Instant Ebook Access, One Click Away – Begin at ebookgate.

com

Treatise on Geophysics Deep Earth Seismology 2nd

Edition Barbara Romanowicz

https://ebookgate.com/product/treatise-on-geophysics-deep-
earth-seismology-2nd-edition-barbara-romanowicz/

OR CLICK BUTTON

DOWLOAD EBOOK

Get Instant Ebook Downloads – Browse at https://ebookgate.com

Click here to visit ebookgate.com and download ebook now
Instant digital products (PDF, ePub, MOBI) available
Download now and explore formats that suit you...

Seismology and Structure of the Earth Treatise on

Geophysics 1st Edition Barbara Romanowicz

https://ebookgate.com/product/seismology-and-structure-of-the-earth-
treatise-on-geophysics-1st-edition-barbara-romanowicz/

ebookgate.com

Treatise on Geophysics Volume 2 Mineral Physics 2nd

Edition Price

https://ebookgate.com/product/treatise-on-geophysics-volume-2-mineral-
physics-2nd-edition-price/

ebookgate.com

Geodesy Treatise on Geophysics 1st Edition Tom Herring

https://ebookgate.com/product/geodesy-treatise-on-geophysics-1st-
edition-tom-herring/

ebookgate.com

An Introduction to Seismology Earthquakes and Earth

Structure 1st Edition Seth Stein

https://ebookgate.com/product/an-introduction-to-seismology-
earthquakes-and-earth-structure-1st-edition-seth-stein/

ebookgate.com
Treatise on Money Joseph Schumpeter

https://ebookgate.com/product/treatise-on-money-joseph-schumpeter/

ebookgate.com

The Mantle and Core Treatise on Geochemistry Volume 2

Treatise on Geochemistry 1st Edition R.W. Carlson

https://ebookgate.com/product/the-mantle-and-core-treatise-on-
geochemistry-volume-2-treatise-on-geochemistry-1st-edition-r-w-
carlson/
ebookgate.com

A Treatise on War Inflation William Fellner

https://ebookgate.com/product/a-treatise-on-war-inflation-william-
fellner/

ebookgate.com

Fundamentals of Geophysics 2nd Edition William Lowrie

https://ebookgate.com/product/fundamentals-of-geophysics-2nd-edition-
william-lowrie/

ebookgate.com

Treatise on Geochemistry Volume 1 Meteorites and

Cosmochemical Processes 2nd Edition Andrew M. Davis (Ed.)

https://ebookgate.com/product/treatise-on-geochemistry-
volume-1-meteorites-and-cosmochemical-processes-2nd-edition-andrew-m-
davis-ed/
ebookgate.com
 1.01 Deep Earth Seismology: An Introduction and Overview
AM Dziewonski, Harvard University, Cambridge, MA, USA
BA Romanowicz, University of California, Berkeley, CA, USA; Collège de France, Paris, France
ã 2015 Elsevier B.V. All rights reserved.

1.01.1 Developments from the Late Nineteenth Century until the Early 1950s 2
1.01.2 Developments from 1950s to the Early 1980s 4
1.01.3 From 1980 to Present: The Era of Tomography and Broadband Digital Seismic Networks 9
1.01.4 Current Issues in Global Tomography 14
1.01.4.1 Resolving Power of Datasets Used for Constructing Models 16
1.01.4.2 Theoretical Assumptions 17
1.01.4.3 Robust Features of Current Global Mantle Models and Their Implications 19
1.01.4.4 Stability of the Planetary-Scale Heterogeneities 23
1.01.4.5 The Need for Consideration of More Complete Modeling of Mantle Flow 24
References 25

Applications of seismology to the study of the Earth’s interior in the measurements of normal-mode and long-period surface
are only a little over 100 years old. Its tools in determining the waves. Two chapters are devoted to the computation of syn-
properties of inaccessible Earth are the most powerful among thetic seismograms in the presence of lateral heterogeneity,
all geophysical methods. The principal reasons are the avail- suitable for the case of body waves (see Chapters 1.05 and
ability of natural (earthquakes) or controlled (explosions and 1.06). Significant progress has recently been made in the com-
vibrators) sources of elastic waves and their relatively low putation of synthetic seismograms in a 3-D Earth using numer-
attenuation with distance. Seismological methods span some ical methods. A review is given in Chapter 1.07. With the
six orders of magnitude in frequency, and the depth of an deployment of dense regional arrays of broadband seismome-
investigated structure may range from a few meters in engineer- ters, another area of rapid progress has been that of the adap-
ing applications to the center of the Earth. Progress in seismol- tation of methodologies first developed in exploration
ogy has been achieved through developments on several seismology to the case of fine structure imaging of the crust
fronts: theory, instrumentation, and its deployment, as well and upper mantle at larger scale. These approaches are
as computational resources. described in Chapter 1.08 for passive-source applications and
Even though the studies of earthquakes and the Earth’s in Chapter 1.15 for the case of active sources. The realization of
structure are closely related, the two subjects are often dis- the importance of anisotropy in the Earth has led to theoretical
cussed separately. This volume is devoted to the Earth’s and methodological developments (see Chapters 1.09 and
structure and Volume 4 to studies of earthquakes. Neverthe- 1.18). Note that the issue of anisotropy is also discussed in
less, the relationship is intimate. For example, it is possible to Chapter 1.19 in the context of the inversion of surface-wave
formulate an inverse problem in which earthquake locations data. Inverse methods, in particular in the context of global
are sought simultaneously with the parameters of the Earth’s and regional tomography, are discussed in Chapter 1.10.
structure, including three-dimensional (3-D) models (see In the second part of Volume 1, reviews of the status of our
Chapter 1.10). knowledge on the structure of the Earth’s shallow layers are
In the past 25 years, important progress has been made on presented, starting with a global review of the Earth’s crustal
several fronts: (1) the development of broadband digital structure (see Chapter 1.11). During the last decade, there has
instrumentation (see Chapter 1.02), which has allowed the been rapid development in using the Earth’s noise as a source
construction of digital seismic databases of unprecedented of the signal. A review of these developments is presented in
quality at both the global and the regional scales; (2) the Chapter 1.12. Two chapters discuss regional structure in the
development of powerful data analysis tools, made possible oceans: Chapter 1.13 for mid-ocean ridges and Chapter 1.14
by ever more efficient computer technology; and (3) theoreti- for hot-spot swells. Chapter 1.18 presents results of studying
cal progress in the forward and inverse computation of the anisotropy in subduction zones with particular attention
effects of strong lateral heterogeneity on seismic-wave propa- devoted to the flow-induced preferential orientation of olivine
gation. The combination of these factors has led to much crystals. Finally, two chapters are devoted to the results of
improved images of structure at the global and regional scale, regional experiments: upper-mantle studies using data from
often helped by the inclusion of constraints from other types of portable broadband experiments (see Chapter 1.16) and
data, primarily from the fields of mineral physics and geody- crustal studies, specifically in Europe, from high-resolution
namics. This volume is thus divided into four parts. The first long-range active-source experiments (see Chapter 1.17).
part principally covers seismic instrumentation, theoretical The third part of this volume concerns the Earth’s deep
developments, and seismic data analysis techniques. Chapter structure, divided into its main units: the upper mantle (see
1.03 discusses the state of the art in the computation of the Chapter 1.19); the transition zone and upper-mantle discon-
Earth’s normal modes, while Chapter 1.04 describes progress tinuities (see Chapter 1.21); regional tomography of

Treatise on Geophysics, Second Edition http://dx.doi.org/10.1016/B978-0-444-53802-4.00001-4 1

2 Deep Earth Seismology: An Introduction and Overview

subducted slabs, with particular attention given their stagna-

tion at the bottom of the transition zone (Chapter 1.20); the
lower mantle; and the highly complex D00 region at the base of
the mantle (Chapter 1.22) as well as the Earth’s core (see
Chapter 1.23). Chapter 1.24 is devoted to the subject of scat-
tering in the Earth and Chapter 1.25 to that of attenuation.
Finally, the fourth part of this volume comprises two chapters,
in which constraints on the Earth’s structure from fields other
than seismology, mineral physics (see Chapter 1.26) and geo-
dynamics (see Chapter 1.27), are discussed.
This volume addresses various aspects of ‘structural seismol-
Figure 1 The historical first recording of a teleseismic event: an
ogy’ and its applications to other fields of Earth sciences. Not all earthquake in Japan recorded in Potsdam on a tiltmeter designed by von
the subjects are covered in comparable detail, even though the Rebeur-Paschwitz. The early seismograms had difficulty with damping
completeness of the coverage was the initial objective of the the pendulum motion and made phase identification difficult.
editors. Compared to the 2007 edition of Volume 1 of Treatise Reproduced from von Rebeur-Paschwitz (1895) Horizontalpendal-
on Geophysics, this edition contains four new (Chapters 1.02, Beobachtungen auf der Kaiserlichen Universitats-Sternwarte zu
1.12, 1.18, and 1.20). Most of the other chapters have been Strassburg 1892–1894. Gerlands Beiträge zur Geophysik 2: 211–536.
significantly updated, except Chapters 1.14, 1.15, and 1.16,
which are reprinted ‘as is’ from the 2007 edition. 1000
In what follows, we briefly describe the developments in 1
seismology from the end of the nineteenth century until the Tg = 90 s

Magnification
present, with the main emphasis on the development of instru- Ts = 15 s
mentation and its deployment, because seismology is a data-
100
driven science. An account of the history of seismology can be
found, among others, in Agnew et al. (2002). We also present Slope = 3
-1
our point of view, which some may consider controversial,
specifically on current issues in global tomography and inter- 10
pretation of the 3-D models not discussed in any of the chapters. 0.001 0.01 0.1 1 10
We justify bringing these issues forward because of our belief Frequency (Hz)
that interpretation of tomographic results in terms of mantle
dynamics does not match the robustness of models built using Figure 2 Plot of the ground-motion (amplitude) response of a World-
data that have good resolution at all depths in the mantle. Wide Standardized Seismograph Network (WWSSN) station with a
seismograph free period (Ts) of 15 s and galvanometer with a free period
(Tg) of 90 s. The segment between these two periods has a flat velocity
response, characteristic of broadband seismometers. The response in
1.01.1 Developments from the Late Nineteenth modern instruments is shaped electronically; a typical FDSN station has a
Century until the Early 1950s flat velocity response from 5 Hz to 360 s. Reproduced from Wielandt
E (2002) Seismic sensors and their calibration. In: Bormann IP (ed.)
The theoretical beginnings of seismology may be traced to the IASPEI New Manual of Seismological Observatory Practice, vol. I,
eighteenth- and nineteenth-century studies of elasticity and pp. 1–46. Potsdam: GeoForschungsZentrum Potsdam, Chapter 1.06.
propagation of elastic waves in solids. Lord Kelvin provided
the first numerical estimate of the period of the fundamental
vibrational mode (0S2) in 1863, but the development of the seismograph with controlled damping was built by Wiechert
proper theory for a homogeneous sphere had to wait nearly in 1904. Soon afterward, Galitzin (1914) developed an
50 years (Love, 1911). Lord Rayleigh solved the problem of electromagnetic seismograph system, where the motion of the
propagation of surface waves in an elastic half-space in 1877. seismometer’s pendulum generated an electric current
This preceded the first mechanical seismographs, which were by motion of a coil in the magnetic field. This current was,
developed in the 1880s. Originally, the seismographs had very in turn, carried to a galvanometer; the rotation of the
low sensitivity and were used for the recording of local galvanometer’s coil in a magnetic field was recorded on photo-
earthquakes. The history of global seismology begins with the graphic paper by a beam of light reflected from a mirror attached
recording of an earthquake in Japan on 19 April 1889 by von to the coil. The response of the system depended on the sensi-
Rebeur-Paschwitz. He associated a disturbance recorded on a tivity and free period of the seismometer and of the galvanom-
tiltmeter, used to study the Earth’s tides, with the reports of a eter and their damping. While the system was more complex, it
great earthquake in Japan. Figure 1 shows a copy of this record- allowed for much more flexibility in selecting a desired
ing as published in Nature (von Rebeur-Paschwitz, 1889, 1895). response. Figure 2 shows the response of the seismograph–
The early seismographs were mechanical pendulums with no galvanometer system and gives an idea of the way it could be
damping, other than friction. Their magnifications (the ratio of shaped by the choice of different free periods of the system’s
the amplitude on a seismogram to the actual ground motion) components. With gradual improvements, the seismometer–
were very low, and because of the lack of damping, the records galvanometer system, recording on photographic paper, was
were very oscillatory and it was difficult to distinguish commonly used during the following 60–70 years, when it
the arrivals of different phases. An improved mechanical was gradually replaced by digital systems (see Chapter 1.02).
 Deep Earth Seismology: An Introduction and Overview 3

With the improvement of the recording systems technol- 14

ogy, phase identification became easier, and it was possible to P
identify P-arrivals (primary) corresponding to compressional
12
waves, S-arrivals (secondary) corresponding to shear waves,

Velocity (km s−1)

and L-arrivals, sometimes called ‘the main phase,’ correspond-
10
ing to surface waves. The surface waves caused some confusion
because there was also a transverse motion, not predicted by
Rayleigh. It was not until 1911 that Love showed that trans- 8
S Seismic velocities
versely polarized surface waves can propagate in a layered
Gutenberg
Earth. 6 Jeffreys-Bullen
Progress in the first decade of the twentieth century was
rapid. Some classical problems such as computation and inver- 4
sion of travel times for the velocity structure were solved by B C D E F G
0
(Benndorf, 1905, 1906; Herglotz, 1907; Knott, 1899; Wiechert, 0 1000 2000 3000 4000 5000 6000
1907; Z€ oppritz, 1907) independently developed equations for Depth (km)
the amplitude of reflected and transmitted waves at the bound-
Figure 3 Comparison of the seismic velocity models of Gutenberg and
ary between two elastic media.
Jeffreys, both built in the 1930s. The principal difference between the
As regards the Earth’s structure, there was a paper by models is the presence of the low-velocity zone in the Gutenberg model
Oldham (1906), in which he proposed the existence of the and the structure near the inner–outer core boundary where the low-
Earth’s core, although there has been some confusion in iden- velocity zone in the Jeffreys model is erroneous and the velocity increase
tification of phases: what he thought to be a delayed S-wave in the inner core is larger than in the Gutenberg model. With the
was actually an SS (e.g., Schweitzer, 2007). Gutenberg (1913) exception of the transition zone (400–650 km depth), the modern models
properly identified reflections from the core–mantle boundary are not very different. Reproduced from Anderson DL (1963) Recent
(CMB) and determined the radius of the core quite accurately, evidence concerning the structure and composition of the Earth’s mantle.
and Jeffreys (1926) showed that the core is liquid. Mohorovičić Physics and Chemistry of the Earth 6: 1–129.
(1910) discovered the boundary between the crust and the
upper mantle, thus beginning the era of studies of the crust
and lithosphere, which greatly accelerated after World War II. and Gutenberg was the existence of a low-velocity zone in the
The first global seismographic networks (GSNs) were estab- depth range 100–200 km in the upper mantle. There were very
lished in the early years of the twentieth century. The first one hot debates on this issue; it can now be explained by the fact
was deployed by John Milne in various countries of the British that they used data from tectonically different regions; there is
Commonwealth with the support of the British Association for a pronounced low-velocity zone in the western United States,
the Advancement of Science and eventually consisted of 30 sta- but not under the Eurasian shield regions.
tions (Adams et al., 2002). The Jesuit Network was established With internal reflections and conversions of waves at the
soon afterward, with a particularly large number of instru- boundaries, seismologists developed a system of phase identifi-
ments in the United States, but also including stations on all cation that reflects a combination of the types of waves (P or S),
continents (Udias and Stauder, 2002). With a global coverage the region in which they propagate (K and I for the P-waves in
sufficient to locate large earthquakes, informal bulletins were the outer and inner core, respectively; PKIKP designates a phase
published using the location method developed by Geiger that travels as P in the mantle, P in the outer core, and P in the
(1910, 1912), which (with many modifications) is still used inner core) and the boundary at which they were reflected (c for
today. In 1922, the International Seismological Summary CMB and i for the inner–outer core boundary). A shear wave
(ISS), with international governance, was established under reflected once from the free surface at the midpoint of its path is
Professor Turner of the University of Oxford with the charge designated by SS; higher multiple reflections, like SSS or SSSSS,
to produce ‘definitive global catalogs’ from 1918 onward. can be observed by sampling a large volume of the Earth along
The slow progress in unraveling the Earth’s structure culmi- their paths. For earthquakes with a finite focal depth, the P- and
nated in the 1930s with the discovery of the inner core by S-waves traveling upward from the source have designation of p
Lehmann (1936) and the compressional velocity, shear or s; following reflection at the surface, they represent the
velocity, and density models by Gutenberg (1913), Jeffreys so-called depth phases (e.g., pP and sP); the travel-time differ-
(1926), and Bullen (1940). The Gutenberg and Jeffreys velocity ence between the arrival of pP and P strongly depends on focal
models are shown in Figure 3; except for the details of the depth.
upper-mantle structure, these models are very similar to the Figure 4 shows examples of various seismic phases, and
modern ones. The low-velocity zone above the inner–outer Figure 5 is the graphic representation of the travel times as a
core boundary in the model of Jeffreys illustrates the some- function of distance computed by Jeffreys and Bullen (1940)
times unfortunate tendency of seismologists to introduce phys- for the model of Jeffreys (1939). It is remarkable that this set of
ically implausible features in the model in order to match the tables, including predictions for different focal depths, was
data; Jeffreys needed to add a 2 s delay to match the inner-core calculated using a manual mechanical calculator! The data
travel times and accomplished it by inserting this feature, used by Jeffreys were extracted from the ISS, the precursor of
which is impossible to reconcile with the chemical and phys- the International Seismological Centre (ISC), which (with
ical properties of materials in this depth range (Birch, 1952). international financial support and governance) resumed the
The other important difference between the models of Jeffreys ISS role in 1964 and continues until today.
4 Deep Earth Seismology: An Introduction and Overview

ScS PS P

P
pP
sP S
SKS pPcP

PP
SKP
SP
Solid
SKKS inner
core SS

sPS
Fluid outer core PPP
SKKP

PPS
PKIKP
Solid mantle
PKJKP SSS

P rays
pPcPSKKP
SKPPKP S rays
Figure 4 Examples of seismic rays and their nomenclature. The most commonly identified phases used in earthquake location are the first arriving
phases: P and PKIKP. Reproduced from Stein S and Wysession M (2003) An Introduction to Seismology, Earthquakes and Earth Structure. Oxford:
Blackwell, ISBN: 0865420785.

Bullen (1949) divided the Earth into a number of concen- surface waves (Rayleigh and Love) in a layered medium
tric shells, designated by letters from A to F; in this division, the with an arbitrary number of layers over a half-space. It
lower mantle was designated by the letter D”. when Bullen involved multiplication of matrices, one for each layer,
recognized that the deepest 150 km of the lower mantle had changing the wave number for a fixed frequency such as
an anomalously flat velocity gradient, he divided the region to match the boundary conditions (vanishing of stresses) at
D into D0 and D00 . More recently, and not entirely correctly, D00 the free surface. Because of the enormous amount of calcu-
came to signify the structure in the deepest 300 km, or so, of lations to be performed, it required application of an elec-
the lower mantle, which is characterized by a still-growing tronic computer, and its application opened yet a new era
collection of structural and compositional complexities (see in seismology. The Haskell’s matrix method has been
Chapter 1.22). adapted to other problems in seismology, such as calcula-
It was recognized relatively early that the dispersion of tion of synthetic seismograms using the ‘reflectivity method’
surface waves was different in the continents than in the (Fuchs and Müller, 1971). Electronic computers were at first
oceans, with an indication that the oceanic crust was signifi- very expensive and rare, and it was not until the 1960s that
cantly thinner. Computing the dispersion of surface waves was they became generally available at universities (Haskell
algebraically and numerically difficult; the correct formulas for worked at the Air Force Cambridge Research Laboratories).
the dispersion of Rayleigh waves in a layer over a half-space Surface-wave dispersion began to be studied intensively in
were formulated by Stoneley (1928), and the case of the two the 1950s principally at the Lamont Geological Observatory of
layers over a half-space could be solved only for a very specific Columbia University, primarily by Ewing and Press, who
set of parameters. observed mantle waves in the 1952 Kamchatka earthquake,
identifying arrivals from R6 to R15 (see Chapter 1.04) and
measuring their group velocities up to a period of 500 s (Ewing
1.01.2 Developments from 1950s to the Early 1980s and Press, 1954). Regional measurements of surface-wave dis-
persion were initiated by Press (1956). A monograph by Ewing
It must have been frustrating for seismologists not to be et al. (1957) summarizes the state of the knowledge on
able to use information about the Earth’s structure con- seismic-wave propagation in layered media at that time.
tained in the most prominent features of the seismograms: Ewing and Press also developed a true long-period seismo-
the dispersed surface waves. This changed when Haskell graph, which was capable of recording mantle waves even for
(1953) adapted to the case of elastic media the method moderately sized earthquakes. This instrument was deployed
first proposed by Thomson (1950) in the acoustics case. at 10 globally distributed International Geophysical Year net-
The approach made it possible to compute dispersion of work stations operated by Lamont.
 Deep Earth Seismology: An Introduction and Overview 5

similar free oscillation periods for the gravest modes but differ
at shorter periods by 1–2%.
When the greatest instrumentally recorded earthquake of
20 May 1960 occurred in Chile, seismologists had all the tools
(theory, computers, and instrumentation) needed to measure
and interpret its recordings. Three back-to-back papers in a
1961 issue of the Journal of Geophysical Research reported the
first correct measurements of free oscillation periods: Benioff
et al., Ness et al. (1961), and Alsop et al. (1961). The observa-
tions were made on strainmeters in Isabella, Ñaña, and
Ogdensburg; seismographs at Pasadena; and a gravimeter at
the University of California, Los Angeles (UCLA). All three
studies agreed in mode identification and found that their
periods were very close to those predicted by the existing
Earth models (Pekeris et al., 1961a). More detailed studies of
the spectra revealed that they are split; the effect of the Earth’s
rotation was shown to explain this effect (Backus and Gilbert,
1961; Pekeris et al., 1961b). Thus, normal-mode seismology
was born. Progress in the theory, particularly considering the
effect of lateral heterogeneities and mode coupling, would
extend over decades to come. First attempts at inversion of
normal-mode data were not particularly successful; the density
model of Landisman et al. (1965) was flat throughout the
lower mantle, implying either a strong radial heterogeneity or
an immensely superadiabatic temperature gradient.
Backus and Gilbert (1967), Backus and Gilbert (1968), and
Backus and Gilbert (1970) provided the formal background for
consideration of geophysical inverse problems, and even
though their own experiments with inversion of normal-
mode periods for the velocities and density, using a subset of
normal-mode data, were discouraging (two very different
models were found fitting the data nearly exactly; Backus and
Gilbert, 1968), the idea of resolving kernels and trade-offs
became part of the geophysical terminology.
Figure 5 Travel times of seismic phases for the surface focus as
computed by Jeffreys and Bullen (1940). Reproduced from Jeffreys
Because seismic methods were considered essential in dis-
H and Bullen KE (1940) Seismological Tables, p. 50. London: British criminating between earthquakes and nuclear explosions
Association for the Advancement of Science. (Berkner et al., 1959), an intensive observational program,
called Vela Uniform, was initiated. One of its components of
great significance to studies of the Earth’s interior was the
It is not often that a mistake leads to favorable results, but World-Wide Standardized Seismograph Network (WWSSN),
this was the case with the free oscillations of the Earth. Benioff consisting of a set of three-component short-period and
(1958) reported an oscillation with a period of 57 min seen in three-component long-period seismographs, with identical
the record of the 1952 Kamchatka earthquake. Even though responses, except for magnification, which depended on local
this observation was eventually attributed to an artifact in the noise levels. At its peak, the WWSSN consisted of 125 stations
functioning of the instrument, it stimulated the theoretical and (Figure 6), with distribution limited by geography and politics:
computational research needed to calculate eigenfrequencies there were no stations in the Soviet Union, China, and poorly
for a realistic Earth model. Some of the calculations preceded developed areas in Africa and South America. The novel aspect
Benioff’s report ( Jobert, 1956, 1957), but the efforts of Pekeris of WWSSN was its standardized response and centralized sys-
and Jarosch (1958) and Takeyuchi (1959) were clearly moti- tem of distribution of copies of seismograms. Individual sta-
vated to explain the observed period. These calculations, using tions were sending the original seismograms to a central
the variational approach and the Jeffreys–Bullen Earth model, location, where they were microfilmed, using a very high-
predicted the period of 0S2 to be 52 min and that of 0T2 resolution camera, and then returned to the stations.
43.5 min. Neither was close enough to Benioff’s ‘observation.’ A seismologist could request copies of seismograms for a par-
The modern approach was developed by Alterman et al. ticular date and receive them as either microfilm chips or
(1959), who recast the system of three second-order partial photographic enlargements to the original size. Several larger
differential equations into a system of six first-order differential institutions had a blanket order on all the records. This data
equations, thus removing the need for differentiation of the accessibility represented major progress with respect to the
elastic constants, and allowed the use of standard numerical earlier procedures, where one had to request copies of seismo-
methods to obtain the solution. Tests using the Gutenberg and grams from individual stations, which greatly limited and
the Jeffreys–Bullen models showed that they predict very delayed the research. WWSSN functioned for 20–25 years,
6 Deep Earth Seismology: An Introduction and Overview

Figure 6 Map of the stations of WWSSN established in the early 1960s, following recommendations of Berkner et al. (1959). Courtesy of US Geological
Survey.

