1
A MULTIDISCIPLINARY APPROACH TO LATE QUATERNARY
PALEOCLIMATOLOGY WITH AN EMPHASIS ON SUB-SAHARAN WEST
AFRICA AND THE LAST INTERGLACIAL PERIOD
by
Nicholas Paul McKay
___________________________
A Dissertation Submitted to the Faculty of the
DEPARTMENT OF GEOSCIENCES
In Partial Fulfillment of the Requirements
For the Degree of
DOCTOR OF PHILOSOPHY
In the Graduate College
THE UNIVERSITY OF ARIZONA
2012
3
STATEMENT BY AUTHOR
This dissertation has been submitted in partial fulfillment of requirements for an
advanced degree at the University of Arizona and is deposited in the University Library
to be made available to borrowers under rules of the Library.
Brief quotations from this dissertation are allowable without special permission, provided
that accurate acknowledgment of source is made. Requests for permission for extended
quotation from or reproduction of this manuscript in whole or in part may be granted by
the head of the major department or the Dean of the Graduate College when in his or her
judgment the proposed use of the material is in the interests of scholarship. In all other
instances, however, permission must be obtained from the author.
SIGNED: Nicholas P. McKay
4
ACKNOWLEDGEMENTS
This research was supported by several funding sources, including National
Science Foundation Grant for Earth Sciences grant 0602355, an Institute for the Study of
Planet Earth Fellowship, an Institute of the Environment Dissertation Improvement
Grant, Dr. H. Wesley Pierce Scholarship, and a Chevron-Texaco Summer Research
Grant. The sediment drilling was supported by the International Continental Scientific
Drilling Program, and the U.S., Austrian, and Canadian National Science Foundations as
well as the Ghanaian Geological Survey.
I am greatly indebted to the drilling team and Lake Bosumtwi project
collaborators, John Peck, Jonathan Overpeck, Tim Shanahan, Chip Heil, John King and
Chris Scholz, who designed and carried out the drilling program, and who have spent
countless hours processing, and analyzing the sediment. Without their efforts none of my
research on Lake Bosumtwi would have been possible. Furthermore, I am thankful for
access to laboratory equipment, sample analysis and insight from Erik Brown, David
Dettman, Bob Downs, and Heidi Barnett.
Discussions with my advisor, Jonathan Overpeck, with my committee members
Julia Cole, Andy Cohen, Erik Brown, and Joellen Russell, as well as my collaborators
Bette Otto-Bliesner, John Peck, Tim Shanahan, Chip Heil, John King, Chris Scholz,
David Dettman, and Bob Downs greatly improved my research. I am especially grateful
for the support and contributions of past and present graduate student colleagues Cody
Routson, Jeremy Weiss, Jessica Conroy, Sarah White, Sarah Truebe, Toby Ault, Diane
Thompson and Suz Tolwinski-Ward.
5
DEDICATION
I dedicate this dissertation to my wife, Amber, who has always supported me and
my decision to pursue this research and degree, even during times when my obligations
required her to manage our household and raise our children without much help from me.
Without her selflessness I would have never completed this dissertation, and she doesn't
get enough credit for all the things she does. I also dedicate this dissertation to my boys,
Carter, Ethan and Landon, who throughout the past five years have helped me to
remember what's important in life, and for the most part, to keep my priorities straight.
6
TABLE OF CONTENTS
LIST OF FIGURES...........................................................................................................10
ABSTRACT.......................................................................................................................11
1. INTRODUCTION.........................................................................................................13
2. PRESENT STUDY........................................................................................................21
3. REFERENCES..............................................................................................................27
4. APPENDICES...............................................................................................................32
APPENDIX A: A HIGH-RESOLUTION PALEOHYDROLOGICAL RECORD FOR
THE PAST 530,000 YEARS FROM LAKE BOSUMTWI, SOUTHERN GHANA........33
Abstract..............................................................................................................................35
1. Introduction...................................................................................................................36
1.1 Setting......................................................................................................................38
1.2 Regional Climatology..............................................................................................39
1.3 Past work on Lake Bosumtwi..................................................................................41
2. Methods.........................................................................................................................43
2.1 Lake Drilling and core processing...........................................................................43
2.2 Geochronology........................................................................................................44
2.3 Scanning X-Ray fluorescence analysis....................................................................45
2.4 Carbonate isotope geochemistry..............................................................................46
2.5 Turbidites.................................................................................................................48
2.6 Laminations ............................................................................................................49
3. Results...........................................................................................................................49
3.1 Geochronology........................................................................................................49
3.1.1 Comparisons with regional and global dust record..........................................51
3.1.2 Tuning to EDC timescale..................................................................................53
3.2 X-ray fluorescence...................................................................................................54
3.2.1 Correcting for biases in the XRF data..............................................................55
3.2.1.1 X-ray absorption and attenuation by water................................................56
3.2.1.2 Sediment compaction with depth..............................................................58
3.2.2 Principal component analysis...........................................................................59
3.2.2.1 Carbonate dilution of XRF-PC1................................................................60
3.3 Lamination indices..................................................................................................61
3.4 Turbidites.................................................................................................................62
7
4. Discussion......................................................................................................................63
4.1 Past 530 ka...............................................................................................................63
4.1.1 400-kyr eccentricity forcing.............................................................................63
4.1.2 100-kyr eccentricity forcing and glacial boundary conditions.........................67
4.1.3 West African contribution to global atmospheric methane concentrations......67
4.2 Latitudinal variability in West African climate over the past 150 ka......................71
4.3 Millenial-scale events during the past 130 ka..........................................................73
4.4 Intervals of extreme aridity......................................................................................75
5. Conclusion and summary..............................................................................................77
6. References.....................................................................................................................80
APPENDIX B: A 12,000-YEAR-LONG, ANNUALLY-RESOLVED VARVE RECORD
SPANNING THE LAST INTERGLACIAL FROM LAKE BOSUMTWI, SOUTHERN
GHANA...........................................................................................................................117
Abstract............................................................................................................................119
1. Introduction.................................................................................................................120
1.1 Site.........................................................................................................................124
1.2 Modern Climate.....................................................................................................125
2. Methods.......................................................................................................................127
2.1 Core Drilling and sampling...................................................................................127
2.2 Sampling and laminae analysis..............................................................................128
2.3 XRF analysis..........................................................................................................128
2.4 Development of the varve chronology..................................................................129
2.5 Geochronology......................................................................................................132
2.6 Carbonate isotope geochemistry............................................................................135
3. Results.........................................................................................................................136
3.1 Generalized Lithology and Laminae structure......................................................136
3.1.1 Unit 1a (60.3 – 60.7 RMCD) .........................................................................137
3.1.2 Unit 1b (60.2 – 60.3 RMCD)..........................................................................137
3.1.3 Unit 1c (59.8 – 60.2 RMCD) .........................................................................137
3.1.4 Unit 1d (59.6 – 59.8 RMCD)..........................................................................138
3.1.5 Unit 2a (59.1 – 59.6 RMCD)..........................................................................138
3.1.6 Unit 2b (58.7 – 59.1 RMCD)..........................................................................139
3.1.7 Unit 2c (58.6 - 58.7 RMCD)...........................................................................139
3.1.8 Unit 2d (58.4 - 58.6 RMCD)..........................................................................140
3.1.9 Unit 3a (58.1 - 58.4 RMCD)...........................................................................140
3.1.10 Unit 3b (57.4 - 58.1 RMCD)........................................................................140
8
3.2 Carbonate isotope geochemistry and mineralogy..................................................141
3.3 Varve thickness......................................................................................................145
3.4 XRF data................................................................................................................146
4. Discussion....................................................................................................................148
4.1 Hydrology during the interglacial..........................................................................148
4.1.1 Early LIG (Varve years 1-5600; 128.6 – 123 ka)...........................................149
4.1.2 Mid-LIG (Varve years 5600-9000; 123 – 119.6 ka).......................................150
4.1.3 Late LIG (Varve years 9000-12100; 119.6 – 116.5 ka)..................................151
4.2 Comparison between the LIG and the Holocene...................................................152
4.2.1 Relationship with solar insolation..................................................................154
4.3 Relationships with sea surface temperature records..............................................156
4.4 An multi-century aridity event 118 ka...................................................................158
4.5 Drought frequency during the LIG........................................................................159
4.5.1 Influence of solar variability...........................................................................163
5. Conclusions and summary...........................................................................................165
5.1 Primary findings....................................................................................................165
5.2 Implications for future hydrology in the region....................................................168
6. References...................................................................................................................169
APPENDIX C: THE ROLE OF OCEAN THERMAL EXPANSION IN LAST
INTERGLACIAL SEA LEVEL RISE.............................................................................198
Abstract............................................................................................................................200
Introduction......................................................................................................................200
Methods............................................................................................................................202
Paleoclimate data.........................................................................................................202
Global Climate Model Simulations.............................................................................203
Paleoceanographic data synthesis....................................................................................204
Global Climate Model Simulations.................................................................................207
The thermosteric component of LIG sea level rise..........................................................210
Conclusion.......................................................................................................................213
Acknowledgements..........................................................................................................213
9
References........................................................................................................................214
Auxiliary Material............................................................................................................217
Error Analysis..............................................................................................................217
Bias Analysis................................................................................................................218
130 ka climate model simulation.................................................................................219
Auxiliary References....................................................................................................220
APPENDIX D: ON THE POTENTIAL OF RAMAN-SPECTROSCOPY-BASED
CARBONATE MASS SPECTROMETRY.....................................................................234
Abstract............................................................................................................................236
Introduction......................................................................................................................236
Experimental Procedure...................................................................................................238
Calcite synthesis...........................................................................................................238
Raman Spectroscopy....................................................................................................239
Isotope notation............................................................................................................240
Water Mass Spectrometry............................................................................................241
Carbonate Mass Spectrometry.....................................................................................242
Results and Discussion....................................................................................................243
Raman-inferred isotope ratios (RRaman).........................................................................243
Potential sources of error and bias...............................................................................244
Potential for improving the accuracy and precision of RRaman estimates......................245
Carbon isotopes............................................................................................................246
Potential for quantifying isotope ratios in other molecules.........................................247
Conclusions......................................................................................................................248
References........................................................................................................................249
APPENDIX E: PERMISSIONS......................................................................................258
10
LIST OF FIGURES
Figure 1, Map of Lake Bosumtwi......................................................................................26
11
ABSTRACT
A primary goal of paleoclimatology is to extend the instrumental record to capture
a wider range of natural variability, documenting the climate system's response to past
changes that have no analog in the historical record. Sediment archives of the recent
geologic past, both marine and lacustrine, offer the opportunity to study how climate
responds to a range of forcings and changing boundary conditions on timescales ranging
from years to millennia. In this dissertation I use lacustrine and marine sediment to
investigate changes late Quaternary climate, with particular focus on the Last Interglacial
period (LIG). First, I use multiple approaches to reconstruct long-term changes in the
West African Monsoon by investigating centennial-scale hydrologic variability recorded
in Lake Bosumtwi sediments over the past 530,000 years. Over this interval, hydrology in
the region is driven by a complex interplay of orbital forcing and glacial-interglacial
boundary conditions. Lake level was generally much lower between 50 and 300 ka, likely
due to the redistribution of rainfall from the tropics to the subtropics, driven by
eccentricity's amplification of precession. Consequently, the Holocene highstand at the
lake was both larger and longer lived than the maximum highstand during the LIG.
Annual layers were continuously deposited through the LIG in Lake Bosumtwi,
and I also present a new, 12,100 year-long, varve record spanning the interval from 128.6
to 116.5 ka. Over the course of the LIG, lake level generally tracks sea surface
temperatures (SST) in Gulf of Guinea, including an abrupt drop in lake level that lasted
about 500 years ca. 118 ka, coincident with cool SSTs in the North Atlantic and severe
aridity in Europe. I find that the despite the generally drier conditions, hydrology varied
12
on similar timescales as the late Holocene, with pronounced multidecadal to centennialscale variability with non-stationary periodicities.
I also investigate the contribution of ocean thermal expansion to sea level rise
during the LIG, using a synthesis of paleoceanographic data and a climate model
simulation. Globally, LIG SSTs were similar to, or slightly cooler than late Holocene
SSTs, with the exception of the North Atlantic, which was several degrees warmer.
Consequently, thermal expansion was likely a minor component of sea level rise during
the interval, explaining between -0.3 and 0.4 m. of the 6 to 8 m highstand. Lastly, I tested
the potential of Raman spectroscopy as a new, non-destructive technique to rapidly
measure oxygen isotopic ratios in carbonates at extremely high resolution. Analyses on a
suite a synthetic calcites indicate that 18O/16O ratios can be measured directly from the
Raman spectra and have a 1:1 correspondence with traditional mass-spectrometry
measurements. At present, the technique does not have the precision necessary to record
natural variability, although there is considerable potential for improving the precision of
the technique.
13
1. INTRODUCTION
One of the greatest challenges of the 21st century will be mitigating and adapting
to global climate change. As concentrations of greenhouse gases continue to build up, ice
caps continue melting, and atmospheric and ocean temperatures continue to rise, Earth's
climate system will be subjected to forcings and boundary conditions outside the range of
observed climate variability. A robust, physically-based understanding of the climate
system is therefore a prerequisite for anticipating global, and more challengingly,
regional climate impacts. The relative brevity of the instrumental record limits our
capacity to understand some aspects of natural climate variability, primarily the nature of
long-term (decadal-scale and longer) variability, and the system's response to changes in
forcing that are outside the limited range of variability in recent decades to centuries.
Herein lies a key role of paleoclimatology: to extend the instrumental record to capture a
wider range of natural variability, and to document the climate system's response to past
changes that have no modern analog. This Earth system forcing and response approach to
paleoclimatology requires continuous, well-dated, and high-resolution records, especially
when we want to know how dramatic changes affect climate on human timescales.
The integration of paleoclimate records are critical for understanding long-term
climate dynamics and providing insight into future change in sub-Saharan West Africa.
The region is prone to abrupt and prolonged drought; the “Sahel Drought” of the late 20th
century began unexpectedly in the late 1960's and lasted for decades (Nicholson, 2001).
