ARTICLE IN PRESS

Quaternary Science Reviews 24 (2005) 1183–1202

Modelling Holocene relative sea-level observations from the

Caribbean and South America
Glenn A. Milnea,, Antony J. Longb, Sophie E. Bassetta
a
Department of Earth Sciences, University of Durham, Durham DH1 3LE, UK
b
Department of Geography, University of Durham, Durham DH1 3LE, UK
Received 27 January 2004; accepted 17 October 2004

Abstract

Holocene data from the Caribbean and the Atlantic coast of South America have been critically assessed and a subset of the best
quality data are presented. These data cover a large north–south geographic extent and display a distinct spatial variation. We make
the first comparisons of this data set to predictions based on a realistic model of glaciation-induced sea-level change with the main
aims of understanding the cause of the observed spatial trend and estimating a eustatic signal for the Holocene. The spatial variation
is dominated by the influence of the ice and ocean mass redistribution on sea-level change, with the ice-induced effect dominating the
observed north–south trend. A best-fitting model is applied to estimate a Holocene eustatic signal from the observations. We find
that the model-corrected data are consistent with a relatively rapid rise of 7–8 mm/yr in the early Holocene with a marked reduction
in this rate around 7 cal. kyr BP. From this time until present, the model-corrected data suggest that the volume of mass transfer
between ice sheets and oceans was no more than
1 m (eustatic sea-level equivalent).
r 2004 Elsevier Ltd. All rights reserved.

1. Introduction describe this climate-induced mass redistribution and

corresponding earth deformation.
Changes in relative sea level (RSL) are caused by Observations of GIA-induced sea-level change can be
vertical motion between the ocean surface and the ocean employed to infer information relating to the geographic
floor. There are a number of earth processes that can form and melt histories of the late Pleistocene ice sheets;
produce vertical deflections of either or both of these the integrated melt signal from global, grounded ice
surfaces over a variety of spatial and temporal scales sheets and glaciers; as well as the viscosity structure of
(e.g., Emery and Aubrey, 1991). During the past
the solid Earth. The largest RSL signal is found in once-
20 kyr; RSL changes have been dominated by the glaciated (or near-field) regions where the ice unloading
transition of the earth system from a glacial to an produced a considerable deformation of the solid Earth.
interglacial stage. The marked global warming during RSL observations from these regions have provided
this transition resulted in
70% of the continental ice useful constraints on local ice deglaciation histories and
budget to melt. This large-scale ice–ocean mass redis- sub-surface radial viscosity structure (e.g., Haskell,
tribution produced a dramatic increase in ocean water 1935; Vening Meinesz, 1937; Wu and Peltier, 1983;
volume as well as a large and ongoing isostatic response Tushingham and Peltier, 1991; Lambeck, 1993; Mitro-
of the solid Earth. The term glacial isostatic adjustment vica, 1996; Lambeck et al., 1998). At increasing
(GIA) is commonly adopted as a general term to distances from the major centers of glaciation, the
ice-induced component of the signal reduces in mag-
Corresponding author. Tel.: +44 191 374 2512; nitude and so the eustatic (or meltwater) signal
fax: +44 191 374 2510. becomes dominant. In addition, the smaller amplitude
E-mail address: [email protected] (G.A. Milne). signal associated with ocean loading and GIA-induced

0277-3791/$ - see front matter r 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.quascirev.2004.10.005
 ARTICLE IN PRESS
1184 G.A. Milne et al. / Quaternary Science Reviews 24 (2005) 1183–1202

perturbations to the Earth’s rotation vector become from four widely distributed data locations and
more evident (e.g., Clark et al., 1978; Milne and concluded that about 3–5 m of eustatic sea-level rise
Mitrovica, 1998; Mitrovica and Milne, 2002). Accord- ocurred between
7 and
2–1 cal. kyr BP. The
ingly, far-field observations are commonly employed to Antarctic ice sheet and temperate mountain glaciers
infer the eustatic component of the RSL signal. This were given as possible sources for this melt signal.
component is a quantity that represents the net global Lambeck (2002), revisited the Australian data to infer
ice–ocean mass flux and so is an important indicator of a eustatic signal of
3 m between approximately 7
climate change (e.g., Clark and Mix, 2000). and 3–2 cal. kyr BP. This result is very similar to that
The change in the eustatic signal during the Holocene obtained from a high quality data set from the French
is of interest for two main reasons. First, it is important Mediterranean coast (Lambeck and Bard, 2000).
to determine the timing and abruptness of the reduction Finally, Peltier (2002) suggests a model that contains
in global melting that ocurred during the early to mid- a decrease in global melt after
7 cal. kyr BP with a
Holocene in order to better understand the driving few meters of melt from this time to
4 cal. kyr BP,
mechanisms for this change. Second, the climatic with no global melt from this time to the pre-
variations that have occurred throughout the mid- to sent. These more recent analyses all suggest a
late-Holocene illustrate the frequency and magnitude of significant decrease in melt rate around 7 cal. kyr BP
natural change within a climatic system that is similar to with a few meters of melt from this time until the late-
the present one. Therefore, observations that are Holocene and no discernable melting during the past
sensitive to climate change during this period provide few thousand years.
key information that can be used to assess the potential In this study, we consider Holocene data from the
impact of anthropogenic influences on the modern Caribbean and the Atlantic coast of South America to
climate system. For example, global present-day sea- achieve three principal aims: (1) critically review the
level rise over the past
100 yr associated with climate published data to produce a high-quality subset that is
change is estimated to be between 1 and 2 mm/yr, or suitable for quantitative modelling purposes; (2) apply a
equivalently 1–2 m/kyr. It is important to compare the GIA model to better understand the distinct spatial
magnitude of this signal to the eustatic change during trends evident in the data; and (3) apply a GIA model to
the current interglacial in order to determine if the determine a preliminary estimate of the Holocene
present-day secular change represents a significant meltwater signal from the highest quality data. Data
anomaly. from this particular region have received little attention
At tectonically stable far-field locations, the RSL from the sea-level modelling community. One recent
signal is commonly characterised by a mid-Holocene exception is the work by Rostami et al. (2000) who
sea-level maximum, or highstand, at the time meltwater compared estimates of palaeo sea levels based on fossil
production decreased (e.g., Pirazzoli and Pluet, 1991). mollusc shells retrieved along the Argentinian Patagonia
The fall in sea-level from this time to the present is a to a single set of predictions generated from a GIA
result of ongoing GIA processes, such as local hydro- model. We have chosen not to include the data described
isostatic loading (termed continental levering) (e.g., by Rostami et al. (2000) in the current analysis since the
Clark et al., 1978) and a global fall in the ocean surface observational uncertainties are too large for our
due to both hydro- and glacio-isostatic loading of the purposes (see next section). The data considered in the
Earth’s surface (termed equatorial ocean syphoning) following analysis are distributed over a relatively large
(e.g., Mitrovica and Milne, 2002). These processes are region (Fig. 1) and they exhibit an interesting spatial
active during the deglaciation period but are not evident variation that provides a stringent test for the GIA
in the far-field sea-level record until the early to mid- model we apply below.
Holocene due to the dominant effect of the eustatic In the following section we present the subset of
signal prior to this time. data from the region that provides the best constraints
Previous authors have inferred the eustatic signal on palaeo sea levels. In Section 3 we compare these
from Holocene sea-level observations. Nakada and data to predictions based on a GIA sea-level model
Lambeck (1989), considered Holocene RSL observa- in order to gauge the success of the model in account-
tions from a variety of locations around the Austra- ing for the observations. The model predictions are then
lian coastline and from a small number of Pacific de-constructed into components associated with the
islands. They concluded that the eustatic melt from different components of the GIA forcing (ice, ocean,
the time of the observed high-stands,
6 cal. kyr BP, and rotation) in order to examine the source of the
was around 1.5–2 m and suggested the Antarctic ice observed spatial trends. Finally, a best-fitting GIA
sheet as a potential source for this signal. More recent model is employed to remove the non-eustatic GIA
studies have provided results that are relatively signal so that the eustatic signal can be inferred from
consistent with the findings of Nakada and Lambeck the data. The main results are then summarised in
(1989). Flemming et al. (1998) considered RSL data Section 4.
 ARTICLE IN PRESS
G.A. Milne et al. / Quaternary Science Reviews 24 (2005) 1183–1202 1185