slowly declining in the quality and number of stations; it

ceased functioning in the late 1980s when data from new
digital stations became available.
Another development of the 1960s was the introduction to
seismology of digital recording, greatly facilitating research and
the development of massive, computerized data-processing
methods. One such facility, the large aperture seismic array
(LASA), shown in Figure 7, was built in the mid-1960s in
Montana. It consisted of six ‘rings,’ for a total of 21 subarrays,
each with 25 short-period seismometers emplaced in bore-
holes, to improve the signal-to-noise ratio. The data were tele-
metered in real time to a central location in Billings, Montana,
where they were processed for detection of a signal. Major
scientific discoveries were made with this tool, particularly
when weak signals were involved, for example, observations
of reflections from the inner core. In practical terms, the array
did not meet the expectations; the site response across this
200 km aperture array varied so much that the signals had
Figure 7 Configuration of the large aperture seismic array (LASA)
limited coherency and signal enhancement by stacking was
and an expanded view of two of its subarrays. Reproduced from Stein
not as effective as expected. A somewhat smaller array was S and Wysession M (2003) An Introduction to Seismology, Earthquakes
installed a few years later in Norway (NORSAR); elements of and Earth Structure. Oxford: Blackwell, ISBN: 0865420785.
this array are still active and have been upgraded recently to
modern, high-dynamic-range band–band response. Most of
the arrays of the International Monitoring System (IMS) used One of the important results obtained from the analysis of
for seismic discrimination purposes have an aperture of only array data was the detection of upper-mantle discontinuities
several kilometers; because on that scale, the coherency at 1 Hz (Johnson, 1967), confirming the result predicted by experi-
frequency can be achieved. mental petrology that there should be two discontinuities at
 Deep Earth Seismology: An Introduction and Overview 7

pressures and temperatures corresponding to depths of about turned out to be an artifact of coupling between toroidal and
400 and 650 km, respectively (Birch, 1952). spheroidal (Russakoff et al., 1998) modes, but the requisite
Surface-wave studies blossomed in the 1960s. At first, mea- theory to consider this effect was not available until 1984.
surements of dispersion involved rather simple ‘analog’ Gilbert and Dziewonski (1975) presented two models based
methods, such as the peak-and-trough approach to measuring on measurements of eigenfrequencies of 1064 modes and
phase and group velocities. Some very important results were mass and moment of inertia for a total of 1066 data. They
obtained in this way, such as the Canadian Shield study of derived two models 1066A and 1066B, with the first being
Brune and Dorman (1963). Digital analysis, however, was smooth through the transition zone and the latter including
soon to take over. Manual digitization of analogue recordings, the 400 and 660 km discontinuities; both models fit the data
WWSSN data in particular, became easier, and with increasing equally well.
availability of computers and decreasing cost of computations, At the 1971 General Assembly of the International Union
various techniques were developed, for the most part involving of Geodesy and Geophysics (IUGG) in Moscow, the need for
applications of the Fourier transform. With the development of a reference Earth model was stated, and a Standard Earth
the fast Fourier transform (FFT) algorithm (Cooley and Tukey, Model Committee formed under the chairmanship of Keith
1965), the cost of time-series analysis declined dramatically; a Bullen. The Committee appointed several subcommittees,
review by Dziewonski and Hales (1972) summarizes the state including one for the radius of the CMB: there were discrep-
of the art at the beginning of the 1970s. Some of these ancies on the order of 10 km at the time. The value recom-
methods, such as the multiple filtration technique to measure mended by the subcommittee was 3484 km (Dziewonski
group velocity dispersion, residual dispersion measurements, and Haddon, 1974), which – within 1 km – withstood the
and time-variable filtration, are still in use today. The 1960s trial of time. Hales et al. (1974) proposed that the seismic
have also seen the first studies of intrinsic attenuation by velocities and density in the standard Earth model should be
Anderson and Archambeau (1964), who developed partial described by a low-order polynomial, with discontinuities at
derivatives (kernels) for Q from mantle-wave attenuation. the appropriate depths. Dziewonski et al. (1975) con-
Also, the first studies of lateral heterogeneity were conducted structed such a model, named parametric Earth model
using the ‘pure path’ approach (Toks€ oz and Anderson, 1966). (PEM), which satisfied the normal-mode, travel-time, and
Seismic experiments with controlled sources were con- surface-wave data. The novelty of this model was that, in a
ducted in a multi-institutional mode. One of the largest exper- single inversion, different structures were obtained for the
iments was ‘Early Rise’ carried out in the July of 1966. A series continental and oceanic crust and upper mantle. The
of 38 explosions in Lake Superior of up to 5 tons of outdated normal-mode periods predicted by these two models
torpedoes were used as the source of signals recorded by hun- (PEM-O and PEM-C) averaged in 2/3 and 1/3 proportion
dreds of seismometers – deployed by 12 governmental and were constrained to match the observed periods and tele-
academic groups – spreading radially in all directions. Signals seismic travel times but separate data sets for continental
were recorded as far as 2500 km, reaching teleseismic distances and oceanic surface-wave dispersion. The differences
and providing a detailed profile of P velocity under a conti- between these two models ceased at the 400 km discontinu-
nental upper mantle (Green and Hales, 1968). A detailed ity, at which depth they became identical with the average
review of crustal and upper-mantle studies with controlled Earth model, PEM-A.
sources is provided in Chapter 1.11. The drawback of the PEM and all the previous models was
With a large new data set, particularly measurements of that they did not consider the physical dispersion due to
previously unreported periods of long-period overtones pro- anelastic attenuation. For a signal propagating in an attenuat-
vided by the analysis of free oscillations generated by the 1964 ing medium to be causal, the wave with higher frequencies
Alaskan earthquake and recorded at WWSSN stations must propagate with higher velocities. Thus, waves with a
(Dziewonski and Gilbert, 1972, 1973), studies of 1-D structure frequency of 1 Hz will propagate more rapidly than waves at
entered a new era. The resolution of this data set was sufficient a frequency of 1 mHz. In order to reconcile the seismic data
to constrain the density profile in the mantle and the core; this that span 3.5 orders of magnitude, it is necessary to consider
turned out to be quite consistent with the behavior, in the the frequency dependence of elastic parameters. This was
lower mantle and outer core, of a homogeneous material pointed out by Liu et al. (1976). The Preliminary Reference
under adiabatic compression. Jordan and Anderson (1974) Earth Model (PREM) constructed by Dziewonski and
were the first to combine the normal-mode and travel-time Anderson (1981), following the idea of parametric representa-
data, including differential travel-time data. tion, considered the frequency dependence using the assump-
Numerous additional overtone data were obtained by tion that Q is constant in the band from 0.3 mHz to 1 Hz.
Mendiguren (1973) and Gilbert and Dziewonski (1975) by This necessitated obtaining the radial profiles of Qm and Qk;
introducing phase equalization techniques, such as ‘stacking’ fortunately, there were new measurements available of normal-
and ‘stripping.’ These methods require the knowledge of the mode and surface-wave Q (Sailor and Dziewonski, 1978) such
source mechanism to predict the proper phase for each seis- that a formal inversion for Q could be conducted simulta-
mogram to be considered; this in itself turned out to be a neously with the inversion for the velocities and density. It
challenging inverse problem. Dziewonski and Gilbert (1974) was recognized earlier that to explain the observed attenuation
derived the spectrum of all six components of the moment- of radial modes, which contain a very high percentage of
rate tensor as a function of time for two deep earthquakes compressional energy (97.5% for 0S0), it was necessary to
(Brazil, 1963; Colombia, 1970). For both events, they introduce a finite bulk attenuation; Anderson and Hart
detected a precursive isotropic component. Eventually, this (1978) preferred to place it in the inner core, and Sailor and
8 Deep Earth Seismology: An Introduction and Overview

Figure 8 The Preliminary Reference Earth Model (PREM) of Dziewonski and Anderson (1981). In addition to the distribution of seismic velocities and
density, PREM contains also the distribution of attenuation of the shear and compressional energy. From the website of Ed Garnero.

Dziewonski (1978) thought that Qk is finite in the upper Two digital seismographic networks were established in the
mantle; unfortunately, the radial modes do not have the req- mid-1970s. One was the International Deployment of Acceler-
uisite depth resolution. Figure 8 shows the seismic velocities ometers (IDA; Agnew et al., 1976, 1986), consisting of 18 glob-
and density as a function of radius; the attenuation in PREM is ally distributed LaCoste–Romberg gravimeters with a feedback
discussed in Chapter 1.25. Another novel aspect of PREM was system that allowed digitization of the signal. It was designed to
its radial anisotropy between the Moho and 220 km depth. record very long-period waves, including the gravest modes of
This feature, at first suspected to be an artifact of the nonli- free oscillations of the Earth: one sample was taken every 20 s
nearity of the inverse problem, has been confirmed by global (later changed to 10 s). Only the vertical component of acceler-
tomographic studies (e.g., Ekstr€ om and Dziewonski, 1998). ation was recorded and the word length was 12 bits; the
The PREM model remains to this day a widely used 1D refer- dynamic range was, therefore, rather limited but still consider-
ence model for seismological studies based on long period ably greater than that of analogue recordings. The sensitivity was
data. For P wave studies, especially those concerning core set such that the scale was saturated for the first surface-wave
sensitive phases, model AK135 (Kennett et al., 1995), which arrivals for events with magnitude 7.0, or so, depending on the
was built to constrain a global dataset of travel times from the station’s distance from the source and radiation pattern. The
ISC bulletins, is often preferred as a 1D reference model. It is to IDA network was operated by the Scripps Institution of
be noted, however, that this model includes a 36 km continen- Oceanography, and the centrally collected data were freely dis-
tal crust, and cannot be used for waveform modeling outside of tributed to the academic community. This later became the
continental areas. future standard in global seismology. An early illustration of
The 1970s have also seen the beginning of seismic tomog- the power of such a global array was the analysis of the splitting
raphy; two studies published simultaneously (Aki et al., 1977; of the gravest modes of free oscillations generated by the 1977
Dziewonski et al., 1977) addressed the problem on different Sumbawa earthquake (Buland et al., 1979; Mw only 8.4).
scales: regional and global. Aki et al. solved for 3-D velocity The other network consisted originally of nine installations
structure under the NORSAR array, while Dziewonski et al. called Seismic Research Observatories (SROs) and five Abbre-
obtained a very low-resolution model of 3-D velocity viated Seismic Research Observatories (ASROs). The SROs
perturbations in the entire mantle and showed significant used borehole instruments, with significantly suppressed
correlation between velocity anomalies in the lowermost man- wind-generated noise levels, particularly on horizontal com-
tle and the gravest harmonics of the gravity field. The study of ponents. The ASROs were placed in underground tunnels or
Dziewonski et al. (1977) was motivated by a paper by Julian mine shafts and the seismographs were protected from the
and Sengupta (1973) who noticed that travel times for rays effects of changing pressure and temperature. The instrumen-
bottoming in the same region of the mantle tend to show tation is described by Peterson et al. (1976). This network was
similar residuals. They interpreted this result qualitatively as designed for monitoring the Nuclear Test Ban Treaty and high
the evidence of lateral velocity variations; no modeling was sensitivity was the main objective. In order to increase the
presented in that paper. The first continental-scale 3-D model dynamic range, the signal was sharply band-pass-filtered, so
of the upper mantle, under North America, was published by that at very long periods (>200 s), the response to acceleration
Romanowicz (1979). was falling as o3, while it was flat for the IDA instruments.
 Deep Earth Seismology: An Introduction and Overview 9

Even so, the SRO and ASRO stations were able to produce from 1976 till present, and this catalog is available online
useful mantle-wave records for events with magnitude greater (http://www.globalcmt.org).
than about 6.5. Later, the network was augmented by Masters et al. (1982) measured center frequencies of spec-
10 WWSSN stations, with the analogue output amplified and tral peaks of the fundamental spheroidal mode from hun-
digitized at 16-bit analog to digital converters. The entire sys- dreds of IDA records and discovered that there are spatially
tem was called Global Digital Seismographic Network distinct patterns in the frequency shifts when plotted at loca-
(GDSN). There was no general data distribution system, but tions of the poles of the great circles corresponding to the
data for selected dates were available upon request from the paths between the source and the receiver. By fitting spherical
Albuquerque Seismological Laboratory. harmonics (even degrees only, because of the symmetry),
Until then, a seismic station typically comprised a set of these authors realized that the pattern is dominated by spher-
seismometers with either ‘long-period’ or ‘short-period’ ical harmonics of degree 2, an observation also made from
responses, or sometimes, as was the case for the WWSSN, one great-circling surface waves by Souriau and Souriau (1983).
of each. This setup was designed at the time of analogue record- Figure 9 shows the pattern of the shifts of spectral peaks and
ing to avoid the microseismic noise peak around 6–7 s period, zero line of the best-fitting spherical harmonics of degree 2
which would have made it impossible to digitize all but the for four groups of 0S‘ modes with different ranges of degree ‘.
largest earthquake signals. With digital recording, and the pos- Note that the modes with the lowest ‘ show a different
sibility of filtering out the microseismic noise by post- pattern than the remaining three groups. Our current under-
processing, this traditional instrument design became standing of this fact is that the low-‘ modes sample deeper
unnecessary. structure (lower mantle) than the higher-‘ groups that pre-
A very important development in seismic instrumentation dominantly sample the upper mantle. The authors performed
took place in Germany in the mid-1970s. An array of a new a parameter search, in which they changed the radii of a shell
kind of instruments with digital recording was deployed near in which the anomaly is located. The best variance reduction
Gräfenberg (Harjes and Seidl, 1978). It used a novel feedback was for the anomaly placed in the transition zone. The lasting
seismograph (Wielandt and Streckeisen, 1982). The system importance of this chapter is that it demonstrated that het-
was rapidly recognized for its linearity and large dynamic erogeneity of very large wavelength and sizeable amplitude
range within a wide band of frequencies – hence the name (1.5%) exists in the Earth’s interior.
‘broadband.’ The Gräfenberg array’s central station had been Following the development of a waveform-fitting technique
colocated with the SRO borehole station GRFO, and the com- that allowed the simultaneous measurement of phase velocity
parisons were very favorable for the broadband instruments, and attenuation along a great-circle path (Dziewonski and
which were capable of reproducing different narrowband Steim, 1982), Woodhouse and Dziewonski (1984) developed
responses using a single data stream. This type of instrumenta- an approach to the interpretation of waveforms that could
tion became the pattern for future developments (e.g., Chapter extract both the even and the odd harmonic coefficients of
1.02). lateral heterogeneity as a function of depth. This method
In addition to the developments in instrumentation, the late involves the ‘path average approximation,’ sometimes called
1970s saw important theoretical developments, related to the PAVA. The seismograms are represented as a sum of all normal
asymptotic properties and coupling of the normal modes. Exam- modes (spheroidal and toroidal) up to a certain frequency
ples of such developments are papers by Woodhouse and Dahlen omax. For a given great circle, each mode is assumed to be
(1978), Jordan (1978), and Woodhouse and Girnius (1982). affected by the average deviation from reference structure
along the great-circle path (which is sensitive only to even-
order harmonics) and along the minor-circle path (sensitive
to both even and odd harmonics). The effect of the great-circle
1.01.3 From 1980 to Present: The Era of Tomography path can be modeled by a shift in eigenfrequency of the mode;
and Broadband Digital Seismic Networks the effect of the minor-arc structure is modeled by a fictitious
shift of the epicentral distance for that mode; this shift depends
The data from both global networks of the 1970s led to results on both even and odd parts of the structure. Woodhouse and
that demonstrated the need for development of a global net- Dziewonski (1984) processed about 2000 mantle-wave seis-
work that would better satisfy the needs of basic research in mograms from the GDSN and IDA networks and obtained a
seismology; three studies are noteworthy. A robust method, model of the upper mantle (Moho – 670 km), M84C, using as
which uses entire segments of (digital) seismograms, was basis functions spherical harmonics up to degree 8 for hori-
developed to obtain reliable mechanisms of earthquakes with zontal variations and Legendre polynomials as a function of
magnitude 5.0 and above (Dziewonski et al., 1981; Ekstr€ om radius up to degree 3. Figure 10 shows a map of shear velocity
et al., 2005). In addition, the method refines the location of the anomalies at a depth of 100 km; there was no a priori infor-
source, which for an event of finite size need not be identical mation used on the location of the plate boundaries. Correc-
with the hypocenter determined from the first arrivals of the tions were made for crustal thickness, recognizing only the
P-waves. This topic is discussed at length in Chapter 4.16. The continental and oceanic structure. An experimental model,
reason why the subject is brought up here is that in most M84A, obtained without applying crustal corrections, showed
aspects of using waveform analysis for the purpose of drawing that not taking crustal thickness into account may result in
inferences about the Earth’s structure, it is necessary to know mapping artificial anomalies at depths as large as 300 km.
the source mechanism. The so-called ‘centroid moment tensor’ Model M84C had a strong degree-2 anomaly in the transition
method has now been applied to over 40 000 earthquakes zone, confirming the results of Masters et al. (1982).
10 Deep Earth Seismology: An Introduction and Overview

Figure 9 Maps of the observed frequency shifts of the fundamental spheroidal modes for four ranges of the order numbers as reported by
Masters et al. (1982). The frequency shifts are plotted at the poles of the individual great-circle paths. It indicates the presence of very large wavelength–
velocity anomalies in the Earth’s interior; the preferred location of the source of the anomaly shown in the figure is the transition zone. However,
the frequency shifts for the largest wavelength panel (top left; modes 0S8 and 0S9) show a different pattern than in the other three panels (modes from
0S27 to 0S34); the low-‘ data have the greatest sensitivity in the lower mantle. Modified from Masters G, Jordan TH, Silver PG, and Gilbert F (1982)
Aspherical Earth structure from fundamental spheroidal mode data. Nature 298: 609–613.

Figure 10 Shear velocity anomalies at a depth of 100 km in the model M84C of Woodhouse and Dziewonski (1984). The scale range is 5% and
the resolving half-wavelength is 2500 km. Plate boundaries are shown as thin yellow lines. Except for the correction for crustal thickness, there was no
other a priori information included in the inversion. Thus, the result demonstrates that the waveform inversion approach is able to distinguish the slow
velocities under the mid-ocean ridges and ancient cratons, for example.
 Deep Earth Seismology: An Introduction and Overview 11

Figure 11 Map of P-velocity anomalies at a depth of 2500 km in model of Dziewonski (1984) derived from inversion of travel-time anomalies from
ISC Bulletins. The resolving half-wavelength of the model is about 3500 km (at the surface). The model, dominated by the harmonics 2 and 3,
clearly shows two large superplumes (African and Pacific) and the ring of fast velocities circumscribing the Pacific Rim. The scale is 0.5%.

Another result also affected future developments, even offset at degree 6 and begins to decrease rapidly, roughly as ‘2;
though this study was not based on waveform analysis, but the ‘corner wave number’ is in this case ‘ ¼ 5. Part (b)
on the ISC bulletin data. With greater computational resources, is the synthesis of this function using degrees from 1 to 8 (81
it was possible to cast the inverse problem for lateral coefficients). The shapes of the continents are clearly recogniz-
heterogeneities in the lower mantle on a larger scale using a able, but details of the coastlines or islands are missing. Panel
substantially greater body of data than in Dziewonski et al. (c) shows the result of a synthesis using harmonics from 1 to
(1977). Unlike in this earlier study, in which blocks were used, 32 (1089 coefficients or 15 times more). There is an obvious
Dziewonski (1984) used global basis functions: spherical har- improvement: Madagascar, New Zealand, Japanese Islands,
monics representing horizontal variations and Legendre poly- Gulf of Mexico, can be identified, but the principal informa-
nomials for radial variations. The degree of expansion was tion on the outline of continental masses is already contained
modest: only degree 6 in harmonics and degree 4 in radius, in panel (b).
with the inversion limited to lower-mantle structure. In many Another point is that power at degree 32 is two orders of
ways, this new study confirmed the earlier one – including the magnitude less than at degree 3. Actually, it is lower by three
correlation of lower-mantle structure with the gravity field orders for the individual Y‘m coefficients, since we define the
(Dziewonski et al., 1977) – but it showed resolution of a truly power as the sum of all squared coefficients of degree ‘. It is
remarkable concentration of the power of heterogeneity in low- impractical to think that such a large range of values could be
order harmonics. Figure 11 shows a map of P-velocity anoma- resolved in an inverse problem, particularly because the strength
lies at a depth of 2500 km. The structure, dominated by degree 2 of the kernels commonly decreases with increasing wave number
(and, to a lesser extent, degree 3), shows two large slow regions, (see Figure 3 a–d in Dziewonski, 1984). However, if the conti-
which came to be known as the African and Pacific ‘super- nents were broken up into many smaller ‘islands,’ then a synthe-
plumes’, and a ring of fast velocities around the Pacific. sis of the spherical harmonic coefficients up to degree 8 might be
Why does global tomography work? Some suspected misleading and provide an aliased picture.
that truncation of the expansion at relatively low harmonic The ‘corner wave number’ seems to be a feature of many
order may lead to aliasing and structures dominated by degrees geophysically important functions: Su and Dziewonski
2 and 3, such as the one shown in Figure 11, are an artifact. The (1992) showed that it occurs in spectra of several global
short answer is that global mantle heterogeneity has a ‘red’ functions, such as free air gravity, SS-S travel-time residuals,
spectrum. That is, the power spectrum of the spherical har- phase velocity, and 3-D models, which have a power spec-
monic expansion begins to drop rapidly above a certain wave trum that is relatively flat up to degrees from 6 to 8, after
number (‘corner wave number’). Su and Dziewonski (1991, which it begins to decrease as ‘2. This seems typical of the
1992) showed that it is clearly visible even in data, such as spectra of two-dimensional functions that are characterized
maps of travel-time anomalies of SS-S or ScS-S. The concept is by a set of large ‘patches,’ such as the large land masses in the
illustrated in Figure 12, using the continent–ocean variations continent–ocean function. If the spherical harmonic expan-
of crustal thickness. sion is truncated at an order number beyond the corner wave
Figure 12 illustrates the point that for functions with a ‘red’ number, the synthesis of such truncated series retains the
spectrum that is flat for small wave numbers but rapidly main character of the original function (e.g., Figure 12). Sim-
decreasing for larger wave numbers, it is possible to extract ilarly, an inversion for a truncated series of spherical har-
most of the signal using the ‘red’ part of the spectrum. Panel monic coefficients does not introduce aliasing if truncation
(a) shows the power spectrum of the continent–ocean func- occurs in the steeply decreasing part of the power spectrum.
tion. It is flat for degrees from 1 to 5, and then, the power is Of course, there are some geophysically important functions
12 Deep Earth Seismology: An Introduction and Overview

Ocean function (l=1–8)

The properties of a
‘super-red’ spectrum

1000
Ocean function
100 (b)

10 Ocean function (l=1–32)

1
1 2 5 10 30 100
(a) Harmonic order

15 times more coefficients

(c) –15 +15

km
Figure 12 An example of a global function with a ‘super red’ spectrum. Crustal thickness under the oceans and continents, with a constant crust
thickness for a given type, is expanded in spherical harmonics and then synthesized using a different cutoff wave number. Panel (a) shows the power
spectrum of the continent–ocean function. Note the nearly flat spectrum for orders from 1 to 5 and then nearly linear decrease (in the log–log scale)
for orders 6 and higher. The slope of that line is about ‘2; the corner wave number is ‘ ¼ 5 in this case. Panel (b) shows the synthesized
continent–ocean function for ‘max ¼ 8. All contours of continents are delineated and much of the variance is explained by this low-order model; this
expansion has been used by Woodhouse and Dziewonski (1984) to introduce corrections for crustal structure. Panel (c) shows the result of a synthesis
using ‘max ¼ 32. The details such as coastline and recovery of larger islands are clear, but they are achieved by using about 15 times more coefficients
than those used to obtain panel (b). Also, the coefficients at higher wave numbers are very small, and it would be unlikely to be correctly derived
through an inversion of observations with errors.

whose spectra have distinctly different characters, for expectation was that this network would be supported by the
example, linear features (slabs) have power spectra that are National Science Foundation (NSF), in analogy to NSF sup-
flat with harmonic order; point-like features (plumes?) have porting astronomy facilities. At the same time, seismologists
power spectra whose values increase with ‘. In these cases, using portable instrumentation came to the realization that
truncation would result in significantly lower amplitudes in they needed a centralized and standardized instrument pool.
the low-pass-filtered image. This led to the formation of a project called Portable Array for
This could also happen if the truncation of the spectrum Seismic Studies of the Continental Lithosphere – PASSCAL.
occurred before the corner wave number. An example of the The GSN and PASSCAL groups merged and formed a consor-
consequences of such an action is given by Mégnin et al. tium known as Incorporated Research Institutions for Seismol-
(1997). In a synthetic test, they limited to degree-6 inversion ogy (IRIS), which incorporated in 1984.
of waveform data calculated for a degree-12 model and showed At the same time, Steim (1986) was developing at Harvard
that the results are aliased by the missing part of the structure his very broadband (VBB) instrument, based on the STS-1
(degrees 7 through 12). But in their test, the power spectrum broadband seismograph (Wielandt and Steim, 1986), and a
had a maximum at degrees 6 and 7, and truncation at the very high-resolution (24-bit) digitizer. The response of a VBB
maximum power is bound to cause aliasing. Boschi and instrument is designed to have a flat response to ground
Dziewonski (1999) demonstrated that there is no aliasing velocity between 5 Hz and 3 mHz, that is, over more than
when travel times computed for a degree 40 model are inverted three orders of magnitude. Such an instrument had to have a
for a synthetic model limited to degree 12 (their plate 6), but very large dynamic range, of about 140 dB to span the range
the truncation was applied at the corner wave number. of ground velocities from the minimum Earth noise to a
Thus, in the 1980s, seismologists demonstrated that they magnitude 9.4 earthquake at 30 epicentral distance. All
can resolve 3-D structure within the Earth interior, giving the these requirements were met, and Steim’s VBB system became
promise of an unprecedented ability to look at a present-day the design goal for the GSN and other networks. Figure 13
snapshot of mantle dynamics. Yet, the observing networks shows the operating range of the system, and Figure 14 illus-
were in decline, with the support for the GDSN likely to be trates the dynamic range of the GSN station in Albuquerque.
discontinued altogether and the original IDA network (limited The high-pass-filtered (75 s) record of the Sumatra–Andaman
to recording of vertical-component mantle waves and free Mw ¼ 9.3 earthquake shows surface waves with an amplitude
oscillations) not meeting the needs of the broader community. of several millimeters and a record of a local microearthquake
In 1983, a plan was put forward to create a GSN of some 100 (M < 1) extracted from the same record; the ratio of the
broadband, three-component seismographic stations sending amplitudes is about 10 000 000! Chapter 1.02 describes in
the data in nearly real time to a central collection facility. The detail the concept and history of the modern broadband
 Deep Earth Seismology: An Introduction and Overview 13

Figure 13 The dynamic range of the VBB channels of a Global Seismographic Network station. The range of the WWSSN short- and long-period
channels are shown for comparison. At some GSN stations, the VBB channels are augmented by very short-period channels and accelerometers.
The response was designed to resolve the ground noise from 5 Hz to tidal frequencies and to record on scale a magnitude 9 earthquake at a distance
of 30 . Courtesy of IRIS.

digital seismograph that has now been implemented in sta- network as of January 2014 (after some 25 years of develop-
tions of the global international network as well as many ment); there are over 200 participating VBB stations, most of
regional seismic networks. which now send data in nearly real time.
Meanwhile, as the US seismologists were organizing them- Similarly impressive progress was made in the area of field
selves, a French effort named GEOSCOPE had already began seismology, where progress in electronics led to overall
taking shape. The objective of GEOSCOPE was the establish- improvement of the portability of the equipment and, in par-
ment of a global network of some 20–25 broadband digital ticular, reduction of power requirements, which makes opera-
seismographic stations utilizing the STS-1 seismometer, and tions much easier. In parallel with the development of the
destined in priority for locations around the world that filled PASSCAL program of IRIS in the United States (see Chapter
gaps in the distribution of seismic stations. The project had 1.16), other portable arrays were developed in other countries
officially begun in 1982, and in 1984, there were already five in the last 20 years, such as the LITHOPROBE cross continental
operational stations (Romanowicz et al., 1984), and 13 in project in Canada (Hammer et al., 2010), Lithoscope program
1986, (e.g., Romanowicz et al., 1991; Roult et al., 2010a,b). in France (Poupinet et al., 1989), or the SKIPPY array in
While telemetry of the data was established later (in 1987), Australia (van der Hilst et al., 1994). Most recently, an ambi-
the state of health of the remote stations was monitored from tious program, USArray, was launched in the United States in
early on using the satellite system Argos, which greatly facili- 2007 as part of the EarthScope project. USArray is aimed at a
tated their maintenance. As it was clear that it was inefficient to systematic investigation of the structure under the contiguous
have two competing global networks, the need for a framework United States with uniform resolution. It consists of three
for international cooperation arose. Also, many countries were parts: a Permanent Array of 100 stations (the ‘backbone’ or
interested in deploying broadband instrumentation for their ‘reference array’), a Transportable Array (TA), and a Flexible
national or regional purposes and were agreeable to share these Array (FA). The FA provides some 300 broadband seismo-
data. The international Federation of Digital Seismographic graphs and over 1000 active-source instruments, for experi-
Networks (FDSN) was formed in 1986 (Romanowicz and ments proposed by individual research groups, aimed at
Dziewonski, 1986), with the purpose of coordinating site elaborating detailed local structure. The TA is the largest com-
selection, data exchange, and standardizing instrument ponent of the program, and it consists of 400 broadband
responses. The FDSN has been very successful in achieving seismograph systems that, starting from the West Coast of the
these goals, despite the fact that it is a purely voluntary, zero- United States, have been moving gradually across the conti-
budget organization. Figure 15 shows a map of the FDSN nental United States to cover the entire area with, roughly,
14 Deep Earth Seismology: An Introduction and Overview

Figure 14 Illustration of the dynamic range of a VBB station (ANMO). A recording of a local microearthquake with magnitude below 1 is extracted
from the record dominated by the minor-arc surface waves generated by the 2004 Sumatra–Andaman magnitude 9.3 event. Reproduced from
Park J, Butler R, Anderson K, et al. (2005) Performance review of the global seismographic network for the Sumatra–Andaman megathrust earthquake.
Seismological Research Letters 76 (3): 331–343.

2000 deployments for up to 2 years in a given location; the well as in Chapters 1.10 and 1.19. However, there are still
average instrument spacing is 70 km. In 2013, it has reached issues that remain unresolved or controversial. Certainly,
the East Coast and preparations are under way to redeploy there are confusing observations related, for example, to aniso-
about half of the instruments in Alaska. The ‘reference array’ tropic properties or differential rotation of the inner core, the
provides the means to relate the waveforms recorded at differ- core–mantle topography, anticorrelation of density and shear
ent stages of TA deployment. Many other portable networks for velocity near the bottom of the mantle, the strength and depth
regional studies of the crust and lithosphere using passive and distribution of anisotropy in the Earth’s mantle, the presence
active sources have been developed and deployed in the last of the postperovskite phase at the base of the mantle, and its
20 years (see Chapters 1.11, 1.16, and 1.17). role in mantle dynamics.
But the foremost issue in our view relates to the derivation
and interpretation of 3-D Earth models. We will mention three
1.01.4 Current Issues in Global Tomography issues in particular: parameterization, datasets used for model
construction, and theoretical assumptions.
In parallel with instrument development, scientific progress in Ever since it was discovered that inversion of ill-
seismic tomography has been rapid during the last 20 years, conditioned matrices can be dealt with by requiring minimi-
and most of the accomplishments are summarized in two zation of the norm or roughness of the model, we have been
reviews by Romanowicz (1991) and Romanowicz (2003), as obtaining models whose reliability is difficult to assess.
 Deep Earth Seismology: An Introduction and Overview 15

IRIS GSN Australia Canada France Germany Italy Japan U.S. Other

Figure 15 Current (January 2014) map of the stations of the Federation of Digital Seismographic Networks. Stations of different member networks are
identified by symbols shown at the bottom. Courtesy of Andy Frassetto (IRIS Instrumentation Services).