The severity of the drought was greater than what was previously thought possible; to
14
illustrate, annual precipitation totals for the region for every year from 1970 to 1997 were
below the 100-yr mean (Nicholson and Webster, 2007). Subsequent research has linked
the decadal- to multidecadal-scale variability in West African hydrology to slow-varying
changes in the meridional structure of sea surface temperatures in tropical and subtropical
Atlantic Ocean (Lamb, 1978; Fontaine and Jacinot, 1996; Janicot et al., 2001; Camberlin
et al., 2001; Nicholson and Grist, 2007). The instrumental record in the region is far too
short (< 100 yrs; Nicholson, 2001) to capture such long term variability, so using highresolution paleoclimate records to extend the instrumental record in region (e.g.,
Shanahan et al., 2009) is necessary to understand the dynamics of drought in the region.
Considerable uncertainty remains in projections of future change for sub-Saharan
West Africa. Precipitation variability in the region is controlled by the complex dynamics
of the West African Monsoon (WAM). Unlike convective precipitation in other equatorial
and monsoonal regions, the West African monsoon is not driven by the convergence of
moist southwesterly monsoonal flow and the Intertropical Convergence Zone (ITCZ), but
rather, by the interaction of upper-level circulation and jets with low level moisture
transport (Nicholson, 2009; Pu and Cook, 2010; 2012). This complexity, and the
importance of small-scale features not resolved in many climate models, such as the West
African Westerly Jet, is part of the reason why most climate models do not simulate a
realistic West African Monsoon, both over the 20th century and in the mid-Holocene
(Cook and Vizy, 2006; Randall et al., 2007; Christensen, 2007; Pu and Cook, 2010,
Rodriguez-Fonseca et al., 2011). Only 4 of the 18 climate models in the Fourth
Assessment Report (AR4) of the Intergovernmental Panel on Climate Change (IPCC)
15
demonstrated reasonably realistic simulations of the WAM during the second half of the
20th century (Cook and Vizy, 2006). Furthermore, Rodriguez-Fonseca et al. (2011)
concluded that, in general, the IPCC AR4 models did not accurately simulate the tropical
SST teleconnections that drive both interannual and decadal variability of the WAM.
Even among models that performed reasonably well in the 20th century, projections for
the 21st century vary wildly (Cook and Vizy, 2006).
Taken together, the short instrumental record in a region dominated by long-term
variability and the uncertainty of future projections for the region highlights the potential
of paleoclimatology to bolster our understanding of climate in the region. Unfortunately,
continental West Africa is one of the most data-poor regions in the world, both over the
past two thousand years (cf., Mann et al., 2008) and during the past glacial period (cf.,
Blome et al., 2012). The primary focus of this dissertation is using lake sediments from
Lake Bosumtwi in southern Ghana to improve our understanding of climate variability in
the region, and to gain insight into future change in sub-Saharan West Africa.
Lake Bosumtwi occupies a 1.08 +/- 0.04 Ma (Koeberl et al., 1997; Jourdan et al.,
2009) meteorite-impact crater located in southern Ghana (6°30' N, 1°25' W)
approximately 150 km north of the Gulf of Guinea (Figure 1). The lake has no outlet and
is isolated from the groundwater table (Turner et al., 1996). Furthermore, the lake surface
occupies about 50% of surface area of the 11-km-diameter impact crater (Turner et al.,
1996). These features make lake level extremely sensitive to changes precipitation and
evaporation, and lake level responds to interannual and decadal-scale changes in
precipitation (Shanahan et al., 2007).
16
Lake Bosumtwi has been recognized as a valuable archive of past hydrologic and
environmental change in West Africa since the 1970's. Seminal work by Talbot and
Delibrias (1977) revealed large changes in lake level during the past 12 kyr recorded by
terraces and sediments exposed by river cuts. This led to a series of studies on sediment
cores taken from the lake and the geomorphology of the lake basin. Pollen studies
revealed that arboreal forests, which were abundant in the region during the Holocene,
were scarce in the region during the end of the last glaciation (Maley and Livingstone,
1983; Talbot et al., 1984; Maley 1987, 1991). Paleoecological studies from the lake also
indicated a decrease in the tropical forests around the lake in the late Holocene (Maley
1997; 2002). These changes in paleoecology are consistent with geochemical evidence
(elemental C and N, as well as δ13C and δ15N), which suggested a much lower lake level
between 30 and 12 ka, and an extremely wet interval between 12 and 3 ka, before
dropping to intermediate conditions in the late Holocene (Talbot and Johannessen, 1992;
Russell et al., 2003). The extremely wet early to mid Holocene period corresponded to a
distinct interval in the sediment, termed the “sapropel” by Talbot et al. (1984). The
interval is extremely rich in organic material (organic carbon 15-20%), and poor in clastic
minerals, and contains an important component of the filamentous blue-green algae
anabeana (Talbot et al., 1984). Terraces, highstand deposits and an overflow notch
indicate that the lake was very high, and occasionally overflowing during the deposition
of this layer (Talbot and Delibrias, 1977, 1980; Shanahan et al., 2006). Lastly, Talbot and
Kelts (1986) investigated the mineralogy and geochemistry of carbonates formed in the
water column or surface sediments of the lake. This work highlighted the extreme
17
variability in both mineralogy and the δ13C and δ18O of carbonates in the sediment
sequence.
Recently, a second phase of research has begun on the lake. One goal of this
research was to conduct further analysis on sediment cores from the the lake, using new
approaches to better understand recent variability. For example, recent studies have taken
advantage of the annual laminations (varves) persistent through the late Holocene to
reconstruct regional hydrology with unprecedented resolution. This culminated in a 2700yr-long varve record that revealed pronounced mulitdecadal to century-scale variability in
regional hydrology associated with changes in Atlantic SSTs and the Atlantic
Multidecadal Oscillation (Shanahan et al., 2009). Peck et al. (2004) examined the mineral
magnetic characteristics of the sediments over a longer time period to reveal intervals
with abundant iron sulfide minerals associated with drought. These intervals are
associated with cold events in the North Atlantic (the Younger Dryas and Heinrich events
1 and 2), indicating that millennial-scale fluctuations in North Atlantic SSTs and glacial
boundary conditions also influence hydrology at the lake. The other goal of this second
phase of research was to use drill cores to extend the sedimentary record through the
entire million-year history of the lake (Koeberl et al., 2005; Peck et al., 2005). Work
towards this goal is the primary focus of this dissertation, and a high-resolution record
from lake extending back 530,000 years is presented in Appendix A.
A particularly useful feature of Lake Bosumtwi is that throughout most of its
history, the lake preserved finely laminated varves. One varved interval of particular
interest corresponds to the Last Interglacial period (LIG). The LIG (roughly equivalent to
18
the Eemian, or Marine Isotope Stage 5e), which occurred from about 128 to 118 ka, is an
interesting period to study for several reasons; first, it's the most recent analog to the
Holocene, with generally similar conditions, although differences in the orbital
configuration resulted in altered insolation forcing, especially seasonally (Berger and
Loutre, 1991). It is often considered a partial analog for future climate, because global
mean temperatures were slightly warmer (0-2°C; Turney and Jones, 2010; McKay et al.,
2011) and sea level was 6-8 m higher than preindustrial conditions (Hearty et al., 2007;
Kopp et al., 2009), although global CO2 (and other GHG) concentrations were far below
modern and projected future concentrations (270-290 ppm; Luthi et al., 2008). The
largest difference in forcing relative to the Holocene during the LIG is the change in
seasonal forcing due to the amplification of the precession cycle by eccentricity (Berger
and Loutre, 1991). This resulted in a more than 30 W m-2 increase in Northern
Hemisphere summer insolation, and a corresponding decrease in Southern Hemisphere
winter insolation at the height of the interglacial, ca. 124 ka. This drove warmer
temperatures in the Northern Hemisphere, especially on land during the summer (Turney
and Jones, 2010). The northern Hemisphere is consistently projected to warm faster in the
future than the southern Hemisphere in all seasons, and especially on land (Meehl et al.,
2007). More regionally, Weldeab et al., (2007) show that the SST in the Gulf of Guinea
was generally warmer during the LIG than the Holocene, which is also a robust feature of
GCM projections for the 21st century (Cook and Vizy, 2006). The Lake Bosumtwi record
offers the rare opportunity to investigate past climate variability with annual resolution
during this particularly interesting interval; these results are presented in Appendix B.
19
Another aspect of the LIG that is relevant to future climate change is that it was
the last interval with substantially higher-than-modern sea level (6-8 m above modern;
Hearty et al, 2007; Kopp et al., 2009). Sea level rise is one of the major socio-economic
hazards associated with global warming, however expected contributions of the different
components of sea level rise (i.e., ocean thermal expansion, melting of glaciers, and
wasting of the Greenland and Antarctic Ice Sheets) are poorly understood. For example,
estimates and projections of the thermal expansion component of sea level rise during the
21st century vary from 20 to 70% of total sea level rise over the 21st century (Meehl et al.,
2007; Pfeffer et al., 2008; Vermeer and Rahmstorf, 2009). The sea level highstand during
the Last Interglacial offers one approach to improve understanding of the relative
contributions of the different sources of sea level rise. Previous work has made it clear
that the majority of the sea level rise originated from melting of the Greenland Ice Sheet
(GIS) and the Antarctic Ice Sheets (Otto-Bliesner et al., 2006; Kopp et al., 2009; Clark
and Huybers, 2009), but the role of thermal expansion has not been carefully examined.
In Appendix C, we combine analyses of paleoceanographic records and global climate
model simulations to better constrain the amount of thermal expansion that occurred
during the LIG sea-level highstand.
High-resolution paleoclimatology is a rapidly-developing field. A critical aspect
of progress in the field is the development of new techniques that allow high-resolution
and rapid acquisition of paleoclimate archive data. Variability in oxygen isotope ratios is
among the most widely used indicators of environmental change, and is a primary
paleoclimate proxy for ice cores, corals, speleothems, soils, and marine and lacustrine
20
sediments. For most of these archives, δ18O is measured on carbonate minerals, typically
using a stable isotope mass spectrometer. Recently, laser-based optical approaches, such
as wavelength-scanned cavity ring-down spectroscopy (WS-CRDS), are allowing for
more rapid, and portable, isotopic analysis. Raman spectroscopy is a technique that relies
on the interaction of light with a molecule to infer information about the vibrational
modes of the molecule. The shifts in energy resulting from excitation of the vibrational
modes result in a corresponding shift in the wavelength of the monochromatic light
(usually a laser) used to excite the sample. This is termed a “Raman shift” and the
resulting spectrum of Raman shifts and their intensities corresponds to the energies of the
stretching modes in the molecule. In carbonates, the symmetric stretching mode of the
oxygen atoms about the carbon atom forms a distinct Raman peak, and the Raman shift is
different for carbonate molecules with one (or more) 18O atom, than for those with three
16
O atoms. It is therefore theoretically possible to quantify the relative abundance of the
different molecule types, and calculate δ18O. If the technique works with sufficient
precision and accuracy, it would be extremely useful in high-resolution paleoclimatology.
Raman spectroscopy is rapid, non-destructive, and can be performed at a resolution on
the order of 1 micron. Such a device could even be made portable. Investigation into the
potential and utility of the technique is presented in Appendix D.
21
2. PRESENT STUDY
The methods, results and conclusions of this study are presented as four papers
formatted for publication in professional journals. All papers are coauthored, and I am
senior author on all four manuscripts. The following is a brief summary of the most
important findings in this dissertation.
The majority of the research presented in this dissertation has stemmed from my
work on the Lake Bosumtwi sediment cores, the primary focus of my work as a student at
the University of Arizona. In the first study (Appendix A), we present the first
continuous, high-resolution terrestrial record of sub-Saharan paleohydrological variability
for the past 530,000 years; a sediment geochemical record from Lake Bosumtwi in
southern Ghana (6°30' N, 1°25' W). In the study, we used the abundance of terrigenous
elements in the sediment, as well as the mineralogy and isotope geochemistry of
carbonates formed in the lake to infer past changes in lake level and hydrology in the
region. Over the past 530 ka, moisture balance at the lake is driven by a complex
interplay of orbital forcing and glacial-interglacial boundary conditions. The 400-kyr
component of orbital eccentricity drove a more arid background state at the lake between
300 and 100 ka, likely due to the amplification of precession and the redistribution of
tropical moisture to more subtropical regions. On glacial-interglacial timescales, lake
level displays a positive relationship with the 100-kyr component of eccentricity, such
that interglacial periods are generally less arid than glacial periods, likely due to the
influence of warmer north Atlantic SSTs during interglacial periods. The influence of
22
Atlantic SSTs is apparent on millennial-timescales as well. Over the past 130 kyr, cold
periods recorded in Greenland ice cores (NGRIP 2004; Wolff et al., 2010), especially
those associated with to Heinrich Events, consistently correspond to low lake level at
Lake Bosumtwi. Six multi-millennial scale intervals of extreme aridity were recorded in
the sediments during the past 530 ka, all during the eccentricity-high interval from 30050 ka. These events generally occurred during times of similar orbital and glacial
boundary conditions, that is, deep glacial conditions and local summer insolation minima.
Two of the intervals represent abrupt drying from wetter-than-average conditions, and
were immediately preceded and followed by intervals of high lake level. The transitions
into and out of these two intervals occurred rapidly, on the order of decades to centuries,
although the origin of these arid excursions remains unclear.
In Appendix B we present a new 12,100-year-long varve record for the Last
Interglacial (LIG) at Lake Bosumtwi, spanning the interval from 128.6 to 116.5 ka. We
use the abundance of terrigenous elements in the sediment, varve thickness, and the
isotope geochemistry and mineralogy of authigenic carbonates in the sediment to infer
changes in lake level during the interval. The LIG peak highstand was both lower and
shorter-lived than the prolonged highstand in the early Holocene, and unlike the
Holocene, the evidence suggests the lake never overflowed during LIG. This result is
consistent with those from Appendix A, which indicated generally drier conditions during
intervals of high eccentricity (on the 400-kyr timescale). In other respects, the LIG and
the Holocene were similar. Both interglacials had two distinct millennial-scale intervals
of higher lake level that occurred during times of precession-driven insolation maxima in
23
July and October, corresponding to the two rainy seasons in the modern climatology.