deposits are assumed to form within
0:5 m of MSL

20°N (e.g., Digerfeldt and Hendry, 1987). Vermitids generally
Jamaica (A) occupy the lower part of the intertidal zone, between
Curacao (B) neap and spring tide levels, and typically have a height
uncertainty of
0:5 m (van Andel and Laborel, 1964).
10°N
Suriname (C) Mollusc samples have a larger uncertainty (up to
2 m
or more) reflecting variations in site exposure, tidal
range and primary depositional context. The effects of
0° compaction is greatest for index points from mangrove
environments, but we minimise these where possible by
using only index points which lie directly above an
Recife (D)
10°S incompressible substrate. Other height uncertainties
relate to unknown changes in palaeotidal range which
may have occurred in the past (we assume here constant
Rio de Janeiro (E) tidal range through time), as well as possible changes in
20°S wave climate (Angulo et al., 1999). Each of the
radiocarbon dates described here has been converted
Santa Catarina (F) to calibrated years before present (cal. kyr BP), using the
Calib programme of Stuiver and Reimer (1998).
30°S
Vermitid and marine molluscs have been corrected for
the standard 400 yr marine reservoir effect before
calibrating with the marine calibration data set. All
40°S
other samples from terrestrial environments are cali-
brated using the bi-decadal atmospheric calibration data
set. Dates are cited with a two sigma age range.
Strait of Magellan (G) Jamaica is the most northerly of the sites examined
Beagle Channel (H) (Fig. 1). Digerfeldt and Hendry (1987) report 55
50°S
radiocarbon dates derived from sedge, mangrove,
100°W 80°W 60°W 40°W 20°W swamp forest and mangrove environments from the
Fig. 1. Map showing the locations of the data considered in the
Black and the Negril rivers. Of these dates, 13 are
present analysis. derived from o1 m above the Pleistocene clays and are
therefore least affected by compaction and form the
basis for our analysis. The present tidal range is
0:5 m
and the index points are thought to have formed less
2. Observations than 0.5 m above present MSL (Digerfeldt and Hendry,
1987). The resulting distribution of data (Fig. 2A) depict
A review of the sea-level data from South America a smoothly rising trend in RSL from
9 m at 8 cal.
demonstrates that a dominant proportion are not kyr BP to
4 m at 6 cal. kyr BP. There is no evidence
suitable for modelling purposes because they have large for a Holocene highstand above present, although data
age or altitude uncertainties. A wide range of indicators from between 4 cal. kyr BP and present are limited. A
are available for sea-level construction, including shells reduction in the rate of RSL rise is evident at
6 cal.
from middens and raised beaches, vermitids (worm kyr BP.
gastropods), mangrove deposits, and wave built and Sea level data from the Venezuala coast are reviewed
lagoonal terraces (e.g., Fairbridge, 1976; Martin and by Rull et al. (1999), who report data from four sites
Suguio, 1978; Martin et al., 1985). In this study we along the Caribbean coast. Most of the data are from
restrict our analyses to what we consider to be the most Cayo Sal and Cayo Sombrero and comprise coastal
reliable sea-level indicators—mangrove peats, vermitids mangrove sediments, corals and massive carbonates.
and, from the southern part of the study area, marine However, the study area has been affected by tectonic
molluscs. Our analysis includes reference to unpublished movements whilst the height relationship of the various
data from Rio de Janeiro State, Brazil (Ireland, 1987, indicators to former sea level is uncertain. An unpub-
1988) and Curacao, Venezuela (Klosowska, 2003). The lished set of mangrove dates from Curacao (Klosowska,
observations we employ in this study are provided in 2003) provide four index points which are likely to have
Appendix A. been lowered from their original elevation by compac-
We estimate height uncertainties from the observed tion (Figs. 1 and 2B). This limited data set shows MSL
relationship between different environments and present rising from
5 m at 7 cal. kyr BP to
2 m by 3 cal.
mean sea level (MSL). Index points from mangrove kyr BP.
 ARTICLE IN PRESS
1186 G.A. Milne et al. / Quaternary Science Reviews 24 (2005) 1183–1202

-10 -8 -6 -4 -2 0 -10 -8 -6 -4 -2 0

5 5

0 0

-5 -5

(A) (E)
-10 -10

5 5

0 0

-5 -5

(B) (F)
RSL (m)

-10 -10

5 5

0 0

-5 -5

(C) (G)
-10 -10
10
5

5
0
0

-5
-5

(D) (H)
-10 -10
-10 -8 -6 -4 -2 0 -10 -8 -6 -4 -2 0
Time (cal. kyr)

Fig. 2. Relative sea-level data obtained from the eight data sites (A–H) shown in Fig. 1. See main text for a discussion of the observations.

Roeleveld and Van Loon (1979) report a series of deposited in coastal mangrove swamps. Limited paly-
radiocarbon dates from the Holocene Younger Coastal nological evidence suggests that the peats formed in
plain of Suriname (Fig. 1) collected from peat layers brackish water coastal Rhizophora swamps at about
 ARTICLE IN PRESS
G.A. Milne et al. / Quaternary Science Reviews 24 (2005) 1183–1202 1187

MSL (Roeleveld and Van Loon, 1979, p. 23). The (1987) has greater confidence in the published dates
ground surface elevations of the sample sites were from vermitids and in situ Ostrea samples, and also on a
estimated (and not surveyed) to MSL and this intro- suite of 18 new radiocarbon dates he obtained from
duces a potentially significant source of error. More- samples of mangrove peat from the Rio de Janeiro
over, it is not always clear whether these samples were region. The latter have good altitudinal control and
obtained from close to the Pleistocene surface or not, so supporting palynological data. In our analysis we use 12
compaction is a further potential source of error. of these mangrove dates, together with 16 vermitid
Twenty-four index points from 11 different sites samples reviewed by Ireland (1987) and listed in Angulo
separated by
130 km in a west–east transect across and Lessa (1997). The resulting RSL graph (Fig. 2E)
the coastal plain are presented in Fig. 2C. The data show shows two discrete data populations; one indicating
MSL rising from
6 m at 8.5 cal. kyr BP to
0 m by MSL at
2 m at
8 cal. kyr BP, and a second group
6 cal. kyr BP, after which the rate of MSL fell sharply. which track a fall in MSL from
3 m at 4 cal. kyr BP to
There is no evidence for a mid-Holocene sea-level present. From these data alone it is not possible to
highstand, although a pronounced deceleration in define the timing of the mid Holocene highstand,
MSL rise occurs at 6 cal. kyr BP. Three index points although it must have occurred sometime between
4
plot above the general spread of data at
7:4 cal. kyr and 8 cal. kyr BP.
BP. Two of these (6410  50; 7427–7249 cal. yr BP (GrN Santa Catarina state, in southern Brazil (Fig. 1), has
7812), and 6580  70; 7585–7330 cal. yr. BP. (GrN yielded a large number of radiocarbon dates from
7813)) are from a single site (Tibiti), which suggest they samples of vermitids, shells and wood (Angulo and
probably record local factors. The third date (6360  70; Lessa, 1997). In our analysis we combine the vermitid
7428–7097 cal. yr BP (GrN 7821), 2:15 m) plots above data from eight sites in Santa Catarina state to generate
an identically aged index point derived from the base of a single composite RSL curve for this state (Fig. 2F).
a mangrove peat immediately above Pleistocene sedi- Twenty-seven dated samples depict a well-defined fall in
ments (6360  70; 7428–7079 cal. yr BP, GrN 4517, MSL from
2 to
3 m at 6 cal. kyr BP to present.
1:6 m). Given these observations we do not place There are no data older than 6 cal. kyr BP, which is the
undue weight to the relatively small differences in age minimum age of the mid-Holocene high-stand in this
and altitude demonstrated by these three index points part of Brazil.
compared to the remaining data set. The presence of It is important for our analysis to identify a RSL
index points during the last 1 cal. kyr BP above MSL curve for the southernmost part of the South American
suggests that the Rhisophora swamps probably formed continent. However, the quality of the RSL data from
slightly above MSL, and not at MSL as suggested by this area is not good, with large age and altitude
Roeleveld and Van Loon (1979). This would mean that uncertainties and added complications provided by the
the data in Fig. 2C over-estimate MSL by
1 m: possible effects of tectonic and isostatic factors asso-
Our fourth site is located on the eastern extremity of ciated, respectively, with Andean uplift and the loading
Brazil. The data comprise eight samples of vermitid effects of the Patagonian and Antarctic ice complexes
reported by van Andel and Laborel (1967) and (see below). Moreover, the complex coastal geometry is
Dominguez et al. (1990). The oldest date is
4 cal. likely to have promoted variations in tidal range
kyr BP, at which time MSL was between þ3 and þ5 m through time. However, sufficient data exist to enable
(Fig. 2D). Despite some scatter to the data, they us to attempt a reconstruction of RSL from the Strait of
nevertheless show a fall in MSL from a mid-Holocene Magellan and the Beagle Channel.
highstand, whose maximum elevation is not known, to Porter et al. (1984) report sea-level data derived from
present. six sites in the Strait of Magellan and Beagle Channel.
The central sector of the Brazilian coast has been They use a combination of peat, shell (middens and
subject to several reviews of Holocene RSL change. beaches) and sand dunes, with elevations cited relative
Martin et al. (1985) present a variety of RSL data to local mean sea level. They combine the data from
derived from shell middens, lagoonal terraces and other these two regions into a single sea level curve, which
marine terraces. They develop RSL curves for eight records a pronounced highstand at
7 cal. kyr BP, with
sectors of the Brazilian coast, each RSL history showing MSL reaching
3:5 m: More recently, Gordillo et al.
MSL 2–3 m above present during the mid-Holocene (1992) report a suite of new shell dates from raised
(
5 cal. kyr BP). Ireland (1987, 1988) provides a critical beaches within the Beagle Channel. The Beagle Channel
assessment of these data and Martin’s analyses, arguing occupies a protected setting, with a microtidal range of
that many of the shell dates (from middens and other 1.24 m on spring tides. Gordillo et al. (1992) suggest
settings) lack defined altitudinal and age relationships to maximum wave set up under storm conditions is 2 m,
a former sea level, whilst samples of fossil wood are not and note that at least three distinct beach terraces
in situ. Most of the data associated with terrace features occur within the channel, at 8–10, 4–6 and 1.5–3 m
also lack precision as sea-level index points. Ireland above MSL.
 ARTICLE IN PRESS
1188 G.A. Milne et al. / Quaternary Science Reviews 24 (2005) 1183–1202

