaumberger, J. 110-11, 115, 117, 129, 151,
155, 184^7, 199 Xenophon, 366
Schroeter, J. F. 50, 52, 69, 383 Xu Zhentao 216, 218, 297, 329
Seidelmann, P. K. 379
Seleucus I, King 115
Yabuuchi Kiyoshi 58
Seneca 54
Yazdigerd III, King 463
Sosigenes 29 Yahya ibn Abi Mansur 458
Standish, E. M. 35, 40 Yau, K. K. C. 214, 222-3, 254-5
Stephenson, F. R. 1, 27-9, 34-5, 46, 57, 74, 81-3, Yoder, C. F. 516
89-91, 113, 115, 117-19, 121, 123-4, 130,
148, 150, 180, 220, 222, 225, 233, 236, 244,
246, 254-5, 287, 297, 334, 411, 432-3, 435-6, Zawilski, M. 377, 392, 405-8, 413-15, 421, 424,
438ff, 452, 457, 464, 466ff, 488fT, 501, 505-6, 426
508, 513-14
 Subject index

Al-Hijrah 434-5, 461ff bureau of astronomy, Chinese 292

Almagest 3, 15, 17, 44, 47, 60, 76, 93-105, 166, 'Burning of the Books' 221
190, 335, 351, 353, 356-72, 458
almanacs, Babylonian 112, 114 calendar/date conversion 29, 59, 222, 230, 293,
animals and birds, effect or total eclipse on 257- 364, 379, 435-7
8, 261, 267, 377, 394, 397, 408-9, 423, 425, calendars
439 Alexandrian 364
altitude measurements 61, 71, 73, 427-8, 464ff, Babylonian 96, 115-16
511 Chinese 222, 229-30, 236, 263, 268, 285
ancient eclipses, early investigations of 1, 7-10, Coptic 438, 462, 468ff
14-17 Gregorian 29, 31-2, 436-7
Anglo-Saxon Chronicle 378 Egyptian 364, 369-72
angular units, Babylonian 118 Islamic 434ff, 461-3, 476
Anno Domini 30 Julian 29-32, 435-7, 463, 470, 488-9
annular solar eclipses 50-1, 62-3, 227, 232-3, Persian 462-3, 468ff
251, 258-9, 334^5, 346, 361, 3 8 2 ^ , 386, Roman republican 29
394, 404, 411, 448, 458, 467-8, 499-500 Syrian 379, 381, 462, 468ff
annularity, zone of 62-4, 69, 238, 382-3, 386, calendar treatises, Chinese 274, 285, 293-4, 297
416, 423 cause of eclipses, early understanding of 123,
anomalistic (apparent) solar day 3-4 254, 342-3, 346-7, 377
Arab astronomers, medieval 8, 456ff central solar eclipses, analysis of 61-70
Arab chroniclers and chronicles, medieval 43Iff ceremonies, eclipse 223, 262
Arab eclipse observations, medieval 61, 71, 73, Chinese astronomers, 213, 230, 274, 289
75-6, 43 Iff, 456ff, 507, 511 Chinese capital, observations at 230-1, 238, 246,
Arabic numerals 474 283
Arabic script 459-60, 469 Chinese eclipse observations 59-60, 62, 67-8,
archons, Athenian 102^, 126, 347, 349, 366, 373 73-5, 213ff, 273ff, 507, 510, 511
artificial satellite observations 34-5, 37, 506, 516 chromosphere, solar 395
Assyrian Chronicle 125-7 chronicles, medieval 48, 60
Assyrian eclipse observations 124-7 chronological eras
astrolabe 464^6, 478-9 Alexander 378
astrology/astrologers 48, 93, 95, 112, 124-5, Arsacid 113, 115, 160
230-1, 236, 439, 446 Byzantine 391, 397
Astrological Treatise (Chinese) 228-9, 232, 241, Christian 29-30, 378
245-6, 248, 251, 262-3, 274, 285, 293 Diocletian 29, 445
Astronomer-Royal, Chinese 247, 291 Foundation of Rome 29-30, 342
astronomical terminology, early 112, 120, 129, Nabonassar 96, 102-4, 364, 371
132-3, 150, 193-202, 217, 223-4, 232, 286, Olympiads 29, 342-3, 345, 359-60
381, 384, 431 Seleucid 29, 113, 115, 121, 160, 378, 395, 423,
atomic clocks 26 489
azimuth, lunar 283-4, 299, 307-10 Yazdigerd 463
chronology, Babylonian 145-6
Babylonian astronomers 2-3, 9, 59, 71-2, 112, Ch'un-ch'iu 220-7
118-24, 343 Classical Chinese 214
Babylonian eclipse observations 8, 59, 64-7, 72, clay tablets
79, 93ff, 128ff, 147ff, 193ff, 335-6, 510, 513 Assyrian, 124
Bamboo Annals 219-20 Babylonian 93, 95, 107-14, 117
bodily tides 36 clepsydra (water clock) 59-60, 71, 121, 276, 278,
British Museum 107, 110-13, 118, 124, 126, 129, 281,466,507,511,513
132, 135, 185 clima (latitude zone) 3 5 3 ^