Over the course of the LIG, lake level generally tracks sea surface temperatures (SST) in
the Gulf of Guinea (Weldeab et al., 2007), including an abrupt drop in lake level that
lasted about 500 years ca. 118 ka, corresponding to cooling in the Gulf of Guinea and
much of the North Atlantic (e.g., Lehman et al., 2002; Oppo et al., 2006) during the
interval. The timing and duration of the event are comparable to the Late Eemian Aridity
Pulse (LEAP) that is observed at several European sites (Sirocko et al., 2005; Seelos and
Sirocko, 2007) and has been interpreted to result from abrupt cooling in the North
Atlantic, and possibly a reduction of Atlantic meridional overturning circulation
(AMOC). A reduction in AMOC, and a cooling of North Atlantic would likely result in
aridity in West Africa (Street-Perrott and Perrot, 1990; Chang et al., 2008; Mulitza et al.,
2008; Shanahan et al., 2009) so the aridity ca. 118 ka is the first indication that the LEAP
occurred in Africa as well as Europe.
Further analysis of the Last Interglacial period is presented in Appendix C. In this
study we produced the first quantitative estimate of the thermosteric component of LIG
sea level rise. To do this, we compiled paleoceanographic data and used a coupled
atmosphere-ocean climate model simulation to examine global SSTs during the height of
the LIG sea level highstand, which was 6 to 8 meter higher than present (Hearty et al.,
2007; Kopp et al., 2009). The paleoceanographic data suggest a peak LIG global sea
surface temperature (SST) warming of 0.7±0.6°C compared to the late Holocene. Our
LIG climate model simulation suggests a slight cooling of global average SST relative to
preindustrial conditions (ΔSST = -0.4°C), associated with a reduction in atmospheric
24
water vapor in the Southern Hemisphere driven by a northward shift of the Intertropical
Convergence Zone, and substantially reduced seasonality in the Southern Hemisphere.
Taken together, the data suggest a small contribution of thermal expansion during the
interval; it is unlikely that thermosteric sea level rise exceeded 0.4±0.3 m during the LIG.
This constraint, along with estimates of the sea level contributions from the Greenland
Ice Sheet (Otto-Bliesner et al., 2006), glaciers and ice caps (Radić and Hock, 2010),
implies that 4.1 to 5.8 m of sea level rise during the Last Interglacial period was derived
from the Antarctic Ice Sheet.
The final study, Appendix D, investigates the potential for using Raman
spectroscopy to measure 18O-16O oxygen isotopic ratios in carbonates. To do this, we
synthesized a suite of isotopically-enriched calcite crystals ranging in composition from
natural (0.2% 18O) to six-times-natural abundance (1.2% 18O). We determined the
Raman-inferred isotopic ratios (RRaman) of individual crystals by fitting pseudoVoigt
curves to the ν1 symmetric stretching peak at 1086 cm-1 and the smaller satellite peak,
associated with the ν1 stretching mode of singly-substituted carbonate groups (C16O218O)
at 1065 cm-1. The ratio of the two numerically-integrated peak areas shows a 1:1
correspondence with the 18O/16O ratios measured by water and carbonate mass
spectrometry, confirming that the relative intensities of the ν1 symmetric stretching peaks
is a direct measure of the isotopic ratio in the carbonates. The 1-sigma uncertainties of the
RRaman values of the individual crystals were 0.00079 (384‰ PDB), and 0.00043 (210‰
PDB) for the 4-crystal sample means. Although this level of uncertainty is too high to
detect natural variability, there are multiple prospects for improving the precision of the
25
technique. Carbon isotope ratios in carbonates cannot be measured by this approach,
however our results do highlight the potential of Raman-based mass spectrometry for C
and other elements in minerals and organic compounds, such as calcium oxalates.
26
90
1°27' W
300
90
90
90
45
6°32' N
30 m
50 m
90
90
45
90
90
45
70 m
45
90
6°27' N
90
90
0
1
2
km
Figure 1. Site map of the Lake Bosumtwi crater, modified from Shanahan (2006). Solid
gray lines show bathymetry (10 m contour). Dashed line indicates crater rim. Inset:
location of Lake Bosumtwi in southern Ghana, West Africa.
27
3. REFERENCES
Berger, A., and M. F. Loutre (1991), Insolation values for the climate of the last 10
million years, Quaternary Science Reviews, 10(4), 297–317.
Camberlin, P., S. Janicot, and I. Poccard (2001), Seasonality and atmospheric dynamics
of the teleconnection between African rainfall and tropical sea-surface
temperature: Atlantic vs. ENSO, International Journal of Climatology, 21(8), 973–
1005.
Chang, P., R. Zhang, W. Hazeleger, C. Wen, X. Wan, L. Ji, R. J. Haarsma, W.-P.
Breugem, and H. Seidel (2008), Oceanic link between abrupt changes in|[nbsp]|
the North Atlantic Ocean and the|[nbsp]|African|[nbsp]|monsoon, Nature
Geoscience, 1(7), 444–448, doi:10.1038/ngeo218.
Christensen, J. H. et al. (2007), Regional climate projections, 2007: The Physical Science
Basis. Contribution of Working group I to the Fourth Assessment Report of the
Intergovernmental Panel on Climate Change, University Press, Cambridge,
Chapter 11, 847–940.
Clark, P. U., and P. Huybers (2009), Global change: Interglacial and future sea level,
Nature, 462(7275), 856–857.
Cook, K. H., and E. K. Vizy (2006), Coupled model simulations of the West African
monsoon system: Twentieth-and twenty-first-century simulations, Journal of
climate, 19(15), 3681–3703.
Fontaine, B., and S. Janicot (1996), Sea surface temperature fields associated with West
African rainfall anomaly types, Journal of climate, 9(11), 2935–2940.
Hearty, P. J., J. T. Hollin, A. C. Neumann, M. J. O’Leary, and M. McCulloch (2007),
Global sea-level fluctuations during the Last Interglaciation (MIS 5e), Quaternary
Science Reviews, 26(17-18), 2090–2112.
Janicot, S., S. Trzaska, and I. Poccard (2001), Summer Sahel-ENSO teleconnection and
decadal time scale SST variations, Climate Dynamics, 18(3), 303–320.
Jourdan, F., P. R. Renne, and W. U. Reimold (2009), An appraisal of the ages of terrestrial
impact structures, Earth and Planetary Science Letters, 286(1), 1–13.
Koeberl, C., R. Bottomley, B. P. Glass, and D. Storzer (1997), Geochemistry and age of
Ivory Coast tektites and microtektites, Geochimica et Cosmochimica Acta, 61(8),
1745–1772.
28
Koeberl, C., J. Peck, J. King, B. Milkereit, J. Overpeck, and C. Scholz (2005), The ICDP
Lake Bosumtwi Drilling Project: A First Report, Scientific Drilling, 1, 23–27.
Kopp, R. E., F. J. Simons, J. X. Mitrovica, A. C. Maloof, and M. Oppenheimer (2009),
Probabilistic assessment of sea level during the Last Interglacial stage, Nature,
462(7275), 863–867.
Lamb, P. J. (1978), Large-scale tropical Atlantic surface circulation patterns associated
with Subsaharan weather anomalies, Tellus, 30, 240–251.
Lehman, S. J., J. P. Sachs, A. M. Crotwell, L. D. Keigwin, and E. A. Boyle (2002),
Relation of subtropical Atlantic temperature, high-latitude ice rafting, deep water
formation, and European climate 130,000-60,000 years ago, Quaternary Science
Reviews, 21(18-19), 1917–1924.
Lüthi, D. et al. (2008), High-resolution carbon dioxide concentration record 650,000–
800,000 years before present, Nature, 453(7193), 379–382.
Maley, J. (1987), Fragmentation de la forêt dense humide africaine et extension des
biotopes montagnards au Quaternaire récent: nouvelles données polliniques et
chronologiques, implications paléoclimatiques et biogéographiques (1987).
Palaeoecology of Africa and the surrounding islands.
Maley, J. (1991), The African rain forest vegetation and palaeoenvironments during late
Quaternary, Climatic change, 19(1), 79–98.
Maley, J. (1997), Middle to Late Holocene changes in tropical Africa and other
continents: paleomonsoon and sea surface temperature variations, Nato ASI Series
I: Global Environmental Change, 49, 611–640.
Maley, J. (2002), A catastrophic destruction of African forests about 2,500 years ago still
exerts a major influence on present vegetation formations, IDS bulletin, 33(1),
13–30.
Maley, J., and D. A. Livingstone (1983), Extension d’un élément montagnard dans le sud
du Ghana (Afrique de l’Ouest) au Pléistocène supérieur et à l’Holocène inférieur:
premières données polliniques, CR Acad. Sci. Paris, Sér, 2(196), 1287–1292.
Mann, M. E., Z. Zhang, M. K. Hughes, R. S. Bradley, S. K. Miller, S. Rutherford, and F.
Ni (2008), Proxy-based reconstructions of hemispheric and global surface
temperature variations over the past two millennia, Proceedings of the National
Academy of Sciences, 105(36), 13252.
McKay, N. P., J. T. Overpeck, and B. L. Otto-Bliesner (2011), The role of ocean thermal
29
expansion in Last Interglacial sea level rise, Geophysical Research Letters,
38(14), L14605.
Meehl, G. A. et al. (2007), Chapter 10: Global Climate Projections. IPCC Fourth
Assessment Report. The Physical Science Basis. Contribution of Working group I
to the Fourth Assessment Report of the Intergovernmental Panel on Climate
Change, University Press, Cambridge, Chapter 10, 747-846.
Mulitza, S., M. Prange, J. B. Stuut, M. Zabel, T. von Dobeneck, A. C. Itambi, J. Nizou,
M. Schulz, and G. Wefer (2008), Sahel megadroughts triggered by glacial
slowdowns of Atlantic meridional overturning, Paleoceanography, 23(4), PA4206.
NGRIP et al. (2004), High-resolution record of Northern Hemisphere climate extending
into the Last Interglacial period, Nature, 431(7005), 147–151.
Nicholson, S. E. (2001), Climatic and environmental change in Africa during the last two
centuries, Climate Research, 17(2), 123–144.
Nicholson, S. E. (2009), A revised picture of the structure of the “monsoon” and land
ITCZ over West Africa, Climate Dynamics, 29, doi:10.1007/s00382-008-0514-3.
Nicholson, S. E., and J. P. Grist (2001), A conceptual model for understanding rainfall
variability in the West African Sahel on interannual and interdecadal timescales,
International Journal of Climatology, 21(14), 1733–1757.
Nicholson, S. E., and P. J. Webster (2007), A physical basis for the interannual variability
of rainfall in the Sahel, Quarterly Journal of the Royal Meteorological Society,
133(629).
Oppo, D. W., J. F. McManus, and J. L. Cullen (2006), Evolution and demise of the Last
Interglacial warmth in the subpolar North Atlantic, Quaternary Science Reviews,
25(23-24), 3268–3277.
Otto-Bliesner, B. L., S. J. Marshall, J. T. Overpeck, G. H. Miller, and A. Hu (2006),
Simulating Arctic climate warmth and icefield retreat in the last interglaciation,
Science, 311, 1751–1753.
Peck, J. A., R. R. Green, T. Shanahan, J. W. King, J. T. Overpeck, and C. A. Scholz
(2004), A magnetic mineral record of Late Quaternary tropical climate variability
from Lake Bosumtwi, Ghana, Palaeogeography, Palaeoclimatology,
Palaeoecology, 215(1-2), 37–57.
Pfeffer, W. T., J. T. Harper, and S. O’Neel (2008), Kinematic Constraints on Glacier
Contributions to 21st-Century Sea-Level Rise, Science, 321(5894), 1340 –1343,
30
doi:10.1126/science.1159099.
Pu, B., and K. H. Cook (2010), Dynamics of the West African westerly jet, Journal of
Climate, 23(23), 6263–6276.
Pu, B., and K. H. Cook (2012), Role of the West African Westerly Jet in Sahel Rainfall
Variations, Journal of Climate, 25(8), 2880–2896, doi:10.1175/JCLI-D-1100394.1.
Radić, V., and R. Hock (2010), Regional and global volumes of glaciers derived from
statistical upscaling of glacier inventory data, Journal of Geophysical Research,
115, 10 PP., doi:201010.1029/2009JF001373.
Randall, D. A. et al. (2007), Climate models and their evaluation: The Physical Science
Basis. Contribution of Working group I to the Fourth Assessment Report of the
Intergovernmental Panel on Climate Change, University Press, Cambridge,
Chapter 8, 589-662.
Rodríguez-Fonseca, B. et al. (2011), Interannual and decadal SST-forced responses of the
West African monsoon, Atmospheric Science Letters, 12(1), 67–74.
Russell, J., M. R. Talbot, and B. J. Haskell (2003), Mid-Holocene climate change in Lake
Bosumtwi, Ghana, Quaternary Research, 60(2), 133–141.
Seelos, K., and F. Sirocko (2007), 14. Abrupt cooling events at the very end of the Last
Interglacial, Developments in Quaternary Sciences, 7, 207–229.
Shanahan, T. M., J. T. Overpeck, W. E. Sharp, C. A. Scholz, and J. A. Arko (2007),
Simulating the response of a closed-basin lake to recent climate changes in
tropical West Africa (Lake Bosumtwi, Ghana), Hydrological Processes, 21(13).
Shanahan, T. M., J. T. Overpeck, C. W. Wheeler, J. W. Beck, J. S. Pigati, M. R. Talbot, C.
A. Scholz, J. Peck, and J. W. King (2006), Paleoclimatic variations in West Africa
from a record of late Pleistocene and Holocene lake level stands of Lake
Bosumtwi, Ghana, Palaeogeography, Palaeoclimatology, Palaeoecology, 242(3),
287–302.
Shanahan, T., J. Overpeck, K. Anchukaitis, J. Beck, J. Cole, D. Dettman, J. Peck, C.
Scholz, and J. King (2009), Atlantic forcing of persistent drought in West Africa,
Science, 324(5925), 377.
Sirocko, F., K. Seelos, K. Schaber, B. Rein, F. Dreher, M. Diehl, R. Lehne, K. Jäger, M.
Krbetschek, and D. Degering (2005), A late Eemian aridity pulse in central
Europe during the last glacial inception, Nature, 436(7052), 833–836.
31
Street-Perrott, F. A., and R. A. Perrott (1990), Abrupt climate fluctuations in the tropics:
the influence of Atlantic Ocean circulation, Nature, 343(6259), 607–612,
doi:10.1038/343607a0.
Talbot, M. R., and G. Delibrias (1977), Holocene variations in the level of Lake
Bosumtwi, Ghana, Nature, 268(5622), 722–724.