Given the limitations to the data from this area, our volume bounded by the geoid and ocean floor must be
analysis of RSL changes in this region may be subject to consistent with the volume of ice lost or gained by the
bias, and a more detailed appraisal must await the adopted ice model. This mass conservation term is
collection of better quality sea-level data. In Fig. 2G we defined as,
present the RSL data from the Strait of Magellan using Z Z t
the peat and shell dates reported by Porter et al. (1984), 1 t M _ I ðtÞ 1
G M ðtÞ ¼ dt 
reducing altitudes to mean sea-level based on present rW 0 AO ðtÞ 0 AO ðtÞ
 
day height relationships between beaches and MSL. The 1 _
 _ c; tÞ dt;
Fðy; c; tÞ  Rðy; ð2Þ
data suggest limited change in MSL during the period g
9–5 cal. kyr BP. The data we use from the Beagle
Channel (Fig. 2H) are mostly derived from shells in which M _ I ðtÞ is the rate of mass increase in grounded
recorded in raised beaches found between Ushuaia and ice per unit time since the beginning of the loading
Rio Varela (a west–east transect of
100 km). The period, rW is the density of water and AO ðtÞ is the area
shells are assumed to have formed up to 2 m above high of the ocean basins, which changes in time as sea levels
tide level, which is itself
0:65 m above MSL. The rise and fall and the grounding lines of marine-based ice
resulting plot of RSL data demonstrate a pronounced sheets advance and retreat. The symbols F _ and R_ are the
high-stand
6 cal. kyr BP at
6 m; followed by a fall to time derivatives of the geopotential and the vertical
present MSL by
3 cal. kyr BP. position of the solid surface, respectively. The angled
brackets denote integration over the ocean basin area
(which is time dependent for the reasons given above).
3. Modelling results The first term on the right-hand side of (2) is the so-
called eustatic sea-level change. The second term
3.1. Introduction describes a globally uniform change in sea level
associated with non-eustatic GIA effects. This term is
Geophysical models that simulate RSL change dominated by the process of equatorial ocean syphoning
associated with earth glaciation have a long history of (e.g., Mitrovica and Milne, 2002) during the Holocene.
development (e.g., Farrell and Clark, 1976; Peltier and In the following, this component shall be referred to as
Andrews, 1976; Nakada and Lambeck, 1987; Mitrovica the non-eustatic spatially uniform signal.
and Peltier, 1991; Johnston, 1993; Milne and Mitrovica, The ocean loading associated with GIA is determined
1996; Milne et al., 1999). These models are based on the by multiplying each incremental change in the global
general relationship field SG with the so-called ocean function which takes
the value of unity over ocean areas and zero over land
1
S G ðy; c; tÞ ¼ Fðy; c; tÞ  Rðy; c; tÞ þ G M ðtÞ; (1) areas and thus depends on the coastal geometry at each
g time step in the calculation. The specific algorithms we
where the parameters y; c and t are, respectively, co- employed to predict the ocean load increments and the
latitude, east longitude, and time relative to the onset of perturbations to the rotational potential have been
surface loading. The functions F and R represent the described in detail elsewhere (e.g., Milne, 1998, 2002;
perturbations to the geopotential and the vertical Milne et al., 1999). These algorithms include a number
position of the solid surface produced by both the of recent improvements such as a time-dependent ocean
surface mass redistribution (ice and ocean components) function (Johnston, 1993) and a correct treatment of
and changes in the rotational potential associated with sea-level change in areas characterised by retreating
glaciation (e.g., Milne, 2002). The F term is normalised marine-based ice (Milne, 1998).
by g, the surface gravitational acceleration, to convert The perturbations to the geopotential and the solid
the perturbation of the geopotential into a height shift of surface in (1) are calculated by convolving the GIA
the equipotential that defines the ocean surface at the forcing functions (ice/ocean loading history and rota-
onset of surface loading. The perturbation to the tional potential history) with the impulse response Love
geopotential is commonly separated into changes numbers for a spherically symmetric, Maxwell viscoe-
produced by the gravitational effect of the surface lastic, self-gravitating and compressible earth model
ice–ocean mass redistribution and the changing rota- (Peltier, 1974). The elastic and density structure of the
tional potential (the so-called ‘‘direct effect’’) and those adopted earth models are based on the seismic model
produced by gravity changes associated with solid earth PREM (Dziewonski and Anderson, 1981) and have a
deformation caused by the surface loading and rota- depth discretisation of 25 km on average. The radial
tional potential (the so-called ‘‘indirect effect’’). viscosity structure is more crudely depth parameterised
The term G M describes a spatially uniform height shift into three regions. To simulate the lithosphere, the
of the geoid. This term is added to ensure that the viscosity in the uppermost region is set to a relatively
system conserves surface ice/water mass. The change in high value of 1043 Pa s so that it acts elastically over
 ARTICLE IN PRESS
G.A. Milne et al. / Quaternary Science Reviews 24 (2005) 1183–1202 1189

typical GIA timescales (
103 yr). The second region between about 7 and 5 cal. kyr BP. The model prediction
constitutes most of the upper mantle and it extends from tracks the four data points that diverge from the general
the base of the lithosphere to the 670 km seismic trend between 7 and 8 cal. kyr BP. However, as
discontinuity. The third region is the lower mantle discussed in Section 2, two of these points are from a
extending from the bottom of the upper mantle to the single site and so may be influenced by local factors.
core-mantle boundary. The thickness of the model Also, the sea-level markers that these data are based on
lithosphere and the viscosity in regions 2 and 3 are were not levelled to mean sea-level and so they may
varied in the following analysis. For convenience, we contain significant error. This is illustrated by the fact
define a reference viscosity model that is broadly that the younger data do not trend to zero at present,
compatible with a number of recent viscosity inferences suggesting that the entire data set are biased 1–2 m too
(e.g., Peltier, 1974; Lambeck et al., 1990, 1998; high, in which case the model fit would be significantly
Lambeck, 1993; Forte and Mitrovica, 1996; Mitrovica, poorer.
1996; Wieczerkowski et al., 1999). This model is The data from Recife (D) show significant scatter and
characterised by a 96 km thick lithosphere, an upper so a high-quality fit is not expected. The general sea-
mantle viscosity of 5  1020 Pa s and a lower mantle level fall is captured relatively well by the model at this
viscosity of 1022 Pa s: site. In contrast to the results for Curacao (B) and
The ice loading model which is the primary input that Suriname (C), the model predicts sea levels that are too
drives the system is based on the global ICE-3G low at this site. The data from Rio de Janiero (E) show
deglaciation model (Tushingham and Peltier, 1991). A good agreement with the model predictions. We do not
number of revisions have been made to the original ICE- believe that the sea-level oscillations suggested by the
3G model. The most significant changes are the addition data during the late-Holocene are robust given the
of a glaciation phase and an increase in the volume and accuracy and precision of the data and so we interpret
melt chronology of the Laurentide component of the the model fit to be close to optimal. The largest misfit is
model in order to produce a good fit to the Barbados found at the two Patagonian sites, Strait of Magellan
sea-level record based on assuming the reference (G) and Beagle Channel (H). As mentioned in Section 2,
viscosity model described above. The Antarctic compo- these data are of relatively low quality and this is
nent of the model has also been revised to produce an reflected in the magnitude of the adopted error bars. The
accelerated melt rate in the early Holocene and to cease data from the Strait of Magellan provide a relatively
melting at 7 cal. kyr BP. sparse temporal record. At this site, the model predicts a
Fig. 3 shows predictions based on the reference model pronounced highstand of
7 m at 7 cal. kyr BP whereas
as well as the data at each of the site locations shown in the observations rule out a Holocene highstand of
Fig. 1. The model captures the general north–south greater than 2–3 m. The data from the Beagle Channel
trend illustrated by the data and described in Section 2. suggest a sea-level highstand around 6 cal. kyr BP as
A monotonic sea-level rise is predicted at both Jamaica large as
6 m: The model captures the timing of the
(A) and Curacao (B) whereas a clear sea-level highstand highstand relatively well but the magnitude is too low by
is predicted for Suriname (C) to Santa Catarina (F) several meters. Also, the earlier data are poorly fit by the
(moving southwards). The predictions for the Strait of model. It is important to note in this regard that the four
Magellan (G) and the Beagle Channel (H) show trends oldest dates are questionable since younger shells occur
that are more complex than at the other sites. (See next at higher elevations in the raised beach deposits that
sub-section for a discussion of the predicted signal at these data are based on, suggesting that some reworking
these sites.) The eustatic curve is also shown on each of older material has occurred.
plot. This illustrates the Holocene melt history of the The misfits at the two most southerly sites could be
reference ice model: a relatively rapid melt rate of due to a number of factors, in addition to poor data
7–8 mm/yr in the early Holocene slowing down to
quality. For example, tectonic processes are known to be
0:5 mm=yr between 7 and 5 cal. kyr BP with zero melt active in this region (Rabassa et al., 2000). Also, these
after 5 cal. kyr BP. sites are in close proximity to the Patagonian and
Even though the reference model has not been tuned Antarctic ice complexes and so the signal could be
to fit this data set the fit is generally of good quality. The significantly affected by errors in these components of
model provides the highest quality fits at the sites with the adopted ice model. These issues are explored further
the most accurate observations, Jamaica (A) and Santa below.
Catarina (F) (see Section 2). The sea-level prediction at
Curacao (B) is generally too high by 1–2 m, however, 3.2. Explaining the spatial trend
this discrepancy may be partly associated with sediment
compaction in the cores since these data are not from The general spatial trend displayed in both the
basal sediment layers (see Section 2). The prediction at observations and model predictions is a result of the
Suriname (C) also appears to be too high, particularly spatial variation in form and magnitude of the
 ARTICLE IN PRESS
1190 G.A. Milne et al. / Quaternary Science Reviews 24 (2005) 1183–1202

-10 -8 -6 -4 -2 0 -10 -8 -6 -4 -2 0

5 5

0 0

-5 -5

(A) (E)
-10 -10

5 5

0 0

-5 -5

(B) (F)
-10 -10
RSL (m)

10
5

0
0

-5
-5

(C) (G)
-10 -10
10
5

0
0

-5
-5

(D) (H)
-10 -10
-10 -8 -6 -4 -2 0 -10 -8 -6 -4 -2 0
Time (cal. kyr)

Fig. 3. Comparison of sea-level predictions to observations at each site location. The model prediction (solid line) is based on the reference model
described in the main text. The dashed lines show the eustatic curve for the adopted ice model. Note the significant deviation of the eustatic signal
from the total predicted signal at each site. The grey lines shows the predictions based on a model that is the same as the reference model except that
the lithospheric thickness is reduced to 71 km. This revised model provides the best fit of those considered in the current analysis to the entire data set
(see Section 3.3).
 ARTICLE IN PRESS
G.A. Milne et al. / Quaternary Science Reviews 24 (2005) 1183–1202 1191

components of the RSL signal associated with the ice

and ocean surface mass flux and the changing rotational 10
potential. In this sub-section we plot these various
components of the RSL signal to illustrate this point.
It is convenient to begin by first separating the total
predicted signal into spatially uniform and spatially 0
varying components. As an example, we do this for the
prediction at Jamaica in Fig. 4. The total sea-level
prediction (equivalent to the solid curve shown in Fig. -10
3A) is the solid black line in the top frame. The other
curves combine to give the total signal. The spatially
uniform component of the signal is a combination of the -20
meltwater or eustatic signal (dashed line) and the non-
eustatic signal (dotted line) (see Eq. (2)). These curves
are, by definition, valid for all locations. As discussed
-30
above, the eustatic signal in the model shows a relatively
rapid sea-level rise until 7 cal. kyr BP at which time the