554
 Subject index 555

clock drift 190-1 equation of time 4-5

clouds, interference of observation by 199-202, equinox and related measurements 44-5
209 estimates of eclipse magnitude at moonrise or
colour of eclipsed Moon 57, 158-9, 177, 375, moonset 208-11
428-9, 453 European astronomers, medieval 60
compendia on astronomy, Arabic 431 European chroniclers and chronicles, medieval
consuls, Roman 126 376ff, 432
core-mantle coupling 23, 27, 516 European history, ancient 344ff
corona, solar 144-5, 219, 226, 360-1, 390, 394 European eclipse records 60, 62-3, 68-9, 334ff,
crescent, solar (in eclipse) 63, 71, 335, 346, 348, 376ff
366, 386, 416-20, 424 explanation of eclipses, ancient 59
crescent visibility (lunar) and calendar 116, 119,
434-46 false sightings of eclipses 246
Crucifixion, darkness at 385, 390, 399 'Five-Element' Treatise, Chinese 228, 231, 233,
culmination of stars (see also ziqpu stars) 188-90 241, 251, 257
cuneiform texts 95, 97, 101, 108-10, 124 fluctuations
in mean motion of Moon 17-23
AT 24, 26-8, passim in mean motion of Sun and planets 17-23
DE102 (lunar ephemeris) 35, 40, 42 fundamental plane 69-70
damaged texts 1 9 3 ^ , 196-7, 208
darkness during large solar eclipse 54, 62, 227, geographical co-ordinates, lists of 44, 97, 225,
234, 237, 242, 251, 257, 259-61, 267-8, 338- 335-6, 386-9, 434
46, 348-50, 367-8, 381-2, 384, 390ff, 438-9, gnomon 437-8, 471
442, 44^-5, 447-8 'goal-year texts' 114-15, 117, 129-30, 134-6, 155,
date determination, astronomical 117-18, 120- 157, 170, 202
1, 133, 138, 142, 148-50, 155-7, 160, 170-2, 'Great Empirical Term' 18, 20-1, 24
177, 199, 338ff, 347, 351ff Greek astronomers, 3, 44, 60, 96
daylight, seasonal variation of 2 Greek history and literature, ancient 334, 337
decade fluctuations 27-9, 513, 516-17 Greek observations, ancient 44, 47, 334ff
decipherment of cuneiform texts 110 Greenwich mean time (GMT) 5-6, 23
delay in detection of a solar eclipse 73, 466, 475
destruction of records 95
Halley's comet 110, 113, 150, 227
dew, fall of during solar eclipse 423
histograms of residuals 511
diaries, Babylonian astronomical 109, 112-15,
horizon, observation of Moon on 1 9 3 ^
117, 129-30, 133-4, 193, 199, 203
horizon profile 72, 78-9, 97, 449, 451
discordant results 507-10
hours, division of day into 2-3
distance of Moon derived from eclipse 351-2,
358-9
double dates for lunar eclipses 97, 451, 476 'identification game' 118
'double-dawn' 219-20 Ides 367, 378-90
double hours, Chinese 247, 252, 274-9, 288-94, Indiction 387
300-8, 312-14 inscriptions, Babylonian 107
duration of eclipse darkness, excessive 385 Indian eclipse records, lack of early 58
duration of totality 398, 406 intercalation 30, 129, 155, 162, 165
dynastic histories, Chinese 228, 245 international atomic time (TAI) 6, 26
dynasties, Chinese 214-15
durations of eclipses 53 j =2 (lunar ephemeris) 34, 40-2
Japanese eclipse observations 266-8, 273
Earth's shadow 151 Jesuit astronomers 284, 286-7
Earth's rotation Juchen 250-1
fluctuations in 2, 6-8, 17-18, 20-1, 26-9, 513- Julian calendar/date 97, 99, 116, 121, 128, 225,
16 230, 254, 260, 338, 364, 369, 378-81, 387
tidal acceleration 6-8, 11-14, 16, 34, 37-9, Julian century 9, 12, 24
45-7, 506 Julian day number 31-2
eclipse tables, Babylonian 149, 152 Julian ephemeris date/days 32, 38
Egypt, ancient 2, 58 Julian year 30-1
epagomenal days 463 Jupiter, occultation by eclipsed Moon 428-30
ephemeris time (ET) 24-6
eponym 126 Kalends 378-80
equal (equinoctial) hours 3, 99-104, 364-5, 370- Kallippic cycle 369-70
2, 466, 478, 481-3, 489-90 Korean eclipse observations 262-5, 273
556 Subject index