Figure 16 Demonstration of the results of inversion of the same data set (40 000 phase-delay data for 75 s Rayleigh waves measured by Ekstro€m
et al., 1997) for different numbers of parameters. The top map shows inversion for spherical harmonic coefficients up to degree 16 (289 parameters).
The bottom map shows the results of inversion for about 10 000 equal-area blocks; in this case, matrix conditioning is necessary. The amplitude of
the anomalies is lower, artifacts of an uneven path distribution are visible (e.g., across the central Atlantic), and it is difficult to find features that
have not been resolved by the top map. The conclusion is that there is a price for unduly increasing the number of unknowns. Courtesy of L. Boschi.
16 Deep Earth Seismology: An Introduction and Overview

Figure 16 illustrates an example of how the results can be number of parameters and information contained in the
altered with the change of parameterization. A set of about model. It is so, partly, because the lower-mantle spectrum is
40 000 phase-delay data for Rayleigh waves with 75 s period dominated by the very low degrees that are well recovered by
(Ekstr€om et al., 1997) are inverted for ‘local’ phase velocities. the degree-6 model.
The top panel of Figure 16 shows the result in which the data
were inverted for a set of basis functions represented by spher-
ical harmonics up to degree 16; this requires solving for 289 1.01.4.1 Resolving Power of Datasets Used for Constructing
unknown coefficients. Because the data set is so large and the Models
global coverage is good, the solution was obtained by an exact Another important issue in assessing global, as well as regional,
matrix inversion. The results look reasonable, without any tomographic models is the data set (or subsets) and the resolv-
indication of instability. The lower panel of Figure 16 shows ing properties that were used to derive them. The mantle
the result obtained using a 2 2 block expansion, which models derived using only teleseismic travel times, for exam-
requires solving for approximately 10 000 unknown values. ple, have very little radial resolution in the upper mantle,
Matrix conditioning was required in this case, and it was because the ray paths do not bottom there. The teleseismic
accomplished by applying combined norm and roughness travel times are sensitive to velocity perturbations in the
damping. What is clear from the ‘high-resolution’ solution is upper mantle, but the variations with radius cannot be satis-
that the amplitudes are generally lower and there are ‘streaks’ factorily resolved above the lower mantle. For example, all
indicating artifacts caused by uneven sampling of the area. maps of upper-mantle velocity anomalies in Figure 18 show
What is difficult to find, however, are any features that appear that a model derived using only teleseismic travel times has
to be significant that are not present in the solution with slow velocities under mid-ocean ridges at all upper-mantle
30 times fewer parameters. Thus, sometimes, less is better. depths, simply from the smearing with depth of the large
There is no absolute rule; as discussed in the previous section, slow anomalies occurring near the surface. This is a result of
the answer depends on the character of power spectra of a norm damping in a situation where sampling is inadequate: a
particular function. compromise between fitting the data and having a low norm
The issue of parameterization is an important one, particu- (low sum, or integral, of squared model perturbations). The
larly if the global behavior of the solution is to be preserved. spreading centers and cratons have largest signal in the upper-
Figure 17 illustrates cross sections of four global P-velocity most 250 km of the upper mantle; the variance in the next
models – derived from the same data source (ISC Bulletins) –
that span three orders of magnitude in the number of param-
eters. Clearly, there is not a simple relationship between the

Figure 17 Comparison of the equatorial cross sections of four

P-velocity models obtained by inversion of travel-time residuals from ISC
Bulletins using a number of unknown parameters that span three Figure 18 A fragment of a figure from Ritsema et al. (2004) illustrating
orders of magnitude. The image of the African and Pacific superplumes is a typical result of an upper-mantle structure obtained using teleseismic
clearly seen in the model derived by using 250 parameters, while it travel times, whose rays do not bottom in the upper mantle. The three
could not be readily inferred from the model that used 250 000 parameters. maps show smeared-out structure from near the top of the mantle
The models obtained using 2500 and 25 000 parameters support the (mid-ocean ridge anomalies). The conclusion drawn from the full figure (see
conclusion drawn from discussion of Figure 16. Modified from Chapter 1.10) is that, to obtain a whole-mantle model, one should use
Boschi L and Dziewonski AM (1999) ‘High’ and ‘low’ resolution images diverse types of data. Modified from Ritsema J, van Heijst HJ, and Wood-
of the Earth’s mantle – Implications of different approaches to tomographic house JH (2004) Global transition zone tomography. Journal of Geophysical
modeling. Journal of Geophysical Research 104: 25567–25594. Research 109: B02302, http://dx.doi.org/10.1029/2003JB002610.
 Deep Earth Seismology: An Introduction and Overview 17

400 km is an order of magnitude less. Thus, the minimum

norm compromise leads to the smearing of the shallow
structure over the entire upper mantle (for positive integers,
l +m + n ¼ 3 has the minimum norm solution (1, 1, 1) with the
norm of 3, the ‘true’ solution (3, 0, 0) has a norm of 9, and an
intermediate solution (2, 1, 0) has a norm of 5). Another way
to accommodate the minimum norm compromise is to make
the upper-mantle model small; there are published models
derived using only teleseismic travel times that have heteroge-
neous structure in the upper mantle with very small amplitude.
Models built using teleseismic travel times and
fundamental-mode dispersion data do not have sufficient Figure 19 Comparison of sensitivity kernels in the vertical plane
resolution in the transition zone to distinguish its unique containing the source and the receiver for SS waves (left) and Sdiff waves
properties. So far, only three research groups involved in (right), using the path average approximation (PAVA, top) and the
whole-mantle modeling (Michigan/Oxford, Berkeley, and Har- nonlinear asymptotic coupling theory (NACT, bottom). PAVA produces
vard) use data allowing sufficient resolution in this region; it is 1-D kernels, which do not represent well the ray character of body waves.
interesting that they obtain these models using different kinds NACT, which includes across-branch mode coupling, produces 2-D
finite-frequency kernels that more accurately represent the sensitivity
of data and theories, therefore adding to the credibility to the
along and around the ray path as well as its variations with position along
results. The use of waveforms in deriving 3-D models was
the ray. Because this is a time domain formalism, the NACT kernels are
pioneered by Woodhouse and Dziewonski (1984), but in the time-dependent and are here represented at a particular point in the
original paper, only data with periods longer than 135 s were waveform, with positive maxima in black and negative ones in white.
used. Long-period body waves were used in inversions by Shadows beyond the source and receiver are due to the truncation in the
Woodhouse and Dziewonski (1986) and Woodhouse and coupling series. Adapted from Li XD and Romanowicz B (1995)
Dziewonski (1989). Dziewonski and Woodward (1991) com- Comparison of global waveform inversions with and without considering
bined waveforms and teleseismic travel times measured by cross branch coupling. Geophysical Journal International 121: 695–709.
Woodward and Masters (1991). The immediate result was
that the two models they derived showed a sudden change in images of phases such as S, SKS, Sdiff, Love, and Rayleigh
the pattern of heterogeneities across the 670 km boundary, and fundamental and overtone waves will represent similar infor-
this was pointed out and discussed by Woodward et al. (1994). mation as the combination of teleseismic travel times and
Later inversions by the Harvard group included also surface- surface-wave phase velocities in the Caltech/Oxford or Harvard
wave dispersion data reported by Ekstr€ om et al. (1997). This models. In addition, unlike travel-time analysis, wave packets
increased resolution near the surface but did not alter the containing several phases with close arrival times but different
behavior across the upper–lower mantle boundary (Gu et al., sampling of mantle structure can be included, improving res-
2001, 2003; Kustowski et al., 2008). All these models were olution. An important element of the data set used by Ritsema
obtained using the path average approximation (PAVA), et al. (1999) is a set of maps of Rayleigh wave overtone disper-
which assumes constant average structure along the great-circle sion from the first to the fifth overtones. These were obtained
path connecting the source and the receiver. The Berkeley by ‘stripping’ the seismograms of subsequent overtones, thus
group pioneered use of a more advanced theory called non- providing the data on the average phase velocity of a partic-
linear asymptotic coupling theory (NACT) first described by Li ular overtone between the source and the receiver (van Heijst
and Romanowicz (1995) and based on the across-branch cou- and Woodhouse, 1997). Since body waves represent super-
pling asymptotic development of Li and Tanimoto (1993). position of overtones, use of complete waveforms is, to a large
This theory allowed them to construct body-wave kernels that extent, equivalent; however, the separation of data for indi-
give good representation of the sensitivity along and around vidual overtones allows assignment of different weights to
the ray path as shown in Figure 19 and therefore opened the different overtones, while direct waveform methods use
way to tomographic inversions for whole-mantle structure them with the weight that is determined by their excitation;
based entirely on waveform data and thus exploit more infor- generally, the amplitude of overtones decreases with the over-
mation in the seismograms than can be achieved with only tone number. On the other hand, waveforms contain informa-
those phases that are well separated from others in teleseismic tion about all the overtones. The importance of the overtones
records. In applying NACT to the development of several gen- (body-wave waveforms) was shown implicitly by Gu et al.
erations of global mantle models (Li and Romanowicz, 1996; (2001) and explicitly by Ritsema et al. (2004). The importance
Mégnin and Romanowicz, 2000), most recently including of using adequate kernels for overtone and body waveforms was
attenuation (Gung and Romanowicz, 2004) and radial anisot- illustrated by Mégnin and Romanowicz (1999) in a comparison
ropy (Panning and Romanowicz, 2006), the Berkeley group of the PAVA and NACT inversion approaches.
divided the seismograms into wave packets containing one or
several body waves or surface-wave overtones, which allowed
them to weigh different phases differently in order to obtain
1.01.4.2 Theoretical Assumptions
uniform sensitivity with depth. The Berkeley group does not
use travel times explicitly, but information on structure is Ray theory vs. finite-frequency theory has been the subject of a
included in the phase of a waveform of, for example, an SS- lively debate within the seismological community during
arrival. Thus, a large collection of waveforms containing the last decade. Much of it has revolved around the
18 Deep Earth Seismology: An Introduction and Overview

‘banana–doughnut’ kernels (Dahlen et al., 2000) and its appli- ð !

X X+ l
l¼L m¼
cation to mantle modeling (Montelli et al., 2004a,b). The dts, r ¼ dvl, m =v0 Yl, m ðy, ’Þx 
implication conveyed by those authors was that the results O l¼0 m¼l
obtained using finite-frequency theory were distinctly superior !
lX
0
¼1 mX
0
¼ + l0
to those obtained using ray theory. The serious implication of Kls0,,rm0 Yl0 , m0 *ðy, ’Þ dO [2]
this was that all tomographic studies to date were deficient. Yet, l0 ¼0 m¼l0
a direct comparison of the finite-frequency and ray theory
models published by Montelli et al. (2004b, Figure 9) demon- However, because of the orthogonality of spherical har-
monics, this reduces to
strates that the two models are nearly identical, save for a
scaling factor of 1.14 (van der Hilst and deHoop, 2005). This X X+ l
l¼L m¼
is because these authors introduced finite volume parameteri- dts, r ¼ dvl, m =v0 Kls,,mr [3]
zation by tetrahedra with unknown velocity perturbations at l¼0 m¼l

the vertices (with linear interpolation within each tetrahedron) with the terms in expansion of the kernel function for l’ > L
without acknowledging that this is equivalent to a low-pass having no effect.
filtration. Figure 20 (top) shows the frequency-dependent kernels for
There have been quite a few papers published on this Love waves at 35 and 200 s period. In the middle, there is the
subject (e.g., Peter et al., 2007). While it is obviously desirable frequency-dependent kernel expanded to degree 18. At the
to use a more accurate theory, it is worthwhile to investigate bottom is the spherical harmonic expansion of a line (‘ray
why the ray theory and finite-frequency theory yield nearly theory’) along the equator K ¼ d(y  p/2) for f > 0 and f < F
identical results. A particularly clear explanation of this effect and zero elsewhere. At the period of 35 s, it is difficult to tell
is presented by Dalton (2007) who used the orthogonal basis the difference between the middle and the bottom ‘effective’
functions (spherical harmonics) for both the kernels and kernel functions. The difference is still quite small at 200 s
velocity perturbation, with the phase velocity model truncated period. Thus, finite parameterization can imitate the finite-
at degree L. A perturbation in phase travel time between source frequency effect, and for relatively low L, the advantage of
located at s and receiver at r can be calculated by integration using the finite-frequency kernels vanishes. Dalton showed
over the surface of the sphere O: that differences are more pronounced at larger L, but the earlier
ð discussion indicates that there is very little power in the higher
dv degrees of dvl,m and the effect on the observed, or computed,
dt s, r ¼ ðy, ’ÞK s, r ðy, ’ÞdO [1]
O v0 dts,r will be insignificant because of ‘natural truncation.’ Thus,
where v0 is the reference velocity and Ks,r(y,f) is the kernel under the circumstances presented here, the effect of using
calculated for a particular location of the source and the finite-frequency kernels is very small. In general, the ratio of
receiver. Expanding dv and K s,r(y,f) in spherical harmonics the width of the finite-frequency kernel to the path length
and truncating expansion of dv at l ¼ L, we have

Phase 35 s 200 s
300⬚ 320⬚ 340⬚ 0⬚ 20⬚ 40⬚ 60⬚ 300⬚ 320⬚ 340⬚ 0⬚ 20⬚ 40⬚ 60⬚

20⬚ 20⬚
Finite-frequency
0⬚ 0⬚
kernels
−20⬚ −20⬚

Expansion of
ff kernels
(deg 18)

Effective
ray-theoretical
kernels
(deg 18)

−0.01 0.00 0.01 −0.0010 −0.0005 0.0000 0.0005 0.0010

Figure 20 Demonstration of the perhaps counterintuitive fact that ‘finite-frequency’ kernels and ‘ray theory’ kernels may lead to very similar models
if the model to be obtained using ‘ray theory’ is parameterized. Parameterization using spherical harmonic expansion up to a maximum order L is
used here, because the result is exact. Top: Born (finite-frequency) kernels for the phase travel-time delay for Love waves at 35 s (left) and 200 s (right).
Middle: Synthesis of the Born kernels above obtained by synthesis of spherical harmonic coefficients up to degree L ¼ 18. The spherical harmonic
coefficients for l > 18 are irrelevant because of the orthogonality of the basis functions. Bottom: Expansion of a 90 long ray up to the same L ¼ 18.
The effective kernels for 35 s waves are difficult to distinguish from the expanded finite-frequency kernels; the differences are more noticeable at 200 s
period but they are still minor.
 Deep Earth Seismology: An Introduction and Overview 19

determines how good the ray theory approximation is; finite- Because the teleseismic wavefield needs to be computed
frequency kernels can be important in regional-scale studies. for a long time interval, in order to include all seismic phases
Indeed, as discussed earlier, the possibility of inverting of interest (up to and including second orbit fundamental-
complete seismograms (i.e., seismic waveforms) without mode Love and Rayleigh waves), the use of the SEM for
selecting individual phases for travel-time measurements global tomography is particularly challenging computation-
opens the way for exploiting more of the information con- ally. The first global shear velocity models developed using
tained in the recorded wavefield, thereby achieving better sam- SEM (French et al., 2013; Lekic and Romanowicz, 2011) are
pling of the mantle. In particular, we note that models based limited to the upper mantle due to the use of relatively long
entirely on the inversion of time domain waveforms are able to periods (longer than 60 s) for computational efficiency. To
achieve comparable results, with an order of magnitude fewer reduce further computation time, they take advantage of
records (e.g., a total of 20 000 records in Panning and efficiencies gained, in the forward step of the computation,
Romanowicz, 2006) as models based on the standard combi- by restricting the numerical computation to the mantle
nation of surface-wave dispersion and body-wave travel times, through coupling with 1-D mode computations in the core
which typically use tens to hundreds of thousands of measure- (CSEM, Capdeville et al., 2003) and replacing the finely
ments. This is because the smaller number of source-station layered crust by an equivalent smooth crust constrained by
paths considered is compensated by the inclusion of waves that surface-wave dispersion data. In the inverse step, the kernels,
sample mantle structure in a richer variety of ways. which do not need to be as accurate as the forward predic-
Since our ability to improve the global distribution of sources tions (e.g., Lekic and Romanowicz, 2011; Tarantola, 2005),
and receivers is limited, once it is recognized that waveform are calculated using a mode-based approximation. The use of
inversion represents potential for better sampling of the Earth’s a Gauss–Newton formulation of the inverse problem leads to
mantle, the attention then shifts to improving the methods of much more rapid convergence than the adjoint approach.
computation of synthetic seismograms in a 3-D Earth. The use of accurate wavefield computations, such as afforded
NACT (Li and Romanowicz, 1995) provided a successful by SEM, is particularly important to improve resolution of
theoretical approach for this purpose and has led to several low-velocity regions, which tend to be hidden from observa-
generations of whole-mantle shear velocity models in the last tion, when ray theory is used, due to wave front healing
20 years. However, asymptotic normal-mode perturbation the- effects (e.g., Nolet and Dahlen, 2000). For example, the
ory is only valid for Earth models for which the wavelength of new-generation tomographic models based on the use of
the structure is large compared to that of the seismic waves SEM resolve the minimum amplitude of the upper-mantle
considered (i.e., smooth models) and heterogeneity is weak low-velocity zone better than previously, even in areas far
(nominally, lateral variations of up to 10%). Yet, in the away from sources and stations, as illustrated in Figure 21,
Earth’s boundary layers, that is, in the upper mantle and in and are starting to resolve finer details of the ‘plumbing’
the D00 region, there is ample evidence for the presence of system in the oceanic upper mantle (e.g., Colli et al., 2013;
stronger heterogeneity, whereas throughout the mantle, het- French et al., 2013; Rickers et al., 2013).
erogeneity at many different scales may be present. First-order
mode perturbation theory is not appropriate in this case. While
higher-order perturbation theory has been worked out (e.g., 1.01.4.3 Robust Features of Current Global Mantle Models
Lognonné and Romanowicz, 1990; Lognonné, 1991), it is and Their Implications
cumbersome for use in practice, in particular because it always
The full impact of the completed FDSN network, progress in
requires a 1-D reference model.
automatic processing of data – measurements of travel times
In the last decade, advances in numerical methods for the
from waveforms or selection of time windows for waveform
computation of the seismic wavefield in arbitrary 3-D structures,
inversion, for example – improved the database by an order of
and increased computational power, have made it possible to
magnitude in comparison with earlier models. This has been
replace normal mode-based synthetics by more accurate ones,
accompanied by better recognition of what is required to
based on the spectral element method (SEM), first introduced in
obtain a stable 3-D model of shear velocity anomalies in the
global seismology by Komatitsch and Vilotte (1998) and later
mantle. Among these are as follows:
perfected by Komatitsch and Tromp (2002). The use of SEM has
now been implemented in tomographic inversions for crustal (i) Assembly of data that are sensitive to structure from the
structure at the local scale (e.g., Tape et al., 2010) and upper- Moho to CMB; usually, this calls for a combination of
mantle structure at regional scales (e.g., Fichtner et al., 2009; surface-wave dispersion, overtones, and teleseismic travel
Rickers et al., 2013; Zhu et al., 2012). The forward numerical times.
computation is generally combined with an ‘adjoint’ numerical (ii) Proper weighting of the subsets of data to assure compa-
computation of the kernels for inversion (e.g., Tromp et al., rable radial resolution.
2005; see Chapter 1.07) or, alternatively, with a ‘scattering (iii) Careful selection of matrix stabilization parameters.
integral formalism’ (e.g., Chen et al., 2007). Still, computational (iv) Even though gaps in coverage due to the uneven distribu-
time increases as the third power of frequency, so the computa- tion of sources and receivers are unavoidable, it is desir-
tional challenge limits the frequency range of computations to able to minimize the impact of this inequality. For
relatively long periods (typically longer than 40 or 50 s), espe- example, Dziewonski (1984) used a form of a ‘summary
cially when using the adjoint framework, which requires many ray’ approach, counting the number of rays connecting a
iterations (sometimes 20–30) due to the slow convergence of particular ‘receiver cell’ and ‘source cell’ (each 5  5 at
the conjugate-gradient method. the equator) and then counterweighed individual
20 Deep Earth Seismology: An Introduction and Overview

East Pacific Rise

100

Depth (km)
200

300

Map view
4.0 4.2 4.4 4.6 4.8 at 70 km depth
VS (km s−1)
SEMum2 dln V S (%) dln ξ (%)
S362ANI
S40RTS −9 −4.5 0 4.5 9 −4 0 4 8 12
(a) Harmon et al. (2009) (b)

S N

100
Depth (km)

200
300
Min/max: −9.8/+2.3%
400
(c)

Figure 21 Isotropic VS structure beneath the East Pacific Rise (EPR). (a) A comparison between the mean 1-D VS profile obtained from the
high-resolution OBS-based tomographic study of EPR structure by Harmon et al. (2009) and that sampled from global models SEMum2 (French et al.,
2013), S362ANI (Kustowski et al., 2008), and S40RTS (Ritsema et al., 2011), in the same location. The 1-D profile of the Harmon et al. study was
obtained from a harmonic mean of the central portion of their model over length scales consistent with the a priori correlation lengths employed in the
SEMum2 inversion (400 km). SEMum2 more closely recovers the strength and depth of the low-velocity zone (LVZ) beneath the EPR inferred
from the local study than either of the other global models. (b) Relative variations in isotropic VS structure at 70 km depth in SEMum2, focused on
the EPR, showing both the location of the comparison in (a) (black arrow) and the extent of the 3750 km line of section in (c) (portion of EPR
highlighted in green). (c) A cross section along the EPR, following the green line and white dots shown in (b), illustrating the strength and extent of the
strong low-velocity anomaly imaged beneath the EPR in SEMum2 (c), nearly 10%. Black arrow and dashed line indicate approximate location of
profiles in (a). Reproduced from Figure S6 of French S, Lekic V, and Romanowicz B (2013) Waveform tomography reveals channeled flow at the base
of the oceanic asthenosphere. Science 342: 227–230.

observations such that impact of a given pair of cells was this inversion is an order of magnitude greater than in their
equal regardless of the number of connections. Since there earlier model S20RTS (Ritsema et al., 1999) and, yet, the differ-
were about 25% of pairs with only one connection and ences between the two models are small, especially at larger
only 25% had more than ten connections, the variance wavelengths. This means that the features within the ‘red’ part
reduction was very modest. of the spectrum are well resolved and further increase in the data
set will not change the image appreciably at long wavelengths.
Figure 22 compares model S40RTS (Ritsema et al., 2011) at There is a fundamental similarity among the models; all
seven different depths with four other relatively recent models: show an order of magnitude decrease in power between 100
S362ANI (Kustowski et al., 2008), SAW642AN (Panning and and 400 km depth (the scale for the top row of maps is 7%,
Romanowicz, 2006), TX2008 (Simmons et al., 2009), and or 3.5 times larger than at other depths). While the dramatic
HMSL-S (Houser et al., 2008). The data sets used to derive difference between the maps at 600 and 800 km is distinct for
each of these models were different; even the theories used the first three models, it is less so for the last two; the explana-
were different. Therefore, their overall similarity (with one tion is that models TX2008 and, especially, HMSL-S did not
exception) indicates congruence of the global 3-D models of use enough data with resolution in this depth range to map
the whole mantle. There are differences, of course, and some of properly this critically important feature. The rms amplitude is
them should be resolved, but the ‘big picture’ is robust and its the lowest between 650 and 1800 km depth but then increases
implications should be considered in modeling the global man- again to reach the maximum at the CMB.
tle flow. Such modeling is necessary to understand mantle The power spectrum as a function of depth and wave num-
dynamics and evolution, since tomography can retrieve only ber is a robust feature (see Figure 9 in Kustowski et al., 2008).
the present-day picture. What makes the work of Ritsema et al. On this basis, it is possible to distinguish several depth regions
(2011) particularly important is that the data set used in with different spectral characteristics; Dziewonski et al. (2010)
 S40RTS S362ANI SAW642AN TX2008 HMSL-S

100 km

400 km

600 km

800 km

Deep Earth Seismology: An Introduction and Overview

1000 km

2000 km

2800 km

Shear velocity variation

–2% 2%
Figure 22 Comparison of shear velocity maps at depths of (from top to bottom) 100, 400, 600, 800, 1000, 2000, and 2800 km for models (from left to right) S40RTS (Ritsema et al., 2011), S362ANI (Kustowski
et al., 2008), SAW642AN (Panning and Romanowicz, 2006), TX2008 (Simmons et al., 2009), and HMSL-S (Houser et al., 2008). The velocity varies from 2% to +2% from the average value. Shear

21
velocity perturbations are between 7% and +7% from the average value for the map at 100 km depth. Reproduced from Ritsema J, Deuss A, van Hejst HJ, and Woodhouse JH (2011) S40RTS: A degree-40
shear-velocity model for the mantle from new Raleigh wave dispersion, teleseismic traveltime and normal-mode splitting function measurements. Geophysical Journal International 184: 1223–1236,
http://dx.doi.org/10.1111/j.1365-246X.2010.04884.x.
22 Deep Earth Seismology: An Introduction and Overview

100 km ( ± 7%)
Heterosphere
200
400 ETZ
600 350 km ( ± 2.5%)

800
1000
1200 600 km ( ± 2.5%)

Depth (km) 1400

1600
1800 Super-plume
1250 km ( ± 1.5%)
zone
2000
2200
2400
2800 km ( ± 3.5%)
2600
2800
0 2 4 6 8 1012 1416 18
Angular degree
Figure 23 Left: Power spectrum of model S362ANI of Kustowski et al. (2008) as a function of harmonic degree and depth. Continuous white lines
indicate inferred major divisions in spectral characteristics; broken white lines show additional subdivisions. Center: Maps of shear velocity anomalies at
depths 100 km (7%), 350 km (2.5%), 600 km (2.5%), 1250 km (1.5%), and 2800 km (3.5%). Right: Schematic division of the mantle
into three zones with profound changes at their boundaries: ‘heterosphere’ with a very high level of heterogeneity; ‘extended transition zone,’ with a
weak, white spectrum at the top and power shifting to degree 2 above the 650 km discontinuity; and ‘superplume zone’ dominated by degrees 2 and 3,
which is strongest at the CMB but continuing with decreasing amplitude throughout the entire lower mantle.