RSL (m)
rate of meltwater addition slows down and is defined to
be zero after 5 cal. kyr BP. The non-eustatic signal is of 9
the opposite sign to the eustatic signal and contributes a
sea-level fall during the Holocene. This component of 6
the spatially uniform signal is dominated by the so-
3
called syphoning mechanism (e.g., Mitrovica and Milne,
2002). This process is driven by deformation of the solid 0
earth and so the signal decreases smoothly with time due
to the viscous component of earth deformation. The -3
geoid fall associated with syphoning and its important
-6
contribution to the prediction of sea-level highstands
during the mid-Holocene has been known for a number -9
of years (e.g., Clark et al., 1978; Nakada and Lambeck,
1989). -12
The spatially varying component of the predicted sea-
-15
level signal at Jamaica is shown by the dash–dotted line.
This signal is represented by the first two terms on the -18
RHS of Eq. (1). This component shows that the local
perturbation to sea-level is that of a monotonic rise of -10 -8 -6 -4 -2 0

20 m during the Holocene. It is the spatial form and Time (cal. kyr)
magnitude of this component of the signal that
Fig. 4. The top frame shows the sea-level prediction at Jamaica (site
determines how the total sea-level signal will vary A) (solid line) based on the reference model and the components that
geographically. sum to give this signal: eustatic (dashed line), non-eustatic spatially
In order to further explore the cause of the local value uniform component (dotted line), and the GIA-induced spatially
of the spatially varying signal at Jamaica, we show in the varying component (dash–dotted line). The lower frame shows the
bottom frame of Fig. 4 the contribution of the changes GIA-induced spatially varying component (solid line) and the
contributions to this from the ice mass redistribution (dashed line),
in ice (dashed line) and ocean (dotted line) mass the ocean mass redistribution (dotted line) and the GIA-induced
distribution and the changing rotational potential perturbation to the rotational potential (dash–dotted line).
(dash–dotted line) to the total local signal (solid line).
From inspection of Fig. 4 (lower frame) it is immediately
apparent that the ice-induced signal is the dominant The rotation- and ocean-induced signals also con-
contributor to the local signal at Jamaica. The rise in tribute a sea-level rise over the Holocene and therefore
sea-level associated with the ice-induced signal is add to the ice-induced effect. At this site the ocean-
dominated by subsidence of the ocean floor. This is induced signal is also largely driven by the solid surface
because the Caribbean is located on the peripheral bulge subsidence caused by the ocean loading in this region
of the Laurentide ice sheet. The change in slope of the following the last glacial maximum. The rotation-
ice-induced signal at 7 cal. kyr BP is caused by the slow induced signal associated with GIA is driven by true
down of global ice melting at this time and the polar wander excited by the ice–ocean mass flux and the
consequent direct effect this has on the geopotential. consequent solid earth response. The spatial and
 ARTICLE IN PRESS
1192 G.A. Milne et al. / Quaternary Science Reviews 24 (2005) 1183–1202

temporal form of this signal have been described in Magellan site (G) gives the best example in this study,
detail elsewhere (e.g., Milne and Mitrovica, 1998). In the with an ocean-induced sea-level fall of almost 9 m
region of study, this GIA forcing produces a sea-level during the Holocene.
rise for sites north of the equator and a sea-level fall for The rotation-induced signal has the smallest magni-
sites south of the equator during the Holocene. The tude and the longest wavelength of the three GIA
largest magnitude of the signal is predicted at mid- component signals (Fig. 5D). At 7 cal. kyr BP, this
latitudes and along the great circle of longitude
75 W: signal has a gradient of 2–3 m over the region under
The spatial pattern of RSL across the study region at consideration. It adds to the ice-induced signal above
7 cal. kyr BP is shown in Fig. 5A. This time was chosen the equator and opposes this signal below the equator.
since the reduction in model meltwater flux during this
period produces a highstand at most sites. Note that 3.3. Estimating the Holocene meltwater signal
Fig. 5A shows a prediction of the total RSL signal and
so includes both spatially uniform and spatially varying Determining the eustatic signal is not straightforward
components. The net spatially uniform signal contri- for a variety of reasons. As mentioned above, there are a
butes
3 m at this time (sum the dotted and dashed number of processes that can contribute to vertical
lines in Fig. 4A). The signal shows a relatively complex motion of both the land and the ocean surfaces and thus
pattern with an amplitude range approaching 30 m. A produce a significant non-eustatic component of RSL
highstand is predicted at this time for the majority of the change (e.g., tectonics, glacio- and hydro-isostasy). The
South American coastline (the white contour marks zero influence of these non-eustatic components must be
RSL), with a few exceptions along the most northerly removed from the observations in order to obtain an
part of the continent and at the very southern tip of the accurate inference of the climatic signal. For example,
Patagonian peninsula. The low sea levels at these the influence of tectonics is minimised by considering
northern and southern locations are due, largely, to data from tectonically stable locations. Of the data
the influence of the Laurentide and Antarctic ice sheets, considered here, those from sites A, B, G and H are the
respectively. most likely to have been influenced by tectonic processes
This is illustrated in Fig. 5B which shows the during the Holocene period.
contribution of the ice-induced signal to the spatially Our first step in estimating the eustatic signal is to
varying signal in Fig. 5A (the spatially uniform determine a set of GIA model parameters that give an
component is not included in Figs. 5B–D). The ice- optimal fit to the observations. The predictions based on
induced signal, which includes perturbations to both the this model can then be adopted to remove the total non-
geoid (direct and indirect effects) and the solid surface, eustatic GIA component of the signal from the data so
produces a relatively long wavelength signal with a that the ‘corrected’ data can be interpreted directly as
dominant north–south gradient. This signal becomes the eustatic signal. The reference model produces a good
progressively more negative as the distance to a major fit at four of the eight sites (see Fig. 3 and above
ice center decreases. Subsequent to 7 cal. kyr BP, during discussion), however, it is not clear if a better fit may be
which there is minimal melt, the influence of peripheral obtained by varying key model parameters. We explore
bulge subsidence dominates the signal. In contrast, note this issue by considering a suite of earth models with
that the relatively small-scale glaciation in Patagonia parameter values that represent a range within which the
results in a sea-level high at this time due to the true average viscosity for that region probably lies. We
dominant crustal uplift in this region from 7 cal. kyr BP define the following minimum and maximum values:
to present. lithospheric thickness (71–120 km), upper mantle visc-
The ocean-induced signal produces spatial gradients osity (0.1–1  1021 Pa s) and lower mantle viscosity
that are perpendicular to the shoreline, due to the (1–50  1021 Pa s). These parameter ranges are loosely
geometry of the ocean loading (Fig. 5C). This loading based on the spread of values inferred for these
leads to a broad uplift of the continents that has been parameters found in the recent literature. In order to
termed continental levering (e.g., Clark et al., 1978). ensure that the predictions for each viscosity model are
Relatively large gradients are predicted near the realistic, the ice model has been tuned in each case to fit
shorelines and so a distinctly different RSL signal can the Barbados sea-level record. This ensures an impor-
be predicted between two sites significantly displaced tant level of consistency for each earth/ice model pair.
relative to this orientation. Sites located on peninsulas The predictions for each of these new models at all
or small islands near a major continental mass (such as eight data sites are shown in Fig. 6. The predictions at
Curacao in this study) commonly experience a sea-level sites A, G and H are most sensitive to variations in earth
rise during the Holocene due to this effect. In contrast, model viscosity structure. This is not surprising since
sites located inland relative to the crude, long-wave- these sites are influenced the greatest by solid surface
length trend of the coastline (e.g. bays or river estuaries), motion in response to surface loading (see Fig. 5). The
can display a significant sea-level fall. The Strait of results for site B show the least sensitivity to variations
 ARTICLE IN PRESS
G.A. Milne et al. / Quaternary Science Reviews 24 (2005) 1183–1202 1193

Fig. 5. Sea-level predictions at 7 cal. kyr BP based on the reference model. (A) shows the total signal and (B–D) show, respectively, the ice-, ocean-
and rotation-induced components of the spatially varying signal only. The locations of data sites are shown by white circles.
 ARTICLE IN PRESS
1194 G.A. Milne et al. / Quaternary Science Reviews 24 (2005) 1183–1202

-10 -8 -6 -4 -2 0 -10 -8 -6 -4 -2 0
5 5 5

0 0 0

-5 -5 -5

(A) (E)
-10 -10 -10
-10 -8 -6 -4 -2 0 -10 -8 -6 -4 -2 0

5 5 5

0 0 0

-5 -5 -5

(B) (F)
-10 -10 -10
RSL (m)

-10 -8 -6 -4 -2 0 -10 -8 -6 -4 -2 0

10 10
5

5 5
0
0 0

-5
-5 -5

(C) (G)
-10 -10 -10
-10 -8 -6 -4 -2 0 -10 -8 -6 -4 -2 0
10 10
5

5 5
0
0 0

-5
-5 -5

(D) (H)
-10 -10 -10
-10 -8 -6 -4 -2 0 -10 -8 -6 -4 -2 0
Time (cal. kyr)