Talbot, M. R., and T. Johannessen (1992), A high resolution palaeoclimatic record for the
last 27,500 years in tropical West Africa from the carbon and nitrogen isotopic
composition of lacustrine organic matter, Earth and Planetary Science Letters,
110(1-4), 23–37.
Talbot, M. R., and K. Kelts (1986), Primary and diagenetic carbonates in the anoxic
sediments of Lake Bosumtwi, Ghana, Geology, 14(11), 912–916.
Talbot, M. R., D. A. Livingstone, P. G. Palmer, J. Maley, J. M. Melack, G. Delibrias, and
S. Gulliksen (1984), Preliminary results from sediment cores from Lake
Bosumtwi, Ghana, Palaeoecology of Africa, 16, 173–192.
Turner, B. F., L. R. Gardner, and W. E. Sharp (1996), The hydrology of Lake Bosumtwi, a
climate-sensitive lake in Ghana, West Africa, Journal of Hydrology, 183(3-4),
243–261.
Turney, C. S. ., and R. T. Jones (2010), Does the Agulhas Current amplify global
temperatures during super-interglacials?, Journal of Quaternary Science, 25(6),
839–843.
Vermeer, M., and S. Rahmstorf (2009), Global sea level linked to global temperature,
Proceedings of the National Academy of Sciences, 106(51), 21527–21532.
Weldeab, S., D. W. Lea, R. R. Schneider, and N. Andersen (2007), 155,000 years of West
African monsoon and ocean thermal evolution, Science, 316(5829), 1303.
Wolff, E. W., J. Chappellaz, T. Blunier, S. O. Rasmussen, and A. Svensson (2010),
Millennial-scale variability during the last glacial: The ice core record, Quaternary
Science Reviews, 29(21–22), 2828–2838, doi:10.1016/j.quascirev.2009.10.013.
32
4. APPENDICES
33
APPENDIX A
A HIGH-RESOLUTION PALEOHYDROLOGICAL RECORD FOR THE PAST
530,000 YEARS FROM LAKE BOSUMTWI, SOUTHERN GHANA
34
A HIGH-RESOLUTION PALEOHYDROLOGICAL RECORD FOR THE PAST
530,000 YEARS FROM LAKE BOSUMTWI, SOUTHERN GHANA
Nicholas P. McKay1, Jonathan T. Overpeck1,2,3, Timothy M. Shanahan4, Erik T. Brown5,
John A. Peck6, Clifford W. Heil7, John W. King7, and Chris A. Scholz8
1
Department of Geosciences, University of Arizona
Department of Atmpospheric Science, University of Arizona
3
Institute of the Environment, University of Arizona
4
Jackson School of Geosciences, University of Texas at Austin
5
Large Lakes Observatory, University of Minnesota Duluth
6
Department of Geology and Environmental Science, University of Akron
7
Graduate School of Oceanography, University of Rhode Island
8
Deptartment of Earth and Environmental Sciences, Syracuse University
2
35
Abstract
The West African Monsoon (WAM) system is an important component of the
global climate system. It is also the primary source of moisture for millions of people in
sub-Saharan Africa. The region is prone to decadal to multi-decadal droughts and pluvials
associated with slow changes in Atlantic sea surface temperatures (SSTs). The
instrumental climate record is too short to understand natural variability in the WAM, let
alone how it responds to large changes in global climate. Here we present the first, highresolution terrestrial record of sub-Saharan paleohydrological variability for the past
530,000 years; a sediment geochemical record from Lake Bosumtwi in southern Ghana
(6°30' N, 1°25' W). The abundance of terrigenous elements in the sediment, as well as the
mineralogy and isotope geochemistry of carbonates formed in the lake, are both sensitive
to changes in lake level, and reveal a dynamic history of hydrologic variability over the
past 530 kyr. Hydrology in the region is driven by a complex interplay of orbital forcing
and glacial-interglacial boundary conditions. The 400-kyr component of eccentricity
drove a more arid background state at the lake between 300 and 100 ka, likely due to the
amplification of precession and the redistribution of tropical moisture to more subtropical
regions. On glacial-interglacial timescales, lake level displays a positive relationship with
the 100-kyr component of eccentricity, such that interglacial periods are generally less
arid than glacial periods, likely due to the influence of warmer north Atlantic SSTs during
interglacial periods. The competing roles of local insolation forcing and North Atlantic
boundary conditions are apparent over the past 150 kyr on a latitudinal transect of records
sensitive to West African hydrology. During that interval, more northerly sites in West
36
Africa are more sensitive to obliquity and North Atlantic temperatures, whereas more
southerly sites respond more to local insolation and precession forcing. The extent of the
competing impacts changes with the amplitude of local insolation forcing, as larger
swings in precession before 70 ka had a more widespread impact on West African
climate. Six multi-millennial scale intervals of extreme aridity were recorded in the
sediments during the past 530 ka, all during the eccentricity-high interval from 300-50
ka. These events are generally associated with deep glacial conditions and local summer
insolation minima, and in two cases, were immediately preceded and followed by
intervals of high lake level. The transitions, both into and out of the arid intervals, appear
to be rapid, occurring on the order of centuries or less, although the underlying climate
dynamics of these transitions remains uncertain.
1. Introduction
The tropics are the heat engine of global climate, and tropical monsoon systems
play a critical role in distributing moisture and energy (Wallace and Hobbs, 2006; Chang
et al., 2011). The tropics also initiate, modulate and interact with both natural and
anthropogenic climate change (Peterson et al., 2000; Visser et al., 2003; Hoerling and
Kumar, 2003; Wang et al., 2007; Lu et al., 2009). Much of observed variability in tropical
monsoon systems has been associated with slow-varying changes in the tropical oceans
(Lamb, 1978; Fontaine et al., 1998; Janicot et al., 2001; Cook and Vizy, 2001; Nicholson
and Webster, 2007); however the terrestrial instrumental records are too short to
understand natural variability in the monsoon systems, let alone their response to abrupt
37
changes. The paleoclimatic history of continental tropics are particularly poorly
understood when compared to the oceans and higher latitudes, both in terms of lowfrequency climate variability and abrupt change. This is largely because of the lack of
long, continuous records comparable to ice and ocean sediment cores, and because of the
lack of well-constrained age models. Recently, records from the Cariaco Basin (Peterson
et al., 2000; Yarinik et al., 2000; Hughen et al., 2004), Lakes Malawi (e.g., Cohen et al.,
2007; Brown et al., 2007; Johnson et al., 2011; Stone et al., 2011) Challa (Verschuren et
al., 2009; Moernaut et al., 2010) and Tanganyika (Tierney et al., 2008), and South
American speleothems (Cruz et al., 2005; Cruz et al., 2009; Kanner et al., 2012) have
broadened our perspective of how tropical climate has varied on long timescales,
responds to (and possibly forces) abrupt change in the North Atlantic, and is modulated
by orbital variability.
New, long drill records from lakes in both East and West Africa, including initial
results from Lake Bosumtwi, suggest that severe aridity was much more common on the
continent between 135 and 75 ka than from 75 to 0 ka (Scholz et al., 2007). Nevertheless,
the long-term climatic history of tropical West Africa, and the past response of the West
African Monsoon (WAM) to extreme climate changes remain poorly understood, which
is of particular concern given the region's vulnerability to drought, and considerable
uncertainty about how continued climate change will affect the hydrology of sub-Saharan
Africa (Cook and Vizy, 2006). The continuous, high-resolution sediment sequence from
Lake Bosumtwi, offers the first opportunity to use a terrestrial record to investigate how
West Africa responds to orbital forcing and glacial boundary conditions over the past
38
million years (Peck et al., in prep).
1.1 Setting
Lake Bosumtwi is a 1.08 +/- 0.04 Ma (Koeberl et al., 1997; Jourdan et al., 2009)
meteorite-impact crater lake located in southern Ghana (6°30' N, 1°25' W; Figure A1).
The crater is the youngest and best preserved complex impact structure on Earth (Koeberl
and Milkereit, 2007). The modern lake is about 8 km in diameter, and fills about half of
crater area, so the majority (80%) of the annual water input to the lake comes from
rainfall directly on the surface (Turner et al., 1996a). The steep sides of the crater and
depth of the lake (76 m) promote semi-permanent stratification of the water column; the
upper 35 m overturn seasonally, however the lake does not mix below 35 m depth
(Almond and Hecky, 2000). Consequently, the bottom waters are anoxic, and sediments
preserve finely laminated varves. Lake level is sensitive to changes precipitation and
evaporation, and responds to intraannual to decadal-scale changes in precipitation,
varying by more than a meter seasonally, and falling more than 2 m during the “Sahel
Drought” from 1970-2000 (Shanahan et al., 2007). Sediment physical properties and the
stable isotope geochemistry of carbonates precipitated in the lake are sensitive to lake
level (Talbot and Delibrias, 1977; Talbot and Kelts, 1986; Peck et al., 2004; Shanahan,
2006; Scholz et al., 2007; Shanahan et al., 2009). This study focuses on the upper 160
meters of the longest sedimentary drill core from the lake.
39
1.2 Regional Climatology
The annual migration of the tropical rainbelt over Guinea Coast drives a bimodal
distribution of annual precipitation, with peak rainfall occurring in May-July and a
smaller peak from September-October (Nicholson, 2009) (Figure A2). Wind speed and
direction in the region is also seasonal, and are controlled by the position of the
Intertropical Convergence Zone (ITCZ). In the boreal summer the ITCZ is positioned
north of the lake over the Sahel and Sahara, and the monsoonal southwesterlies peak in
August. The winds drive upwelling off the Guinea Coast, cooling the Gulf of Guinea and
the Guinea Coast region to its annual minimum (Shanahan, 2006). This reduces
evaporation and stabilizes the atmosphere, suppressing convection in the region (Moron,
1994, Opoku-Ankomah, 1994). During the boreal winter the ITCZ is positioned near the
Guinea Coast, wind speed is reduced, and warm, aerosol-rich northeasterly 'Harmattan'
winds are common. Temperatures are highest during this season (~28°C at Kumasi;
Shanahan, 2006). Dust supplied to the region by these winds has been shown to be a
substantial component of Ghanaian soils (He et al., 2007) and solute load in Lake
Bosumtwi (Turner et al., 1996b).
Precipitation variability in West Africa is strongly influenced by the distribution of
sea surface temperatures (SSTs), particularly in the tropical Atlantic (Lamb, 1978; Janicot
et al., 2001), but also in the eastern Mediterranean (Rowell, 2003), and tropical Pacific
(Rowell, 2001; Janicot et al., 2001). Two modes of variability in the tropical Atlantic
have been linked to West African precipitation. The equatorial zonal mode, termed
“Atlantic Niños” (Zebiak, 1993) is primarily associated with interannual variability, and
40
the warm phase of the mode drives increased convection and precipitation over the
Guinea Coast (Moron, 2001). The pronounced decadal variability in West African rainfall
has been associated with the meridional mode, which controls the cross-equatorial SST
and pressure gradient during the boreal summer (Fontaine and Jacinot, 1996; Camberlin
et al., 2001; Nicholson and Grist, 2007). This sea-level pressure gradient influences the
strength and position of the mid-level African Easterly Jet (AEJ), which along with the
strength and position of the upper level Tropical Easterly Jet (TEJ), control the intensity
and position of deep convection during the monsoon season (Nicholson and Webster,
2007; Nicholson, 2008; 2009a; Dezfuli and Nicholson, 2011). Upper atmospheric
circulation modulates interannual variability in the convective conditions, whereas
moisture is primarily derived from lower level in the atmosphere. There are two sources
of low-level moisture for the WAM: the southerly monsoonal flow, and a distinct, lowlevel westerly jet originating off the coast of West Africa near the Atlantic ITCZ, which
was identified by Grodsky et al. (2003). This low-level jet has been termed the West
African Westerly Jet (WAWJ), and recent work suggests that it is critical in driving
interannual and decadal variability in precipitation in the Sahel (Pu and Cook, 2010;
2012).
Past hydrologic changes in the region have been related to interhemispheric
Atlantic SST patterns as well. Shanahan et al., (2009) showed that lake level at Lake
Bosumtwi was tied to the Atlantic Multidecadal Oscillation (AMO) over the past 2700
years. Cold intervals in the North Atlantic have been associated with extremely dry
conditions in much of West Africa, most notably during Heinrich 1 (Stager et al., 2011);
41
although similar conditions have been observed in both terrestrial and marine proxies
throughout the past 100 ka (Peck et al., 2004; Weldeab et al., 2007; Mulitza et al., 2008;
Tjallingii et al., 2008; Castañeda et al., 2009; Itambi et al.; 2009).
1.3 Past work on Lake Bosumtwi
Lake Bosumtwi has been recognized as a valuable archive of past hydrologic and
environmental change in West Africa since the 1970's. Seminal work by Talbot and
Delibrias (1977) revealed large changes in lake level during the past 12 kyr recorded by
terraces and sediments exposed by river cuts. This led to a series of studies on sediment
cores taken from the lake and the geomorphology of the lake basin. Pollen studies
revealed that arboreal forests, which were abundant in the region during the Holocene,
were scarce in the region during the end of the last glaciation (Maley and Livingstone,
1983; Talbot et al., 1984; Maley 1987, 1989, 1991). Further studies suggested that the
tropical forests of the early Holocene abruptly declined ca. 3 ka (Maley, 1997; 2002).
These changes in paleoecology are consistent with geochemical evidence (elemental C
and N, as well as δ13C and δ15N), which suggested much lower lake levels between 30 and
12 ka, and an extremely wet interval between 12 and 3 ka, before dropping to
intermediate conditions in the late Holocene (Talbot and Johannessen, 1992; Russell et
al., 2003). The extremely wet early to mid Holocene period corresponded to a distinct
interval in the sediment, termed the “sapropel” by Talbot et al. (1984). The interval is
extremely rich in organic material (organic carbon 15-20%), and poor in clastic minerals,
and contains an important component of the filamentous blue-green algae anabeana
42
(Talbot et al., 1984). Terraces, highstand deposits and an overflow notch indicate that the
lake was very high, and occasionally overflowing during the deposition of this layer
(Talbot and Delibrias, 1977, 1980; Shanahan et al., 2006). Lastly, Talbot and Kelts (1986)
investigated the mineralogy and geochemistry of carbonates formed in the water column
or surface sediments of the lake. This work highlighted the extreme variability in both
mineralogy and the δ13C and δ18O of carbonates in the sediment sequence.