proposed five such regions; more recent analysis (Dziewonski recently subducted slab, which appears to lay flat at the bottom
et al., 2013) finds that these can be combined into three of the transition zone for sometimes as far as 2000 km. This is
principal zones. consistent with the already discussed pattern of abrupt change
Figure 23 shows the power spectrum of model S362ANI with of shear velocity anomalies at depths of 600 and 800 km for
maps of shear velocity anomalies at five different depths. The the first three models in Figure 22.
principal divisions are between the ‘heterosphere’ (HS, Moho – At the top of the ‘superplume planet’ in the mid-mantle, the
250 km depth), the ‘extended transition zone’ (ETZ; signal is weak and has a somewhat flat spectrum, even though
250–650 km), and the lower mantle or ‘superplume zone’ the degrees 2 and 3 have the largest power. The amplitude of
(SPZ). The spectrum changes gradually within the ETZ from anomalies begins to increase below 1800 km depth. In the
white to being dominated by degree 2, but this change is not ‘abyssal layer’ (2400 km – CMB), degrees 2 and 3 are clearly
sudden and is not associated with the 410 km discontinuity. In dominant and show the pattern of two superplumes (African
the SPZ, there is a change in the gradient of heterogeneity at about and Pacific) separated by a wide circumpolar ring of higher
2400 km depth (e.g., Lekic et al., 2012), but the pattern does not than average velocities. This pattern of zero line is very robust,
change. These two more subtle spectral changes are shown with as shown by Lekic et al. (2012).
broken lines; the fundamental ones - with continuous lines. Figure 24 shows the result of cluster analysis of the deepest
Comparison of the five maps shows that the heterogeneity 1000 km of the mantle of the five models shown in Figure 22.
is distinctly different in each zone or subzone. ‘Heterosphere’ is On a constant area 2  2 grid, the average velocity anomaly
dominated by surface tectonic features: slow ridges, cooling for a particular model is classified as either ‘slow’ or ‘fast’; if it is
plates, and fast cratons. Between 150 and 250 km depth, the slow, integer 1 is added to the voting map and 0 if it is fast. If all
power of heterogeneity decreases by an order of magnitude, models are ‘slow’ at a given grid point, the corresponding color
hence the name. is deep red; if all models are fast, then deep blue. The robust-
Below 250 km, the upper part of the ETZ is characterized by ness of this classification is demonstrated by the fact that over
a nearly white spectrum. Oceans, particularly Pacific, are slow 95% of the area is either deep red or dark blue. The other
but do not show the age signature; continental structure tends important inference is that even though the individual models
to be faster than average, but the location of the anomalies is show variations in amplitude, these do not lead to the change
not as closely associated with the cratons as at 100 km depth. in sign of the anomaly: the dark red and dark blue regions are
Between 450 and 650 km depth, the degree-2 anomalies contiguous. One exception is a distinctly separated ‘slow’
become dominant. Regional-scale studies such as Fukao et al. anomaly, which Lekic et al. (2012) named the ‘Perm
(1992, 2001, 2009), Fukao and Obayashi (2013), and Chapter anomaly’ because of the proximity of its surface projection to
1.20 as well as global scale of Gu et al. (2001, 2003) explain the city of Perm (Russia). Another possible exception is a larger
this by stagnation above the 650 km discontinuity of the area beneath the East Pacific Rise, but the models are not
 Deep Earth Seismology: An Introduction and Overview 23

becomes much shallower. It is possible to speculate that the

steep gradient in the lowermost mantle is due to a chemical
heterogeneity whose strength decreases with radius to reach
that of the ambient mantle 500 km above the CMB. Further, it
would be possible to modify the ‘average’ Earth model by
adding a small 1-D linear increase of velocities between
1600 km depth and the CMB, where it would amount to
0.5%. If this were then subtracted from the 3-D model, the
map of anomalies in the abyssal zone region would be mostly
limited to slow anomalies, such that with this modification,
the principal anomalies would be limited to superplumes.
5 4 3 2 1 0 Thus, the chemical heterogeneity and the abyssal layer could
Figure 24 Cluster analysis of shear velocity (Vs) profiles (1000– be limited to the base of the superplumes.
2800 km depth range) from five recent global tomographic models
(Houser et al., 2008; Kustowski et al., 2008; Mégnin and Romanowicz,
2000; Simmons et al., 2010; Ritsema et al., 2011) in the lower mantle
1.01.4.4 Stability of the Planetary-Scale Heterogeneities
defines two regions whose geographic extents are consistent across Richards and Engebretson (1992) mapped four geophysical
models, tracing out the African and Pacific superplumes, as well as a functions (their Figure 1): geoid, hot-spot locations, velocity
single, globally contiguous faster-than-average region. Pixels that make
anomalies near the CMB, and a sum of the mass of all slabs
up this map are color-coded according to how many models assign the
subducted during the last 120 Ma. Correlation is significant
Vs profile beneath that point to the slow cluster. The models are spatially
filtered to exclude power at spherical harmonic degrees ‘ >18. Recon- only for degrees 2 and 3. Of the four functions, the first three
structed location (Torsvik et al., 2008) of the center of Siberian Trap are surface observables; only the seismic velocity anomalies
eruptions is indicated by the white cross. Reproduced from Lekic V, correspond to a particular depth – the lowermost mantle.
Cottaar S, Dziewonski A, and Romanowicz B (2012) Cluster analysis of Even though Ricard et al. (1993) and Lithgow-Bertelloni and
global lower mantle tomography: A new class of structure and implica- Richards (1998) proposed a model with the sinking slabs
tions for chemical heterogeneity. Earth and Planetary Science Letters distributed at different depths according to their age, correla-
357–358: 68–77. tion between the velocity and the density anomalies (sinking
slabs) is poor in the lowermost mantle. This may be called the
unanimous here. The white cross indicates the location of the ‘slab sinking paradox’: how the velocity anomalies near the
Siberian Traps 250 Ma ago by Torsvik et al. (2008). CMB may “know” about slabs that are still suspended higher
In order to put the Perm Anomaly on the map, it is necessary in the mantle? Dziewonski et al. (2010) proposed that this
to synthesize the first 12 harmonics of the expansion of the map paradox could be explained if the long-term (100 Ma) slab
shown in Figure 24 (see Figure 7 in Lekic et al., 2012). The subduction cyclically repeats itself and the anomalies in the
significance of the unanimity of all five models is that we can lowermost mantle reflect a long-term average of these cycles.
resolve the structure near the CMB to about 1000 km half- The important degrees are 2 and 3; for each of these degrees,
wavelength. This confirms the conclusion that degree-2 and the correlation is very high, but the relative power of degree 2
degree-3 anomalies are real features rather than a result of of seismic anomalies is three times greater than that of degree
insufficient data to resolve smaller-scale structures. Another con- 3; in integrated slab expansion, degree 3 has higher power than
clusion is that there is a dichotomy between the planetary-scale degree 2. This may mean that degree 2 may be more stable in
structures (superplumes) and smaller-scale anomalies, such as time than degree 3. Because of its apparent longevity,
ULVZs or hypothetical narrow plumes. It is not clear, therefore, Dziewonski et al. (2010) named the degrees 2 and 3 the
that the latter structures play a predominant role in whole- ‘mantle anchor structure’ (MAS).
mantle dynamics. On the other hand, the superplume pattern The degree-2 anomaly, which dominates the abyssal layer,
continues upward to the middle mantle. Figure 3 in Lekic et al. has geometric properties that have led Dziewonski et al. (2010)
(2012) shows that the voting pattern of Figure 24 continues to a speculation – which some thought ‘outrageous’ – that
into the middle mantle; even though the signal-to-noise ratio is degree-2 anomalies may be stable over very long times. What
considerably lower in the 800–1800 km depth range. The is special about degree-2 is that when translated into density
superplume pattern is distinct and contains all the surface loca- anomalies, it may affect the Earth’s nonhydrostatic moment of
tions of hot spots; also, the band of faster-than-average velocities inertia tensor. Figure 25 shows a comparison of the degree-2
continues to maintain its width, indicating that the fundamental anomaly at 2800 km depth in 3-D models S362ANI
pattern of velocity anomalies continues throughout the lower (Kustowski et al., 2008), SAW24B16 (Mégnin and
mantle, although with decreasing amplitude. Romanowicz, 2000) and S20RTS (Ritsema et al., 1999). It
A by-product of the voting procedure is the derivation of shows nearly identical images that are very close to the har-
average fast and slow velocity anomalies for each of the all five monic Y20 rotated so that its symmetry axis lies in the equato-
models (see Figure 5 of Lekic et al., 2012). There are two rial plane. The random test shows that to obtain such a result
important conclusions that can be drawn from this compari- by chance is less than 1 in 1000. If the negative velocity
son. First, the slow anomalies are about three times stronger anomalies are associated with higher effective density, then
than the fast ones; this is consistent with the fact that in the minimum moment of inertia axis would be crossing the
Figure 24, the area of ‘blue’ is much larger than that of ‘red.’ equator at the center of the red anomalies (about 10 east or
Second, there is a sharp change in the gradient of the velocity 170 west) as required by the rotation dynamics. The solution
anomalies at about 500 km from the CMB, above which it is degenerate for the other two principal components: any axis
24 Deep Earth Seismology: An Introduction and Overview

(a) (b) (c)

Figure 25 The nearly indistinguishable degree-2 structures at 2800 km depth of three global tomographic S-velocity models: (a) S362ANI
(Kustowski et al., 2008), (b) SAW24B16 (Mégnin and Romanowicz, 2000), (c) S20RTS (Ritsema et al., 1999). Pink circles indicate paleopole locations
from the true polar wander reconstructions of Besse and Courtillot (2002). Reproduced from Dziewonski A, Lekic, and Romanowicz A (2010)
Mantle anchor structure; an argument for bottom up tectonics. Earth and Planetary Science Letters 299: 69–79.

passing through the polar great circle crossing the equator at The ‘ETZ’ is the critical region in which the communication
100 east or 80 west will have the maximum moment of between the heterosphere and the superplume zone is taking
inertia. Thus, this great circle would be the preferred direction place. It is clear that some slabs become stagnant above the
of ‘true polar wonder’ (TPW) due to mass redistribution. The 650 km discontinuity, but the degree-2 velocity anomaly is
orange dots in Figure 25 are the TPW positions during the last well correlated with the integrated slabs during the last
200 Ma as determined by Besse and Courtillot (2002); the 15–20 Ma and does not show evidence of earlier accumula-
agreement is good even though no TWP observations were tion. Thus, it is not clear what happens with the material
included in building the seismic models. Modeling of TPW subducted earlier. It is also not clear how the superplumes
using geodynamic data was initiated by Steinberger and O- interact with the transition zone. The Pacific is slow both
’Connell (1997) and continues till the present (c.f. Conrad below and above the 650 km discontinuity, but this is not as
et al., 2013; Rouby et al., 2010; Steinberger and Torsvik, obvious for the African superplume. Romanowicz and Gung
2010). The TPW is increasingly difficult to determine for (2002) argued that attenuation in the transition zone better
times before the magnetic stripes on the oceanic floor have reflects its thermal state than the velocity anomalies and that
been destroyed. Some paleomagnetic evidence exists that the the Q anomalies in the transition zone are well correlated with
MAS hypothesis is consistent with the data as old as 1.3 Ga the velocity anomalies at the CMB.
(Jean Besse, 2013, personal communication). If this is correct, The paradigm of three-layer mantle dynamics, each of which
then there is no reason not to test an assumption that it was acts on a different spatial and temporal scale and has been
formed not long after the Earth’s accretion. derived mostly from seismic data providing a present-day snap-
shot of 3-D velocity anomalies, must be tested by geodynamic
modeling. From the published results of numerical modeling of
1.01.4.5 The Need for Consideration of More Complete
mantle flow, it is clear that such a structure with three boundary
Modeling of Mantle Flow
layers has yet to be realized and match the observations.
A quadrupole convection essentially identical to the pattern in With the recent advances in numerical modeling, a clearer
Figure 25 has been proposed by Busse (1983) and its stability picture of the connection between lower-mantle and transition
implications pointed out by Le Pichon and Huchon (1984) on zone structure on the one hand and structure in the
the basis of correlation of the degree-2 geoid with velocity asthenosphere on the other hand is starting to emerge.
anomalies in the lowermost mantle (Dziewonski et al., 1977). Figure 26 shows a 3-D rendering of the central Pacific portion
With the incomparably higher quality of current global models of upper-mantle model SEMum2 (French et al., 2013), viewed
of heterogeneity, we do not have a mantle flow model that from the east to the west, which shows a number of vertically
reproduces large-scale features revealed by mantle tomography. elongated low-velocity columns (plumes?), many of them
The concepts of ‘top-down’ or ‘bottom-up’ tectonics are not rooted in the lower mantle and extending up to about 350 km
contradictory, but rather complementary. The heterosphere of depth. At shallower depths, this structure is replaced by a series
Figure 23 is highly energetic – it contains above 80% of the of horizontally elongated low-velocity channels or ‘fingers,’
volume integrated power of mantle heterogeneity – and acts on aligned parallel to the absolute plate motion and extending
a relatively short timescale: the pattern of subductions has into the low-velocity zone right below the lithosphere. These
changed considerably during the last 200 Ma (Lithgow- low-velocity channels, which are present in all major ocean
Bertelloni and Richards, 1998, Figure 6). The lower-mantle basins, are spaced regularly with a wavelength of about 1800–
heterogeneity structure (superplume zone) shows great stabil- 2000 km and manifest the dynamic interaction in the
ity, as indicated by the ‘slab paradox’ and TPW results, and may ‘heterosphere’ between the upwellings from the ‘superplume
impose the long time influence on the heterosphere’s behavior. zone’ and ETZ (e.g., Figure 23) and the tectonic plates.
 Deep Earth Seismology: An Introduction and Overview 25

Macdonald Samoa Marquesas Hawaii

Pitcairn Tahiti
dln Vs
3.00

2.00

1.00 250

Depth (km)
0.00
500
–1.00 Low-velocity
fingers
750
–2.00

–3.00 1000

Low-velocity
plumes
North

Figure 26 3-D rendering of the low-velocity parts of a portion of model SEMum2 (French et al., 2013) in the central Pacific, shown down to
1000 km depth. The range of velocity anomalies shown is 1% to 3%. The view is from the east, standing in the vicinity of the East Pacific Rise
and looking towards the west. Green cones indicate major hot spots. Note the regularly spaced low-velocity fingers right below the very slow
low-velocity zone. These fingers extend for several thousand kilometers into the Pacific basin, in the direction of the Pacific Plate absolute plate
motion. Similar structures are seen in the south and North Atlantic Ocean and in the Indian Ocean.

We expect that, in the next 5–10 years, these images, as well as Backus GE and Gilbert F (1970) Uniqueness in the inversion of gross Earth data.
those of the lower mantle, will become even sharper, as the Philosophical Transactions of the Royal Society of London, Series A 266: 169–205.
Benioff H (1958) Long waves observed in the Kamchatka earthquake of November 2,
frequency range accessible to global tomography based on
1952. Journal of Geophysical Research 63: 589–593.
numerical methods increases, providing definitive clues on such Benndorf H (1905) Über die Art der Fortpflanzung der Erdbebenwellen im Erdinneren. 1.
fundamental questions as the relation of hot spots to plumes, and Mitteilung. Sitzungsberichte der Kaiserlichen Akademie in Wien. Mathematisch-
plumes to superplumes, and the role of superplumes in the global Naturwissenschaftliche Klasse 114, Mitteilungen der Erdbebenkommission. Neue
mantle circulation and the fate of subducted slabs. Folge 29: 1–42.
Benndorf H (1906) Über die Art der Fortpflanzung der Erdbebenwellen im Erdinneren.
2. Mitteilung. Sitzungsberichte der Kaiserlichen Akademie in Wien. Mathematisch-
Naturwissenschaftliche Klasse 115, Mitteilungen der Erdbebenkommission. Neue
Folge 31: 1–24.
References Berkner LV and Members of the panel (1959) Report of the Panel on Seismic
Improvement. The Need for Fundamental Research in Seismology, p. 212.
Adams RD, et al. (2002) International seismology. In: Lee WHK, Kanamori H, Washington, DC: Department of State.
Jennings PC, and Kisslinger C (eds.) International Handbook of Earthquake and Besse J and Courtillot V (2002) Apparent and true polar wander and the geometry of the
Engineering Seismology Part A, pp. 29–37. San Diego, CA: Academic Press. geomagnetic field over the last 200 Myr. Journal of Geophysical Research
Agnew D, Berger J, Buland R, Farrell W, and Gilbert F (1976) International deployment 107: 2300. http://dx.doi.org/10.1029/2000JB000050.
of accelerometers: A network for very long period seismology. EOS. Transactions of Birch AF (1952) Elasticity and constitution of the Earth’s interior. Journal of Geophysical
the American Geophysical Union 57: 180–188. Research 57: 227–286.
Agnew DC, Berger J, Farrell WE, Gilbert JF, Masters G, and Miller D (1986) Project IDA; a Boschi L and Dziewonski AM (1999) High’ and ‘low’ resolution images of the Earth’s
decade in review. EOS. Transactions of the American Geophysical Union 67: 203–212. mantle – Implications of different approaches to tomographic modeling. Journal of
Agnew DC, et al. (2002) History of seismology. In: Lee WHK, Kanamori H, Jennins PC, Geophysical Research 104: 25567–25594.
and Kisslinger C (eds.) International Handbook of Earthquake and Engineering Brune J and Dorman J (1963) Seismic Waves and the Structure of the Crust and
Seismology, pp. 3–12. San Diego, CA: Academic Press. Mantle in the Canadian Shield: Geological Society of America, pp. 28–29, Special
Aki K, Christofferson A, and Husebye E (1977) Determination of the three-dimensional Paper.
structure of the lithosphere. Journal of Geophysical Research 82: 277–296. Buland R, Berger J, and Gilbert F (1979) Observations from the IDA network of
Alsop LE, Sutton GH, and Ewing M (1961) Free oscillations of the Earth observed attenuation and splitting during a recent earthquake. Nature 277: 358–362.
on strain and pendulum seismographs. Journal of Geophysical Research Bullen KE (1940) The problem of the Earth’s density variation. Bulletin of the
66: 631–641. Seismological Society of America 30: 235–250.
Alterman Z, Jarosch H, and Pekeris CL (1959) Oscillations of the Earth. Proceedings of Bullen KE (1949) Compressibility-pressure hypothesis and the Earth’s interior.
the Royal Society of London, Series A 259: 80–95. Geophysical Journal of the Royal Astronomical Society 5: 355–368.
Anderson DL (1963) Recent evidence concerning the structure and composition of the Busse F (1983) Quadrupole convection in the lower mantle. Geophysical Research
Earth’s mantle. Physics and Chemistry of the Earth 6: 1–129. Letters 10: 285–288.
Anderson DL and Archambeau CB (1964) The anelasticity of the Earth. Journal of Capdeville Y, Chaljub E, Vilotte JP, and Montagner JP (2003) Coupling the spectral
Geophysical Research 69: 2071–2084. element method with a modal solution for elastic wave propagation in global earth
Anderson DL and Hart RS (1978) Q of the Earth. Journal of Geophysical Research models. Geophysical Journal International 152: 34–67. http://dx.doi.org/10.1046/
83: 5869–5882. j.1365-246X.2003.01808.x.
Backus GE and Gilbert F (1961) The rotational splitting of free oscillations of the Earth. Chen P, Jordan TH, and Zhao L (2007) Full three-dimensional tomography: A
Proceedings of the National Academy of Sciences of the United States of America comparison between the scattering-integral and adjoint-wavefield methods.
47: 362–371. Geophysical Journal International 170: 175–181.
Backus GE and Gilbert F (1967) Numerical applications of formalism for geophysical Colli L, Fichtner A, and Bunge H-P (2013) Full waveform tomography of the upper
inverse problems. Geophysical Journal of the Royal Astronomical Society mantle in the south Atlantic region: Imaging a westward fluxing shallow
13: 247–276. asthenosphere? Tectonophysics 604: 26–40.
Backus GE and Gilbert F (1968) The resolving power of gross Earth data. Geophysical Conrad CP, Steinberger B, and Torsvik TH (2013) Stability of active mantle upwelling
Journal of the Royal Astronomical Society 16: 169–205. revealed by net characteristics of plate tectonics. Nature 498: 479–482.
26 Deep Earth Seismology: An Introduction and Overview

Cooley JW and Tukey JW (1965) An algorithm for machine computation of complex Fukao Y, Widiyantoro S, and Obayashi M (2001) Stagnant slabs in the upper and lower
Fourier series. Mathematics of Computation 19: 297–301. mantle transition region. Reviews of Geophysics 39: 291–323.
Dahlen F, Hung S-H, and Nolet G (2000) Fréchet kernels for finite-frequency Galitzin BB (1914) Vorlesungen über Seismometrie, deutsche Bearbeitung unter
traveltimes – I. Theory. Geophysical Journal International 141: 157–174. Mitwirkung von Clara Reinfeldt (also Golicyn). Herausgegeben von Oskar Hecker,
Dalton CA (2007) The global attenuation structure of the upper mantle. Ph.D. thesis. vol. VIII, p. 538. Berlin: Verlag Teubner.
Harvard University. Geiger L (1910) Herdbestimmung bei Erdbeben aus den Ankunftszeiten. Nachrichten
Dziewonski AM (1984) Mapping the lower mantle: Determination of lateral von der K€oniglichen Gesellschaft der Wissenschaften zu G€ottingen. Mathematisch-
heterogeneity in P-velocity up to degree and order 6. Journal of Geophysical physikalische Klasse. pp. 331–349 (Translated into English by Peebles FWL,
Research 89: 5929–5952. Anthony H, Corey SJ).
Dziewonski AM and Anderson DL (1981) Preliminary Earth model (PREM). Physics of Geiger L (1912) Probability method for the determination of earthquake epicenters from
the Earth and Planetary Interiors 25: 297–356. the arrival time only. Bulletin of St. Louis University 8: 60–71.
Dziewonski AM, Chou T-A, and Woodhouse JH (1981) Determination of earthquake Gilbert F and Dziewonski AM (1975) An application of normal mode theory to the
source parameters from waveform data for studies of global and regional seismicity. retrieval of structural parameters and source mechanisms from seismic spectra.
Journal of Geophysical Research 86: 2825–2852. Philosophical Transactions of the Royal Society of London, Series A
Dziewonski AM and Gilbert F (1972) Observations of normal modes from 84 recordings 278: 187–269.
of the Alaskan earthquake of 28 March 1964. Geophysical Journal of Royal Green RWE and Hales AL (1968) Travel times of P waves to 30 degrees in central United
Astronomical Society 27: 393–446. States and upper mantle structure. Bulletin of the Seismological Society of America
Dziewonski AM and Gilbert F (1973) Observations of normal modes from 84 recordings of 58: 267–289.
the Alaskan earthquake of 28 March 1964, Part II: Spheroidal overtones from 285–100 Gu YJ, Dziewonski AM, and Ekstr€om G (2003) Simultaneous inversion for mantle shear
seconds. Geophysical Journal of the Royal Astronomical Society 35: 401–437. velocity and topography of transition zone discontinuities. Geophysical Journal
Dziewonski AM and Gilbert F (1974) Temporal variation of the seismic moment tensor International 154: 559–583.
and the evidence of precursive compression for two deep earthquakes. Nature Gu JY, Dziewonski AM, Su W-J, and Ekstr€om G (2001) Shear velocity model of the
247: 185–188. mantle and discontinuities in the pattern of lateral heterogeneities. Journal of
Dziewonski AM and Haddon RAW (1974) The radius of the core–mantle boundary Geophysical Research 106: 11169–11199.
inferred from travel time and free oscillation data: A critical review. Physics of the Gung Y, Panning M, and Romanowicz B (2003) Global anisotropy and the thickness of
Earth and Planetary Interiors 9: 28–35. continents. Nature 422: 707–711.
Dziewonski AM, Hager BH, and O’Connell RJ (1977) Large scale heterogeneities in the Gung Y and Romanowicz B (2004) Q tomography of the upper mantle using three
lower mantle. Journal of Geophysical Research 82: 239–255. component long period waveforms. Geophysical Journal International
Dziewonski AM and Hales AL (1972) Numerical analysis of dispersed seismic waves. 157: 813–830.
In: Adler B, Fernbach S, and Bolt BA (eds.) Methods in Computational Physics, Gutenberg B (1913) Über die Konstitution des Erdinnern, erschlossen aus
vol. 11, pp. 39–85. New York: Academic Press. Erdbebenbeobachtungen. Physikalische Zeitschrift 14: 1217–1218.
Dziewonski AM, Hales AL, and Lapwood ER (1975) Parametrically simple Earth models Hales AL, Lapwood ER, and Dziewonski AM (1974) Parameterization of a spherically
consistent with geophysical data. Physics of the Earth and Planetary Interiors symmetrical Earth model with special references to the upper mantle. Physics of the
10: 12–48. Earth and Planetary Interiors 9: 9–12.
Dziewonski AM, Lekic V, and Romanowicz B (2010) Mantle anchor structure: An Hammer PTC, Clowes RM, Cook FA, van der Velden AJ, and Vosudevan K (2010) The
argument for bottom up tectonics. Earth and Planetary Science Letters 299(2010): Lithoprobe trans-continental lithospheric cross-sections imaging the internal
69–79. structure of the North American continent. Canadian Journal of Earth Sciences
Dziewonski AM, Lekic V, and Romanowicz B (2013) Planet within a planet: 47: 821–857.
Implications of principal component analysis of global tomographic models. EOS. Harjes H-P and Seidl D (1978) Digital recording and analysis of broad-band seismic
Transactions of the American Geophysical Union 94(Fall Meeting Supplement), data at the Gräfenberg (GRF) array. Journal of Geophysics 44: 511–523.
Abstract DI32-01. Harmon N, Forsyth D, and Weeraratne DS (2009) Thickening of young Pacific
Dziewonski AM and Steim JM (1982) Dispersion and attenuation of mantle waves lithosphere from high-resolution Rayleigh wave tomography: A test of
through waveform inversion. Geophysical Journal of the Royal Astronomical Society the conductive cooling model. Earth and Planetary Science Letters
70: 503–527. 278: 96–106.
Dziewonski A, Su W-J, and Woodward R (1991) Global structures of the Earth’s interior. Haskell NA (1953) The dispersion of surface waves on multilayered media. Bulletin of
EOS, Transactions American Geophysical Union, fall meeting: 72: 451. the Seismological Society of America 43: 17–43.
Ekstr€om G, Dziewonski AM, Maternovskaya NN, and Nettles M (2005) Global seismicity Herglotz G (1907) Über das Benndorfsche Problem der Fortpflanzungsgeschwindigkeit
of 2003: Centroid-moment-tensor solutions for 1087 earthquakes. Physics of the der Erdbebenwellen. Physikalische Zeitschrift 8: 145–147.
Earth and Planetary Interiors 148: 327–351. Houser C, Masters G, Shearer P, and Laske G (2008) Shear and compressional velocity
Ekstr€om G and Dziewoski AM (1998) The unique anisotropy of the Pacific upper mantle. models of the mantle from cluster analysis of long-period waveforms. Geophysical
Nature 394: 168–172. Journal International 174: 195–212. http://dx.doi.org/10.1111/j.1365-
Ekstr€om G, Tromp J, and Larson EW (1997) Measurements and models of global 246X.2008.03763.x.
surface wave propagation. Journal of Geophysical Research 102: 8137–8157. Jeffreys H (1926) The rigidity of the Earth’s central core. Monthly Notices of the Royal
Ewing M, Jardetzky WS, and Press F (1957) Elastic Waves in Layered Media. New York: Astronomical Society 1(Geophysical Supplement): 371–383.
McGraw-Hill. Jeffreys H (1939) The times of P, S and SKS, and the velocities of P and S, month not.
Ewing M and Press F (1954) An investigation of mantle Rayleigh waves. Bulletin of the Royal Astronomical Society, Geophysical Supplement 4: 498–533.
Seismological Society of America 44: 127–148. Jeffreys H and Bullen KE (1940) Seismological Tables, p. 50. London: British
Fichtner A, Kennett B, Igel H, and Bunge H (2009) Full seismic waveform tomography Association for the Advancement of Science.
for upper mantle structure in the Australasian region using adjoint methods. Jobert N (1956) Evaluation de la période d’oscillation d’une sphere hétérogène, par
Geophysical Journal International 179: 1703–1725. application du principe de Rayleigh. Comptes Rendus De L Academie Des Sciences,
French S, Lekic V, and Romanowicz B (2013) Waveform tomography reveals channeled Paris 243: 1230–1232.
flow at the base of the oceanic asthenosphere. Science 342: 227–230. Jobert N (1957) Sur la période propre des oscillations sphèroı̈dales de la Terre.
Fuchs K and Müller G (1971) Computation of synthetic seismograms and comparison Comptes Rendus De L Academie Des Sciences, Paris 244: 242–258.
with observations. Geophysical Journal of the Royal Astronomical Society Johnson LR (1967) Array measurements of P velocity in the upper mantle. Journal of
23: 417–433. Geophysical Research 72: 6309–6325.
Fukao Y and Obayashi M (2013) Subducted stagnant slabs above, penetrating through Jordan TH (1978) A procedure for estimating lateral variations from low-frequency
and trapped below the 660 km discontinuity. Journal of Geophysical Research eigenspectra data. Geophysical Journal of the Royal Astronomical Society
118: 1–19. 52: 441–455.
Fukao Y, Obayashi M, Inoue H, and Nenbai M (1992) Subducting slabs stagnant in the Jordan TH and Anderson DL (1974) Earth structure from free oscillations
mantle transition zone. Journal of Geophysical Research 97: 4809–4822. and travel times. Geophysical Journal of the Royal Astronomical Society
Fukao Y, Obayashi M, and Nakakuki T (2009) the Deep Slab Project Group. Annual 36: 411–459.
Review of Earth and Planetary Sciences 37: 19–46. http://dx.doi.org/10.1146/ Julian BR and Sengupta MK (1973) Seismic travel time evidence for lateral
annurev.earth.36.031207.124224. inhomogeneity in the deep mantle. Nature 242: 443–447.
 Deep Earth Seismology: An Introduction and Overview 27