Fig. 6. Sea-level predictions at all eight data sites for a suite of seven models that sample a range of earth model parameters. Each pair of lines
represent a prediction based on a model in which one parameter has been varied in the reference model. The reference model prediction is shown by
the solid line. The dashed lines indicate a change in lithospheric thickness from 71 km (grey) to 120 km (black). The dotted lines indicate a change in
upper mantle viscosity from 1020 Pa s (grey) to 1021 Pa s (black). The dash–dotted lines indicate a change in lower mantle viscosity from 1021 Pa s
(grey) to 5  1022 Pa s (black). For each earth model, the ice-model was tuned so that the sea-level prediction at Barabdos remained consistent with
the observations. The observational data are also shown (in light gray shading) to facilitate comparison with the model predictions.
 ARTICLE IN PRESS
G.A. Milne et al. / Quaternary Science Reviews 24 (2005) 1183–1202 1195

in the viscosity structure. This is unexpected since the eustatic rise included in the model is consistent with the
amount of deformation at this location is comparable to data from sites E and H. The match is less good at site
that at site A. However, the sensitivity at this site has A, where the observations suggest a lower melt rate
been somewhat reduced due to the joint variation of between 8 and 7 cal. kyr BP. The data from site C also
viscosity and melt model. In terms of ice loading, this suggest a slower melt rate during this period but the
site experiences a similar signal to that at Barbados and scatter in the observations limit this interpretation. Also,
so the changes made to the ice model in order to the model fit at this site is poor and so the results are less
maintain the quality of fit at Barbados have reduced the reliable. We conclude that there is no strong evidence to
variation in the predicted signal at Curacao. suggest that the model eustatic signal for the early
The results shown in Fig. 6 can be employed to choose Holocene is significantly incorrect.
an optimal GIA model with which to estimate the Data from sites A and C provide the only cohe-
eustatic signal. The primary criteria we are seeking at rent record of sea-level change during the period of
this stage is a model that can account for the spatial dramatic melt-rate decline. The observational con-
trend evident in the data. A spatially uniform discre- straints at each site are relatively poor due to scatter
pancy for a given earth viscosity model can be in the data or a lack of data points. However, at both
accounted for by changing the eustatic component of sites, a marked reduction in melt rate at 7 cal. kyr BP, or
the ice model. We first focus on the model fits at the best slightly earlier, is suggested and so is consistent with
quality data sites, A and F. Three models produce a the model.
good fit to the data at these sites: the reference model From the mid-Holocene to present, there is no strong
and those with a relatively thick (120 km) or thin (71 km) evidence to suggest that there has been significant net
lithosphere relative to the reference model. The earth melt (or growth),
1 m eustatic equivalent, of ice
models with values of upper mantle viscosity that reservoirs. The best constraint in this regard comes
deviate from the reference model lie beneath the data from data at site F. These data are of high quality (see
at site A and above the data at site F, indicating that Section 2) and the corrected data trace out a remarkably
these models do not capture the spatial trend in the data flat signal from
6 cal. kyr BP to present. This trend is
and so can be ruled out. A similar argument can be quite robust as it does not significantly vary for the
made to exclude the viscosity models that exhibit range of ice/earth model pairs considered. Data from
different values of lower mantle viscosity compared to sites A, B, D, E and H are consistent with this
the reference model. result, although the data at these sites exhibit greater
Of the remaining models, the reference model scatter. We interpret the oscillations in the signal at site
produces the lowest w2 results at sites A and F. However, E to be within the precision of the data and so not
the model with a relatively thin lithosphere improves the significant (see Section 2). There is no substantial
fit at sites B and H without significantly reducing the evidence for or against the small melt (
0:5 mm=yr)
quality of fit at sites A and F. In fact, this model gives between 7 and 5 cal. kyr BP in the model. Results from
the lowest overall w2 results for all eight sites (see Fig. 3). sites A, C and H are limited due to scatter. There is
We therefore choose this model to correct the observa- certainly no evidence for a significantly larger melt rate
tions for non-eustatic GIA effects so that the remaining during this period.
signal can be interpreted as the eustatic signal. These There are no regionally consistent deviations of the
results are shown in Fig. 7. Note that, in contrast to the corrected data from the model eustatic curve and so we
predictions shown in Figs. 3 and 6, which were conclude that the model is approximately correct for the
calculated at locality averaged positions, the specific Holocene period based on these data and the range of
location for each index point was used to produce the earth and ice models considered. Data from sites C and
model-corrected data. At sites where the model predicts G are the least well accounted for by the model.
the data well, the resulting corrected trend is more Inspection of the results in Fig. 6 indicate that data from
coherent with less data scatter. The results is Fig. 7F are sites C and G would be best fit by, respectively, the
a good example (compare to Fig. 3F). model with a high lower mantle viscosity and a model
A key point to discern from the results in Fig. 7 is with a relatively thick lithosphere. However, of the suite
whether there is a regionally consistent discrepancy of model predictions considered, there is no single
between the model eustatic and that suggested by the earth–ice model combination that can fit all of the
data. Such a residual could be interpreted directly as the observations well. The large residual at these sites could
required correction to the eustatic signal of the model. be associated with a number of limitations related to
As expected, the corrected data mark out a eustatic both the observations and the model predictions. For
trend similar to that of the model at sites where the example, the influence of lateral variations in earth
model-data comparison is good (A, B, E, F and, to a structure or the influence of vertical tectonics, which are
lesser extent, D and H). Of the sites that include data not accounted for the in model, may be important at
before 7 cal. kyr BP (A, C, E and H), the relatively rapid these locations.
 ARTICLE IN PRESS
1196 G.A. Milne et al. / Quaternary Science Reviews 24 (2005) 1183–1202

-10 -8 -6 -4 -2 0 -10 -8 -6 -4 -2 0
5 5

0 0

-5 -5

-10 -10

-15 -15
(A) (E)

5 5

0 0

-5 -5

-10 -10

-15 -15
(B) (F)
RSL (m)

5 5

0 0

-5 -5

-10 -10

-15 -15
(C) (G)

5 5

0 0

-5 -5

-10 -10

-15 -15
(D) (H)

-10 -8 -6 -4 -2 0 -10 -8 -6 -4 -2 0

Time (cal. kyr)

Fig. 7. Sea-level observations at each site corrected for the non-eustatic GIA contribution of the best-fitting model (see main text). The eustatic
component associated with the best-fitting model is shown as a dashed line.
 ARTICLE IN PRESS
G.A. Milne et al. / Quaternary Science Reviews 24 (2005) 1183–1202 1197

4. Summary due to the continental levering effect. This process also

plays a key role in governing the spatial variation
We have critically assessed both published and between sites but this is less apparent given the more
unpublished Holocene sea-level data from the Carib- complex spatial form and shorter wavelength of the
bean and the Atlantic coast of South America in order predicted signal. The rotation-induced component has
to produce a regional data set of sufficient quality for the smallest magnitude of the three GIA signals but
quantitative modelling purposes. A large portion of the contributes significantly to the observed trend due to its
data considered are of relatively poor quality and so long wavelength and relatively large north–south
were not used in the modelling analysis. The edited data gradient along the longitude meridian the data are
set includes observations obtained at 8 localities, from distributed.
Jamaica in the north to the Beagle Channel in the south. Predictions of the non-eustatic GIA signal were
The observations have a large north–south extent and, removed from the observations so that the residual
consequently, exhibit a clear spatial trend in the signal could be used to infer a Holocene eustatic trend.
Holocene sea-level signal. The observations were com- The results indicate that the eustatic component of the
pared to predictions based on a realistic model of reference model is consistent with the bulk of the data.
glaciation-induced sea-level change in order to under- A relatively rapid and uniform sea-level rise of 7–8 mm/
stand the cause of the observed spatial trend and to yr during the early Holocene is consistent with the small
estimate a eustatic signal for the Holocene. amount of data available for this period. The data also
The data were compared to 7 different GIA models suggest a decrease in melt rate around 7 cal. kyr BP, with
that defined a range of viscosity parameters consistent a global ice volume change of
1 m eustatic
with the significant uncertainty in these values. Each equivalent from this time to the present.
model was tuned to fit the Barbados sea-level record to
ensure a level of consistency for each set of predictions.
The model predictions successfully capture the general Acknowledgements
RSL trend at most sites to within data scatter and
precision. An earth model with a 71 km thick litho- We would like to thank Jerry Mitrovica and Kurt
sphere, and upper and lower mantle viscosities of 5  Lambeck for constructive reviews of the original manu-
1020 and 1022 Pa s; respectively, produces the best fit to script. We also thank Bogusia Klosowska and Orson
the data at all 8 sites. These values are consistent with van de Plasche for making the data from Curacao
those inferred in a number of recent analyses. available to us. This paper is a contribution to IGCP
Separating the predicted signal into spatially varying Project 495 ‘‘Quaternary Land–Ocean Interactions’’ and
components associated with the ice, ocean and rota- the INQUA working group on ‘‘Coastal and Marine
tional GIA forcings indicates that the ice-induced signal Processes’’.
is the dominant contributor to the observed spatial
trend. In particular, crustal subsidence in the Caribbean
caused by the melting of the North American ice sheets Appendix A
results in no predicted Holocene highstand in this
region, consistent with observations at Jamaica and The relative sea-level index points used in this analysis
Curacao. Also, the distinct signal observed in Patagonia are shown in Table A1. The ages of the index points are
is a result of the direct effect of the melting West given in radiocarbon years before present and are listed
Antarctic ice sheet on the geoid during the early with the maximum and minimum age range for the
Holocene combined with the dominant role of crustal calibrated date (two standard deviations). MSL refers to
subsidence associated with this ice sheet subsequent to the reconstructed altitude of the index point with respect
this time. These effects combine to produce a Holocene to present-day MSL. Note that the latitude and long-
highstand
2 kyr later than predicted elsewhere. The itude values for the data from the Strait of Magellan and
ocean-induced signal produces large gradients perpen- Beagle Channel are estimated from maps in Rabassa et
dicular to and in the vicinity of continental shorelines al. (2000) and Porter et al. (1984).
 1198
Table A1

Site Latitude Longitude Laboratory 14C age Age error Cal max Cal min MSL MSL error  (m) Material dated Reference
Code (yr) (yr) (yr) (yr) (m)