A second phase of research on the lake began around the turn of the 21st Century,
with two major emphases. One goal of this research was to conduct further analysis on
sediment cores from the the lake, using new approaches to better understand recent
variability. For example, taking advantage of the annual laminations (varves) persistent
through the late Holocene to reconstruct regional hydrology with unprecedented
resolution. This culminated in a 2700-yr-long varve record that revealed pronounced
mulitdecadal to century-scale variability in regional hydrology associated with changes in
Atlantic SSTs and the Atlantic Multidecadal Oscillation (AMO; Shanahan et al., 2009).
Peck et al. (2004) examined the mineral magnetic characteristics of the lake to reveal
intervals with abundant iron sulfide minerals associated with drought. These intervals are
associated with cold events in the North Atlantic (the Younger Dryas and Heinrich events
1 and 2), indicating that millennial-scale fluctuations in North Atlantic SSTs and glacial
boundary conditions also influence hydrology at the lake. High-resolution seismicreflection data from the lake revealed pronounced, dense layers corresponding to lake
lowstands (Scholz et al., 2002; Brooks et al., 2005). The primary goal of this second
phase of research was to use drill cores to extend the sedimentary record through the
43
entire million-year history of the lake (Koeberl et al., 2007). Initial results from the long
drill cores identified several intervals of severe aridity between 75 and 135 ka, including
the presence of a thick, dense unit at 32 m in the drill cores that corresponds to a lake
wide seismic unconformity (Brooks et al., 2005), and likely resulted from nearly
complete dessication of the lake (Scholz et al., 2007). In this study, we examine the
geochemistry of the upper 160 m of the 296-m sediment sequence at cm-scale, extending
the drill core record to 530 ka.
2. Methods
2.1 Lake Drilling and core processing
In July and August of 2004, the global lake drilling system (GLAD-800) was used
to recover 12 sediment cores from five core sites along a depth transect in the lake
(Figure A3). This study focuses on the 296-m-long core from deep-water (76 m) site 5
(core BOS04-5B), which extends from the present-day lake floor to the brecciated
bedrock, and includes a basal sediment layer of impact-glass bearing accretionary lapilli.
This layer corresponds to the final fallback of material following the meteorite impact,
indicating that the record spans the entire 1.08 Myr history of lacustrine sedimentation at
the lake. After drilling, the cores were shipped to the University of Rhode Island, where
they were split, and described, then imaged and analyzed using a Geotek® multi-sensor
core logger for a suite of magnetic and physical parameters. The paleomagnetic
parameters of the sediments were measured on U-channels taken central axis of the split
core sections (Heil et al., in prep.). The cores were sampled at 4-cm-resolution for the
44
analysis of sediment magnetic hysteresis, x-ray diffraction mineralogy, total organic and
inorganic carbon content, bulk organic carbon and nitrogen isotopes, and grain size. Core
logging and physical sediment properties were used to integrate the depth scales for the
three drill cores from site 5 (BOS04-5A,B&C) into a relative mean core depth (RMCD;
pers. comm. C. Heil).
2.2 Geochronology
The uppermost 22 m (RMCD) of the drill core have been dated by 155
radiocarbon and low-blank radiocarbon ages, as well as varve counts in both late
Holocene and late Pleistocene intervals (Shanahan et al., submitted). The majority of
these ages were measured on piston cores that have been tied into the RMCD scale
(Shanahan et al., submitted). Beyond the limit of radiocarbon dating the age model is
more uncertain; however, we have employed multiple, independent dating techniques to
constrain the timing of sediment deposition in the lake. Optically stimulated
luminescence (OSL) ages were determined in collaboration with S. Forman at the
University of Illinois, Chicago (Shanahan et al., in prep). To avoid age underestimates
that can be associated with anomalous fading of feldspars, we relied on the post-infrared,
blue-light response that is mainly associated with quartz (Banerjee, 2001). Furthermore,
the ages were also corrected for downcore variability in porosity, organic matter and
biogenic silica content, which affect the dose rate (Forman et al., 2007). OSL ages
between 0 and 50 ka were generally in agreement with radiocarbon age estimates (Figure
A4). Below about 60 RMCD (130 ka), measured OSL samples were at or near saturation,
45
and provide only minimum ages, making this the effective limit of the technique for the
drill core.
U-Th dating of authigenic carbonates in the sediments provide additional age
control; however the challenges associated with cleaning and mechanically separating the
fine layers of carbonate to remove detrital Th limits both the precision and capacity of
measuring sediments from deep in deep-water core 5B. Carbonate nodules found in
shallow-water cores are easier to analyze, and more consistently yield usable ages. U-Th
ages from core site 1 are tied into the site 5 chronology by correlating the down-hole
natural gamma logs between the sites (Shanahan et al., in prep).
Paleomagnetic properties of the sediments were used to identify the
Brunhes/Matuyama and Upper Jaramillo geomagnetic reversal boundaries (Heil et al., in
prep). The ages of these events are well known (781 +/- 3 and 988 +/- 3 ka; Horng et al.,
2002), and constrain the chronology deep in the drill core. Finally, the timing of the
meteor impact, and therefore the base of the sediment sequence, has been dated by 40Ar39
Ar and fission track techniques (Koeberl et al., 1997) to 1.08 +/- 0.04 Ma (Jourdan et
al., 2009).
2.3 Scanning X-Ray fluorescence analysis
The bulk elemental abundance of the upper 159 meters of core BOS04-5B was
analyzed using the Itrax® scanning x-ray fluorescence (XRF) analyzer at the Large Lake
Observatory at the University of Minnesota, Duluth. The archive halves of split sediment
cores were analyzed at 1-cm-resolution, with 60 sec count times and a Mo x-ray source
46
operated at 30 kV and 20 mA. The Itrax was also used to generate digital x-radiographs
with 200-μm-resolution in the downcore direction for each of the core segments.
2.4 Carbonate isotope geochemistry
Carbonate mass spectrometry samples have not been measured systematically in
core 5B, however, high-resolution samples in varved intervals between 60.7 - 57.6
(Appendix B) and 1.5 – 0 RMCD (Shanahan, 2006; Shanahan et al., 2009), as well as
bulk sediment samples taken from carbonate-rich intervals, primarily above 50 RMCD
have been analyzed for δ18O and δ13C isotope geochemistry. Dried sediments were
heated under vacuum at 100°C for several hours to remove volatile organic compounds
prior to measurement. The δ18O and δ13C of these carbonates were measured at the
University of Arizona, using an automated carbonate preparation device (KIEL-III)
coupled to a gas-ratio mass spectrometer (Finnigan MAT 252). Powdered samples were
reacted with dehydrated phosphoric acid under vacuum at 70°C. The isotope ratio
measurement is calibrated based on repeated measurements of NBS-19 and NBS-18 and
precision is ±0.10‰ for δ18O and ±0.08‰ for δ13C (1 sigma).
The δ18O of authigenic carbonates in Lake Bosumtwi is a function of the lake
water δ18O (δ18Olw), the water temperature, and the carbonate mineralogy. Whereas
changes in lake water temperature probably exert a minor influence, the extremely
variable carbonate mineralogy in Lake Bosumtwi complicates the interpretation of past
changes in the δ18Olw. At least six carbonate minerals are present in the sediments (calcite,
high Mg-calcite, aragonite, manganosiderite, dolomite and ankerite), which all fractionate
47
oxygen differently during formation. Furthermore, in the cases of calcite and high Mgcalcite, and dolomite and ankerite, the composition of the carbonates exist on a
continuum, varying between two end members, resulting in variable fractionation factors
even within individual mineralogies. To account for this, we calculate δ18Olw for each
sample based on its mineralogy, over a wide (10°C) range in temperatures. This results in
a fair amount of uncertainty, however the variability in carbonate δ18O in Lake Bosumtwi
is extremely large (>10‰), so large-scale trends and rhythms are still apparent. That said,
there are still several problems with the approach. First, the mineralogy of each sample,
much less its precise elemental composition, has not been determined. Furthermore,
whereas fractionation for most of these carbonates have been studied empirically, only
theoretical estimates are available for ankerite, and minerals along the ankerite-dolomite
series (Chacko and Deines, 2008). For this study, we have adopted a simplified approach
for adjusting for mineralogy.
First, manganosiderite-rich intervals are readily identified by high Mn peaks in
XRF data. No other mineral has been identified in the sediments that corresponds to Mn
intensities higher than 1000 cps. Dolomite and ankerite are identified by their very high
δ18O and δ13C values (Appendix B; Talbot and Kelts, 1986). No other carbonate mineral
in the lake has been identified that has δ18O higher than 7‰ and δ13C higher than 15‰.
Moreover, neither dolomite or ankerite has been found with δ18O or δ13C lower than those
values. Samples which do not appear to be either manganosiderite or dolomite/ankerite
are classified as calcite/Mg-calcite/aragonite. There are differences in the fractionation
factors of these three Ca-carbonates, however they most likely result in a maximum offset
48
of ~1‰ (Tarutani et al., 1968; Kim and O'Neill, 1997). Because we cannot reliably
distinguish the minerals in the samples that have been analyzed, we treat calcite, high
Mg-calcite and aragonite as calcite. The differences in fractionation between dolomite
and ankerite are larger, likely up to 4‰, but are extremely uncertain (Chacko and Deines,
2008). We treat all of these samples as dolomite, likely resulting in an underestimation of
δ18Olw for the more ankeritic carbonates. The dolomites and ankerites are consistently
formed in the most 18O-enriched lake waters, so this misrepresentation results in a more
conservative estimate of δ18Olw during lowstands, and ankerite-rich intervals are
interpreted cautiously. Ultimately, δ18Olw is estimated using the simplified mineralogies
are used to account for the different phosphoric acid fractionations (Rosenbaum and
Shephard, 1986; das Sharma, 2002), and water-carbonate fractionations (Carothers et al.,
1988; Kim and O'Neill, 1997; Vasconcelos et al., 2005) for each mineral over a 10°C
range in water temperature.
2.5 Turbidites
The depths and thicknesses of turbidites greater than 1 cm in thickness were
visually identified for the entire length of core BOS04-5B. For the upper 159 meters of
the core, both core face images and high-resolution x-radiographs were used to identify
turbidites, which are generally apparent as fining upward packages of homogenized
sediment with a distinct, dense, light-gray clay cap (Figure A5).
49
2.6 Laminations
Two approaches were used to estimate downcore variability in the presence and
quality of laminations in the sediment. First, core face images were used to visually rank
the laminations for consecutive 10-cm intervals throughout all of core BOS04-5B. Each
interval was subjectively deemed well-laminated (lamination code {LC} = 3),
moderately-laminated (LC = 2), poorly-laminated (LC=1), or unlaminated (LC=0). A
second, objective approach used the x-radiographs that span the upper 159-m of core
BOS04-5B to estimate changes in the presence of laminations by calculating the ratio of
vertical to horizontal variance in consecutive 5-cm intervals of the x-radiograph (Figure
A6). Well-laminated intervals have a lot of vertical variance in the x-radiograph, and very
little horizontal variance, resulting in high vertical to horizontal variance (V:H) ratios.
Alternatively, massive intervals have little variance in both directions, resulting in much
lower V:H ratios (Figure A6). Gaps in the sediment cores can greatly affect this
calculation, and were removed prior to calculation.
3. Results
3.1 Geochronology
To model the age-depth relationship in the drill core and assign ages to depths
over the million-year record we used the “CLassical Age Modelling” (CLAM) approach
of Blaauw (2009). In this technique an ensemble of splines are fit to the ages in the
record, with the the ages for each ensemble member selected stochastically from their
probability distribution functions. For 14C ages, the full calibrated probability
50
distributions are used, for all other ages normal distributions are used with their analytical
standard deviations (Figure A4). There is no evidence for reversals in core BOS04-5B,
therefore we removed all of the ensemble members with age reversals. Because turbidites
make up nearly 25% of core BOS04-5B, and they are deposited extremely rapidly (on the
order of hours to days) when compared to deep-water sedimentation at the site, we
removed turbidites from the depth scale before developing the age model. By fitting
smooth splines through the ages, a built-in assumption with CLAM is that sedimentation
rate changes gradually between age-control points. This approach certainly
underestimates variability in sedimentation rate, but by removing turbidites from the
depth scale, we avoid the biases that would result from the gradual decrease in turbidite
abundance with depth.
The ages and uncertainties of all the dates, as well as the ensemble envelope and
median estimate from CLAM are shown in Figure A5. The density and precision of
geochronometers decreases with depth in the core, and this is reflected in the spread of
model ensembles. The age model is well-constrained for the upper 30 m (RMCD) of the
core by radiocarbon and OSL ages, however below 30 m there are far fewer ages, with
greater uncertainties. Below ~ 75 m (RMCD) the few U/Th ages in the interval have
large error bars, and this uncertainty is reflected in a wide range of age model ensembles.
Geomagnetic reversals and the impact age reduce this uncertainty at the base of the core,
however, the range of age model ensembles exceeds 100 kyr for much of the middle of
the drill core. This range of uncertainty most likely precludes the possibility of useful
millennial-scale comparison between paleohydrological indicators at Lake Bosumtwi and
51
other regional and global records of climate variability during the late Pleistocene.
Therefore, we examined the potential of “tuning” the poorly-dated interval of the record
to another record, and ultimately used a proxy for dust content in the sediments to tune
the record to glacial-interglacial changes in the EPICA Dome C dust record. To minimize
overfitting or confirmation bias, we rely on the well-dated past 200 kyr to validate any
potential matching between records. To be acceptable, any match must be mechanistically
sound, and clearly covarying and persistent over the most recent 200 kyr.
3.1.1 Comparisons with regional and global dust record
Dust export from the Sahara and Sahel varied dramatic during the Pleistocene,
generally in phase with other records of dust variability throughout the world in response
to large-scale changes in global climate (Tiedemann et al., 1994; deMenocal, 1995;
Moreno et al., 2002; Larrasoaña et al., 2003; Moreno 2012). It is reasonable to assume
that glacial-interglacial-scale changes in Saharan dust flux are recorded in the Lake
Bosumtwi sediments as well, and provide a potential means of synchronizing the records.
The ratio of coercivity (Hc) to the coercivity of remanence (Hcr) in the Lake Bosumtwi
sediments is indicative of the magnetic grain size of the sediment (Peck et al., 2004).