Kennett BLN, Engdahl ER, and Buland R (1995) Constraints on seismic velocities in the Panning M and Romanowicz B (2006) A three dimensional radially anisotropic model
Earth from traveltimes. Geophysical Journal International 122: 108–124. of shear velocity in the whole mantle. Geophysical Journal International
Knott CG (1899) Reflection and refraction of elastic waves, with seismological 167: 361–379.
applications. The London, Edinburgh, and Dublin Philosophical Magazine and Park J, Butler R, Anderson K, et al. (2005) Performance review of the global
Journal of Science, Series 5 48, 64–97, 567–569. seismographic network for the Sumatra–Andaman megathrust earthquake.
Komatitsch D and Tromp J (2002) Spectral-element simulations of global seismic wave Seismological Research Letters 76(3): 331–343.
propagation: I. Validation. Geophysical Journal International 149: 390–412. Pekeris CL, Alterman Z, and Jarosch H (1961a) Comparison of theoretical with
Komatitsch D and Vilotte J-P (1998) The spectral element method: An efficient tool to observed values of the periods of the free oscillations of the Earth. Proceedings of
simulate the seismic response of 2D and 3D geological structures. Bulletin of the the National Academy of Sciences of the United States of America 47: 91–98.
Seismological Society of America 88: 368–392. Pekeris CL, Alterman Z, and Jarosh H (1961b) Rotational multiplets in the spectrum of
Kustowski B, Ekstr€om G, and Dziewonski AM (2008) The shear-wave velocity structure the Earth. Physical Review 122: 1692–1700.
in the upper mantle beneath Eurasia. Geophysical Journal International Pekeris CL and Jarosch H (1958) The free oscillations of the Earth. In: Benioff H,
174: 978–992. Ewing M, HowellJun BF, and Press K (eds.) Contributions in Geophysics in Honor
Landisman M, Sato Y, and Nafe JE (1965) Free vibrations of the Earth and the properties of Beno Gutenberg, pp. 171–192. New York: Pergamon.
of its deep interior regions: Part 1. Density. Geophysical Journal of the Royal Peter D, Tape C, Boschi L, and Woodhouse JH (2007) Surface wave tomography:
Astronomical Society 9: 439–502. Global membrane waves and adjoint methods. Geophysical Journal International
Le Pichon X and Huchon P (1984) Geoid, Pangea and convection. Earth and Planetary 171: 1098–1117. http://dx.doi.org/10.1111/j.1365-246X.2007.03554.x.
Science Letters 67: 123–135. Peterson J, Butler HM, Holcomb LT, and Hutt C (1976) The seismic research
Lehmann I (1936) P’. Publ. Bureau Cent. Séism. Inter., Série A, Trav. Scient 14: 87–115. observatory. Bulletin of the Seismological Society of America 66: 2049–2068.
Lekic V, Cottaar S, Dziewonski AM, and Romanowicz B (2012) Cluster analysis of Poupinet G, Frechet J, and Thouvenot F (1989) Portable short period vertical seismic
global lower mantle tomography: A new class of structure and implications for stations transmitting via telephone or satellite. In: Cassinis R and Nolet G (eds.)
chemical heterogeneity. Earth and Planetary Science Letters 357–358: 68–77. Digital Seismology and Fine Modelling of the Lithosphere, pp. 9–26. London:
Lekic V and Romanowicz B (2011) Inferring upper-mantle structure by full waveform Plenum.
tomography with the spectral element method. Geophysical Journal International Press F (1956) Determination of crustal structure from phase velocity of Rayleigh waves
185: 799–831. – Part I. Southern California. Bulletin of Geological Society of America
Li XD and Romanowicz B (1995) Comparison of global waveform inversions with and 67: 1647–1658.
without considering cross branch coupling. Geophysical Journal International Ricard Y, Richards M, Lithgow-Bertelloni C, and LeStunff YL (1993) A geodynamic
121: 695–709. model of mantle density heterogeneity. Journal of Geophysical Research
Li X-D and Romanowicz B (1996) Global mantle shear velocity model developed using 98: 21895–21910.
nonlinear asymptotic coupling theory. Journal of Geophysical Research Richards M and Engebretson D (1992) Large-scale mantle convection and the history of
101: 22245–22272. subduction. Nature 355: 437–440.
Li XD and Tanimoto T (1993) Waveforms of long-period body waves in a slightly Rickers F, Fichtner A, and Trampert J (2013) The Iceland–Jan Mayen plume system and
aspherical Earth model. Geophysical Journal International 112: 92–102. its impact on mantle dynamics in the North Atlantic region: Evidence from full-
Lithgow-Bertelloni C and Richards MA (1998) The dynamics of Cenozoic and Mesozoic waveform inversion. Earth and Planetary Science Letters 367: 39–51.
plate motions. Reviews of Geophysics 36: 27–78. Ritsema J, Deuss A, van Heijst HJ, and Woodhouse J (2011) HS40RTS: A degree-40
Liu H-P, Anderson DL, and Kanamori H (1976) Velocity dispersion due to anelasticity: shear-velocity model for the mantle from new Rayleigh wave dispersion, teleseismic
Implications for seismology and mantle composition. Geophysical Journal of the traveltime and normal-mode splitting function measurements. Geophysical Journal
Royal Astronomical Society 47: 41–58. International 184: 1223–1236.
Lognonné P (1991) Normal modes and seismograms in an anelastic rotating Earth. Ritsema J, van Heijst HH, and Woodhouse JH (1999) Complex shear wave velocity
Journal of Geophysical Research 96: 20309–20319. structure imaged beneath Africa and Iceland. Science 286: 1925–1928.
Lognonné P and Romanowicz B (1990) Modelling of coupled normal modes of the Ritsema J, van Heijst HJ, and Woodhouse JH (2004) Global transition zone
Earth: The spectral method. Geophysical Journal International 102(2): 365–395. tomography. Journal of Geophysical Research 109. B02302: http://dx.doi.org/
Love AEH (1911) Some Problems of Geodynamics. Cambridge: Cambridge University 10.1029/2003JB002610.
Press. Romanowicz B (1979) Seismic structure of the upper mantle beneath the United States
Masters G, Jordan TH, Silver PG, and Gilbert F (1982) Aspherical earth structure from by three-dimensional inversion of body wave arrival times. Geophysical Journal of
fundamental spheroidal mode data. Nature 298: 609–613. the Royal Astronomical Society 57: 479–506.
Mégnin C, Bunge HP, Romanowicz B, and Richards MA (1997) Imaging 3-D spherical Romanowicz B (1991) Seismic tomography of the Earth’s mantle. Annual Review of
convection models: What can seismic tomography tell us about mantle dynamics? Earth and Planetary Science 19: 77–99.
Geophysical Research Letters 24(11): 1299–1302. Romanowicz B (2003) Global mantle tomography: Progress status in the last 10 years.
Mégnin C and Romanowicz B (1999) The effects of the theoretical formalism and data Annual Review of Geophysics and Space Physics 31: 303–328.
selection on mantle models derived from waveform tomography. Geophysical Romanowicz BA, Cara M, Fels JF, and Rouland D (1984) Geoscope: A French initiative
Journal International 138: 366–380. in long-period three-component seismic networks. EOS. Transactions of the
Mégnin C and Romanowicz B (2000) The three-dimensional shear velocity structure of American Geophysical Union 65: 753–754.
the mantle from the inversion of body, surface and higher-mode waveforms. Romanowicz BA and Dziewonski AM (1986) Towards a federation of broadband seismic
Geophysical Journal International 143(3): 709–728. networks. EOS. Transactions of the American Geophysical Union 67: 541–542.
Mendiguren J (1973) Identification of free oscillation spectral peaks for 1970 July 31, Romanowicz B and Gung Y (2002) Superplumes from the core–mantle boundary to the
Colombian deep shock using the excitation criterion. Geophysical Journal of the lithosphere: Implications for heat flux. Science 296(2002): 513–516.
Royal Astronomical Society 33: 281–321. Romanowicz B, Karczewski J-F, et al. (1991) The GEOSCOPE program: Present
Mohorovičić A (1910) Potres od 8. X 1909. God. Izvjesće Zagr. met. Ops. Zag. 1909, status and perspectives. Bulletin of the Seismological Society of America
Zagreb (Das Beben vom 8. X 1909. Jahrbuch des meteorologischen Observatoriums 81: 243–264.
in Zagreb f ür das Jahr 1909), 9, Part 4, 1–63. Rouby H, Greff-Lefftz M, and Besse J (2010) Mantle dynamics, geoid, inertia and TPW
Montelli R, Nolet G, Dahlen F, Masters G, Engdahl E, and Hung SH (2004a) Global since 120 Myr. Earth and Planetary Science Letters 292: 301–311.
P and PP travel time tomography: Rays vs. waves. Geophysical Journal International Roult G, Montagner JP, Romanowicz B, et al. (2010a) The GEOSCOPE program:
158: 637–654. Progress, and challenges during the past 30 years. Seismological Research Letters
Montelli R, Nolet G, Dahlen F, Masters G, Engdahl E, and Hung SH (2004b) Finite- 81: 428–452.
frequency tomography reveals a variety of plumes in the mantle. Science Roult G, Montagner J-P, Romanowicz B, et al. (2010b) The GEOSCOPE Program:
303: 338–343. Progress and challenges during the past 30 years. Seismological Research Letter
Ness NF, Harrison JC, and Slichter LB (1961) Observations of the free oscillations of the 81: 429–452.
Earth. Journal of Geophysical Research 66: 621–629. Russakoff D, Ekstr€om G, and Tromp J (1998) A new analysis of the great 1970
Nolet G and Dahlen FA (2000) Wave front healing and the evolution of seismic delay Colombia earthquake and its isotropic component. Journal of Geophysical Research
times. Journal of Geophysical Research 105: 19043–19054. 102: 20423–20434.
Oldham RD (1906) The constitution of the interior of the Earth, as revealed by Sailor RV and Dziewonski AM (1978) Observations and interpretation of attenuation of
earthquakes. Quarterly Journal of the Geological Society 62(1–4): 456–475. normal modes. Geophysical Journal of the Royal Astronomical Society 53: 559–581.
28 Deep Earth Seismology: An Introduction and Overview

Schweitzer J (2007) In: Gubbins D and Herrero-Bervera E (eds.) Encyclopedia of van der Hilst R, Kennett B, Christie D, and Grant J (1994) Project Skippy explores the
Geomagnetism and Paleomagnetism: Kluwer-Springer, ISBN:1-4020-3992-1 (in lithosphere and upper mantle below Australia. EOS. Transactions of the American
print). Geophysical Union 75(177): 180–181.
Simmons NA, Forte AM, and Grand SP (2009) Joint seismic, geodynamic and mineral van Heijst HJ and Woodhouse JH (1997) Measuring surface-wave overtone phase
physical constraints on three-dimensional mantle heterogeneity: Implications for the velocities using a mode-branch stripping technique. Geophysical Journal
relative importance of thermal versus compositional heterogeneity. Geophysical International 131: 209–230.
Journal International 177: 1284–1304. von Rebeur-Paschwitz E (1889) The earthquake of Tokyo, April 18, 1889. Nature
Simmons NA, Forte A, Boschi L, and Grand S (2010) GyPSuM: a joint tomographic 40: 294–295.
model of mantle density and seismic wave speeds. Journal of Geophysical Research von Rebeur-Paschwitz E (1895) Horizontalpendal-Beobachtungen auf der Kaiserlichen
115(B12310). http://dx.doi.org/10.1029/2010JB007631. Universitats-Sternwarte zu Strassburg 1892–1894. Gerlands Beiträge zur Geophysik
Souriau A and Souriau M (1983) Test of tectonic models by great-circle Rayleigh waves. 2: 211–536.
Geophysical Journal of the Royal Astronomical Society 73: 533–551. Wiechert E (1907) Über Erdbebenwellen. Theoretisches über die Ausbreitung der
Steim JM (1986) The very-broad-band seismograph. Doctoral Thesis, Cambridge, MA: Erdbebenwellen. Nachrichten von der K€oniglichen Gesellschaft der Wissenschaften
Department of Geological Sciences, Harvard University, p. 184. zu G€ottingen, Mathematisch-physikalische Klasse, pp. 413–529.
Stein S and Wysession M (2003) An Introduction to Seismology, Earthquakes and Earth Wielandt E (2002) Seismic sensors and their calibration. In: Bormann IP (ed.) IASPEI
Structure. Oxford: Blackwell, ISBN: 0865420785. New Manual of Seismological Observatory Practice, vol. 1, pp. 1–46. Potsdam:
Steinberger B and O’Connell RJ (1997) Changes of the Earth’s rotation axis owing to GeoForschungsZentrum Potsdam, Chapter 5.
advection of mantle density heterogeneities. Nature 387: 169–173. Wielandt E and Steim JM (1986) A digital very-broad-band seismograph. Annales
Steinberger B and Torsvik TH (2010) Toward an explanation for the present and past Geophysicae 4: 227–232.
locations of the poles. Geochemistry, Geophysics, Geosystems. http://dx.doi.org/ Wielandt E and Streckeisen G (1982) The leaf-spring seismometer: Design and
10.1029/2009GC002889. performance. Bulletin of the Seismological Society of America 72: 2349–2367.
Stoneley R (1928) A Rayleigh wave problem. Proceedings of the Leeds Philosophical Woodhouse JH and Dahlen FA (1978) The effect of a general aspherical perturbation on
and Literary Society (Science Section) 1: 217–225. the free oscillations of the Earth. Geophysical Journal of the Royal Astronomical
Su W-J and Dziewonski AM (1991) Predominance of long-wavelength heterogeneity in Society 53: 335–354.
the mantle. Nature 352: 121–126. Woodhouse JH and Dziewonski AM (1984) Mapping the upper mantle: Three
Su W-J and Dziewonski AM (1992) On the scale of mantle heterogeneity. Physics of the dimensional modeling of earth structure by inversion of seismic waveforms. Journal
Earth and Planetary Interiors 74: 29–54. of Geophysical Research 89: 5953–5986.
Takeyuchi H (1959) Torsional oscillations of the Earth and some related problems. Woodhouse J and Dziewonski A (1986) Three dimensional mantle models based on
Geophysical Journal of the Royal Astronomical Society 2: 89–100. mantle wave and long period body wave data. EOS. Transactions of the American
Tape C, Liu Q, Maggi A, and Tromp J (2010) Seismic tomography of the southern Geophysical Union 67: 307.
California crust based on spectral-element and adjoint methods. Geophysical Woodhouse JH and Dziewonski AM (1989) Seismic modelling of the Earth’s large-scale
Journal International 160: 195–216. three dimensional structure. Philosophical Transactions of the Royal Society of
Tarantola A (2005) Inverse Problem Theory and Methods for Model Parameter London, Series A 328: 291–308.
Estimation. Philadelphia, PA: Society for Industrial and Applied Mathematics. Woodhouse JH and Girnius TP (1982) Surface waves and free oscillations in a
Thomson ST (1950) Transmission of elastic waves through a stratified solid medium. regionalized Earth model. Geophysical Journal International 68: 653–673.
Journal of Applied Physics 21: 89–93. Woodward RL, Dziewonski AM, and Peltier WR (1994) Comparisons of seismic
Toks€oz MN and Anderson DL (1966) Phase velocities of long-period surface heterogeneity models and convective flow calculations. Geophysical Research
waves and structure of the upper mantle. Journal of Geophysical Research Letters 21: 325–328.
71: 1649–1658. Woodward RL and Masters G (1991) Lower-mantle structure from ScS-S differential
Torsvik TH, Steinberger B, Cocks LRM, and Burke K (2008) Longitude: Linking Earth’s travel times. Nature 352: 231–233.
ancient surface to its deep interior. Earth and Planetary Science Letters Zhu H, Bozdag E, Peter D, and Tromp J (2012) Structure of the European upper mantle
276: 273–282. revealed by adjoint tomography. Nature Geoscience 5: 493–498.
Tromp J, Tape C, and Liu Q (2005) Seismic tomography, adjoint methods, time Z€oppritz K (1907) Über Erdbebenwellen. II: Laufzeitkurven. Nachrichten von der
reversal and banana doughnut kernels. Geophysical Journal International K€oniglichen Gesellschaft der Wissenschaften zu G€ottingen, Mathematisch-
160: 195–216. physikalische Klasse: 529–549.
Udias A and Stauder W (2002) The Jesuit contribution to seismology. In: Lee WHK,
Kanamori H, Jennins PC, and Kisslinger C (eds.) International Handbook of
Earthquake and Engineering Seismology, pp. 19–27. San Diego, CA: Academic Press. Relevant Website
van der Hilst RD and deHoop MV (2005) Banana-doughnut kernels and mantle
tomography. Geophysical Journal International 163: 956–961. http://www.globalcmt.org – Global CMT Web Page.
 1.02 Theory and Observations - Instrumentation for Global and Regional
Seismology
JM Steim, Quanterra Inc., Shirley, MA, USA; Harvard University, Cambridge, MA, USA
ã 2015 Elsevier B.V. All rights reserved.

1.02.1 Introduction 29
1.02.2 Seismic Signals and Noise 31
1.02.2.1 Representations: Time Series and Spectra 31
1.02.2.1.1 Spectral analysis 31
1.02.2.1.2 Power spectral density: power and amplitude 32
1.02.2.1.3 Bandwidth 32
1.02.2.1.4 Minimum earth noise 33
1.02.2.1.5 Measuring signal amplitudes 35
1.02.2.1.6 Observed signal amplitudes and noise 35
1.02.2.1.7 Saturation (clip) level 36
1.02.2.1.8 Operating range 36
1.02.2.2 Representations: Instrumentation Systems 37
1.02.2.2.1 Seismograph systems 38
1.02.2.2.2 Noise, signals, and sensor response 39
1.02.2.3 Actual GSN Operating Range 41
1.02.2.4 Dynamic Range and Bandwidth 42
1.02.3 Seismometers and Systems 42
1.02.3.1 Mechanical Pendulum 42
1.02.3.2 Electronic Force Feedback 42
1.02.3.2.1 Limits on seismometer resolution and operating range 46
1.02.3.3 Systems 51
1.02.3.3.1 A/D conversion 51
1.02.3.3.2 Digital filters 56
1.02.3.3.3 Timing 59
1.02.3.3.4 Ancillary data 61
1.02.3.3.5 Archival methods: impact on quality and longevity of data set 63
1.02.3.4 Calibration 64
1.02.3.4.1 Sensor 64
1.02.3.4.2 Orientation and location 66
1.02.3.4.3 Network calibration 67
1.02.3.5 Quality Assessment and System Characterization 67
1.02.3.5.1 Noise 67
1.02.3.5.2 Signals 68
1.02.3.5.3 Instrument nonidealities 69
1.02.3.6 Instruments in Global and Regional Networks 70
1.02.3.7 Instrumentation Directions 73
1.02.3.7.1 Renewed interest in boreholes 73
1.02.3.7.2 Lower gain and greater bandwidth 74
1.02.3.7.3 Downhole digitization: integrated with a sensor 74
1.02.3.7.4 Complementary technologies 74
Acknowledgments 74
References 74

1.02.1 Introduction ground displacement with low mechanical magnification.

By 1903, Galiztin’s seismographs incorporated the important
Dewey and Byerly (1969) explained that by 1900, the first features of instrumentation used for global and regional obser-
global network of observatories equipped with seismographs vation for much of the twentieth century (Reid, 1912).
designed by the British seismologist John Milne regularly sent Galiztin’s design used an inertial pendulum having an electro-
records to Milne’s laboratory (Hoover, 1912) and contributed dynamic transducer and damping, with photographic record-
to the development of the first global travel-time tables. ing of ground velocity over a broad range of frequencies using a
Milne’s broadband seismograph photographically recorded beam of light deflected by a sensitive galvanometer. Such

Treatise on Geophysics, Second Edition http://dx.doi.org/10.1016/B978-0-444-53802-4.00023-3 29

30 Theory and Observations - Instrumentation for Global and Regional Seismology

Z LP
0S4 0S5 0S6 0S7 0S8 0S9 0S10

Michoacan, Mexico
Sep 19,1985
Mw 8.0
HRV
HAL
84/12/16 New Ireland mb 5.5 dep 65 km dist 125°

Simulated SRO LP

P diff PP PPP SS
10 min

0.60 0.74 0.88 1.02 1.16 1.30 1.44 1.58 1.72

Frequency (mHz)

Figure 1 Two examples showing low-frequency information obtained from very broadband (VBB) recordings. Resolution of signals such as these
previously required instrumentation specialized for low-frequency measurement. The left panel shows extraction of a long-period seismogram
from VBB data that appears to contain only noise. The upper trace is dominated by 3–4 s microseisms. The trace below has been digitally filtered to
reveal a small, distant event. The right panel shows observed (solid) and synthetic (dashed) free-oscillation amplitude spectra from the Michoacan,
Mexico, Mw 8.0 event on 19 September 1985 recorded on the prototype VBB system at HRV (blue) and very long-period IDA station HAL (red) in Halifax,
Nova Scotia. Signal-to-noise and synthetic agreements are comparable at the two sites located at similar distances and azimuths from the event.
Horizontal axis is frequency from 0.60 to 1.8 mHz. Adapted from Steim, J., 1986. The very broadband seismograph, PhD Thesis, Harvard University.

instrumentation evolved during the early twentieth century twentieth century was delayed by the detection mission. The
(Benioff, 1932; Press et al., 1958) and remained a standard Graefenberg (GRF) array project (Harjes and Seidl, 1978) dur-
for sensitive observation of long-period and short-period sig- ing the 1970s, however, illustrated the unrealized potential of
nals well into the 1980s in the World-Wide Standard Seismo- digital broadband data.
graph Network (WWSSN) (Oliver and Murphy, 1971). Although the available digitally recorded low-frequency
Early studies using hand-digitized photographic records data (Agnew et al., 1976; Peterson et al., 1976) were limited
(e.g., Bloch and Hales, 1968; Dziewonski et al., 1969; Iyer, in the early 1980s, the use of whole waveforms (Dziewonski
1964; Satô, 1958) showed the information that could be har- and Steim, 1982) began to elucidate features of global earth
vested through digital processing of seismograms. At the same structure, culminating in derivation of the first Earth models
time, the need to detect and quantify the yield of nuclear (Woodhouse and Dziewonski, 1984) using information con-
explosions began to shape much of the development of instru- tained in the entire waveform. This type of analysis was not
mentation for global observation. The resulting focus for sev- possible before the availability of substantial amounts of
eral decades during the second half of the twentieth century high-quality digital data made possible by the pioneering
was on the detection of the smallest signals. Because of the digital networks. Both the great potential and considerable
limitations of recording technology then available, instrument limitations of the digital instrumentation deployed during
designers were driven to shape the frequency response of seis- the 1970s became apparent (Dziewonski et al., 1981;
mographs to amplify, selectively, segments of the seismic spec- Peterson, 1982).
trum where ambient Earth noise is least, above and below the Desire for better, higher dynamic range broadband instru-
frequencies of marine microseisms (Longuet-Higgins, 1950), mentation as well as simultaneous rapid development in
roughly 0.02–0.04 Hz, and in a narrowband around 1 Hz. electronics finally enabled the realization of an instrumenta-
The potential advantages of a seismograph having a broad, tion technique that could embody in a practical way most of
flat frequency response with direct digital recording in a com- the requirements of global seismological research. Building
puter were recognized as early as 1961 in Caltech’s broadband upon the works of Plešinger and Horalek (1976) and
digital seismograph project (Miller, 1963): Wielandt and Streckeisen (1982) and the first broadband
array at GRF (Harjes and Seidl, 1978), the very broadband,
The broad-band system can be used to simulate any desired standard or VBB, system was developed at ETH, Zürich, and Harvard
seismograph and provides greater dynamic range and more flexibil- University in the early 1980s (Steim, 1986; Steim and
ity than most existing seismographs. (Smith, 1965) Wielandt, 1985; Wielandt and Steim, 1985, 1986). The objec-
tive was a single high-quality digital broadband data stream
Some efforts during the 1970s concentrated on idealized that included low frequencies previously measured only with
wideband accelerometers (e.g., Usher et al., 1977, 1978, 1979), specialized instrumentation that tended to saturate on large
including the venerable Teledyne–Geotech KS-36000-01 – an earthquakes (Agnew et al., 1976). Results from experimental
anachronistic combination of new electronic force-feedback VBB systems (Figure 1) showed that the objective was
(EFFB) and narrowband filtration – used in the Seismic achievable.
Research Observatory (SRO) stations (Peterson et al., 1976). Except for implementation, the technology now in global
Evolution of the broadband seismographs of the early and regional networks is largely unchanged from the system
 Theory and Observations - Instrumentation for Global and Regional Seismology 31

design developed at Harvard in the early 1980s (Peterson and exhibits a variety of characteristics, appearing as
and Hutt, 1989; Steim, 1986; Steim and Wielandt, 1985), impulses, steps, transient wavelets, quasioscillatory wave
using a single sensor (Wielandt and Steim, 1986) to acquire trains, or apparently random waveforms, and may originate
the seismic spectrum from tidal frequencies to about 10 Hz, as ground motion or artifacts of instrumentation. Some con-
a high-linearity analog-to-digital (A/D) converter, contem- tinuous background activity, for example, marine micro-
poraneous digital data reduction with compact data streams seisms, may be considered noise in many contexts, however,
at several sampling rates, and remote data access (Steim, may contain significant information (Shapiro and Campillo,
1987). The initial goals of instrumentation for new global 2004). Noise for an instrument designer is the background
and regional networks (Lay et al., 2002, after IRIS, 1985) level of actual or instrument-generated equivalent ground
had been achieved in the Harvard system, enabling rapid motion observed at the output of a seismograph. Instrument
establishment and growth of broadband networks during designers, sometimes to the consternation of seismologists,
the past 25 years: devote considerable effort to the study and minimization of
noise. Noise defines the smallest detectable signal in an
(1) On-scale broadband recordings of earthquakes as large as instrumentation system and, therefore, its usefulness for
Mw ¼ 9.5 (equivalent to the 1960 Chile earthquake) at 30 . many requirements. Studies, for example, of small, low-
On-scale low-gain recordings of all earthquakes at 1 sps (sam-
ples per second)
frequency signals, such as normal modes from moderate
(2) Noise below ambient Earth noise earthquakes or ‘hum’ (Kurrle and Widmer-Schnidrig, 2008;
(3) Bandwidth spanning all solid Earth free oscillations and Tanimoto et al., 1998), require instrumental resolution at
regional body waves (up to 15 Hz or higher as regional wave the level of minimum Earth noise. Understanding instrument
propagation considerations dictate)
behavior requires some representations appropriate to
(4) Linearity sufficient to record signals near ambient noise in the
presence of signals near clipping at well-separated frequencies describe simultaneously the amplitude and frequency of
(5) Response known to 1% across the bandwidth signals and noise to be measured.
(6) Sensor cross axis coupling less than about 1%
(7) Timing adequate to measure teleseismic body wave arrivals to
0.01 s
1.02.2.1 Representations: Time Series and Spectra
In 1987, the addition to the data system of strong motion 1.02.2.1.1 Spectral analysis
for the first time in a station of the global network, Caltech’s Spectral analysis (e.g., Bracewell, 1978; Oppenheim and
PAS (Pasadena, CA) station (Kanamori, 1990; Kanamori Schafer, 2009; Scherbaum, 2001) is a primary tool for charac-
et al., 1990), achieved the goal of measurement at a single terization of ambient seismic noise and instrument behavior
station of more than 220 dB dynamic range in ground and is possible because the Earth and seismographic instru-
acceleration. This chapter will focus on the technology of mentation systems can be considered within some range of
global and regional teleseismic observation as currently prac- input to be linear time-invariant, or LTI (Oppenheim and
ticed in the widely used facilities comprising the continuously Schafer, 2009). Although spectral analysis is an essential tool
recording research networks typical of contributors to for instrumentation, seismic signals occur and are measured in
the International Federation of Digital Seismic Networks the time domain. Some phenomena – for example, ‘clipping’ –
(Dziewonski, 1994). can be described adequately only in the time domain, since
This instrumentation comprises sensitive inertial broad- these violate the LTI assumption. This chapter therefore
band seismometers and associated recording equipment and develops a representation to characterize the range of ampli-
provides the majority of data for studies of earthquakes and tudes and frequencies measured by instrumentation from the
deep earth structure. Complementary instrumentations, perspective of behavior in the time domain.
including, ‘strong motion,’ arrays, strainmeters (Barbour and Let a real sequence {xm} represent M samples at
Agnew, 2012), or rotational (Forbriger, 2009) sensors, are not discrete times mDt of a continuous function of time x(t)
treated here. The ‘broadband revolution’ was approached in containing seismic noise or signals. Define a sequence
the 1980s mainly from the perspective of standards appropri- {Xk} of length M:
ate for global science-driven seismology. Over the past
30 years, VBB design elements have diffused throughout most 1 M1 X
aspects of modern earthquake instrumentation for regional Dfxm g ¼ Xk ¼ pffiffiffiffiffi xm ei2pkm=M [1]
M m¼0
and global-scale observation and will continue to influence
new instrumentation developed over the next decade. as the discrete Fourier transform (DFT) of the sequence {xm}.
The corresponding inverse DFT represents a sequence in time
as the sum of its DFT spectral coefficients Xk:

1.02.2 Seismic Signals and Noise 1 M X

1
D1 fXk g ¼ xm ¼ pffiffiffiffiffi Xk ei2pkm=M [2]
M k¼0
Seismologists are primarily interested in seismic signals. Seis-
mic signals may appear as transient wavelets lasting seconds In practice, the DFT is computed by one of many fast
to minutes or oscillatory wave trains lasting as long as days or Fourier transform (FFT) (Cooley and Tukey, 1965) algorithms
months and, in terms of acceleration, span eleven orders of for efficiency. We assume familiarity with operations with the
magnitude. Signals occur in the presence of background activ- DFT but will review some topics relevant to the description of
ity that also spans a wide range in frequencies and amplitudes seismic signal and noise spectra.
32 Theory and Observations - Instrumentation for Global and Regional Seismology

1.5

0.5
1 0.45
Normal
0.4
LHZ data

Probability density
0.5 0.35
Gaussian white
0.3
Uniform white
0 0.25
1/f
0.2
±3.3s
-0.5 0.15
0.1
-1 0.05
0
-1.5 -1 -0.5 0 0.5 1 1.5
-1.5
Ampltiude

Figure 2 Left: Three types of noise, all with identical variance ¼ rms2 ¼ s2 ¼ 1/12: white noise having a normal distribution (blue), white noise with
pffiffiffi and 1/f noise (green). The crest factor for 1/f and normally distributed white noise is 3–4, while the crest factor for
uniform distribution (red),
uniform white noise is 3: The crest factor 3.3 includes most peaks for normally distributed noise. The noise of quantization in digital instrumentation
(Bennett, 1948) is uniformly distributed and can be minimized by design. Right: Amplitude probability density distributions for the three noise sources at left.
The scattered dots show actual (scaled) amplitudes of a broadband velocity seismic noise sample with a bandwidth of 0.4 Hz, having a normal distribution.
Stationary seismic noise not of cultural origin typically is normally distributed; the time-domain amplitude can be estimated with a crest factor 3.3.