Santa Catarina, Brazil

Ponta de Itapiruba 1 281200 481410 CENA 181 450 65 254 0 0.2 0.5 Vermitid Angulo et al. (1999)
Ponta de Itapiruba 1 281200 481410 CENA 182 3920 70 4085 3694 1.05 0.5 Vermitid Angulo et al. (1999)
Ponta de Itapiruba 1 281200 481410 CENA 183 4340 70 4650 4281 1.75 0.5 Vermitid Angulo et al. (1999)
Ponta de Itapiruba 1 281200 481410 CENA 185 570 65 316 0 0.2 0.5 Vermitid Angulo et al. (1999)
Ponta do Ji 281250 481440 CENA 189 1840 100 1600 1212 0.55 0.5 Vermitid Angulo et al. (1999)
Ponta do Ji 281250 481440 CENA 190 980 65 658 482 0.4 0.5 Vermitid Angulo et al. (1999)
Ponta do Ji 281250 481440 CENA 191 4060 70 4286 3885 2 0.5 Vermitid Angulo et al. (1999)
Pontada Passagem da Barra 281300 481440 CENA 192 4600 70 4968 4617 1.95 0.5 Vermitid Angulo et al. (1999)
Ponta da Passagem da Barra 281300 481440 CENA 193 2570 70 1356 2062 1 0.5 Vermitid Angulo et al. (1999)

G.A. Milne et al. / Quaternary Science Reviews 24 (2005) 1183–1202

Ponta da Passagem da Barra 281300 491440 CENA 194 1200 70 902 635 0.2 0.5 Vermitid Angulo et al. (1999)
Ponta da Ilhorta 281320 481450 CENA 186 1210 70 908 641 0.2 0.5 Vermitid Angulo et al. (1999)
Ponta da Ilhorta 281320 481450 CENA 187 1580 70 1273 965 1.05 0.5 Vermitid Angulo et al. (1999)
Ponta da Ilhorta 281320 481450 CENA 188 2060 70 1807 1474 1.1 0.5 Vermitid Angulo et al. (1999)
Ponta da Galheta 281340 481470 CENA 176 1690 90 1404 1052 0.8 0.5 Vermitid Angulo et al. (1999)
Ponta da Galheta 281340 481470 CENA 177 2210 70 1964 1625 1.15 0.5 Vermitid Angulo et al. (1999)
Ponta da Galheta 281340 481470 CENA 178 2820 70 2737 2341 0.75 0.5 Vermitid Angulo et al. (1999)
281340 481470

ARTICLE IN PRESS
Ponta daGalheta CENA 179 5410 80 5931 5595 2.1 0.5 Vermitid Angulo et al. (1999)
Ponta da Galheta 281340 481470 CENA 180 2910 70 2795 2470 2 0.5 Vermitid Angulo et al. (1999)
Cape of Santa Marta 281360 481480 CENA 169 1610 110 1360 928 1.05 0.5 Vermitid Angulo et al. (1999)
Cape of Santa Marta 281360 481480 CENA 170 2430 520 3349 928 0.85 0.5 Vermitid Angulo et al. (1999)
Cape of Santa Marta 281360 481480 CENA 171 2340 80 2144 1777 0.95 0.5 Vermitid Angulo et al. (1999)
Cape of Santa Marta 281360 481480 CENA 172 1730 70 1629 928 1.55 0.5 Vermitid Angulo et al. (1999)
Cape of Santa Marta 281360 481480 CENA 173 1180 70 888 624 0.2 0.5 Vermitid Angulo et al. (1999)
Cape of Santa Marta 281360 481480 CENA 174 2080 60 1807 1513 0.35 0.5 Vermitid Angulo et al. (1999)
Cape of Santa Marta 281360 481480 CENA 175 1370 60 1042 780 0.25 0.5 Vermitid Angulo et al. (1999)
Praia de Armacao 251470 481360 GX-14061 1045 75 725 506 2.55 0.5 Vermitid Angulo (1989)
Sao Fco. Do Sul 261130 481290 GX-14060 4015 90 4279 3805 3.05 0.5 Vermitid Angulo (1989)

Rio de Janiero
Lagoa do Padre 221570 421450 KL 2222.01 6800 110 7832 7432 2.5 0.25 Herbaceous peat Ireland (1988)
Lagoa do Padre 221570 421450 KL 2222.02 7150 120 8177 7683 2.75 0.25 Herbaceous peat Ireland (1988)
Itaipu 221570 431020 KL 2375.01 6840 100 7838 7488 1.95 0.25 Herbaceous peat Ireland (1988)
Itaipu 221570 431020 KL 2376.01 6860 110 7928 7488 1.3 0.25 Herbaceous peat Ireland (1988)
Itaipu 221570 431020 KL 2226.02 7110 110 8154 7675 1.55 0.25 Herbaceous peat Ireland (1988)
Itaipu 221570 431020 KL 2374.01 7140 110 8168 7684 1.65 0.25 Herbaceous peat Ireland (1988)
Itaipu 221570 431020 KL 2226.03 7810 75 8929 8404 1.77 0.25 Herbaceous peat Ireland (1988)
Itaipu 221570 431020 KL 23745/02 7970 100 9062 8481 2.4 0.25 Herbaceous peat Ireland (1988)
Lagoa do Padre 221560 421460 KL 2223.02 2590 65 2780 2363 1.05 0.25 Herbaceous peat Ireland (1988)
Lagoa do Padre 221560 421460 KL 2223.01 2270 55 2351 2120 1.02 0.25 Herbaceous peat Ireland (1988)
Itaipu-Acu 221570 421540 KL 2225.01 2460 55 2736 2346 0.42 0.25 Herbaceous peat Ireland (1988)
Itaipu-Acu 221570 421540 KL 2225.02 2700 60 2918 2737 0.32 0.25 Herbaceous peat Ireland (1988)
Tarituba 231020 441360 Bah-478 975 80 672 460 0.7 0.4 Vermitid Martin and Suguio (1978)
Parati-Mirim 231140 441370 Bah-482 1490 80 1223 894 1 0.4 Vermitid Martin and Suguio (1978)
Mangaratiba 221580 441020 Bah-499 1630 65 1295 1046 0.8 0.4 Vermitid Martin and Suguio (1978)
Ilha Grande 231100 441100 Gif-1060 1670 100 1403 999 1.5 0.5 Vermitid Delibras and Laborel (1969)
Praia de Mambucaba 231020 441330 Bah-468 1840 90 1569 1229 1.5 0.5 Vermitid Martin and Suguio (1978)
Ilha do Araujo 231090 441410 Bah-470 2300 85 2115 1705 1.4 0.4 Vermitid Martin and Suguio (1978)
Cabo Frio 221540 421000 Gif-1935 2400 95 2295 1812 2 0.5 Vermitid Delibras and Laborel (1969)
Corao Grande 221550 431500 Bah-473 2595 132 2683 1938 1.6 0.4 Vermitid Martin and Suguio (1978)
Arraial do Cabro 221560 421010 Bah-1298 2680 180 2782 1925 1.8 0.4 Vermitid Martin and Suguio (1978)
Frade 221580 441260 Bah-465 2695 130 2734 2069 1.5 0.4 Vermitid Martin and Suguio (1978)
Mangaratiba 221570 441020 Bah-472 3255 100 3338 2799 1.7 0.4 Vermitid Martin and Suguio (1978)
Buzios 221460 411530 Bah-1306 3360 180 3390 2984 1.8 0.4 Vermitid Martin and Suguio (1978)
Buzios 221440 411510 Bah-1304 3420 180 3697 2810 2.1 0.4 Vermitid Martin and Suguio (1978)
Ilha Grande 231100 441100 Gif-1059 3420 110 3545 2985 3 0.5 Vermitid Delibras and Laborel (1969)
Ponta do Retiro 211210 401570 Bah-1008 3620 150 3888 3168 3 0.4 Vermitid Martin and Suguio (1978)
Cabo Frio 221540 421000 Gif-1934 3900 110 4156 3583 3 0.5 Vermitid Delibras and Laborel (1969)

Recife
Cabo San Agostinho 81200 341560 Shell-A17 1190 130 976 523 1.6 0.5 Vermitid van Andel and Laborel (1964)
Cabo San Agostinho 81200 341560 Shell-A21 1750 170 1681 943 1.4 0.5 Vermitid van Andel and Laborel (1964)
Cabo San Agostinho 81200 341560 Bah-1236 2010 160 1938 1254 4.83 0.5 Vermitid Dominguez et al. (1990)
Gaibu 81190 341560 Bah-1237 2570 160 2702 1856 2.56 0.5 Vermitid Dominguez et al. (1990)
Gaibu 81190 341560 Bah-1222 2670 170 2761 1936 1.16 0.5 Vermitid Dominguez et al. (1990)
Cabo San Agostinho 81200 341560 Shell-A16 2790 150 2843 2132 2.2 0.5 Vermitid van Andel and Laborel (1964)
Cabo San Agostinho 81200 341560 Shell-A22 3660 170 3986 3168 2.6 0.5 Vermitid van Andel and Laborel (1964)

G.A. Milne et al. / Quaternary Science Reviews 24 (2005) 1183–1202

Gaibu 81190 341560 Bah-1221 3870 170 4296 3400 4.36 0.5 Vermitid Dominguez et al. (1990)

Suriname and Guyana

Costeri 1 51210 551240 GrN 7810 545 35 634 513 1.11 1 Peat Roeleveld and Van Loon (1979)
Costeri 1 51210 551240 GrN 7811 7620 60 8538 8340 6.12 1 Peaty clay Roeleveld and Van Loon (1979)
Tibiti 1 51300 551530 GrN 7812 6410 50 7427 7249 0.29 1 Peat Roeleveld and Van Loon (1979)
51300 551530