Late Babylonian astronomical texts 107, 109-20, non-tidal variations in length of day 7, 12, 23,
128ff, 147ff 27-9, 37-8, 506, 508-9, 514
Latin, most medieval European records in 376,
387 observatories, early 93, 230, 262, 295
latitude, early determination of terrestrial 353-4 occultations of stars, lunar 1, 21, 27, 34, 41,
leap years 379 46-7, 505, 510
length of day oceanic tides 36-7, 39-40
standard reference 25, 28 octave, of a festival 403, 407, 419-20
variation 1-2, 7, 11-12, 16, 20, 27-9, 37, 43, oracle bones, Chinese 215-19
48, 501, 513-16 origin of Moon, resonance theory of 13
light intensity changes during solar eclipse 50
local apparent time (LT) 3-4, 6, passim parallax, solar 351-2
local mean time (LMT) 4-5 partial lunar eclipse, duration of maximal phase
longitude, early determination of terrestrial 48, 151, 178-81
57, 354, 358, 459, 461, 491-2 partial solar eclipse, analysis of 70-1, 223, 233ff,
lunar limb profile 410-11 348, 366, 386, 404, 416-21, 431, 442
lunar eclipse timings 147ff, 274, 276-7, 279, 281, Peloponnesian War 346-7
295, 297, 335, 358, 368-72, 476ff penumbral lunar eclipses 55, 122-3, 185, 476
lunar eclipses photosphere, solar 410
cause of 55-7 poems/poets, ancient Greek 338-9, 344-5
frequency of 55, 57 planetary conjunctions, close 45
'population bias' 80-1
lunar ephemeris/theory 17-18, 20, 25
portents/omens, eclipses as 124, 144, 222, 230-2,
lunar laser ranging 35
lunar secular acceleration 2, 7-17, 21-2, 24-5, 234, 237, 242, 432-3
post-glacial uplift 23, 37, 516
33, 132, 337
prayers, Islamic 437-8, 442, 446-8
'lunar sixes' 119-20
pre-telescopic observations 43-4
lunar tidal acceleration (h) 33-7, 40-2
prediction of eclipses, early attempts at 73, 101,
112, 114, 122-4, 156-7, 165-6, 196-7, 246,
magnitude, early units of 59, 74, 76, 77, 100, 252-3, 284, 342-4, 439, 446, 459, 470, 511
102, 133-4, 139, 142, 202, 209-10, 311, 252- preservation, degree of textual 113-14
3, 288-93, 303, 311-14, 353, 371, 404, 421, printing, block 213, 238-9, 248-9
463, 466ff printed texts 377
magnitude, eclipse 51-3, 56-7, 137-41, 241, 243, prominence, solar 399
253, 257, 288-93, 303, 310-14, 351ff, 382, provincial observers, Chinese 230-1, 283
387, 422, 427-8, 442-3, 463, 470ff, 494-8
manuscripts, 377, 432 recession of Moon (tidal) 12, 16, 24, 36
Mayan eclipse records, lack of 58-9 redundant solutions 503
mean longitude refraction 72, 470
of Moon 23-5 rejection of records 167-8, 185, 385
of Sun 2 3 ^ , 40 Religious Chronicle, Babylonian 143-6
mean solar day 3, 6, 16 right ascension, solar 231-2, 234ff, 246ff
Mercury transits 21, 34 ring phase (of annular eclipse) 62-3, 232-3, 258-
meridian 446, 464, 471, 480, 482-3, 485 9, 348, 361, 382, 386, 394, 411, 423
Metonic cycle 116, 369 Roman history 334
'minute of day' 461 romanisation (systems of) 214
monasteries, medieval European 376, 378, 39Iff
Mongols 250 saros cycle 114
month, change in length of synodic 36 Scriptores Rerum Italicarum 377
Monumenta Germaniae Historica, Scriptores 377 sea, total eclipse observations at, 348-50
Moon rising or setting eclipsed 102, 193ff, 303- sea-level variations 36-7, 517
seasonal (unequal) hours 3, 99-104, 370-1, 381,
4, 308, 335, 357, 369, 372, 432, 436, 450-2,
466, 470-2, 480-3, 486, 489
463, 479-80, 485, 496-9
second, definition of 25-6
sexagenary cycle 216, 218, 222, 293
negative (BC) years 30, 40, 99, 104 sexagesimal arithmetic 461
'Newtonian time' 23-5 Shou-shih-li I 285-7, 293-5, 297
night-watches sign convention for latitude and longitude 63
Babylonian 164 solar eclipse magnitudes, analysis of 74-7
Chinese 274, 279-83, 297-300, 301-2, 305 solar eclipse timings, analysis of 71-4, 131-7,
Nones 378-80 247, 252, 274, 277-9, 288-97, 335, 364-5,
'Normal stars' 129-30, 150 427-8, 468ff
 Subject index 557