1.02.2.1.2 Power spectral density: power and amplitude distributed random noise (Peterson, 1993). The time-domain
Power spectral density (PSD) is defined as power per unit of amplitude of seismic noise can be estimated by the product of
bandwidth. This can be written for the discrete series Xk as the crest factor for normally distributed noise times the rms
 2  2 amplitude within a defined bandwidth. Figure 2 shows the
Xk  =M Xk  =M
PSDk ¼ ¼ ¼ Dt jXk j2 time-domain amplitudes of several noise sources and their
width of frequency bin k 1=MDt amplitude probability density distributions compared with
jXk j2 the (scaled) distribution for a sample of broadband velocity
¼ [3]
r data representing marine microseisms. The choice of crest
where r ¼ 1/Dt ¼ the sample rate in samples per second. PSD has factor is arbitrary. A peak will typically be distinguishable
units amplitude2/Hz or power/Hz. PSD contains no phase infor- above noise using a crest factor of 3.3.
mation – only the distribution of power over frequency – and
therefore cannot be inverted to infer time-domain amplitudes 1.02.2.1.3 Bandwidth
for arbitrary waveforms. For a real sequence {xm}, Xk is Signal and noise amplitudes are measured in the time domain
conjugate-symmetric about M/2. PSD may therefore be reported over a finite bandwidth. Bandwidth is important when consid-
as twice the value for 0 < k  M/2; the value for k ¼ 0 is the mean ering seismic signals that span more than five decades in fre-
of {xm}. Spectra are often represented as dB(P) ¼ 10log (P), quency. The integrated power of a constant PSD over the
where P represents power density PSDk, or power PSDkDf. decade from 10 to 100 Hz contains 10 000 times the power
Df ¼ 1/MDt ¼ r/M is the frequency bin width. in a decade from 1 to 10 mHz. Development of a representa-
The types of stationary noise typically encountered in seis- tion to describe, simultaneously, large transient signals and
mic systems comprise both white and 1/f noise with amplitudes small stationary noise therefore requires an appropriate con-
having uniform or normal (Gaussian) probability density sideration of bandwidth.
functions (PDFs). White noise is defined as noise in which A seismograph responds to both signals and noise in a band
|Xk| ffi constant in eqns [1] and [2]. 1/f noise typically includes of frequencies in which the signals are contained. An objective
a large class of physical behaviors in which |Xk|2 grows with in instrument design is to maximize the signal-to-noise (SNR)
inverse frequency raised to some power, often as f1. ratio in the band of frequencies of interest. Seismic signals –
The crest factor for a waveform is defined as the ratio of peak body and surface waves at frequencies higher than a few
amplitude
pffiffiffi to rms (root mean squared). For a sine wave, this is mHz – comprise wave groups of widely varying amplitude
2. For a sequence of normally distributed noise of reasonable (more than 11 orders of magnitude) recognized in frequency
length for analysis, the crest factor is typically 3–4. The crest bands of constant relative width (width proportional to fre-
factor 4 corresponds to a statistical confidence level of quency) spanning perhaps a small fraction to several octaves.
99.99% that the maximum amplitude in a sequence is con- Below several mHz, seismic noise represented as any typical
tained within four standard deviations from the mean of the function of ground motion (e.g., acceleration, velocity, or dis-
sequence. PSD is a correct measure of stationary noise power as placement) increases. Over the large bandwidth of more than
a function of frequency, but not transient signals, and can be four decades in frequency from 0.3 mHz to 10 Hz or above,
used to approximate the amplitude in the time domain of SNR can be maximized by a sensor with an output having
stationary noise with a known crest factor. Minimum ambient roughly equal noise (and large signal) power in bands of
seismic noise has been shown to behave as normally constant relative width across the entire frequency range of
 Theory and Observations - Instrumentation for Global and Regional Seismology 33

-10

Power per octavce

-20

-30 White
Constant PSD
-40
1/f
-50

-60
0.0001 0.001 0.01 0.1 1 10
Frequency (Hz)
Figure 3 Power spectra of synthetic random noise representing white noise (blue) and 1/f noise (green) sampled at 20 sps (samples per second)
having identical variance 1/12, summed over bands of constant relative width of one octave. For reference, an isoline of PSD ¼ variance/sample rate is
also shown (dashed red). Constant PSD produces increasing power with increasing frequency in bands of constant relative width, as appropriate for
signal measurement. 1/f noise produces constant power when summed over bands of constant relative width. 1/f noise in bands of constant relative
width cannot be reduced by filtering to reduce bandwidth. 1/f noise is typical of physical systems below some corner frequency.

interest. Under this condition, at the sensor output, the band, instrumental disturbances and earthquakes were omitted, the
for example, from 0.005 to 0.01 Hz, would contain the same intrinsic noise and the sensitivity of the instruments to environ-
noise and signal power as the band from 5 to 10 Hz. While this mental inputs are included in these models. These models,
cannot be exactly achieved, it is approximated with VBB seis- therefore, tend to be biased toward high estimates of actual
mograph systems. Earth noise. The models could be interpreted as a behavioral
Discrete PSD spectra represent power density in bands of model of the performance expected from a well-installed,
constant, finite, width, for example, the bin width equal to r/M properly running instrument of similar technology. Generally,
in previous examples. The power in a band of constant relative the progression of models has been toward lower minimum
width centered at frequency f is computed by integrating (sum- noise over a wider frequency range, although recent estimates of
ming) PSD over bands from frequency f1 to f2 that are propor- noise >1 Hz based on actual global and regional stations are
tional to f. For example, the band limits pfor
ffiffiffi a one-octave pffiffiffiband somewhat higher than some earlier, perhaps optimistic, models.
where f2 ¼ 2f1 may be defined as f1 ¼ f =p2ffiffiffiand f2p¼ ffiffiffi f 2. pThe
ffiffiffi The general rise in observed levels >1 Hz may reflect the
bandwidth, therefore, is f2  f1 ¼ f 2  1= 2Þ ¼ f = 2. encroachment of man-made noise sources on many of the
Constant PSD in bands of constant relative width produces observation sites that comprise existing networks. Figure 4 com-
power increasing proportional to f. Processes, including many pares several important vertical models. The GSNtypZ trace is
instrumentation and physical noise sources, characterized by the visually estimated low percentile of the most often observed
1/f noise produce constant power in bands of constant relative quiet levels in the Global Seismographic Network (GSN)
width (Figure 3). low-noise model (Berger et al., 2004). At low frequencies,
To appreciate the contrast in consideration of power in GSNtypZ is roughly equivalent to the Peterson (1993) new
bands of constant width or constant relative width, the partic- low-noise model (NLNM). From the perspective of instrument
ular relative band size, for example, 1/3 octave or one octave, is design, low vertical noise levels present a greater challenge than
not important. The important aspect is the dependence on f of higher horizontal noise levels. Horizontal noise levels between
the power over a large range of frequencies. The choice of 100 and 250 s period in the GSN and GEOSCOPE (Roult et al.,
bandwidth for signals to determine the frequency dependence 2010) networks typically are between 160 and 165 dB accel-
of their time-domain amplitudes is more significant. eration PSD, more than 20 dB above vertical levels.
Heightened cold-war detection work during the 1980s
drove efforts to find unusually quiet sites at frequencies
1.02.2.1.4 Minimum earth noise >1 Hz, remote from man-made noise sources (Bache et al.,
A number of parametric models of ambient seismic noise have 1986; Berger et al., 1988; Bungum et al., 1985; Herrin, 1982;
been constructed over the past 50 years (Agnew and Berger, Li et al., 1984; Rodgers et al., 1987; Gurrola et al., 1990). More
1978; Berger et al., 2004; Brune and Oliver, 1959; Fix, 1972; recent efforts have found quiet sites in Antarctica (Anderson
Melton, 1976; Murphy and Savino, 1975; Peterson, 1980, et al., 2006) and Central Asia (Hutt, personal communication).
1993; Peterson and Tilgner, 1985; Wielandt and Streckeisen, High-frequency levels at exceptionally quiet sites can be as low
1982). All models are based on a collection of observations as 30–40 dB below typical levels at many global and regional
and hence limited by instrumentation of the era and represent stations, although these low levels are observed infrequently
minimum noise levels in observed spectra. The minimum level during no-wind periods and at culturally quiet time times
observed at any particular frequency is not necessarily observed (nights and weekends). Resolution of the lowest high-frequency
simultaneously at another frequency. Although obvious levels constitutes a weak goal for general broadband
34 Theory and Observations - Instrumentation for Global and Regional Seismology

Acceleration power spectral density dB re 1 (m s-2)2 Hz-1

Earth noise models
–120

–130

–140

–150

–160

–170

–180

–190

–200
100 000 10 000 1000 100 10 1 0.1 0.01
Period (s)

GSN PLNM GSNtypZ Lajitas, 1984

Figure 4 The two most commonly used vertical minimum Earth noise models: the Peterson model (Peterson, 1993; green) and the global
seismographic network (GSN) noise model (Berger et al., 2004; blue) as acceleration PSD. The trace-labeled GSNtypZ (light blue) is the visually
estimated lower limit of the most dense areas (not the lowest percentile) of the probability distribution in the GSN model, representing typical
observations at quiet sites. The Lajitas, 1984 trace (Li et al., 1984; pink) shows the low-percentile, no-wind measurements in boreholes at a then-quiet
remote site in Lajitas, TX, United States. Current levels at that site are above the levels seen three decades ago. The Peterson (trace-labeled PLNM)
model represents a reasonable estimate of achievable minimum noise at the quietest sites at frequencies near 10 Hz (Hutt, personal communications) and is
not typically observed at current operating sites of global and regional networks. Above 1 Hz, seismic noise is mainly man-made and wind-generated
(Carder, 1963). The range from 0.05 Hz to 1 Hz is dominated by marine microseisms. Minimum levels below 0.01 Hz are observable only on lowest-noise
instruments in thermally stable installations.

BFO Pa STS 1/Z Sep 20–30, 1994

102
0.6 -180
0.5 Coherency IBPM
Transferluriction magnitude (nm s-1 hPa-1)

-181
0.4

0.3
-182
0.2
Vertical acceleration (db m2 s-3)

0.1 -183
101
0
10-4 10-3 10-2 -184

-185
BPM

-186

-187
100

AGW WG77 -188

70 80 90 100 200 300 400 500
10 -4
10 -3
10 -2 Period (s)
Frequency (Hz)

Figure 5 Left: Admittance of several models between local atmospheric pressure and vertical acceleration versus frequency for 10 days
data from September 1994. The models show minima near 3 mHz where effects of Newtonian attraction and surface loading approximately cancel.
Reprinted by permission from W. Zürn and E. Wielandt. On the minimum of vertical seismic noise near 3 mHz Geophysical Journal International (2007)
168 (2): 647–658, Figure 8. Right: Stack of quiet-time noise spectra from 32 of the lowest-noise GSN stations, showing ‘hum’ (Nawa et al., 1998)
in the frequency range of the Earth noise minimum near 3 mHz. With permission from Berger et al. (2004).

instrumentation for earthquake studies where low gain is The vertical noise minimum at frequencies less than several
required to acquire large signals on-scale. Instrumentation for mHz may not be fully described by existing models. Figure 5
detection purposes may require very low intrinsic sensor noise shows that the low levels frequently observed near 3 mHz com-
and may also exchange dynamic range for increased gain to prise clear signals (‘hum’) above the background minimum
overcome A/D noise. (Zürn and Wielandt, 2007) in atmospherically generated noise.
 Theory and Observations - Instrumentation for Global and Regional Seismology 35

content, therefore appropriately describes the largest signals

encountered by instrumentation.
-160 75%
dB (relative to 1 (m s-2)2 Hz-1
50% Like noise, signals span a broad range of frequencies, and
25% their time-domain amplitudes depend on their bandwidth and
5% frequency content. Definition of a bandwidth over which to
-180 consider, simultaneously, signals and noise enables descrip-
1%
NLNM
tion of the dependence on frequency of the largest and smallest
signals observed and measurable by a system. The particular
-200 choice of bandwidth is arbitrary; however, the bandwidth
GSN 50%-tile should be sufficiently wide to make a reasonable estimate of
GSN 25%-tile the peak time-domain amplitude of coherent signals within
GSN 5%-tile the band while not so wide as to obscure the dependence on
-220
GSN 1%-tile
B1 mean 5 days frequency. Wave groups tend to be dominated by a small range
of frequencies comprising, for example, often on the order of
10-4 10-3 10-2 an octave or two (Steim, 1986). Clinton and Heaton (2002)
(Hz) chose one-octave bands to estimate the frequency dependence
Figure 6 Observed gravity residuals (1st, 5th, 25th, 50th, and 75th of time-domain amplitudes using 1000 recordings. One
percentiles, gray lines) from the OSG-056 (B1) superconducting octave may be appropriate for most dispersed waveforms,
gravimeter (SG) at BFO (Schiltach, Germany), compared with the 1st, that is, typically for events beyond 100 km. One octave is not
5th, 25th, and 50th percentiles for the vertical GSN noise model (Berger adequate for nondispersed waveforms, for example, from
et al., 2004 – dashed blue lines) and USGS NLNM (red). The 75th nearby events, potentially containing high frequencies in
percentile of the mean PSD of the five most quiet SG days (small black phase, and hence tightly grouped in time, creating brief peaks
stars) is also plotted. The GSN noise model is derived from STS-1/VBB considerably higher than would be observed in an arbitrarily
sensors and is not corrected for barometric pressure. Adapted from
limited band. Figure 7 (Clinton and Heaton, 2002), for exam-
Rosat S and Hinderer J (2011). Noise levels of superconducting
ple, shows that the one-octave filtered acceleration of a large
gravimeters: updated comparison and time stability. Bulletin of the
Seismological Society of America 101(3), 1233–1241. close-by event has 20% of the actual peak amplitude of the
signal energy over a wide bandwidth.
The peak time-domain amplitude of signals is used here
Hum (Nawa et al., 1998) is at the current limit of instru- without arbitrary bandwidth restriction. To estimate the band
mental resolution (Widmer-Schnidrig, 2003) and can be of frequencies comprising a signal peak, a band-pass filter is
observed only in spectra stacked to reduce uncorrelated noise. used that is sufficiently wide to retain, say, at least 80% of the
Observation on horizontal sensors is unusual (Kurrle and peak amplitude. This may require two or more octaves,
Widmer-Schnidrig, 2008). Minimum ambient seismic noise depending on frequency content. Such an approach is roughly
may be below current models at frequencies <1 mHz. Spectra equivalent to estimation of frequency based on measuring the
(Figure 6) from superconducting gravimeters (SG) show levels reciprocal time of zero crossings of a peak oscillatory signal
(Rosat and Hinderer, 2011; Widmer-Schnidrig, 2003) below cycle. This approach to represent the frequency dependence of
the current noise models at frequencies <1 mHz when the SG peak signal amplitudes differs from some previous analyses
data are corrected for barometric pressure (see later this chap- (Bormann and Wielandt, 2013; Clinton and Heaton, 2002;
ter), suggesting future model revisions. Recent increased avail- Evans et al., 2010; Steim, 1986) in which signal amplitudes
ability to seismologists of SG data (Van Camp et al., 2008; Xu may be somewhat underestimated by arbitrary band limiting.
et al., 2008; GWR-Instruments, 2011) may encourage re- A seismometer never ‘sees’ a peak signal limited to an arbitrary
examination of the low-noise models for frequencies band such as 1/3 octave. The seismometer sees all frequencies
<1 mHz. With the deployment of digital networks over the comprising a wave group. We seek an approximation of the
last 25 years, long observation times and millions of observa- frequency content of signal peaks in order to localize, on a plot
tions are now possible. of amplitude versus frequency, a data point that may be com-
pared with the saturation level of instrumentation.

1.02.2.1.5 Measuring signal amplitudes 1.02.2.1.6 Observed signal amplitudes and noise
Signals generated by earthquakes span a range of 11 orders of Figure 8 shows noise models and selected signals, represented
magnitude in amplitude from the level of minimum Earth as measured peak amplitudes (for signals) and equivalent peak
noise, roughly 1010 m s2, to greater than 10 m s2 near a amplitudes for the GSN LNM (Berger et al., 2004) and new
significant earthquake. The goal for instrumentation is to high-noise model (Peterson, 1993). Some peak signal mea-
sense and record the largest possible range of amplitudes across surements are shown as one-octave-wide bars, although the
the broadest frequency range. The peak time-domain ampli- actual measurement bandwidth may be wider. Spectra extend-
tude of signals defines the ability to measure the largest signals ing from <0.1 Hz to >10 Hz are based on one-octave filtered
linearly. Every instrumentation system has a maximum ampli- seismograms of large events from Clinton and Heaton (2002).
tude limit, its saturation or clip level, above which the LTI In addition, the amplitude of solid Earth tides and the ampli-
assumption is invalid, and all frequencies are affected. Clip tude of free mode 0S0 during the first days following the Mw
level expressed as ground motion is dependent on the fre- 8.8 Chile earthquake of 2010 are shown. The range in mea-
quency of signals comprising a transient wave group. Peak sured peak accelerations exceeds 230 dB, which no wideband
time-domain signal amplitude, with an estimated frequency accelerometer has ever come close to resolving.
36 Theory and Observations - Instrumentation for Global and Regional Seismology

4
2 Raw acceleration
0
Imaxl = 5.20m s-2
–2
4
2 Bandpass 0.125–0.25 Hz
0
Acceleration - (m s-2)

–2 Imaxl = 0.836m s-2

4
2 Bandpass 0.25–0.5 Hz
0
–2 Imaxl = 1.23m s-2
4
2 Bandpass 0.5–1 Hz
0
–2 Imaxl = 0.927m s-2
4
2 Bandpass 1–2 Hz
0
–2 Imaxl = 0.597m s-2
10 20 30 40 50 60 70
Figure 7 Sample of vertical acceleration data band-passed for M7.5 earthquake at 10 km distance (from Clinton and Heaton, 2002). The actual peak
acceleration is approximately five times higher than viewed in one-octave bands.

1.02.2.1.7 Saturation (clip) level The most striking feature is the extent in amplitude and fre-
Within some defined band of frequencies, the saturation level quency of the operating range required of a broadband system
of a sensor in terms of ground motion in principle is the meeting the goals for global network operation. Dynamic range,
maximum output (voltage or digital representation) divided like operating range, is frequency-dependent and can be con-
by the transfer function relating the output to ground motion. sidered to be the frequency- and bandwidth-dependent ratio of
For an EFFB sensor, the power supply voltage and maximum maximum/minimum operating range limits. One may also
feedback force define the actual saturation level across some consider definition of a companion useful dynamic range for a
range of frequencies. There may be additional design- seismometer, as the interval between a seismometer’s clip level
dependent physical mechanical or electrical limits that reduce and the level of ambient seismic noise at the seismometer
the actual input-equivalent saturation level. For example, the output. The difference in definition can be quite large. The
saturation level of a Streckeisen STS-2 seismometer is specified KS36000-01 seismometer used in the SRO system (Herrin,
as 13 mm s1 velocity at 20 Hz, decreasing linearly to 1982), for example, was described as having 120 dB dynamic
5.3 mm s1 at 50 Hz. This is because the maximum input range, whereas the useful dynamic range was typically only
acceleration that can be balanced by the feedback system is about 50 dB at 6 s period, because high gain caused much of
1.67 m s2. Outside the nominal signal passband, there may the operating range to be occupied with microseismic noise.
be additional limitations. Using peak or rms descriptions of amplitudes, the required
useful dynamic range for broadband systems reaches a maxi-
mum of order 107, or 140 dB.
1.02.2.1.8 Operating range The frequency dependence of a system’s operating range is
Hutt et al. (2009b) defined operating range as readily visualized as a bounded area in a two-dimensional
amplitude–frequency plot. Steim (1986) introduced such a rep-
The portion of the ground motion spectrum between the clip level resentation as a readily interpretable method to visualize the
and the self noise of a seismometer, [where the latter is represented
as the root mean square (rms) of a sliding half-octave window.]
dependence of seismograph operating range on frequency and
Operating range depends on the bandwidth in which the noise is to describe the optimum approach to realize design goals estab-
measured. lished for new networks. Such a representation is approximate
because of the arbitrary choices necessary to represent simulta-
The present analysis considers peak amplitudes; however, neously the amplitude of stationary noise and transient signals as
this results in small differences compared with appropriate use a function of frequency. The approach, however, has become
of rms values. The overall dependence on frequency of operat- frequently used in the past three decades (Hutt et al., 2009a).
ing range as the interval between noise and clip level describes Evans et al. (2010) recently proposed this approach to represent
the important aspect of a sensor’s ability to measure signals. the operating range of instrumentation in a standardized way.
 Theory and Observations - Instrumentation for Global and Regional Seismology 37

Noise models and observed single peak amplitudes

20

0
Peak acceleration per octave dB re 1 m/s2 −20

−40

146dB
−60

230dB
−80

146dB
−100

130dB
−120 Tides

−140

−160

−180 0S0

−200

−220
100 000 10 000 1000 100 10 1 0.1 0.01
Period (s)

GSN LNM Lajitas HNM M8.8 Δ79 M2.5 9 km

M4 Δ1.3 M6.9 Δ6 M6.9 Δ37 M8 Δ36 Tides
M7.5 near M6.9 near M7.5 100 km M6.5 100 km M4.5 100 km
−180 dB −160 dB
Figure 8 Noise models and signals as equivalent peak acceleration in a one-octave bandwidth. The peak amplitude of transient signals and
stationary noise cannot be represented exactly in a common representation. The equivalent peak amplitude of stationary noise is approximated as
3.3 times the rms value in a one-octave band in order to compare with peak earthquake signal amplitudes. Peak levels (GSN stations HRV and PFO,
1985–2010) of six events, for example, M6.9D37 (an earthquake of undifferentiated magnitude 6.9 at epicentral distance 37 ; thick blue), are
represented as a one-octave wide bar at the estimated frequency of the wave group having the maximum peak amplitude. The peak amplitude of these
signals is not band-limited. Four spectra spanning the frequency range 0.1–10 Hz, for example, M6.5 100 km (light blue), are peak amplitudes
measured in one-octave wide bands (from Clinton and Heaton, 2002). The GSN (dark blue) and Lajitas (violet) noise models from Figure 4 are included
along with the Peterson (1993) new high-noise model (HNM; brown). Additionally, the amplitude of solid Earth tides and mode 0S0 (0.8 nm s2 peak)
following the 2010 Chile Mw 8.8 earthquake are shown. The ratio of maximum acceleration to minimum noise level exceeds 230 dB, while the
amplitude range from just below the LNM to the amplitudes of moderate events at global and regional distances exceeds 140 dB. Relatively small
nearby events produce accelerations exceeding that of the greatest earthquakes seen at distances >30 . Isolines of constant acceleration PSD
appear with increasing power per octave. Power of the HNM near 10 Hz exceeds minimum Earth noise at 3 mHz by 130 dB. The octave-filtered
(M4.5 100 km; tan) spectra are much lower than the peak amplitude (M4 D1.3; thick brown) over a larger bandwidth.

While the bandwidth used to estimate the frequency con- 1.02.2.2 Representations: Instrumentation Systems
tent of signal peaks may be variable, the choice of an arbitrary
The problem of optimization over a large operating range of
bandwidth for the estimation of the time-domain amplitude of
SNR of a seismograph system requires a consideration of seis-
stationary noise affects only the interval between signals and
mograph system frequency response. The transfer function of a
noise. Various noise contributions are compared exactly to
continuous-time LTI system, such as a seismometer, is typically
each other, when the conversion to time-domain amplitude
described using Laplace transforms (e.g., Bracewell, 1978). The
for all noise sources is consistent. Signal peaks also compare
Laplace transform F(s) of the function f(t) is given by
with instrument clip level.
Clinton and Heaton (2002) partially compensated for ð1
bandwidth limitation in peak signal estimation by reducing F ðsÞ ¼ Lff ðt Þg ¼ est f ðt Þdt [4]
0
the representation of instrumental clip level by a bandwidth-
dependent factor. The approach here is simply to plot actual where F(s) is defined as the Laplace transform of f(t), a func-
peak signal amplitudes at an estimated frequency, along with tion of time t  0, and s is complex. In the case of s ¼ io ¼ i2pf,
instrument peak clip level and an empirically justified repre- the Fourier transform results. f(t) and F(s) define a Laplace
sentation of noise as an equivalent peak amplitude in the time transform pair. When f(t) is convolved with the impulse
domain. response h(t) of a system assumed to be LTI, such as a
38 Theory and Observations - Instrumentation for Global and Regional Seismology

seismometer, the resultant output y(t) may be represented Equation [7] represents a class of digital filter called a finite
in the time domain as the convolution f(t)*h(t) or in the impulse response (FIR) filter used universally in the digitizing
frequency domain as Y(s), the product of F(s)∙H(s), where and processing systems of broadband seismographs. FIR filters
H(s) is the transfer function describing the LTI system with have the desirable property of unconditional stability, since
impulse response h(t). The transfer function H(s) of an LTI their evaluation depends solely on current and prior samples
system is defined as the ratio of the Laplace transforms of the of an input sequence, and not on the output of the filter. Low-
system output Y(s) divided by the system input F(s). pass FIR filters are typically used as digital anti-alias filters
An analogous Z-transform is defined for discrete-time causal because of the ability to construct very sharp filters of high
systems (which all digital seismographs can be considered): rejection of unwanted frequencies while preserving a maxi-
X
1 mum passband.
X ðzÞ ¼ Zfxm g ¼ xm zm [5] The general topic of discrete-time signal processing is beyond
m¼0 the scope of this section. The reader is referred to the general
where {xm} is a real sequence and z is complex. A particularly literature on discrete signal processing (e.g., Welch, 1967;
useful feature of the Z-transform is that a unit delay in the time Bracewell, 1978; Cooley et al., 1970; Oppenheim and Schafer,
domain is represented by multiplication by z1 in the z-domain, 2009; Rabiner et al., 1972; Rader and Gold, 1967) and specific
that is, seismological applications (e.g., Haney et al., 2012; Harjes and
Seidl, 1978; IRIS, 2012; Scherbaum, 2001; Seidl, 1980). This
Zfxmn g ¼ zn X ðzÞ [6] chapter will use Laplace and z-transforms, however, to describe
A digital filter may be described by important aspects of seismometer and system behavior.
X
N
ym ¼ dn xmn [7] 1.02.2.2.1 Seismograph systems
n¼0 The seismograph systems providing global and regional data
comprise one or more sensors with a voltage output that is
where a sequence {dn} of N real constant coefficients is con-
digitized in an A/D converter. Additional digital processing
volved with the input sequence {xm} to produce an output
follows to record and transmit digitized data. Other
sequence {ym}. The operation may be expressed as multiplica-
approaches to convert ground motion to numbers are in prin-
tion of rational polynomials in the z-domain using the transfer
ciple possible, but until now, and likely during at least the next
function D(z) of the filter:
decade, sensors having a voltage output followed by an A/D
X
N
will continue to be dominant.
DðzÞ ¼ dn zn [8]
The major elements of a system are represented in Figure 9:
n¼0
a sensor with transfer function H(s) ¼ E(s)/V(s) relating the
and transforms of, say, input ground velocity v(t) to output voltage
Y ðzÞ ¼ Zfym g ¼ DðzÞ XðzÞ [9] e(t); a digitizer converting samples of voltage e(m’Dt) at

Sampling rate clock

Positive power supply

1/Δτ
Counts, for
Volts, for counts, for example, ±225
Velocity,for example, example, 20ksps
±8mm s-2 example,±20 20sps
Digital anti-
v(t) e (t) X m’ ym
Seismometer Digitizer alias filter
H(s)
X m’ = e (m’Δτ) and
V(s) E (s)
decimator 1/KΔτ
Limits
Limits