ARTICLE IN PRESS
Tibiti 1 GrN 7813 6580 70 7585 7330 0.79 1 Peat Roeleveld and Van Loon (1979)
Ricanau 1 51370 541210 GrN 7814 5810 60 6775 6450 0.665 1 Clayey peat Roeleveld and Van Loon (1979)
Ricanau 1 51370 541210 GrN 7815 5660 60 6623 6305 0.87 1 Peaty clay Roeleveld and Van Loon (1979)
Ricanau 1 51370 541210 GrN 7816 6090 70 7207 6747 1.6 1 Peaty clay Roeleveld and Van Loon (1979)
Ricanau 1 51370 541210 GrN 7817 6580 80 7608 7321 2.12 1 Clayey peat Roeleveld and Van Loon (1979)
Ricanau 1 51370 541210 GrN 7818 6845 50 7786 7587 3.325 1 Peaty clay Roeleveld and Van Loon (1979)
Meursweg 1 51370 551120 GrN 7819 560 35 647 517 1.65 1 Peat Roeleveld and Van Loon (1979)
Meursweg 1 51370 551120 GrN 7820 4190 60 4857 4530 1.23 1 Clayey peat Roeleveld and Van Loon (1979)
Meursweg 1 51370 551120 GrN 7821 6360 70 7426 7097 2.135 1 Peat Roeleveld and Van Loon (1979)
Meursweg 1 51370 551120 GrN 7822 7140 60 8109 7794 3.825 1 Peaty clay Roeleveld and Van Loon (1979)
Bigi Poika II 51270 551280 GrN 7823 2020 45 2114 1874 1.36 1 Clayey peat Roeleveld and Van Loon (1979)
Bigi Poika II 51270 551280 GrN 7824 6350 60 7422 7100 0.58 1 Clayey peat Roeleveld and Van Loon (1979)
Coppename West 1 51330 561000 GrN 7825 6650 70 7660 7426 2.43 1 Clayey peat Roeleveld and Van Loon (1979)
Coppename West 1 51330 561000 GrN 7826 6680 110 7717 7340 2.725 1 Peat Roeleveld and Van Loon (1979)
Coppename West 1 51330 561000 GrN 7827 7040 80 8009 7687 4.145 1 Clayey peat Roeleveld and Van Loon (1979)
Burnside 1 51530 561240 GrN 7878 2140 50 2310 1953 1.075 1 Peat Roeleveld and Van Loon (1979)
Ogle Bridge ? ? GrN 3058 8590 65 9700 9486 19.4 1 Peaty clay Roeleveld and Van Loon (1979)
Cupido 51390 561370 GrN 4517 6360 70 7426 7079 1.6 1 Peat Roeleveld and Van Loon (1979)
Cupido 51390 561370 GrN 4518 5130 70 6165 5720 0 1 Peat Roeleveld and Van Loon (1979)
Matawaribo 51330 551110 GrN 4847 7240 100 8284 7844 4.62 1 Peaty clay Roeleveld and Van Loon (1979)
Baboenhol 51160 551040 GrN 5731 6310 45 7317 7096 0.5 1 Peaty clay Roeleveld and Van Loon (1979)

Jamaica
Negril 181010 781100 Lu-2084 3560 60 4063 3690 1.1 1 Sedge peat Digerfeldt and Hendry (1987)
Negril 181010 781100 Lu-2049 4370 60 5262 4833 1.75 1 Sedge peat Digerfeldt and Hendry (1987)
Negril 181010 781100 Lu-2050 4450 60 5305 4865 2.1 1 Sedge peat Digerfeldt and Hendry (1987)
Negril 181010 781100 Lu-2041 5100 60 5987 5664 3.35 1 Mangrove peat Digerfeldt and Hendry (1987)
Negril 181010 781100 Lu-2028 5680 60 6638 6310 4.12 1 Swamp forest peat Digerfeldt and Hendry (1987)
Negril 181010 781100 Lu-2026 6960 70 7939 7664 8.8 1 Sedge peat Digerfeldt and Hendry (1987)
Negril 181010 781100 B-2399 6300 80 7420 7002 8.2 1 Peat Digerfeldt and Hendry (1987)
Negril 181010 781100 Lu-2068 6220 70 7317 6809 6.1 1 Peat Digerfeldt and Hendry (1987)
Black River 181080 771380 Lu-2088 480 45 617 471 0.35 1 Sedge peat Digerfeldt and Hendry (1987)
Black River 181080 771380 Lu-2069 3590 60 4086 3698 1.05 1 Swamp forest peat Digerfeldt and Hendry (1987)
Black River 181080 771380 Lu-2077 4140 60 4835 4446 1.65 1 Swamp forest peat Digerfeldt and Hendry (1987)

1199
Black River 181080 771380 Lu-2071 4410 60 5291 4847 1.7 1 Transitional peat Digerfeldt and Hendry (1987)
 1200
Table A1 (continued )

Site Latitude Longitude Laboratory 14C age Age error Cal max Cal min MSL MSL error  (m) Material dated Reference
Code (yr) (yr) (yr) (yr) (m)

Black River 181080 771380 Lu-2070 5470 60 6401 6116 3.2 1 Swamp forest peat Digerfeldt and Hendry (1987)

Beagle Channel
Lago Roca 1 541460 681340 AC-1060 5920 90 5861 5473 5.78 2 Marine shells Rabassa et al. (2000)
Alakush 541560 681340 AC-0937 4440 120 4868 4272 2.38 2 Marine shells Rabassa et al. (2000)
Lago Roca 2 541460 681340 NZ-7730 7518 58 8104 7845 1.33 2 Estuarine marine shells Rabassa et al. (2000)
Rio Ovando 541460 681340 SI-6735 4425 55 4774 4434 0.48 2 Marine shells Rabassa et al. (2000)
Rio Ovando Camping 541460 681340 Pta 7691 7500 80 8123 7784 0.38 2 Estuarine marine shells Rabassa et al. (2000)
Nacientes Rio Ovando 541460 681340 Pta 7573 4160 45 4358 4093 0.38 2 Marine shells Rabassa et al. (2000)
Lapataia 3 541460 681340 SI-6739 5800 65 6342 6033 0.67 2 Marine shells Rabassa et al. (2000)

G.A. Milne et al. / Quaternary Science Reviews 24 (2005) 1183–1202

Lapataia 2 541460 681340 SI-6738 7620 70 8211 7925 0.72 2 Estuarine marine shells Rabassa et al. (2000)
Lapataia 1 541560 681340 SI-6737 8240 60 8910 8575 0.97 2 Estuarine marine shells Rabassa et al. (2000)
Bahia Ensenada 541530 681200 Pa-1012 2120 45 1813 1585 0.62 2 Marine shells Rabassa et al. (2000)
Bahia Golondrina 541530 681200 AECV 877 Cc 5460 110 6082 5596 7.38 2 Marine shells Rabassa et al. (2000)
Punta Pinguinos 1 541530 681200 L-1016C 5430 270 6400 5220 5.88 2 Marine shells Rabassa et al. (2000)
Punta Pinguinos 2 541530 681200 L-1016B 1400 300 1553 445 0.12 2 Marine shells Rabassa et al. (2000)
Ushuaia 541530 681200 AECV 876 Cc 5160 130 5835 5246 5.38 2 Marine shells Rabassa et al. (2000)

ARTICLE IN PRESS
Playa Larga 3 541490 681140 Pa-1015 4190 60 4412 4098 2.38 2 Marine shells Rabassa et al. (2000)
Playa Larga 2 541490 681140 Pa-1016 3095 60 3002 2742 0.38 2 Marine shells Rabassa et al. (2000)
Punta Parana 541530 671530 Pta. 7686 4370 70 4768 4333 3.38 2 Marine shells Rabassa et al. (2000)
Bahia Brown 541530 671310 Pa-1010 2970 70 2896 2456 0.38 2 Marine shells Rabassa et al. (2000)
Isla Gable 541530 671240 AECV 648 Cc 4790 100 5552 5027 2.38 2 Marine shells Rabassa et al. (2000)
Cutalataca 541530 671240 Pa-1009 2770 50 2662 2349 1.62 2 Marine shells Rabassa et al. (2000)
Peninsula Gusano 541530 671280 QL-1652 4600 30 4865 4732 0.93 2 Marine shells Porter et al. (1984)
Punta Piedra Buena 541530 671280 QL-1653 1470 30 1071 941 1.97 2 Marine shells Porter et al. (1984)
Rio Varela 541530 671170 Pta 7581 6290 70 6903 6573 3.88 2 Estuarine marine shells Rabassa et al. (2000)
Playa Larga 541490 681140 Pa-1018 5615 120 6274 5726 5.38 2 Marine shells Rabassa et al. (2000)
Bahia Brown 541530 671310 Pa-1011 985 135 784 332 0.82 2 Marineshells Rabassa et al. (2000)

Strait of Magellan (Punta Arenas)

Puerto del Hambre 531240 701560 QL-1468 7980 50 9006 8647 1.34 1 Peat Porter et al. (1984)
Puerto del Hambre 531240 701560 QL-1467 3978 70 4804 4160 2.74 1 Peat Porter et al. (1984)
Bahia San Gregorio 521530 701090 QL-1479 3860 40 3914 3693 0.29 2 Marine shells Porter et al. (1984)
Bahia Gente Grande 521350 701140 QL-1657 5860 40 6369 6185 0.65 2 Marine shells Porter et al. (1984)
Bahia Gente Grande 521350 701140 QL-1658 5625 30 6105 5927 0.29 2 Marine shells Porter et al. (1984)
Bahia Gente Grande 521350 701140 QL-1659 4600 40 4962 4696 0.71 2 Marine shells Porter et al. (1984)

Curacao
JB1/310 121070 681490 UtC-10409 5346 47 6269 6108 3.1 0.25 Mangrove peat Klosowska (2003)
JB1/5/4 121070 681490 UtC-10411 6220 50 7246 7098 4.5 0.25 Mangrove peat Klosowska (2003)
JB2/1/4 121070 681490 UtC-10408 1692 37 1694 1558 0.4 0.25 Mangrove peat Klosowska (2003)
JB3/2/1 121070 681490 UtC-12179 1745 37 1708 1625 0.3 0.25 Mangrove peat Klosowska (2003)
JB3/2/4 121070 681490 UtC-10407 2659 40 2845 2779 1.1 0.25 Mangrove peat Klosowska (2003)
 ARTICLE IN PRESS
G.A. Milne et al. / Quaternary Science Reviews 24 (2005) 1183–1202 1201