Intervals with high (Hc/Hcr) ratios are associated with hematite and goethite, which are
delivered to the lake as aerosols from the Sahara and the Sahel (Peck et al., 2004). Over
the past 200 ka, Hc/Hcr generally covaries with dust flux deposited in the Gulf of Guinea
(ODP 663; Figure A7; deMenocal et al., 1993) and off the west coast of North Africa
(ODP 659; deMenocal, 1995). However, it is difficult to establish robust tie points
52
between the records because of the low resolution of marine cores. Interestingly, dust flux
recorded in the EPICA Dome C (EDC) Antarctic ice core matches fairly well with Lake
Bosumtwi dust content, especially when comparing the general presence and absence of
dust, rather than the amplitude of the peaks (Figure A7). This similarity continues over
the entire 800 kyr of EDC core (Figure A8). Obviously, the source areas and winds that
deliver dust to Antarctica and Lake Bosumtwi are in different hemispheres, and may
respond differently to even global-scale climate changes. Glacial intervals, however, are
characterized by much higher dust flux in both hemispheres (deMenocal et al., 1993,
deMenocal, 1995, Lambert et al., 2008). In West Africa, the increase in dust flux is
associated with drier conditions in North Africa, and an intensification of the winter trade
winds, both of which occur during cold intervals in the North Atlantic (deMenocal and
Rind, 1993; Moreno, 2012). At EDC, the dramatic increase in dust during glacial periods
is believed to be associated with increased aridity in South American source regions, and
a longer lifetime of aerosols in a drier atmosphere (Lambert et al., 2008). The southern
westerlies were likely slower during glacial intervals, however they were also shifted
northward, potentially enlarging the South American source areas (Toggweiler and
Russell, 2008). Dust flux in both regions is sensitive to glacial-interglacial changes in
Atlantic SSTs and probably underwent similar changes at similar times during the
Pleistocene. This, combined with the good visual match between the high-resolution
records allows us to “tune” the Lake Bosumtwi chronology to the commonly-used EDC
chronology.
53
3.1.2 Tuning to EDC timescale
To roughly synchronize Bosumtwi dust record (Peck et al., in prep) to the EDC
dust record (Lambert et al., 2008), we picked 14 tie points at glacial-interglacial
transitions over the past 800 kyr (Figure A8). Because the uppermost sediments are well
dated, we did not allow any adjustment to the chronology between 0 and 70 ka. Given the
tie points, there appears to be no drift or systematic bias in the Lake Bosumtwi
chronology. Five of the tie points were younger in the Lake Bosumtwi chronology, and
eight were older (the tie point at 70 ka is the same in both records). The age model was
then adjusted by stretching, or condensing the curve to fit the tie points, no other tuning
was performed. This approach preserves any relative changes in sedimentation rate
inferred by the original model (Figure A4). The tuned age model falls within the bounds
of the ensemble envelope until near the Brunhes/Matuyama reversal where the ensemble
narrows, and the tuned model slightly exceeds the envelope (Figure A4). Generally, the
tuning increased sedimentation rates during glacial intervals and reduced sedimentation
rates during interglacials. This is consistent with the hypothesis that interglacials are
generally associated with higher lake level at Lake Bosumtwi, since the basin area, and
sedimentation rates are reduced when the lake is high (Shanahan et al., 2008; 2009).
After tuning on glacial-interglacial transitions, the power spectra of lake level
indicators show significant variability at orbital (i.e., eccentricity, obliquity, and
precession) timescales (Figure A9). The presence of 100-kyr cycles could be a function of
the tuning, however the occurrence of significant peaks at the independent obliquity and
precession timescales provide support for the tuning procedure (cf., Imbrie et al., 1984;
54
EPICA, 2004).
3.2 X-ray fluorescence
The relative abundance of Si, K, Ti, and to a certain extent, Fe (Figure A10) are
interpreted as terrigenous elements, following Shanahan et al., (2009). The primary
source of these elements in the lake sediments is terrigenous material that washes in from
the drainage basin during the rainy season, and they comprise a major component of
annual sedimentation in varves in both the late Holocene (Shanahan et al., 2008; 2009)
and the Last Interglacial (Appendix B). Because the terrigenous elements are washed in
from the lake basin, their relative abundance is strongly influenced by the size of the
drainage basin and the proximity of the core site to the shoreline, both of which are
primarily functions of lake level. Therefore, the terrigenous elements are expected to vary
inversely with lake level; most abundant when the lake is low, and least when it is high.
The abundance of terrigenous elements is a robust lake level indicator for the late
Holocene, covarying with carbonate oxygen isotope values, geomorphic lake level
indicators, and a variety of other forms evidence (Shanahan et al., 2009).
Fe is also a terrigenous element, and is an important component of the sediment
that washes into the lake. However, Fe is also present in minerals precipitated in the lake,
or post-depositionally in the sediment, primarily manganosiderites and various species of
iron-sulfide minerals. This complicates the interpretation of the Fe intensities, and we
consider them separately from the other terrigenous elements. Mn in the sediment record
is present in several minerals, but is predominately associated with manganosiderites. Mn
55
is also associated with iron-sulfide minerals and ankerite, but the intensity of the Mn
peaks associated with these minerals is much lower (<100 cps) than with the
manganosiderites (>1000 cps). Therefore, we consider Mn to be a robust indicator of the
presence of manganosiderite.
Ca XRF intensities primarily indicate the presence, and relative abundance of
various species of calcium carbonates in the sediments. There is very little Ca in the
bedrock of the drainage, which is predominantly schist with small outcrops of granite
(Koeberl et al., 1998), and the low intensities of Ca in the terrigenous layers of the
sediment are consistent with this observation. The Ca that is incorporated in authigenic
carbonates in the sediments is most likely originally delivered to the lake as dust. Dust
deposited by northerly Harmattan winds is an important component of Ghanaian soils
(He et al., 2007), and is the likely origin of much of the dissolved Na in Lake Bosumtwi
(Turner et al., 1996b).
Finally, the relative abundance of S in the sediment shows pronounced variability,
and is present in three distinct components of the sediment. For most of the sediment, S
XRF intensities are low, and somewhat noisy, but generally track the abundance of
organic matter in the sediment. Higher S intensities are associated with various the
various iron-sulfide minerals (pyrrhotite), and less frequently, diagenetic gypsum.
3.2.1 Correcting for biases in the XRF data
Scanning XRF analysis of wet, split sediment cores allows for rapid and highresolution determination of relative elemental abundance, and is a particularly useful tool
56
for long sediment sequences. However, XRF intensities can be affected by a number
factors, especially when applied over highly variable lithologies, such as those present in
the core BOS04-5B. One favored approach in dealing with these uncertainties is to
examine elemental ratios; however, in Lake Bosumtwi sediments, the absolute abundance
of terrigenous elements is closely related to lake level (Shanahan et al., 2009). In this
section, we correct for changes in sediment water content and compaction of the core
with depth.
3.2.1.1 X-ray absorption and attenuation by water
One challenge associated with XRF scanning wet sediments is evaluating the
effect of x-ray absorption by interstitial water and water condensed on the film used to
reduce dessication of the cores during analysis (Kido et al., 2006, Tjallingii et al., 2007).
Intervals with elevated water content have consistently lower XRF counts, and the
relative change in counts varies between elements such that lighter elements are more
greatly affected. Both empirical and theoretical corrections have been proposed and
applied to this problem, in both cases the goal is to estimate the counts that would have
been collected from dry sediment. Because we do not have the XRF measurements of
packed, dry sediment necessary to calibrate the record, we use the theoretical correction
of Kido et al. (2006) which uses the known mass attenuation constants of the different
elements to calculate the impact of x-ray absorption and scattering by water. The
correction requires knowledge of the bulk sediment composition and the water content.
Both of these variables can calculated from the XRF data themselves, so we employ the
57
correction iteratively, first calibrating the data, then correcting the XRF intensities, then
recalibrating with the adjusted intensities and reapplying the correction until both the
calibrations and adjusted counts stabilize.
The bulk sediment composition is estimated by simple 3-point calibrations with
standard reference materials (SRMs) that are routinely measured on the Itrax core scanner
at the University of Minnesota, Duluth. The resulting compositional estimates are
probably not very accurate, however uncertainty in the composition has only a minor
effect on the correction of Kido et al. (2006), because the differences in mass attenuation
values for the primary sedimentary constituents are small. On the other hand, the
determination of water content results in more uncertainty in the correction. The water
content of core BOS04-5B has been calculated from the dry and wet bulk densities
measured using on the Geotek Core Scanner, and by loss-in-ignition, respectively. The
ratio of coherent (Rayleigh) to incoherent (Compton) scattering (coh/inc) is strongly
related to bulk density, and is calibrated to sediment water content. Unfortunately, the
water content values used for calibration were measured immediately after the core was
split, whereas the XRF data were collected several years later. This means that the
calibration likely overestimates the water content at the time of measurement, since the
cores had dried somewhat during storage. We estimate the amount of drying as about
30% based on comparison of the cores at the time of XRF analysis with core photos taken
after splitting, and therefore multiply the original water content values by 0.7. This choice
is somewhat arbitrary, however allowing this value from 60 to 80% results in only subtle
changes, especially with regard to relative variability within each estimate (Figure A11).
58
3.2.1.2 Sediment compaction with depth
Unconsolidated sediment compacts as a function of the increasing weight of
overlying material with depth. In marine and lacustrine sediments, this results in a
downcore increase in density and decrease in porosity that changes most rapidly near the
surface (Bahr et al., 2001). Because XRF intensities increase with density, it is
worthwhile to account for the downcore compaction component of changes in wet bulk
density (WBD) of the sediment. This is generally done empirically, by fitting an
exponential (Bahr et al., 2001), logarithmic (Sheldon and Retallack, 2001), or low-order
polynomial (Huybers and Wunch, 2004) to measured WBD or porosity data. These
approaches work reasonably well for cores with relatively little downcore variability in
WBD, such as most marine sediment cores. Core BOS04-5B, however, contains greatly
variable lithologies and WBD throughout the 300 m interval (Pers. comm. J. King).
Attempting to fit an exponential, logarithmic or polynomial to the WBD results in a
compaction curve that either overfits the data, resulting in reversals in the compaction
curve, or fits poorly in the upper 50 meters of the core, the section most strongly affected
by differential compaction. Therefore we adopt a piecewise approach, and fit a piecewise
2nd and 1st order polynomial to the data. To do this, we use a least-squares fitting
algorithm to fit the uppermost part of the curve with a quadratic polynomial, and then at a
depth determined by the algorithm, the function becomes linear with the slope of the
quadratic at the depth where the function switches. This approach has several advantages.
First, it allows the uppermost part of the curve to be fit well, while maintaining the
structure of a simple compaction curve with no negative slopes. Second, the algorithm is
59
non-parametric, in that the only user input was the type (2nd and 1st-order polynomials) of
the function. Secondly, all of the parameters for the equations, including the depth at
which they switch, are chosen by the algorithm to best fit the WBD data. The relationship
between density and XRF counts is complex, varying among different elements and
sediment compositions, however over the relatively small range in density associated
with the core compaction curve, the relationship is probably nearly-linear. Therefore, we
adjust the XRF intensities for downcore compaction by dividing the intensities by the fit
compaction curve.
3.2.2 Principal components analysis
Principal components analysis (PCA) is a technique that can be used to identify
and consolidate the leading modes of variability shared between multiple time series into
several, uncorrelated records (Ghil et al., 2002). The first principle component (PC) of
late Holocene XRF data is strongly associated with the abundance of terrigenous
material, and tracks lake level over the past 2700 years (Shanahan et al., 2009). We
performed PCA on the water-content and core-compaction adjusted XRF intensities of
Al, Si, K, S, Ca, Ti, Mn, Fe, Sr, and Rb. The first PC of the XRF data spanning the past
530 ka (XRF-PC1) explains 39% of the variance, and is strongly associated with the
terrigenous elements Al, Si, K, Ti and Rb (Table 1; Figure A12). The other elements are
primarily loaded on other PCs, except for Mn, which loads negatively on PC1. PC2
explains 28% of the variance, and is dominated by variability in carbonate mineralogy,
with Ca and Sr associated with the various Ca-carbonates in the sediment, and Mn and Fe
60
associated with manganosiderite. The results of the PCA are very similar to those for the
late Holocene varve record (Shanahan et al., 2009), and the whole Holocene XRF record
(Shanahan, 2006) from Lake Bosumtwi.
3.2.2.1 Carbonate dilution of XRF-PC1
XRF-PC1, because it is so strongly driven by changes in the terrigenous content
of core BOS04-5B, likely responds to changes in the availability of terrigenous sediment
supply, and terrigenous drainage area, which at Lake Bosumtwi are primarily functions of
lake level (Shanahan et al., 2008; 2009). Because we interpret XRF-PC1 primarily as a
lake level indicator, it is important to identify and account for processes that affect XRFPC1 that are not necessarily related to changes in the influx of terrigenous sediment.
Dilution of terrigenous material by carbonate is one such process. Figure A13 shows the
correlation between XRF-PC1 and Ca, and Mn, both of which track the abundance of
different carbonate minerals in the sediments. The precipitation of these carbonates can
be related to changes in lake level, however, it is much more likely that anti-correlation of
terrigenous material with Ca and Mn (Figure A13) is due to dilution, rather than a
prolonged reduction of terrigenous influx into the lake. To avoid misinterpreting these
decreases in XRF-PC1, we remove this dilution effect by fitting a line to the Ca-PC1 and
Mn-PC1 trends where Ca and Mn are particularly abundant. The relationship between Ca
and Mn and XRF-PC1 is not significant at low Ca and Mn abundances (i.e., where there
is little or no carbonate). Consequently, we only adjusted XRF-PC1 above intensities
where the anti-correlation is statistically significant (above 260 and 160 cps, respectively;
61
Figure A13). The corresponding XRF-PC1 data are then adjusted, following the fit line.
The adjusted correlations are shown in Figure A13. This adjustment has a minor impact
on most of the data points in the record. Forty-one percent of the XRF data are adjusted,
however the adjustment is very small for most points, only 12% are adjusted by more
than one unit, and 5% by more than two.