D(z), K

Negative power supply

Analog Digital
continuous-time discrete-time

Figure 9 A simplified digital seismograph. The elements include a sensor with Laplace transfer function H(s) to produce a voltage, a digitizer to
sample the voltage, and low-pass digital filters to produce a digital sequence {ym} representing the physical ground motion input, for example,
ground velocity v(t). The transfer function H(s) maps the strong frequency dependence of ground motion amplitude, in terms of ground acceleration,
onto a weaker frequency dependence of voltage measured by the digitizer. The transfer function H(s) is designed to maximize SNR (signal-to-noise)
across the frequency range of operation, not to enhance specific signals. The system saturation (clip) level is limited partially by power supply voltages,
although additional limits may exist.
 Theory and Observations - Instrumentation for Global and Regional Seismology 39

discrete times m’Dt to an output digital sequence xm0 at a rate corresponds to the maximum voltage measurable by the A/D.
1/Dt much higher than the highest signal frequency of interest; The frequency response of the sensor maps the sensor’s oper-
and a digital anti-alias filter with transfer function D(z) and ating range as a function of frequency onto the operating range
decimator of factor K. The final stage decimates the filtered of the A/D and ideally preserves maximum SNR. While a
sequence variety of sensor frequency responses are possible, particularly
in an EFFB seismometer, only a small class of frequency
0
X
N
y m0 ¼ dn xm0 n [10] responses and gains similar to VBB are useful in optimally
n¼0 delivering a voltage representing seismic signals from
<1 mHz to >10 Hz so matched to a digitizer.
at a high rate 1/Dt to the reduced-bandwidth output sequence
Frequency response and gain are not strictly related to a
{ym} delivered for analysis at a lower sample rate 1/Dt ¼ 1/KDt:
seismometer’s SNR. For example, although a strong-motion
0
ym ¼ y Km [11] accelerometer has a frequency response extending to zero fre-
quency, its SNR is many orders of magnitude less than a
This description is greatly simplified to focus on the effect of modern broadband seismometer at frequencies of, say,
sensor frequency response on system operating range and 1 mHz, despite the fact that such a seismometer has a high-
omits important features: noise in the sensor and digitizer, pass response to acceleration. It is also possible to design an
analog anti-alias filtration in the digitizer, time labeling, and electronic seismometer with very high sensitivity resulting
additional stages of digital filtration and decimation. Although from high gain and a flat frequency response extending from
anti-alias filtration in the digitizer is an essential element, its 10 s to 100 Hz, but this could result in having an SNR and
impact on the overall system behavior seen at seismic frequen- dynamic range worse than a simple geophone.
cies is generally only a small time delay that can be compen-
sated in a time label, since oversampling A/D converters operate
with a high rate 1/Dt. 1.02.2.2.2.2 Synthesis of a VBB system response
Figure 10 shows spectra of the seismic noise models on land
1.02.2.2.2 Noise, signals, and sensor response shown in Figure 8 represented as ground velocity. The ampli-
The central issue for broadband system design is the choice of tude in bands of constant relative width from 3 mHz to
response H(s) that both optimizes SNR in the output voltage of >10 Hz is roughly constant and increases approximately
the sensor E(s) and allows the greatest operating range to be 40 dB/decade at frequencies <3 mHz. Let V(s) represent
‘mapped’ onto the available capability of the digitizer, defined ground velocity, E(s) the voltage output of a seismometer,
by its intrinsic noise and maximum signal limits. The limits and G a generator constant, having units volts/velocity, for
may be mechanical and electrical in the seismometer and are example, V m1 s. A second-order high-pass transfer function
electrical in the A/D. Power supply voltages partially define H(s) relating E(s) to V(s) of the form
these limits.
EðsÞ s2
¼ HðsÞ ¼ G 2 [12]
V ðsÞ s þ 2xo0 s þ o0 2
1.02.2.2.2.1 Significance of frequency response
The question of an ‘optimum’ frequency response for a broad- where x is the fraction of critical damping, is proportional to
band seismic sensor was not frequently considered before the velocity above a corner at angular frequency o0. The choice of
1980s (e.g., Berckhemer, 1970; Lake, 1964). Because of the o0 ffi 2p/360 tends to equalize noise and, coincidentally, signal
limitations of photographic recording and then-current A/D power in bands of constant relative width from 0.01 mHz to
and digital data storage technology, considerable efforts in 10 Hz. H(s) can be considered the frequency response of a VBB
networks such as the WWSSN, high-gain long-period seismometer. An aside: on the sea floor surface, the seismic noise
(HGLP), and SRO programs were devoted to shaping the fre- distribution is different, resulting in a better match over
quency response of instruments to enhance signals suitable for 0.02–10 Hz to the higher overall noise level there with a response
detection of small nuclear explosions and reject frequencies proportional to acceleration (Webb, 2002; Webb et al., 2001).
considered to be dominated by noise (Murphy and Savino, Sea floor low-frequency noise levels are high relative to sea floor
1975; Oliver and Murphy, 1971; Peterson et al., 1976; Sutton high-frequency levels, although all sea floor levels are above
and Oliver, 1959). quiet levels on land (Webb and Crawford, 2010).
The objectives for sensor frequency response in a broad- Figure 11 shows the match between seismic noise levels
band system instead are to include many frequencies, to max- at the output of a VBB seismometer described by eqn [12] and
imize SNR, and to minimize gain. It is not surprising that such actual representative noise of a high-resolution A/D converter
a system was not the focus of attention and in fact was consid- (Quanterra Q330HR) used in global and regional networks.
ered cumbersome (Smith, 1964) until digital computing and The constant G ffi 2200 V m1 s has been chosen to adequately
storage became widely available during the 1980s. resolve minimum noise without loss of large signals. In such
Digital VBB instrumentation shapes sensor frequency a system, both seismic noise and A/D noise are described by
response H(s) to include most signals seen at teleseismic dis- roughly flat noise amplitude in frequency bands of constant
tances at the sensor output in a manner approximately opti- relative width from <0.1 mHz to >10 Hz. The noise of
mally matched with the A/D. Matching means that the noise 24-bit quantization (Bennett, 1948) at 100 sps with a range of
voltage at the sensor’s output is only slightly larger than the 20 V is also shown for comparison. The typical vertical noise
noise voltage of the A/D as a function of frequency and that the levels observed at quiet GSN sites are well above actual
maximum voltage produced by a large signal roughly A/D noise. Only high-frequency noise during the quietest
 Noise models, peak velocity per octave
20
0

Peak velocity per octave dB re 1 m s-1

-20
-40
-60
-80
-100
-120
-140
-160
-180
-200
-220
100 000 10 000 1000 100 10 1 0.1 0.01
Period (s)

GSN LNM Lajitas HNM -180 dB -160 dB

Figure 10 This figure shows the noise models of Figure 8 represented as peak ground velocity (crest factor 3.3) in one-octave bands. The noise
models are approximately ‘rotated’ clockwise compared to acceleration per octave, reducing the ratio of maximum to minimum amplitude over the
frequency range from 3 mHz to >10 Hz. Noise levels at frequencies below 3 mHz increase 40 dB per decade. The amplitude in bands of constant
relative width from 0.01 mHz to 10 Hz or higher can be roughly equalized with a second-order high pass at 3 mHz. 180 dB (dashed light blue)
and 160 dB (dashed violet) are acceleration PSD isolines.

Noise models in volts at 360s VBB sensor (2200 V m-1 s) output

40

20

0
Peak volts per octave dB re 1 V

-20

-40

-60

-80

-100

-120

-140
10 000 1000 100 10 1 0.1 0.01
Period (s)
GSN LNM HNM Clip(20 V)
Q330HR Noise ideak 24 bits Lajitas
GSN typ quiet -180 dB

Figure 11 The noise models of Figure 10 represented as peak voltage (crest factor 3.3) at the output of a very broadband (VBB) seismometer having
a response flat to ground velocity with a two-pole high-pass corner at 360 s period and gain 2200 V m1 s. The noise models are compared with
the actual noise of A/D systems (thick yellow dashed line) used in global network stations. Except for the peak containing marine microseisms,
the amplitude of the noise models as a function of frequency is nearly equalized with this frequency response. The amplitude variation of minimum
Earth noise with frequency is 50 dB over the entire frequency range. The margin between low Earth noise and A/D noise is positive and roughly
uniform. The interval from noise to 20 V clip level (red dotted line) is similarly nearly constant. 24-bit, 100 sps quantization noise is also shown (dashed
green). The isoline 180 dB acceleration PSD is ‘bent’ and ‘tilted’ by this approach showing VBB’s frequency-dependent map of constant
acceleration PSD to voltage at the A/D converter (dashed blue). The frequency response described is VBB, originally implemented in the STS-1/VBB
seismometer for frequencies <10 Hz.
 Theory and Observations - Instrumentation for Global and Regional Seismology 41

periods at exceptional sites, such as Lajitas, TX, during the 1980s illustrated by the M6.9, 6 , are at the maximum signal-
(current levels; there are higher), is not well resolved by such a handling limit. Small, nearby, felt events, illustrated by M4,
system. 1.3 , are well within the capability. Large events nearby and at
A system as represented in Figure 11 with flat response to regional distances are well above the maximum signal-
velocity extending from 360 s to arbitrarily high frequency handling capability and require a lower-gain sensor to measure
does not quite exist. Existing Streckeisen STS-1/VBB and on scale. Although the interval between the sensor’s intrinsic
Geotech KS54000I sensors used in global networks, however, noise and saturation level is 146 dB, this range is never fully
provide an approximation, with an upper second-order corner exercised, since Earth noise defines the minimum output level.
frequency at 5–10 Hz. A number of existing broadband sensors Figure 11 or 12, based on peak levels as described earlier,
with a lower corner period of 120 s or more have nearly flat shows that a maximum observed h idynamic range of
clip level
frequency responses extending well above 10–40 Hz or higher 135–140 dB defined as 20log noise level using peak ampli-
but do not typically resolve minimum Earth noise at frequen- tudes  (the pffiffiffi more often defined rms ratio is
cies <4–5 mHz. The development of sensors better approxi- 20log 3:3= 2 ¼ 7 dB higher) is achieved in an octave
mating the wide response depicted in Figure 11 with lowest band of frequencies near 30 s period, where surface wave sig-
noise is an area of current research. nals may be large and seismic noise low.
Such an operating range representation helps illustrate the
difference between a sensitive instrument and a low-gain or
1.02.2.3 Actual GSN Operating Range
strong-motion instrument. Strong-motion instruments
Figure 12 shows the signals and noise of Figure 8 and includes designed for large signals as seen in the near field of an earth-
the actual saturation and estimated noise level of the STS-1/ quake are defined from the ‘top down’ by their maximum peak
VBB seismometer to represent the frequency dependence of the signal levels, usually 2 g; their noise level is subordinate,
operating range of global network systems. A similar figure defined by their electronics and mechanical system at a level
would result with other types of velocity broadband sensors, somewhere below their defined maximum signal level. Strong-
such as the Geotech KS54000I, the Streckeisen STS-2, and the motion sensors in current use achieve a typical useful dynamic
Nanometrics T240, with differences in the clip level and low- range approaching 150 dB, diminishing at frequencies <1 Hz.
frequency noise level (discussed later). With the STS-1/VBB, Instruments intended for the most sensitive low-frequency
such a system is capable of high-SNR recording of free oscilla- measurements, however, are designed from the ‘bottom up’
tions and on-scale recordings of most events to at least M8 at for a target noise level and achieve some maximum signal level
30 distance. Moderately large events at regional distances, dependent on electronics.

Very broad band operating range

20
Peak acceleration per octave dB re 1 m s-2

0
-20
-40
-60
-80 T = 5 °C
146 dB

-100
-120 Tides
-140
-160
-180 0S0

-200
-220
100 000 10 000 1000 100 10 1 0.1 0.01
Period (s)
GSN LNM Lajitas HNM M8.8 Δ79
M2.5 9 km M4 Δ1.3 M6.9 Δ6 M6.9 Δ37
M8 Δ36 Tides T=5 °C Clip STS1
M7.5 near M6.5 near M7.5 100 km M6.5 100 km
M4.5 100 km -180dB GSN typ quiet STS1 noise

Figure 12 The frequency-dependent operating range of a VBB system. The seismic noise models and signals of Figure 8 are combined with the
instrumentation noise (dashed green) and clip level (dashed blue) of a VBB system using the frequency response illustrated in Figure 11. Many large
signals at regional and teleseismic distances are within the operating range. The largest signals at regional and near distances are above the
maximum capability. The actual maximum useful dynamic range of a VBB system exceeds 140 dB at some frequencies. Also shown is the approximate
acceleration equivalent of a 5  C temperature change on the effective mass position. Thermal stability in the installation is required.
42 Theory and Observations - Instrumentation for Global and Regional Seismology

1.02.2.4 Dynamic Range and Bandwidth acceleration. The transfer function of the mechanical system
M(s) is defined as
Dynamic range is a number often used to describe the ratio of
the largest to smallest time-domain signal handled by an XðsÞ 1
MðsÞ ¼ ¼ [15]
instrumentation system. Dynamic range, however, is meaning- AðsÞ s2 þ s mb þ mk
less without specification of the frequency band over which the
measurement is made. In the case of a broadband seismic Equation [15] can be written as
system spanning nearly seven decades in frequency, the partic- X ð sÞ 1
ular choice of band limits greatly affects the resulting ratio. MðsÞ ¼ ¼ [16]
AðsÞ s2 þ 2xo0 s þ o0 2
A few examples illustrate this point.
The useful dynamic range observed over the full bandwidth where x represents the fraction
pffiffiffiffiffiffiffiffiof
ffi critical damping of the oscil-
of the GSN VBB sensors is not particularly impressive when lator and o0 ¼ 2p=T0 ¼ k=m is the (angular) resonant fre-
deployed at a noisy site. The interval between short-period quency of the undamped mechanical oscillator having period
ambient noise and clipping may be only 80 dB or less: this is T0. Above resonance o0, the displacement X(s) of the mass
the ratio of clip level to the voltage produced by the sensor that relative to the frame represents ground displacement D(s):
one would measure with a voltmeter at such a site. The maxi- AðsÞ
mum dynamic range defined as an rms ratio, say in one octave X ðsÞ ¼ DðsÞ [17]
s2
near 30 s, on the other hand, exceeds 140 dB. Dynamic range is Below resonance, X(s) is proportional to ground
usually specified in instrumentation as the rms ratio of a max- acceleration:
imum amplitude sine wave to the noise within a specified band
of frequencies. For the system shown in Figure 11 where actual AðsÞ
XðsÞ [18]
A/D noise is 75 nV rms/octave and the A/D maximum rms o0 2
level for a sine wave is 14.14 V, actual A/D rms dynamic range The addition of an electrodynamic transducer, that is, a coil
is roughly a constant 164 dB in one-octave bands below the 1/f of wire moving through a magnetic field, with generator con-
corner and 147 dB in a one-octave band at 50 Hz. stant G to measure X(s) creates the familiar seismometer pro-
ducing a voltage E(s) proportional to mass velocity:

sAðsÞ
1.02.3 Seismometers and Systems EðsÞ ¼ G sXðsÞ ¼ G
s2 þ 2xo0 s þ o0 2
s2 V ð sÞ
¼G 2 [19]
Measurement conventions, models of Earth noise and signals, s þ 2xo0 s þ o0 2
and a system matching frequency response to noise and signals
to achieve design goals of global and regional networks have Well above resonance o0, E(s) is proportional to ground
been introduced. This section will discuss the design and oper- velocity V(s). This is the form of response of a conventional
ation of sensors and systems to achieve the result depicted in electrodynamic geophone:
Figure 12. The discussion partially follows Steim (1986). EðsÞ G V ðsÞ [20]
The frequency response of eqn [19] is simulated by a force-
feedback broadband velocity sensor. Compare eqn [12]. Tra-
1.02.3.1 Mechanical Pendulum
ditional seismographs such as the WWSSN with photographic
An inertial seismometer sensitive to vertical motion consists of recording are described by eqn [19] with an additional second-
a mass suspended elastically against gravity from a frame that order low-pass function describing the galvanometer
rests on the Earth’s surface. The inertial mass and viscoelastic (Hagiwara, 1958; Wielandt and Mitronovas, 1976).
suspension create a mechanical oscillator that for small
motions can be described by a second-order linear differential
1.02.3.2 Electronic Force Feedback
equation:
EFFB has been used in seismometry since at least the mid-
b k
x€ðt Þ þ x_ ðt Þ þ xðt Þ ¼ aðtÞ [13] 1950s (Benioff, 1955). EFFB was originally used primarily to
m m generate a restoring force to maintain the inertial pendulum
where m is the inertial mass, x(t) is the displacement of the within its mechanical operating limits, typically under varia-
mass as a function of time in a fixed coordinate system, b de- tion of position caused by temperature change, tilt change, and
scribes frictional damping of the mass motion and has units mechanical creep (Sutton and Latham, 1964). At least one
force/velocity, k is the spring constant of the elastic suspension early effort (DeBremaecker et al., 1962) included direct digiti-
and has units force/displacement, and a(t) is the acceleration zation having a claimed dynamic range of 100 dB, and a system
of the seismometer frame as a function of time. If the seismom- with EFFB featuring laser displacement sensing of the inertial
eter frame is coupled firmly to the ground, a(t) also represents pendulum and a dynamic range up to 140 dB (Hemelrick,
ground acceleration. Laplace transformation of eqn [13] gives 1964) was proposed, but apparently not constructed.
EFFB sensors first reached wide deployment in the Interna-
b k
s2 XðsÞ þ s XðsÞ þ XðsÞ ¼ AðsÞ [14] tional Deployment of Accelerometers (IDA) network (Agnew
m m et al., 1976) with the Block and Moore (1966) conversion of
where s is the complex frequency variable and X(s) and A(s) gravimeters (LaCoste, 1934) by the addition of capacitive dis-
represent Laplace-transformed mass displacement and ground placement sensing. The weak electrostatic feedback limited the
 Theory and Observations - Instrumentation for Global and Regional Seismology 43

saturation level. The low-frequency noise, however, was quite The combined frequency response of the mechanical seismom-
low and these systems provided valuable data. eter and electrical transducer is given byO(s) ¼ M(s)R(s), the
The present era of high-dynamic-range EFFB sensors began open-loop response, that is, without feedback.
with the deployment of relatively small sensors having strong In the model of Figure 13, a feedback force is generated by
feedback to shape response using electromagnetic transducers. an electrical frequency response – shaping network that drives
The SRO (Melton, 1976; Peterson et al., 1976) program using a force transducer acting on the inertial mass. The network is
the Geotech KS36000 and the GRF array (Harjes and Seidl, described as an admittance, or inverse impedance, network
1978) using the newly developed Streckeisen STS-1 (Wielandt that converts a voltage to a current I(s) for driving a transducer
and Streckeisen, 1982) provided the first significant volumes of that generates a feedback force proportional to applied current.
high-quality digital data from broadband EFFB sensors. The The force operates on the sensor’s suspension to produce an
KS36000 operated as an accelerometer, which unnecessarily input to the mechanical sensor that is equivalent to an accel-
restricted its potential operating range (Peterson, 1982; Steim, eration acting on the mass of the sensor.
1986). The STS-1’s application of EFFB for the first time enabled The admittance network contains the electrical components
a dynamic range of 140 dB in a broadband sensor. The STS-1/ or subcircuits that determine the overall response of the seis-
VBB version with a 360 s corner period enabled by 1986 mometer in the closed loop, that is, with feedback applied. An
(Wielandt and Steim, 1986) the first realization of digital VBB admittance network to convert a voltage to a current for the
seismometry in global and regional networks. At the same time, force transducer may consist simply of resistors and capacitors,
miniaturized versions of EFFB sensors (Usher et al., 1977, 1978, or it may include active circuits. An electromagnetic force
1979) were developed throughout the 1980s and 1990s. The use transducer is typically a coil of wire in a magnetic field, equiv-
of EFFB allows a relatively compact sensor that is comparatively alent to an electromagnetic velocity transducer used to sense
easy to isolate against environmental inputs to achieve the motion in, for example, a geophone. The coil produces a
response of a much larger equivalent open-loop sensor. The force proportional to the current flowing through it. The
stable 360 s corner period of the STS-1/VBB, in fact, is not admittance network has response N(s) having units current/
practically achievable in an open-loop sensor. For a review of voltage ¼ 1 O1. The force transducer and inertial mass may be
EFFB seismometers during this period of rapid development, see considered together to have a response G(s) that can be simply
Steim (1986). a constant, say, g (not to be confused with the acceleration of
Figure 13 shows the mechanical seismometer of eqn [16], gravity), having units acceleration/current. In this formulation,
with response M(s) converting ground acceleration A(s) into a current applied to the force transducer produces an equiva-
inertial mass displacement X(s). EFFB requires an electrical trans- lent acceleration of the mass of the mechanical seismometer.
ducer that converts the inertial mass displacement X(s) into an The combined response of the feedback from voltage to
electrical voltage E(s). The electrical transducer may, in general, ground acceleration is then F(s) ¼ N(s)G(s).
have any frequency response R(s), although the primary types of The voltage E(s) may be written as the sum of ground and
transducer would be an electromagnetic velocity transducer, as (negative) feedback accelerations acted upon by the open-loop
in a geophone, or a displacement transducer, in which R(s) transfer function O(s):
would equal a constant, say, r having units volts/displacement.
The choice of a displacement transducer is most appropriate for
EðsÞ ¼ ðAðsÞ  EðsÞFðsÞÞOðsÞ ¼ AðsÞOðsÞ  EðsÞFðsÞOðsÞ [21]
resolution of the inertial mass motion at very low frequencies.

Ground
acceleration Open-loop electrical response to ground acceleration O(s) = M(s)R(s)

A (s) X(s) E(s)

Inertial sensor Displacement
S M(s) transducer
+ Displacement R(s)

- Volts, for
example,Ê 20

Force transducer I(s)

and effective mass Admittance network
G(s) N(s)
Acceleration F(s)E(s) Current

Transfer function of voltage to effective feedback acceleration F(s) = G(s)N(s)

Figure 13 A generalized Laplace-transformed description of an electronic force-feedback (EFFB) seismometer, with acceleration input A(s), and
voltage output E(s). Feedback reduces the displacement of the inertial mass X(s) by a frequency-dependent factor O(s)F(s) (the loop gain)
compared to the open-loop case. Smaller physical displacement improves linearity in the closed loop. Useful dynamic range can be very high in
terms of acceleration over a wide frequency range if the feedback path F(s) is controlled by passive electrical components. Active components in
the feedback path introduce noise that appear in the output signal E(s). At frequencies where O(s)F(s) 1, the frequency response E(s)/A(s) of the
closed loop is 1/F(s), controlled mainly by the feedback network.
44 Theory and Observations - Instrumentation for Global and Regional Seismology

 
which may be rewritten to give the output voltage E(s) as a r M0 ðsÞ
OðsÞ ¼ RðsÞMðsÞ ¼ MðsÞ
function of ground acceleration or the closed-loop transfer   sTD þ 1 Meq ðsÞ
function: r
¼ M0 ðsÞ [24]
EðsÞ OðsÞ sTD þ 1
¼ [22]
AðsÞ 1 þ OðsÞFðsÞ
At frequencies where O(s)F(s) 1 (true for frequencies up  
g sC þ 1=Rp þ 1=stR
to about 5 Hz in the STS-1/VBB), the closed-loop response is F ðsÞ ¼ GðsÞN ðsÞ ¼ [25]
approximately sCRC þ 1

EðsÞ 1 where RC is the resistance of the feedback transducer coil and

[23]
AðsÞ F ð sÞ the transfer function R(s) of the displacement transducer
In this case, the closed-loop response of the seismometer is includes a single-pole time delay TD and the response of an
controlled entirely by the feedback network and force inverse filter (Wielandt and Steim, 1986). The inverse filter has
transducer and is independent of the mechanical response response M’(s)/Meq(s). If Meq(s) M(s), the inverse filter con-
of the seismometer without feedback. The quantity O(s)F(s) verts the second-order mechanical response M(s) into a
is called the loop gain and is a measure of the effectiveness response of the same form M’(s) describing an effective
of feedback in determining the seismometer response. mechanical sensor with a much longer free period T0’ ¼ 2p/
Loop gain is a fundamental descriptor of any feedback o0’ than the actual sensor. The purpose is to boost the loop
system. gain at frequencies where the condition O(s)F(s) 1 is not
In a high-dynamic-range VBB sensor, such as the STS-1/ otherwise met. The identical match Meq(s) ¼ M(s) is not
VBB, the admittance network has the form illustrated in required, and the exact choice of parameters of the inverse
Figure 14. The three feedback paths describe a system with filter has little effect on the overall closed-loop response.
proportional, integral, and derivative (PID) feedback. This Higher loop gain enforces the approximation in eqn [23] and
architecture is common in many feedback control systems desensitizes the closed-loop response to variations in the free
(Ang et al., 2007). At frequencies where the feedback force period and damping of the mechanical suspension.
is strong (loop gain 1), the feedback current delivered to Substitution of eqns [24] and [25] into eqn [22], and some
the force transducer by the admittance network balances approximations, gives the closed-loop response to acceleration
ground acceleration. Each of the three parallel paths has the explicitly including the effects of finite RC0 TD0 and o0’:
highest admittance over different frequency ranges. The volt-
age E(s) is proportional to the integral of ground acceleration, EðsÞ 1
¼
or ground velocity in the broad range where the path across C AðsÞ gC
sðsCRCþ1Þ
is dominant, and to the derivative of ground acceleration or h i h 02
i
the second derivative of ground velocity where the path s5 RCrgTDþs4 Kþs3 rgC
1
þs2 Ko0 0 2þrgC
2x
o0 0þ1 þs CR1 pþorgC
0
þtRC
1

through the integrator is dominant.