References Lambeck, K., 2002. In: Mitrovica, J.X., Vermeersen, L.L.A. (Eds.),
Sea Level Change from Mid-Holocene to Recent Time: An
Angulo, R.J., 1989. Fossil Vermitetdae between latitudes 25 34 S Australian Example with Global Implications in Glacial Isostatic
and 27 09 S, State of Parana and Santa Catarina—Brazil. Adjustment and the Earth System: Sea Level, Crustal Deforma-
International Symposium on Global Changes in South America tion, Gravity and Rotation. AGU Monograph, Geodynamics
During the Quaternary, Special Publication 1, São Paulo, Series, vol. 29. American Geophysical Union, Washington, DC,
pp. 263–268. pp. 33–50.
Angulo, R.J., Lessa, G.C., 1997. The Brazilian sea-level curves: a Lambeck, K., Bard, E., 2000. Sea-level change along the French-
critical review with emphasis on the curves from the Paranaguá and Mediterranean coast for the past 30,000 years. Earth and Planetary
Cananéia regions. Marine Geology 140, 141–166. Science Letters 175, 203–222.
Angulo, R.J., Giannini, P.C.F., Suguio, K., Pessendam, L.C.R., 1999. Lambeck, K., Johnston, P., Nakada, M., 1990. Holocene glacial
Relative sea-level changes in the last 5500 years in southern Brazil rebound and sea-level change in NW Europe. Geophysical Journal
(Laguna-Imbituba region, Santa Catarina State) based on vermitid International 103, 451–468.
14 Lambeck, K., Smither, C., Johnston, P., 1998. Sea-level change, glacial
C ages. Marine Geology 159, 33–229.
Clark, P.U., Mix, A.C., 2000. Ice sheets by volume. Nature 406, rebound and mantle viscosity for northern Europe. Geophysical
689–690. Journal International 134, 102–144.
Clark, J.A., Farrell, W.E., Peltier, W.R., 1978. Global changes in Martin, L., Suguio, K., 1978. Excursion route along the coastline
postglacial sea level: a numerical calculation. Quaternary Research between the town of Cananéia (State of São Paulo) and Guaratiba
9, 265–287. outlet (State of Rio de Janeiro), International Symposium on
Delibras, C., Laborel, J., 1969. Recent variations of the sea level along Global Changes in South America during the Quaternary, Special
the Brazilian coast. Quaternaria 14, 45–49. Publication 2, ABEQUA-INQUA, São Paulo, 136pp.
Digerfeldt, G., Hendry, M.D., 1987. An 8000 year Holocene sea-level Martin, L., Flexor, J.M., Blitzkow, D., Suguio, K., 1985. Geoid
record from Jamaica: implications for interpretation of Caribbean change indications along the Brazilian coast during the last 7,000
reef and coastal history. Coral Reefs 5, 165–169. years. Proceedings of the Vth International Congress on Coral
Dominguez, J.M.L., Bittencourt, A.C.S.P., Leão, Z.M.A.N., Azaeve- Reefs, vol. 3, Tahiti, pp. 85–90.
do, A.E.G., 1990. Geologia do Quaternário costeiro do estado de Milne, G.A., 1998. Refining models of the glacial isostatic adjustment
Pernambuco. Review Brasilian Geoscience 20, 208–215. process. Ph.D. Thesis, University of Toronto, Toronto, Canada.
Dziewonski, A.M., Anderson, D.L., 1981. Preliminary Reference Milne, G.A., 2002. In: Mitrovica, J.X., Vermeersen, L.L.A. (Eds.),
Earth Model (PREM). Physics of the Earth and Planetary Interiors Recent Advances in Predicting Glaciation-Induced Sea-level
25, 297–356. Changes and Their Impact on Model Applications in Glacial
Emery, K.O., Aubrey, D.G., 1991. Sea levels, land levels, and tide Isostatic Adjustment and the Earth System: Sea Level, Crustal
gauges. Springer-Verlag, New York, pp. 237. Deformation, Gravity and Rotation. AGU Monograph, Geody-
Farrell, W.E., Clark, J.A., 1976. On postglacial sea level. Geophysical namics Series, vol. 29. American Geophysical Union, Washington,
Journal of the Royal Astronomical Society 46, 647–667. DC, pp. 157–176.
Fairbridge, R.W., 1976. Shellfish-eating Preceramic Indians in coastal Milne, G.A., Mitrovica, J.X., 1996. Postglacial sea-level change
Brazil. Science 191, 353–359. on a rotating Earth: first results from a gravitationally self-
Flemming, K., Johnston, P., Zwartz, D., Yokoyama, Y., Lambeck, K., consistent sea-level equation. Geophysical Journal International
Chappell, J., 1998. Refining the eustatic sea-level curve since the 126, F13–F20.
Last Glacial Maximum using far- and intermediate-field sites. Milne, G.A., Mitrovica, J.X., 1998. Postglacial sea-level change on a
Earth and Planetary Science Letters 163, 327–342. rotating Earth. Geophysical Journal International 133, 1–19.
Forte, A.M., Mitrovica, J.X., 1996. A new inference of mantle Milne, G.A., Mitrovica, J.X., Davis, J.L., 1999. Near-field hydro-
viscosity based on a joint inversion of post-glacial rebound data isostasy: the implementation of a revised sea-level equation.
and long-wavelength geoid anomalies. Geophysical Research Geophysical Journal International 139, 464–482.
Letters 23, 1147–1150. Mitrovica, J.X., 1996. Haskell [1935] revisited. Journal of Geophysical
Gordillo, S., Bujalesky, G., Pirazzoli, P., Rabassa, J., Saliège, J., 1992. Research 101, 555–569.
Holocene raised beaches along the northern coast of the Beagle Mitrovica, J.X., Milne, G.A., 2002. On the origin of ocean syphoning.
Channel, Tierra del Fuego, Argentina. Palaeogeography Palaeo- Quaternary Science Reviews 21, 2179–2190.
climatology Palaeoecology 99, 41–54. Mitrovica, J.X., Peltier, W.R., 1991. On post-glacial geoid subsidence
Haskell, N.A., 1935. The motion of a viscous fluid under a surface over the equatorial oceans. Journal of Geophysical Research 96,
load. Physics 6, 56–61. 20,053–20,071.
Ireland, S., 1987. In: Tooley, M.J., Shennan, I. (Eds.), The Holocene Nakada, M., Lambeck, K., 1987. Glacial rebound and relative sea-
Sedimentary History of the Coastal Lagoons of Rio de Janeiro level variations: a new appraisal. Geophysical Journal of the Royal
State, Brazil in Sea-Level Changes, Institute of the British Astronomical Society 90, 171–224.
Geographers, Basil. Blackwell, Oxford, pp. 25–66. Nakada, M., Lambeck, K., 1989. Late Pleistocene and Holocene sea-
Ireland, S., 1988. Holocene coastal changes in Rio de Janeiro State, level change in the Australian region and mantle rheology.
Brazil. Ph.D. Thesis, University of Durham, Durham, UK. Geophysical Journal International 96, 497–517.
Johnston, P., 1993. The effect of spatially non-uniform water loads on Peltier, W.R., 1974. The impulse response of a Maxwell Earth.
predictions of sea level change. Geophysical Journal International Reviews of Geophysics 12, 649–669.
114, 615–634. Peltier, W.R., 2002. On eustatic sea level history: last glacial maximum
Klosowska, B., 2003. Late Holocene embayment and salina record of to Holocene. Quaternary Science Reviews 21, 377–396.
Curac- ao (Dutch Antilles): criteria to monitor environmental Peltier, W.R., Andrews, J.T., 1976. Glacial isostatic adjustment–I. The
change and biodiversity. Ph.D. Thesis, Vrije Universiteit, Am- forward problem. Geophysical Journal of the Royal Astronomical
sterdam, The Netherlands. Society 46, 605–646.
Lambeck, K., 1993. Glacial rebound of the British Isles—II. A high Pirazzoli, P.A., Pluet, J., 1991. World Atlas of Holocene Sea-
resolution, high-precision model. Geophysical Journal Interna- Level Changes. Elsevier Oceanography Series, vol. 58. Elsevier,
tional 115, 960–990. Amsterdam.
 ARTICLE IN PRESS
1202 G.A. Milne et al. / Quaternary Science Reviews 24 (2005) 1183–1202

Porter, S.C., Stuiver, M., Heusser, C.J., 1984. Holocene sea- Implications for sea-level rise and disturbance history. Journal of
level changes along the Strait of Magellan and Beagle Coastal Research 15, 496–504.
Channel, southernmost South America. Quaternary Research Stuiver, M., Reimer, P., 1998. CALIB Version 4.0 (Computer
22, 59–67. program), URL http://depts.washington.edu/qil/
Rabassa, J., Coronato, A., Bujalesky, G., Salemme, M., Roig, C., Tushingham, A.M., Peltier, W.R., 1991. ICE-3G: a new global model
Meglioli, A., Heusser, C., Gordillo, S., Roig, F., Borromeir, A., of late Pleistocene deglaciation based on geophysical predictions of
Quattrocchio, M., 2000. Quaternary of Tierra del Fuego, South- post-glacial relative sea level change. Journal of Geophysical
ernmost South America: an updated review. Quaternary Interna- Research 96, 4497–4523.
tional 68–71, 217–240. van Andel, T.H., Laborel, J., 1964. Recent high relative sea level stand
Roeleveld, W., Van Loon, A.J., 1979. The Holocene development of near Recife, Brazil. Science 145, 580–581.
the young coastal plain of Suriname. Geologie en Mijnbouw 58, Vening Meinesz, F.A., 1937. The determination of the earth’s plasticity
21–28. from the post-glacial uplift of Scandinavia: isostatic adjustment.
Rostami, K., Peltier, W.R., Mangini, A., 2000. Quaternary marine Koninklijke Akademie van Weten-schapen 40, 654–662.
terraces, sea-level changes and uplift history of Patagonia, Wieczerkowski, K., Mitrovica, J.X., Wolf, D., 1999. A revised
Argentina: comparisons with predictions of the ICE-4G (VM2) relaxation-time spectrum for Fennoscandia. Geophysical Journal
model of the global process of glacial isostatic adjustment. International 139, 69–86.
Quaternary Science Reviews 19, 1495–1525. Wu, P., Peltier, W.R., 1983. Glacial isostatic adjustment and the free air
Rull, V., Vegas-Vilarrúbia, T., de Pernı́a, N., 1999. Palynological gravity anomaly as a constraint on deep mantle viscosity. Geophy-
record of an early–mid-Holocene mangrove in Eastern Venezuela. sical Journal of the Royal Astronomical Society 74, 377–449.