The net effect of all of the corrections applied to the XRF data can be examined
by comparing Figures A10a and b. The primary difference is the removal of the long-term
decrease in counts towards the top of the core, the smaller-scale and deeper variability in
each element is generally unaffected. For the most part, the corrections have little
influence on the main conclusions. However, they do influence the interpretation of the
inferred wetness of the past 75 kyr relative to the rest of record. This interval is most
influenced by the lack of sediment compaction, and lake level appears artificially high in
the uncorrected data.
3.3 Lamination indices
The subjective lamination index and the logarithm of the V:H density variance
ratio are generally consistent over the upper 160 m of core BOS04-5B (Figure A14).
Taken together, the two indices reveal significant variability in the presence of well
preserved laminations, which are present for most of the upper half of the sediment
sequence. Laminations are formed when the lake floor is anoxic, limiting bioturbation.
XRF Mn and Ca intensities, indicative of carbonate abundance in the sediment, generally
track the lamination indices, and are more abundant when laminations are present.
62
Carbonates often form fine laminations in the sediment, so the two indicators are not
independent. Much of the carbonate found in the sediments from the deep-water core
sites is interpreted to have formed in anoxic water, at or near the sediment-water interface
(Talbot and Kelts, 1986). Covariance between carbonate abundance and laminations
indices supports the hypothesis that the indices record variability in the oxygenation state
of the lake bottom waters.
3.4 Turbidites
Turbidite abundance generally decreases upcore, from an average of 41 cm per
meter over the bottom 50 m to an average of 3 cm per meter for the upper 50 m (Figure
A15). This likely reflects lake basin evolution the past million years (Peck et al., 2006; in
prep). Early in the lake's history, the crater sides were steeper, and the crater floor was
300 m deeper. As the margins eroded and filled the basin with lake sediment, the relief
decreased, as did the probability of mass wasting events. Considerable shorter-term
variability is overprinted on the gradual decreasing trend, suggesting that lake level and
climate variability also influenced the turbidite abundance. Over the past 530 ka, this
shorter-term variability in turbidite abundance shows a remarkable correspondence with
Mn abundance (Figures A16 and A17), as well as organic matter content in the core (pers.
comm C. Scholz). Mn indicates the presence of manganosiderite, in the interval, which
along with organic carbon, is an indicator of high lake level. It is important to note that
the manganosiderite does not occur in the turbidites, nor is it consistently deposited
immediately before or after a turbidite. Therefore, we do not interpret this co-occurrence
63
as a direct cause and effect relationship. That is, the manganosiderite is not being
deposited because a turbidite was deposited. A lag correlation analysis (Figure A17)
suggests that on average, Mn deposition leads turbidite deposition by 1000-3000 years.
We hypothesize that turbidites are associated with terminations of highstands. During a
sustained highstand, deltas and terraces build up high on the crater walls, as lake level
falls, these deposits become unstable and prone to mass wasting, leading to increased
turbidite abundance.
4. Discussion
4.1 Past 530 ka
4.1.1. 400-kyr eccentricity forcing
The magnetic hysteresis parameter Hc:Hcr, which is primarily a proxy for dust
content in Lake Bosumtwi, is shown along with EDC dust concentration in Figure A16
(Lambert et al., 2008, Peck et al., in prep). The similarity between these records were
used to “tune” the Bosumtwi chronology at glacial-interglacial boundaries, so leads or
lags between the records cannot be evaluated. The two records often covary ten thousand
year timescales, likely tracking changes in dust flux between stadials and interstadials.
The greatest mismatch between the records is found between 250 and 150 ka, where
Hc:Hcr is generally much higher than EDC dust. This suggests increased dust flux in
West Africa, relative to the high latitudes of the southern Hemisphere, during most of
Marine Isotope Stages (MIS) 6 and 8 relative to other glacial intervals during the past 530
ka. During particularly arid conditions greigite forms in Lake Bosumtwi sediments,
64
which also increases Hc:Hcr (Peck et al., 2004). Regardless, the mineral magnetic
indicators suggest particularly arid conditions during this interval.
The XRF-PC1 record, an indicator of terrigenous content and lake level at Lake
Bosumtwi, also suggests generally dry conditions between 300 and 100 ka. This interval
of lower lake levels, combined with wetter conditions from 530 to 300 ka, and since 100
ka suggests changes in background aridity in the region with a periodicity of ~400-kyr
(Figure A16). To compare the XRF-PC1 records to the 400-kyr of orbital eccentricity, we
isolate the 400- and 100-kyr components of eccentricity over the past 1.1 Myr using
singular spectrum analysis (SSA), a non-parametric technique that extracts the primary
modes of periodic or semi-periodic variability from a timeseries (Ghil et al., 2002). The
400- and 100-kyr components of eccentricity make up 41 and 54% of the variance in
eccentricity over the past 1.1 Myr, and are shown in Figure A16. XRF-PC1 shows a close
correspondence to 400-kyr component of orbital eccentricity; the dry interval corresponds
to the interval of highest eccentricity over the past 530 kyr. Deeper in the core, variability
in the presence of laminations (Figure A18) and organic carbon content (not shown; pers.
comm C. Scholz) suggest a similar trend during the penultimate 400-kyr eccentricity
cycle.
Superimposed on the 400-kyr change in background state is shorter-term
variability that corresponds to the 100-kyr component of eccentricity, but with the
opposite sign (Figure A16). Whereas wet conditions are associated with low eccentricity
on the 400-kyr timescale, they correspond with high-eccentricity on the 100-kyr
timescale. This complex relationship likely results from the interplay of local orbital
65
forcing with glacial boundary conditions. We suggest that the inverse relationship
between Lake Bosumtwi hydrology and 400-kyr-scale eccentricity is driven by
eccentricity's modulation of precession, following the hypothesis of Scholz et al. (2007).
During intervals of high eccentricity (e.g. 100-300 ka), the amplitude of seasonal
insolation changes on precessional timescales is substantially increased. Both the West
African and Asian Monsoon systems intensified and expanded northward during the
high-eccentricity Last Interglacial (Miller et al., 1991; Szabo et al., 1995; Wunneman et
al., 2007; Wang et al., 2007, Drake et al., 2011). An intensified and northward shifted
monsoon would impact the hydrology at Lake Bosumtwi (and more generally, the Guinea
Coast) in several ways. Lake Bosumtwi presently has a double peak in precipitation,
associated with the northward passage of the rainbelt in May, June and July, and the
southward passage during September and October (Figure A2). July and August are dry;
the core of the rainbelt is situated to the north and intensified southeasterly trades
increasing upwelling along the Guinea Coast, lowering SSTs in the Gulf of Guinea, and
suppressing convection over the Guinea Coast (Moron, 1994; Opoku-Ankomah, 1994;
Nicholson, 2009b). The WAM was probably most intensified and north-shifted when an
eccentricity-amplified precession cycle drove insolation maxima during July and August.
Under such conditions, the rainbelt would most likely spend more time north of the
Guinea Coast, shifting the first rainy season earlier in the year and the second rainy
season later in the year. This would move rainy seasons away from the peak monsoon
season, and intensify the southeasterly trades during July and August, lengthening the
summer dry season at Lake Bosumtwi. This longer dry interval occurring at the same
66
time as the season of highest solar insolation would also drive increased evaporation from
the lake.
Redistribution of tropical moisture from equatorial regions towards the subtropics
on 400-kyr timescales is also observed in anti-phased dust records from the Atlantic and
Mediterranean (Tiedemann et al., 1994; Larrasoaña et al., 2003), consistent with our
observations from Lake Bosumtwi. The hypothesis described above could explain dry
conditions during precession minima (JJA insolation maxima), but not the generally dry
conditions throughout the rest of a precession cycle. High-resolution studies of the
Holocene (Shanahan, 2006) and the Last Interglacial (Appendix B) suggest that the
wettest conditions at the lake occur when the annual insolation maximum occurs during
one of the two rainy seasons (i.e, June-July or September-October). However, if the mean
position of the rainbelt moves north and south in response to precessional forcing,
seasonality of precipitation would vary in counterpoint with seasonality of maximum
insolation maximum (Figure A19). For example, when insolation peaks in midsummer,
the rainbelt would cross the Guinea Coast earlier and later in the year, moving rainy
seasons away (temporally) from the season of highest insolation. When insolation peaks
in May or October, the northward extent of the WAM is probably reduced, and the
rainbelt would cross the Guinea Coast later in the season, moving the rainy season closer
to mid-summer, but once again, away from the season of highest insolation. Following
this model, twice during each precession cycle the season of insolation maximum will
coincide with one of the rainy season, consistent with what is observed in the Last
Interglacial and the Holocene. However when eccentricity is higher, seasonal differences
67
in insolation are amplified, and the period when insolation maxima and the rainy seasons
coincide may be much shorter (Figure A19). If true, this implies that for most of a
precession cycle, the season of highest insolation occurs during a dry season, and
maximum seasonal insolation is higher at higher eccentricities, potentially resulting in
increased evaporation from the lake.
4.1.2 100-kyr eccentricity forcing and glacial boundary conditions
The 100-kyr periodicity of Lake Bosumtwi lake level is most likely due to
changes in glacial boundary conditions. The West African Monsoon is sensitive to the
interhemispheric temperature gradient in the Atlantic Ocean, and sea surface
temperatures (SSTs) in the North Atlantic on timescales ranging from annual to
millennial (e.g., Mulitza et al., 2008; Shanahan et al., 2009; Stager et al., 2011). Sustained
periods of cold SSTs in the North Atlantic, most notably Heinrich Events, have a
dramatic influence on African hydrology, driving drought at Lake Bosumtwi (Peck et al.,
2004) and many other sites in Africa (Collins et al., 2010; Stager et al., 2011). Periods of
low-eccentricity on the 100-kyr timescale correspond with deep glacial conditions in the
North Atlantic, and drought at Lake Bosumtwi, most likely through the influence of cold
Atlantic SSTs.
4.1.3 West African contribution to global atmospheric methane concentrations
Obliquity and precession-scale periods are also observed in the power spectra of
XRF-PC1, Mn and Ca (Figure A9). The role of precession can be observed in the
68
similarity between atmospheric methane and Mn and Ca abundance in the sediments
(Figure A16d). Tropical West Africa is an important methane-producing region (Lelieveld
et al., 1998; Frankenberg et al., 2005, Singarayer et al., 2011), and variability in the West
African Monsoon drives changes in the extent and moisture balance of African wetlands.
Precession-scale variability in ice-core methane records is commonly attributed to
changes in Northern Hemisphere monsoon systems (e.g., Ruddiman and Raymo 2003,
Loulergue et al., 2008) and this hypothesis is consistent with records of the Asian
Monsoon inferred from speleothems over the past 380 ka (Wang et al., 2008; Cheng et
al., 2009). The Lake Bosumtwi record offers the first opportunity to evaluate the role of
West Africa with a moisture-sensitive record form a terrestrial site over the past 530 kyr.
XRF-PC1, and especially the abundance of Mn and Ca in the sediments show a close
correspondence with atmospheric methane. Mn and Ca track carbonate mineralogy in
core BOS04-5B, Mn is almost exclusively associated with manganosiderite, and Ca is
associated with the calcium carbonate minerals calcite, aragonite and dolomite.
Manganosiderite and the calcium carbonates are not deposited at the same time, rather,
the ratio of dissolved Fe2+ and Ca in the lake water controls which type of carbonate is
deposited (Berner, 1971). We interpret manganosiderite as a wet indicator for several
reasons. To reach sufficiently high Fe/Ca ratios (i.e., > 0.05; Berner 1971), most likely
requires a sustained anoxic conditions at the lake floor, supported in part by increased
deposition of organic material, and a reduction of eolian Ca supply to the lake. These
conditions are consistent with high lake level and a generally wetter climate. This is also
supported by other evidence: during the Holocene manganosiderite was only deposited
69
when the lake was very high, and other lake level indicators, such as varve thickness
(Appendix B), δ18O of carbonates (Figure A20) and δD leafwax (pers. comm. T.
Shanahan) all confirm that manganosiderite is only deposited during relatively high lake
level. Manganosiderite was not deposited when the lake was overflowing during the early
Holocene, probably due to a reduction in bicarbonate driven by a freshening of the lake.
The majority of carbonates in Lake Bosumtwi sediments are formed near or just
below the sediment-water interface under anoxic conditions, their formation partially
driven by methanogenesis (Talbot and Kelts, 1986). Consequently, carbonates are rarely
found in unlaminated intervals when the bottom waters were at least seasonally
oxygenated. Perhaps, during intervals of lower lake level, calcium carbonate deposition
primarily occurs nearer the shoreline. This is consistent with the observation of abundant
calcite nodules in intervals of the shallow-water drill cores (pers. comm. J. Peck).
Therefore, the range of calcium carbonate minerals comprising the XRF Ca record are
interpreted as indicating intermediate lake level. Calcite and aragonite are probably
indicative of higher lake levels than Mg-calcite and dolomite, but these cannot be easily
distinguished in the XRF data.
Intervals with abundant Mn are consistently associated with high atmospheric
methane concentrations, and peaks in Mn often occur during precessional-scale peaks in
CH4 (Figure A16d). Typically, smaller peaks in CH4 correspond to peaks in Ca, although
some of the high methane peaks during MIS 7 and 5 are associated with Ca, rather than
Mn. This is part of the general trend of lower lake level during the high-eccentricity
interval from 300-100 ka, where Mn is less common. The influence of the 400-kyr
70
eccentricity component is clear during low-CH4 intervals as well. Calcium carbonates
were deposited throughout most of MIS 12, 10 and 2, however, there are prolonged
intervals of unlaminated sediment with no carbonates during parts of MIS 8 and 6,
suggesting very dry conditions at the lake.
The close correspondence between the Mn and Ca and EDC CH4 over the past
530 ka suggests that the contribution of CH4 from sub-Saharan Africa is generally in
phase with other Northern Hemisphere methane and sources. This conclusion could be
biased by the fact the chronologies are not independent, however the two appear in phase
during the well-dated and untuned most recent 50 ka, and it seems improbable that the
close correspondence of high-frequency peaks is spurious, given that the dust records
were only synchronized on glacial-interglacial boundaries. On longer timescales,
however, tropical West Africa seems to have had a variable contribution to atmospheric
methane, as the conditions at the lake were consistently drier between 300 and 100 ka
than at other times. The 400-kyr eccentricity cycle is not an important component of
atmospheric methane variability over the past 800 kyr (Lambert et al., 2008). This likely
means one of two things. First, it is possible that this pronounced 400-kyr variability is
restricted to the Guinea Coast region, which provides a relatively small contribution to
the global methane budget; the modelling study of Singarayer et al. (2011) suggests that
sub-Saharan North Africa is responsible for about 5-10% of global emissions. If true,
low-frequency changes in regional methane production might be overwhelmed by the rest
of the global production. However, an expansion of the rainbelt, and increased
precipitation in the subtropics at the expense of the tropics would affect a much greater
71
proportion of the global methane producing regions. This possibility is supported by
other long equatorial paleohydrological records that record drier conditions in past higheccentricity intervals (i.e., the Last Interglacial Period), in both Africa (Stone et al., 2011)
and South America (Bush et al., 2004, Hanselman et al., 2005). In this case, the reduced
contribution of tropical wetlands during high-eccentricity intervals may be compensated
by increases in subtropical moisture.