[26]
The open-loop O(s) and feedback F(s) transfer functions
can be generalized to better model a typical sensor: where K ¼ (RC þ TD/C)/rg

Rp = 7.96 MW

Where loop gain >>1, current l is proportional to ground

acceleration. Over a broad range where loop gain >> 1, voltage E
across C is therefore proportional to the time integral of l, or
I (s) ground velocity. E(S) ≈ I(s) /sC E(S)
Rc < 1 KΩ
Capacitor C = 5.08 mF
Force transducer
coil resistance
Feedback current I varies from 0.6 μV/656 K ≈ 1 pA at low
frequencies to 20 V/1850 Ω @ 30 Hz ≈11 mA,
or > 200 dB dynamic range.
⌠ τ = 985
R = 656 KΩ ⌡

Figure 14 An admittance network in an EFFB seismometer converts an output voltage to a frequency-dependent current applied to a force
transducer. The feedback force is equivalent to ground acceleration with a maximum useful dynamic range exceeding 200 dB. The feedback path
controlling closed-loop response is frequency-dependent and dominated by the capacitive path over a broad range. Feedback current is limited at
high frequency by force transducer internal resistance. The network is a simplified version of the network used in the STS-1/VBB. In the actual
sensor, two force transducers are used: one for the Rp and C path, and one with a smaller transducer constant and opposite polarity for the integral
path, which is inverted. The values of feedback components are representative of the median of a population of 200 vertical and horizontal
sensors, and hence do not represent a specific instrument.
 Theory and Observations - Instrumentation for Global and Regional Seismology 45

Consider first the case where RC ¼ TD ¼ 0, that is, there is no EðsÞ 1 sðsCRC þ 1Þ
¼ h i [29]
feedback transducer coil resistance or time delay in the dis- AðsÞ gC s5 RC TD þ s4 RC þTD =C þ s3 1 þ s2 þ s 1 þ 1
rg rg rgC CRp tRC
placement transducer. Even for relatively large values of o0’,
2x 0
the term rgC o0 1, allowing eqn [26] to be written: For RC ¼ 800, TD ¼ 102, r ¼ 100 000 V m1,
2 1
g ¼ 82 m s A , and other values as in Figure 14, eqn [29]
EðsÞ 1 s
¼ h i [27] gives the nominal response of the STS-1/VBB, flat in ground
AðsÞ gC s3 1 þ s2 þ s 1 þ o0 0 2 þ 1 velocity with closed-loop generator constant:
rgC CRp rgC tRC

The coefficient of s3 affects frequencies >1 Hz. The remain- Gc ¼ 1=gC [30]
ing second-order polynomial defines the low-frequency behav- 1
where with the nominal parameters, Gc ¼ 2400 V m s from
ior of the transfer
pffiffiffiffiffiffiffiffi
ffi function, with closed-loop free period
360 s to 10 Hz. The behavior above 10 Hz is strongly depen-
TC ¼ 2p tRC, and the fraction of critical damping:
dent on the loop gain, defined by r, g, and C, and the poles
" # 
1 1 o0 0 2 tR 0:5 created by RC ¼ 6 0 and TD ¼
6 0. While a flat response to higher
xc ¼ þ [28] frequency is possible in principle by increasing r alone with
2 Rp rg C
electronic gain, increasing r alone creates a sharply peaked
Equation [28] shows the importance of the inverse filter response. A flat response to higher frequency without peaking
in desensitizing the closed-loop damping to changes in is possible only if TD is minimized, and the loop gain is
the mechanical period of the inertial suspension. With the increased using a higher value of r and g/RC. This is the approach
inverse filter, o’0 o0, and o0’2/rg 1/Rp, allowing damping taken in later-generation sensors such as the STS-2, STS-2.5, and
to be set by Rp. Having made this approximation, the closed- T240. Table 1 lists a few parameter variations whose effect on
loop response including the effects of RC, TD can be amplitude response is shown in Figure 15. Increasing displace-
considered: ment transducer gain r has no effect on the closed-loop gain,

Table 1 Parameters of the STS-1/VBB closed-loop transfer function, varied to produce the five cases shown in Figure 15

Param\case Nominal STS-1 r" g" g", r", TD#, RC# No inverse filter

r (V m1) 100 000 400 000 100 000 400 000 100 000
RC (O) 800 800 800 100 800
TD (s) 0.01 0.01 0.01 0.0001 0.01
g (m s2A1) 82 82 328 328 82
T0 (s) 30 30 30 30 3

VBB reponse parameter dependence

1.E+05 4

3
1.E+04
2
1.E+03
V-1 m-1 s-1

0 Phase (rad)
1.E+02
g,r, TD¯, RC¯ g,r, TD¯, RC¯
-1
1.E+01 r r
g Nominal STS-1/VBB -2
1.E+00 Nominal STS-1/VBB
No inverse filter -3
No inverse filter

1.E-01 -4
0.0001 0.001 0.01 0.1 1 10 100 1000 0.0001 0.001 0.01 0.1 1 10 100 1000
Frequency Frequency

Figure 15 The effect on the closed-loop bandwidth, flatness, gain, and phase of the STS-1/VBB for various parameter changes in Table 1. Other
parameter values as in Figure 14. The up and down arrows indicate changes in the parameters relative to the nominal response. Bandwidth may be
extended by increasing r or gain reduced by increasing g, but peaking occurs at the upper corner unless TD and RC are minimized. The closed-loop
gain is controlled by the force transducer constant g and feedback capacitor C. The inverse filter reduces dependence of the low-frequency corner on
parameters of the mechanical suspension. The actual sensor uses a smaller TD, and includes a single-pole low-pass filter external to the feedback network
to define the two-pole asymptotic response above 10 Hz. In addition, the effective RC is higher and is adjusted at the time of manufacture to control
uniformity of the response at the 10 Hz corner. These parameter values are representative of the median of a population of 200 horizontal and vertical
sensors. Actual vertical sensors have lower g; horizontal sensors have higher g, resulting in corresponding differences in other feedback components in
actual sensors to achieve the nominal response. Actual feedback component values are trimmed to produce a uniform response on all instruments.
Exploring the Variety of Random
Documents with Different Content
ses hoquets s'entendaient deux salles plus loin. Au dernier moment,
une religieuse lui tenait une bougie allumée dans la main ; la
servante, de l'autre côté du lit, racontait le plaisir qu'elle venait
d'avoir à la kermesse de son village ; la sœur écoutait, amusée ;
toutes deux se penchaient au-dessus du lit en riant, sans se
préoccuper de la mourante, dont le regard intelligent allait de l'une à
l'autre. La cire de la bougie coulait sur la main de la jeune femme et
la brûlait. Ses hoquets se précipitaient ; elle fit une grimace ridicule
en se mordant la langue, et ce fut tout. La sœur enleva la bougie,
regarda négligemment la morte, et s'éloigna avec la servante, en
poursuivant la conversation.
Une couturière tuberculeuse avait accouché en agonisant, sans
pousser un gémissement ; mais, quand elle fut délivrée et qu'on
emporta l'enfant pour le laver, elle s'efforça de lever les bras et
bégaya :
— Je ne le verrai pas.
Elle devint livide, sa tête ballotta de droite et de gauche : elle
était morte.
J'irai mourir ainsi, moi! jamais!!
J'en ai pour cinq ans, si je ne guéris pas : j'aurais alors vingt-
quatre ans, Klaasje seulement quatorze, et je ne serais plus là! Ah!
non, non! je ne veux pas. Il me faut ces médicaments qui me
guériront. Le docteur se les fait donner à la pharmacie de l'hôpital :
j'en aurai donc toujours.
Quand mes bouteilles étaient vides, j'allais chez le chef de service
qui, chaque fois, poussait le verrou.
 PROSTITUÉE

«Ma fille a le billet jaune».

Dostoïevsky.

Encore une fois, nous étions sans manger. Hein frappait depuis
deux jours sur l'enclume, avec les lourds marteaux de son métier de
forgeron, sans avoir pris aucune nourriture ; il était affalé sur une
chaise, pâle, la tête baissée, les bras pendants, engourdis le long du
corps, et répétait :
— Je ne peux plus, je ne peux plus. Les petites jambes de
Klaasje s'étaient dérobées sous lui, et il gisait à terre, contre le mur ;
les autres enfants étaient dispersés, ici et là, dans la chambre, tous
malades de faim. Ma mère avait le visage enfiévré, et des
clignotements d'yeux précipités qui accusaient son affolement ; moi,
des vertiges me faisaient chanceler.
Ma sœur aînée nous avait quittés, et nous attendions mon père,
parti dès le matin à la recherche de quelque chose à gagner. Il
rentra ivre et demanda à manger.
Je regardais autour de moi, sentant qu'un malheur allait arriver,
si on ne trouvait immédiatement une issue. Ma décision fut prise.
J'allongeai ma jupe en traîne ; je tirai mes cheveux sur le front ; je
m'ajustai le mieux que je pus, en regrettant de n'avoir pas de fard,
comme j'en avais vu aux prostituées, et dis à ma mère que j'allais
sortir. Elle voulut m'accompagner, pour rapporter plus vite les
victuailles.
Une fois au centre de la ville, je lui recommandai de rester à
distance. Bientôt un homme me fit signe de le suivre, et m'emmena
dans une maison de rendez-vous. Quand, après, je lui réclamai mon
salaire, il me demanda si je me moquais de lui.
— Pour cinq francs, je puis avoir une femme chic, et tu es fichue
comme une mendiante et sale en proportion. Ouste! laisse-moi
passer.
En bas, il refusa de payer la chambre. La tenancière nous
menaça de la police, et il finit par régler. A la sortie, la femme me
cria :
— Sale guenille, je te ferai «carter», si tu oses revenir.
Ma mère m'attendait au boulevard ; quand je lui racontai la
chose, elle resta pétrifiée.
— Que pouvais-je faire? Que pouvais-je faire? J'ai risqué d'être
enceinte d'un inconnu, d'attraper la sale maladie, on m'a insultée, et
pour rien, pour rien! et les enfants, mon Dieu, les enfants!
— Si nous ne rapportons rien, ils mourront, dit ma mère.
Je pleurais, la figure contre un arbre. Mais la vision de nos
enfants qui nous attendaient, me rendit toute mon énergie.
— Je vais continuer, dis-je ; mais tenez-vous donc plus loin : vous
me suivez sur les talons.
Je n'avais pas de mouchoir et, en essuyant mes larmes de mes
mains, je me barbouillais la figure.
J'entendis bientôt murmurer derrière moi :
— Petite, petite…
Je me retournai et vis un géant qui me suivait.
— Petite, viens avec moi.
Je le suivis.
Il me conduisit dans une autre maison, et me donna quelques
francs d'avance.
 Il me mania avec grande précaution : il avait manifestement peur
de me casser. Il riait de ma figure noire, il riait de ma maigreur, tout
mon être minime le mettait en joie, et il répétait sans cesse :
— Petite, petite!
Après quelque temps, on vint frapper à la porte en criant :
— Dites donc, vous autres, le temps est passé ; du monde
attend ; il nous faut la chambre.
Croyant que c'était la police, je m'étais jetée, terrifiée, contre le
géant, ce qui le mit encore en joie. Il m'entoura de ses bras, et riant
doucement, murmura :
— Allons, petite! Allons, petite!
Comme j'étais bien sur cette immense poitrine! pour la première
fois de ma vie, je me sentis protégée. Tous les sbires de la ville
n'auraient pu dénouer les bras qui m'enserraient : il leur aurait dit,
amusé :
— Voyons, c'est une petite, une petite.
Une fois à la rue, je galopai vers ma mère. Nous achetâmes de
pauvres vivres, et, dès le bas de l'escalier, nous criâmes aux
enfants :
— Nous avons du pain! nous avons du pain!
Au bout de quelques jours, notre ménage marcha régulièrement,
comme jamais il n'avait marché. Les enfants mangeaient aux heures,
étaient lavés, allaient à l'école ; ma mère vaquait au ménage ; mon
père ne buvait plus : il faisait le café et pelait les pommes de terre.
Seule, je rageais et pleurais, accroupie sur le vieux canapé qui me
servait de lit.
La simplicité avec laquelle mes parents s'adaptaient à cette
situation, me les faisait prendre en une aversion qui croissait chaque
jour. Ils en étaient arrivés à oublier que moi, la plus jolie de la
nichée, je me prostituais tous les soirs aux passants. Sans doute, il
n'y avait d'autre moyen pour nous de ne pas mourir de faim, mais je
me refusais à admettre que ce moyen fût accepté sans la révolte et
les imprécations qui, nuit et jour, me secouaient.
 J'étais trop jeune pour comprendre que, chez eux, la misère avait
achevé son œuvre, tandis que j'avais toute ma jeunesse et toute ma
vigueur pour me cabrer devant le sort.
 TABLE DES MATIÈRES

Vision 1
Mes parents 5
Quand je me réveillai, c'était le soir 17
Premier Exode 21
Reliefs et Oripeaux 25
Têtes et Peaux d'Anguilles 29
Deuxième Exode 33
Non! Non! 37
A l'École catholique 47
La Soupe aux Pois 53
Catéchisme et Première Communion 59
J'entends les puces marcher 71
Déception 79
Mon père propose de nous abandonner 83
Je fais des visites 87
Toupie et Cerf-volant 101
Une Expulsion 107
Ma Robe de Première Communion 115
Jours de fête 119
Nous vivons de charité 123
 Ah! vous aviez des «kwartjes»! 129
L'Usurière 133
Baâtje 137
Si nous étions riches 145
Je fais pipi dans mes jupes 151
Les deux Grenadiers 155
Le Village Rouge 163
Marchande de Rue 169
Une leçon de vie pratique 181
Je quitte ma place 191
Ma fille, Monsieur Cabanel 199
Troisième Exode 207
Fabrique de Chapeaux 213
Ils pèlent des oignons 223
Une nuit au parc de Bruxelles 227
La variole 241
Les pommes de terre 245
Un pain pour des timbres 249
Kees acrobate 253
Symphonie de la faim 261
Klaasje condamné 267
A l'hôpital 277
Prostituée 291

DIJON, IMP. DARANTIERE.

Extrait du Catalogue de la BIBLIOTHÈQUE-CHARPENTIER à

3 fr. 50 le volume
 EUGÈNE FASQUELLE, ÉDITEUR, 11, RUE DE GRENELLE

DERNIÈRES PUBLICATIONS

MARGUERITE AUDOUX
Marie-Claire 1 vol.
ÉMILE BERR
Sonia et ses Amis 1 vol.
JULES CLARETIE
Quarante ans après (Impressions d'Alsace et de Lorraine, 1 vol.
1870-1910)
La Vie à Paris. 1910 1 vol.
ÉTIENNE COROT
De Loin 1 vol.
LÉON DAUDET
La Mésentente 1 vol.
LUCIE DELARUE-MARDRUS
Tout l'Amour 1 vol.
MARTHE FIEL
Sur le Sol d'Alsace 1 vol.
FERNAND GAVARRY
Pièces et Morceaux 1 vol.
CHARLES-HENRY HIRSCH
Parfieu et Martin 1 vol.
JULES HURET
En 1 vol.
Allemagne : La Bavière et la Saxe
— Berlin 1 vol.
En Argentine : De Buenos Aires au Gran Chaco 1 vol.
HENRY KISTEMAECKERS
Lord Will Aviateur 1 vol.
VICTOR MARGUERITTE
L'Or 1 vol.
OCTAVE MIRBEAU
La 628-E8 1 vol.
NEEL DOFF
Jours de Famine et de Détresse 1 vol.
MICHEL PROVINS
Celles qu'on brûle. Celles qu'on envoie 1 vol.
ÉDOUARD QUET
Les Épaves 1 vol.
LOUIS DE ROBERT
Le Roman du Malade 1 vol.
J.-H. ROSNY Aîné
La Guerre du Feu 1 vol.
EDMOND ROSTAND
Chantecler 1 vol.
ÉMILE ZOLA
Correspondance. — Les Lettres et les Arts 1 vol.

ENVOI FRANCO PAR POSTE CONTRE MANDAT

4279. — L.-Imprimeries réunies, rue Saint-Benoît, 7, Paris.

*** END OF THE PROJECT GUTENBERG EBOOK JOURS DE FAMINE
ET DE DÉTRESSE: ROMAN ***

Updated editions will replace the previous one—the old editions will
be renamed.

Creating the works from print editions not protected by U.S.

copyright law means that no one owns a United States copyright in
these works, so the Foundation (and you!) can copy and distribute it
in the United States without permission and without paying
copyright royalties. Special rules, set forth in the General Terms of
Use part of this license, apply to copying and distributing Project
Gutenberg™ electronic works to protect the PROJECT GUTENBERG™
concept and trademark. Project Gutenberg is a registered trademark,
and may not be used if you charge for an eBook, except by following
the terms of the trademark license, including paying royalties for use
of the Project Gutenberg trademark. If you do not charge anything
for copies of this eBook, complying with the trademark license is
very easy. You may use this eBook for nearly any purpose such as
creation of derivative works, reports, performances and research.
Project Gutenberg eBooks may be modified and printed and given
away—you may do practically ANYTHING in the United States with
eBooks not protected by U.S. copyright law. Redistribution is subject
to the trademark license, especially commercial redistribution.

START: FULL LICENSE

THE FULL PROJECT GUTENBERG LICENSE
 PLEASE READ THIS BEFORE YOU DISTRIBUTE OR USE THIS WORK

To protect the Project Gutenberg™ mission of promoting the free

distribution of electronic works, by using or distributing this work (or
any other work associated in any way with the phrase “Project
Gutenberg”), you agree to comply with all the terms of the Full
Project Gutenberg™ License available with this file or online at
www.gutenberg.org/license.

Section 1. General Terms of Use and

Redistributing Project Gutenberg™
electronic works
1.A. By reading or using any part of this Project Gutenberg™
electronic work, you indicate that you have read, understand, agree
to and accept all the terms of this license and intellectual property
(trademark/copyright) agreement. If you do not agree to abide by all
the terms of this agreement, you must cease using and return or
destroy all copies of Project Gutenberg™ electronic works in your
possession. If you paid a fee for obtaining a copy of or access to a
Project Gutenberg™ electronic work and you do not agree to be
bound by the terms of this agreement, you may obtain a refund
from the person or entity to whom you paid the fee as set forth in
paragraph 1.E.8.

1.B. “Project Gutenberg” is a registered trademark. It may only be

used on or associated in any way with an electronic work by people
who agree to be bound by the terms of this agreement. There are a
few things that you can do with most Project Gutenberg™ electronic
works even without complying with the full terms of this agreement.
See paragraph 1.C below. There are a lot of things you can do with
Project Gutenberg™ electronic works if you follow the terms of this
agreement and help preserve free future access to Project
Gutenberg™ electronic works. See paragraph 1.E below.
1.C. The Project Gutenberg Literary Archive Foundation (“the
Foundation” or PGLAF), owns a compilation copyright in the
collection of Project Gutenberg™ electronic works. Nearly all the
individual works in the collection are in the public domain in the
United States. If an individual work is unprotected by copyright law
in the United States and you are located in the United States, we do
not claim a right to prevent you from copying, distributing,
performing, displaying or creating derivative works based on the
work as long as all references to Project Gutenberg are removed. Of
course, we hope that you will support the Project Gutenberg™
mission of promoting free access to electronic works by freely
sharing Project Gutenberg™ works in compliance with the terms of
this agreement for keeping the Project Gutenberg™ name associated
with the work. You can easily comply with the terms of this
agreement by keeping this work in the same format with its attached
full Project Gutenberg™ License when you share it without charge
with others.

1.D. The copyright laws of the place where you are located also
govern what you can do with this work. Copyright laws in most
countries are in a constant state of change. If you are outside the
United States, check the laws of your country in addition to the
terms of this agreement before downloading, copying, displaying,
performing, distributing or creating derivative works based on this
work or any other Project Gutenberg™ work. The Foundation makes
no representations concerning the copyright status of any work in
any country other than the United States.

1.E. Unless you have removed all references to Project Gutenberg:

1.E.1. The following sentence, with active links to, or other

immediate access to, the full Project Gutenberg™ License must
appear prominently whenever any copy of a Project Gutenberg™
work (any work on which the phrase “Project Gutenberg” appears,
or with which the phrase “Project Gutenberg” is associated) is
accessed, displayed, performed, viewed, copied or distributed:
 This eBook is for the use of anyone anywhere in the United
States and most other parts of the world at no cost and with
almost no restrictions whatsoever. You may copy it, give it away
or re-use it under the terms of the Project Gutenberg License
included with this eBook or online at www.gutenberg.org. If you
are not located in the United States, you will have to check the
laws of the country where you are located before using this
eBook.

1.E.2. If an individual Project Gutenberg™ electronic work is derived

from texts not protected by U.S. copyright law (does not contain a
notice indicating that it is posted with permission of the copyright
holder), the work can be copied and distributed to anyone in the
United States without paying any fees or charges. If you are
redistributing or providing access to a work with the phrase “Project
Gutenberg” associated with or appearing on the work, you must
comply either with the requirements of paragraphs 1.E.1 through
1.E.7 or obtain permission for the use of the work and the Project
Gutenberg™ trademark as set forth in paragraphs 1.E.8 or 1.E.9.

1.E.3. If an individual Project Gutenberg™ electronic work is posted

with the permission of the copyright holder, your use and distribution
must comply with both paragraphs 1.E.1 through 1.E.7 and any
additional terms imposed by the copyright holder. Additional terms
will be linked to the Project Gutenberg™ License for all works posted
with the permission of the copyright holder found at the beginning
of this work.

1.E.4. Do not unlink or detach or remove the full Project

Gutenberg™ License terms from this work, or any files containing a
part of this work or any other work associated with Project
Gutenberg™.

1.E.5. Do not copy, display, perform, distribute or redistribute this

electronic work, or any part of this electronic work, without
prominently displaying the sentence set forth in paragraph 1.E.1
with active links or immediate access to the full terms of the Project
Gutenberg™ License.

1.E.6. You may convert to and distribute this work in any binary,
compressed, marked up, nonproprietary or proprietary form,
including any word processing or hypertext form. However, if you
provide access to or distribute copies of a Project Gutenberg™ work
in a format other than “Plain Vanilla ASCII” or other format used in
the official version posted on the official Project Gutenberg™ website
(www.gutenberg.org), you must, at no additional cost, fee or
expense to the user, provide a copy, a means of exporting a copy, or
a means of obtaining a copy upon request, of the work in its original
“Plain Vanilla ASCII” or other form. Any alternate format must
include the full Project Gutenberg™ License as specified in
paragraph 1.E.1.

1.E.7. Do not charge a fee for access to, viewing, displaying,

performing, copying or distributing any Project Gutenberg™ works
unless you comply with paragraph 1.E.8 or 1.E.9.

1.E.8. You may charge a reasonable fee for copies of or providing

access to or distributing Project Gutenberg™ electronic works
provided that:

• You pay a royalty fee of 20% of the gross profits you derive
from the use of Project Gutenberg™ works calculated using the
method you already use to calculate your applicable taxes. The
fee is owed to the owner of the Project Gutenberg™ trademark,
but he has agreed to donate royalties under this paragraph to
the Project Gutenberg Literary Archive Foundation. Royalty
payments must be paid within 60 days following each date on
which you prepare (or are legally required to prepare) your
periodic tax returns. Royalty payments should be clearly marked
as such and sent to the Project Gutenberg Literary Archive
Foundation at the address specified in Section 4, “Information
 about donations to the Project Gutenberg Literary Archive
Foundation.”

• You provide a full refund of any money paid by a user who

notifies you in writing (or by e-mail) within 30 days of receipt
that s/he does not agree to the terms of the full Project
Gutenberg™ License. You must require such a user to return or
destroy all copies of the works possessed in a physical medium
and discontinue all use of and all access to other copies of
Project Gutenberg™ works.

• You provide, in accordance with paragraph 1.F.3, a full refund of

any money paid for a work or a replacement copy, if a defect in
the electronic work is discovered and reported to you within 90
days of receipt of the work.

• You comply with all other terms of this agreement for free
distribution of Project Gutenberg™ works.

1.E.9. If you wish to charge a fee or distribute a Project Gutenberg™

electronic work or group of works on different terms than are set
forth in this agreement, you must obtain permission in writing from
the Project Gutenberg Literary Archive Foundation, the manager of
the Project Gutenberg™ trademark. Contact the Foundation as set
forth in Section 3 below.

1.F.

1.F.1. Project Gutenberg volunteers and employees expend

considerable effort to identify, do copyright research on, transcribe
and proofread works not protected by U.S. copyright law in creating
the Project Gutenberg™ collection. Despite these efforts, Project
Gutenberg™ electronic works, and the medium on which they may
be stored, may contain “Defects,” such as, but not limited to,
incomplete, inaccurate or corrupt data, transcription errors, a
copyright or other intellectual property infringement, a defective or
damaged disk or other medium, a computer virus, or computer
codes that damage or cannot be read by your equipment.

1.F.2. LIMITED WARRANTY, DISCLAIMER OF DAMAGES - Except for

the “Right of Replacement or Refund” described in paragraph 1.F.3,
the Project Gutenberg Literary Archive Foundation, the owner of the
Project Gutenberg™ trademark, and any other party distributing a
Project Gutenberg™ electronic work under this agreement, disclaim
all liability to you for damages, costs and expenses, including legal
fees. YOU AGREE THAT YOU HAVE NO REMEDIES FOR
NEGLIGENCE, STRICT LIABILITY, BREACH OF WARRANTY OR
BREACH OF CONTRACT EXCEPT THOSE PROVIDED IN PARAGRAPH
1.F.3. YOU AGREE THAT THE FOUNDATION, THE TRADEMARK
OWNER, AND ANY DISTRIBUTOR UNDER THIS AGREEMENT WILL
NOT BE LIABLE TO YOU FOR ACTUAL, DIRECT, INDIRECT,
CONSEQUENTIAL, PUNITIVE OR INCIDENTAL DAMAGES EVEN IF
YOU GIVE NOTICE OF THE POSSIBILITY OF SUCH DAMAGE.

1.F.3. LIMITED RIGHT OF REPLACEMENT OR REFUND - If you

discover a defect in this electronic work within 90 days of receiving
it, you can receive a refund of the money (if any) you paid for it by
sending a written explanation to the person you received the work
from. If you received the work on a physical medium, you must
return the medium with your written explanation. The person or
entity that provided you with the defective work may elect to provide
a replacement copy in lieu of a refund. If you received the work
electronically, the person or entity providing it to you may choose to
give you a second opportunity to receive the work electronically in
lieu of a refund. If the second copy is also defective, you may
demand a refund in writing without further opportunities to fix the
problem.

1.F.4. Except for the limited right of replacement or refund set forth
in paragraph 1.F.3, this work is provided to you ‘AS-IS’, WITH NO
OTHER WARRANTIES OF ANY KIND, EXPRESS OR IMPLIED,
INCLUDING BUT NOT LIMITED TO WARRANTIES OF
MERCHANTABILITY OR FITNESS FOR ANY PURPOSE.

1.F.5. Some states do not allow disclaimers of certain implied

warranties or the exclusion or limitation of certain types of damages.
If any disclaimer or limitation set forth in this agreement violates the
law of the state applicable to this agreement, the agreement shall be
interpreted to make the maximum disclaimer or limitation permitted
by the applicable state law. The invalidity or unenforceability of any
provision of this agreement shall not void the remaining provisions.

1.F.6. INDEMNITY - You agree to indemnify and hold the Foundation,

the trademark owner, any agent or employee of the Foundation,
anyone providing copies of Project Gutenberg™ electronic works in
accordance with this agreement, and any volunteers associated with
the production, promotion and distribution of Project Gutenberg™
electronic works, harmless from all liability, costs and expenses,
including legal fees, that arise directly or indirectly from any of the
following which you do or cause to occur: (a) distribution of this or
any Project Gutenberg™ work, (b) alteration, modification, or
additions or deletions to any Project Gutenberg™ work, and (c) any
Defect you cause.

Section 2. Information about the Mission

of Project Gutenberg™
Project Gutenberg™ is synonymous with the free distribution of
electronic works in formats readable by the widest variety of
computers including obsolete, old, middle-aged and new computers.
It exists because of the efforts of hundreds of volunteers and
donations from people in all walks of life.

Volunteers and financial support to provide volunteers with the

assistance they need are critical to reaching Project Gutenberg™’s
goals and ensuring that the Project Gutenberg™ collection will
remain freely available for generations to come. In 2001, the Project
Gutenberg Literary Archive Foundation was created to provide a
secure and permanent future for Project Gutenberg™ and future
generations. To learn more about the Project Gutenberg Literary
Archive Foundation and how your efforts and donations can help,
see Sections 3 and 4 and the Foundation information page at
www.gutenberg.org.

Section 3. Information about the Project

Gutenberg Literary Archive Foundation
The Project Gutenberg Literary Archive Foundation is a non-profit
501(c)(3) educational corporation organized under the laws of the
state of Mississippi and granted tax exempt status by the Internal
Revenue Service. The Foundation’s EIN or federal tax identification
number is 64-6221541. Contributions to the Project Gutenberg
Literary Archive Foundation are tax deductible to the full extent
permitted by U.S. federal laws and your state’s laws.

The Foundation’s business office is located at 809 North 1500 West,

Salt Lake City, UT 84116, (801) 596-1887. Email contact links and up
to date contact information can be found at the Foundation’s website
and official page at www.gutenberg.org/contact

Section 4. Information about Donations to

the Project Gutenberg Literary Archive
Foundation
Project Gutenberg™ depends upon and cannot survive without
widespread public support and donations to carry out its mission of
increasing the number of public domain and licensed works that can
be freely distributed in machine-readable form accessible by the
widest array of equipment including outdated equipment. Many
small donations ($1 to $5,000) are particularly important to
maintaining tax exempt status with the IRS.

The Foundation is committed to complying with the laws regulating

charities and charitable donations in all 50 states of the United
States. Compliance requirements are not uniform and it takes a
considerable effort, much paperwork and many fees to meet and
keep up with these requirements. We do not solicit donations in
locations where we have not received written confirmation of
compliance. To SEND DONATIONS or determine the status of
compliance for any particular state visit www.gutenberg.org/donate.

While we cannot and do not solicit contributions from states where

we have not met the solicitation requirements, we know of no
prohibition against accepting unsolicited donations from donors in
such states who approach us with offers to donate.

International donations are gratefully accepted, but we cannot make

any statements concerning tax treatment of donations received from
outside the United States. U.S. laws alone swamp our small staff.

Please check the Project Gutenberg web pages for current donation
methods and addresses. Donations are accepted in a number of
other ways including checks, online payments and credit card
donations. To donate, please visit: www.gutenberg.org/donate.

Section 5. General Information About

Project Gutenberg™ electronic works
Professor Michael S. Hart was the originator of the Project
Gutenberg™ concept of a library of electronic works that could be
freely shared with anyone. For forty years, he produced and
distributed Project Gutenberg™ eBooks with only a loose network of
volunteer support.
Project Gutenberg™ eBooks are often created from several printed
editions, all of which are confirmed as not protected by copyright in
the U.S. unless a copyright notice is included. Thus, we do not
necessarily keep eBooks in compliance with any particular paper
edition.

Most people start at our website which has the main PG search
facility: www.gutenberg.org.

This website includes information about Project Gutenberg™,

including how to make donations to the Project Gutenberg Literary
Archive Foundation, how to help produce our new eBooks, and how
to subscribe to our email newsletter to hear about new eBooks.
Welcome to Our Bookstore - The Ultimate Destination for Book Lovers
Are you passionate about books and eager to explore new worlds of
knowledge? At our website, we offer a vast collection of books that
cater to every interest and age group. From classic literature to
specialized publications, self-help books, and children’s stories, we
have it all! Each book is a gateway to new adventures, helping you
expand your knowledge and nourish your soul
Experience Convenient and Enjoyable Book Shopping Our website is more
than just an online bookstore—it’s a bridge connecting readers to the
timeless values of culture and wisdom. With a sleek and user-friendly
interface and a smart search system, you can find your favorite books
quickly and easily. Enjoy special promotions, fast home delivery, and
a seamless shopping experience that saves you time and enhances your
love for reading.
Let us accompany you on the journey of exploring knowledge and
personal growth!

ebookgate.com