solar eclipses, cause of 48-54 time units, early 59, 60, 72, 118-19, 247, 252,
solar ephemeris 40 275-9, 288-94, 300-8, 312-14
solar secular acceleration 14-17, 22-3, 33-4, 132, timed and untimed observations, comparison of
337 502ff
sources of historical records 58, 245, 274, 284-6 timing accuracy, early 121, 124, 147-8, 278
spline fitting, cubic 509-14 torque, electromagnetic 516-17
stars and/or planets, visibility of during solar total solar eclipses 50-1, 62-70, 129-31, 143-4,
eclipse 50, 62, 64, 129-31, 227, 237, 241-3, 221, 223, 233ff, 334-5, 338^6, 348-64, 377,
247, 254, 258-61, 267-8, 346, 348, 359-60, 385, 387ff, 431, 438-41, 443-5, 448, 475
364, 368, 381-4, 387, 390ff, 438-9, 442, 444- totality, zone of 50, 52, 6 2 ^ , 66-9, 226, 235, 238,
5 242, 247-8, 251, 255-7, 263-65, 268, 350-1,
stone inscription describing total eclipse, Italian 362, 390, 392-3, 398, 404, 408
401, 403 twilight 280
Sun rising or setting eclipsed 76-9, 131, 133,
141ff, 236, 368, 425-6, 432, 436, 438, 449- universal time (UT) 6, passim
50, 463, 472, 496-9
sundial 3 ^ visibility of stars during annular eclipses 227,
sunrise and sunset 346, 348, 381-4
accuracy of measured times 278 visual acuity 467
as reference moments 71-2, 129ff, 147ff Vulgate Bible 385
early definition of 140
surface (area) magnitude of eclipse 353, 463, 475, water, observation of eclipse by reflection in 417,
482, 489, 494, 497 419, 421, 463-4, 472
Syriac script 395-6 water clock (see clepsydra)
weekdays 379, 381, 387ff, 436
tables of eclipses, Babylonian 114-15, 117 weighting of observations 192, 502, 510
techniques for observation of solar eclipses, early Wen-hsien T'ung-k'ao 253, 274, 277, 285-6, 288-
54 91, 294, 297-8, 300, 302-5
telescopic observations 1, 17-23, 46, 506-8, 513
temperature fall during solar eclipse 360-1 Yuan-shih 250, 252, 276, 285, 293-4, 297-300,
terrestrial dynamical time (TDT) 26 302-5
terrestrial time (TT) 26, passim
textual errors 173-5
ziggurat, Babylonian 72, 93, 97-8
tides and tidal friction 7, 11-14, 16, 23, 27-8, 33,
Zij (astronomical handbook) 456-9, 471, 479,
34-5, 36-9, 506, 513
481-2, 488, 496
'time degrees' 59, 433, 447, 466
ziqpu stars 188-90
time of moonrise or moonset as a function of
zodiac/zodiacal signs 130, 138, 150, 157, 159,
AT 86-7, 194 194, 203