Despite the fact that precession-scale variability in moisture balance at Lake
Bosumtwi appears to be in phase with global CH4 concentrations, the influence of
eccentricity and glacial-interglacial boundary conditions on regional hydrology highlights
the complexity of the influence of orbital precession on the WAM. Global CH4 emissions
are partly driven by hydrological variability in disparate regions, that each have their
own, often complex relationship with orbital forcing (Singarayer et al., 2011), and for
more equatorial regions, these relationships remain largely unknown. Therefore, it seems
premature to assume that an in-phase relationship between Northern Hemisphere
monsoon systems and precession was present throughout the late Quaternary, or to use
that assumption to adjust a geochronology (cf., Ruddiman and Raymo, 2003), especially
since the late Holocene (6ka to preindustrial) increase in methane does not appear to have
resulted from anthropogenic activities, invalidating the relationship between methane and
Northern Hemisphere insolation in the Holocene (Singarayer et al., 2011).
4.2 Latitudinal variability in West African climate over the past 150 ka
Over the past 150 kyr, moisture balance at Lake Bosumtwi, as recorded by XRF-
72
PC1 and δ18Olw, is generally consistent with ocean core records of terrestrial moisture
variability from West Africa (Figure A20). A latitudinal profile of the available records
that have been interpreted to record West African moisture variability over the past 150
kyr indicates a latitudinally varying response to North Atlantic boundary conditions (i.e.,
temperature and ice volume) and local insolation forcing. Strong similarities with the
NGRIP δ18O record in the northern-most West African records gradually become more
similar to precessional-scale changes in insolation at lower latitudes (Figure A20). The
variable response of the records during two high-obliquity and high-precession (low NH
insolation) intervals from about 100 to 85 ka and 55 to 40 ka are particularly interesting.
During both intervals obliquity was high and Greenland temperatures were relatively
warm, whereas summer insolation at 10°N was low. During the interval between ca. 50
and 40 ka, the most northerly records (> ~10°N) appear near-average in terms of moisture
balance, whereas the Sahelian records suggest that effective moisture during this interval
was high, in some cases comparable with the Holocene (deMenocal, 1995; Zabel et al.,
2001; Castañeda et al., 2009). Records sensitive to moisture in more southerly areas, such
as the Congo and Sanaga river basins (Schneider et al., 1997; Weldeab et al., 2007), show
very dry conditions from ~50-40 ka, tracking low-latitude NH summer insolation. Lake
Bosumtwi lies on the boundary between these two responses; XRF-PC1 suggests wet
conditions from about 48-43 ka, right through the minimum in local summer insolation.
This is, however superimposed drier conditions for the rest of the interval.
A similar orbital configuration occurred ca. 90 ka, however the West African
response was quite different (Figure A20). The influence of increased obliquity is evident
73
in several of the more northerly records, but, all of the records suggest below average or
very dry conditions during the interval, especially centered on the low latitude Northern
Hemisphere summer insolation minimum ca. 95 ka. Extreme aridity during this general
interval (115 – 90 ka) is a common feature of paleohydrological records from both East
and West Africa, to latitudes as far a south as the Northern Kalahari Desert (Scholz et al.,
2007; Blome et al., 2012). AMOC during the interval appears to have been comparable to
50 ka (Castañeda et al., 2009; Lopes dos Santos et al., 2010), so the most likely cause of
the differing responses is the difference in amplitude of the precessional forcing. The
very low summer insolation (15-20 W m-2 less than 45 ka) between the two responses
seems to have overwhelmed the influence of the North Atlantic between 100 and 85 ka
(Figure A20).
4.3 Millenial-scale events during the past 130 ka
In the North Atlantic, the last glacial period was characterized by abrupt,
millennial-scale changes in temperature; most notably, Dansgaard-Oeschger (D-O) events
(Severinghaus and Brook 1999; Masson-Delmotte et al., 2005). Many of the D-O warm
events occurred in sequences of progressively cooler peaks, terminating in a Heinrich
event, periods of iceberg discharge that were the among the coldest intervals the North
Atlantic, and are believed to have resulted in a weakening of AMOC (McManus et al.,
2004). D-O and Heinrich events have a global signature (Overpeck and Cole, 2006), and
are often associated with past arid intervals in the tropics (Peterson et al., 2000; Lea et al.,
2003) including West Africa (Peck et al., 2004; Mulitza et al., 2008; Collins et al., 2010;
74
Stager et al., 2011). The climate dynamics underlying this connection remain uncertain;
in parts of the tropics it seems likely that the aridity was associated with a southward shift
in the ITCZ, a hypothesis that is supported by climate models (Broccoli et al., 2006).
Whereas there is strong evidence that ITCZ shifted south of northern Lake Malawi during
Heinrich Events (Brown et al., 2007; Johnson et al., 2011), there is little evidence of
wetter conditions in southern Africa associated with this shift (Stager et al., 2011). In fact,
virtually all records from both hemispheres suggest drier conditions during Heinrich 1,
suggesting a weakening or contraction, instead of, or at least in addition to, a shift in the
rainbelt (Collins et al., 2010; Stager et al., 2011).
Cold intervals in the North Atlantic consistently coincide with intervals of low
lake at Lake Bosumtwi over the past 130 ka (Figure A21). Furthermore between 60 and
30 ka, each arid Heinrich interval is preceded by gradually decreasing millennial-scale
wet peaks that may correspond with D-O cycles. This link between West African
hydrology and North Atlantic temperatures is also found in records to the north (off the
coast of Senegal; Mulitza et al., 2008) and south (Sanaga River discharge; Weldeab et al.,
2007) for at least the past 60 kyr. This provides support for the hypothesis that for
Heinrich Events 1 – 5, West Africa experienced a contraction, rather than a shift, of the
rainbelt. The Younger Dryas (YD; or Heinrich Event 0), appears to have had a much
more limited impact on the region. At Lake Bosumtwi, the interval was a relative
lowstand, but lake level was still ~30 m higher than modern levels (Peck et al., 2004;
Shanahan et al., 2006). XRF-PC1 suggests that the interval was particularly wet, although
this result is likely biased by the presence of manganosiderite during the interval, and not
75
immediately after in the early Holocene, when the lake was overflowing. Nevertheless,
the presence of manganosiderite, the δ18Olw and the low terrigenous content all agree that
the Younger Dryas was wet in the context of the past 130 ka (Figure A21), however it
was still drier than the preceding Bølling-Allerød period and the early Holocene (Peck et
al., 2004; Shanahan et al., 2006).
4.4 Intervals of extreme aridity
One pronounced feature of core BOS04-5B is the presence of several thick, dense,
organic-poor clay intervals. These intervals are massive, and are almost exclusively
comprised of fine-grained terrigenous material; organic carbon content drops below 2%
in the intervals, and they are typically carbonate-free (Scholz et al., 2007; Peck et al., in
prep.). Root casts, and filled mud cracks are also observed in some intervals (pers. comm.
J. Peck). We interpret these deposits as evidence for extremely low lake-level. The lack of
evaporite minerals (neither halites nor evaporitic carbonates are found in the sediments)
suggest that the lake did not completely dry up, but the drill site may have been exposed
seasonally. The upper most of these layers, Arid Interval 1 (AI-1), was described by
Scholz et al., (2007) and is fairly well dated by nearby OSL and U/Th ages, suggesting
that the layer was deposited between 74 to 70 ka. The duration of the event is not well
constrained, as sedimentation rates may have been significantly higher than the mean
during the interval. Given the observed relationship between North Atlantic temperatures
and hydrology at Lake Bosumtwi, it seems most likely the event coincides with Heinrich
7. The OSL and U/Th ages, as well as this correlation with Heinrich 7 suggest that the
76
event also occurred at about the same time as the eruption of the Toba super-volcano,
74+/-2 ka (Oppenheimer, 2002). The Toba eruption is the largest known eruption in the
Quaternary, and is thought to have had a severe impact on global climate for at least
decades, and possibly accelerated global cooling into glacial conditions (Rampino and
Self, 1992; Zielinksi et al., 1996; Jones et al., 2005). It remains unclear, however, the
extent of global change resulting from the eruption (Oppenheimer, 2002). We cannot rule
out the possibility that AI-1 resulted in part from the Toba eruption. However, the
strongest argument against such a possibility is the presence of five other similar units
deposited over the past 530 ka, which were, for the most part, deposited during intervals
with similar orbital forcing and glacial boundary conditions (Figure A22).
All six intervals of extreme aridity of the past 530 kyr occurred between 300 and
50 ka, when the 400-kyr component of orbital eccentricity was highest (Figure A16). Five
of the six intervals also occurred during glacial conditions, and they all occurred when
the global temperatures were cooler than average, and when global CH4 production was
low (Figure A22). AI-1, 2, 3, and 6 occurred during precession-driven insolation minima;
however that does not appear to be a prerequisite, since AI-4 and 5 were deposited near
insolation maxima. That said, uncertainty in the age model prevents more certain analysis
of the phasing between the events and precession. AI-2 was deposited during an interval
when severe aridity was common throughout Africa (Scholz et al., 2007; Blome et al.,
2012). The lower section of AI-2 contains unlaminated deposits of ankerite, and is the
most ankerite-rich interval of core BOS04-5B. Ankerite is indicative of very dry
conditions, and the carbonate geochemistry of those ankerites suggest a lake water δ18O
77
of at least 10-13‰ (SMOW; Figure A21), although the oxygen fractionation during
ankerite precipitation is uncertain, and is probably a minimum value (Chacko and Deines,
2008). AI-3 and 4, deposited during MIS 6, are the thickest arid intervals, and correspond
to the longest period of low organic and inorganic deposition in the past 530 kyr. AI-5
occurred during MIS 7, and is most similar to AI-1. AI-1 and 5 are particular interesting
in that they both appear to have occurred soon after, and immediately preceding wet
conditions at the lake. In addition to the presence of manganosiderite and organic rich
sediment, low lake water δ18O values and the occurrence of rain forest pollen taxa (pers.
comm. W. Gosling) further suggest that the the change from immediately before to AI-1
was dramatic, and probably occurred on the order of centuries. Further research into the
nature and timing of the transitions into and out of these severe arid intervals is a subject
of ongoing investigation.
5. Conclusion and summary
In this study, we presented the first high-resolution, terrestrial record from West
Africa spanning the last 530 kyr.

The chronology of the record is well constrained over the past 150 kyr by
14
C, OSL and U/Th ages, and is constrained deeper in the core by
geomagnetic reversal ages and the age of meteor impact. However, the
chronology of the middle of the sediment sequence is less wellconstrained. To allow comparisons with regional records, climatic forcings
and boundary conditions we use the strong relationship between mineral-
78
magnetic dust proxies in Lake Bosumtwi and dust content in Antarctic ice
cores to tune the record on glacial-interglacial transitions, similar to the
approach of EPICA (2004).

XRF-inferred elemental abundance and carbonate geochemistry and
mineralogy are indicative of lake level and hydrologic balance at the lake,
and indicate an interplay of orbital forcing and Atlantic SSTs boundary
conditions over the past 530 kyr.

Lake Bosumtwi hydrology tracked the 400-kyr component of orbital
eccentricity, with substantially drier background conditions during the
interval of high eccentricity between 300 and 100 ka.

On 100-kyr timescales, lake level had a positive relationship with
eccentricity, with generally wetter conditions occurring during
interglacials and eccentricity highs, although the relative wetness of each
glacial and interglacial period varied substantially, modulated by the 400kyr component of eccentricity.

On shorter timescales, Lake Bosumtwi hydrology appears to have been a
balance between high- and low-latitude forcing. Over the past 150 kyr,
West African hydrology shows a latitudinally varied response to these two
forcings; more northerly sites are more sensitive to obliquity and
temperatures in the North Atlantic, whereas more southerly sites respond
more to local insolation and precession forcing. Lake Bosumtwi typically
lies between these two responses. Interestingly, the amplitude of the
79
forcing influences the latitudinal extent of the response: only more
southerly sites responded to the local insolation low ca. 45 ka; whereas all
of West Africa dried during the more severe decrease ca. 90 ka.

Abrupt, millennial-scale climate changes in the North Atlantic (i.e.,
Dansgaard-Oeschger and Heinrich Events) are recorded as abrupt changes
in lake level in Lake Bosumtwi. As is common at several sites in both East
and West Africa, Heinrich events consistently correspond to dry intervals
at Lake Bosumtwi. There is some evidence to suggest that D-O cycles may
correspond to progressively drier wet-intervals preceding each Heinrich
Event, but uncertainty in the chronology precludes conclusive evidence of
this coupling. Millennial-scale variability is greatly reduced during the
penultimate glacial period, suggesting a variable influence of abrupt events
in the North Atlantic, probably driven by changes in the amplitude of
orbital forcing.

Six multi-millennial scale intervals of extreme aridity were recorded in the
sediments during the past 530 ka, all during the eccentricity-high interval
from 300-50 ka. These events share common features, typically occurring
during deep-glacial conditions, and often coinciding with local summer
insolation minima. Two of the events (AI-1 and 5), appear to have began
and ended abruptly, with wetter-than-average conditions occurring
immediately before and after the arid intervals. Sedimentation rates into
and out of the transitions are uncertain, but the wet-to-dry shift most likely
80
occurred on the order of centuries or less. The amplitude, duration, and
most importantly, the climate dynamics underlying these abrupt changes is
an area of continuing investigation.
6. References
Almond, M. J. R., and R. E. Hecky (2000), Water Column Properties of Lake Bosumtwi,
Ghana, Proceedings of the 45 th Conference on Great Lakes Research., pp. 2–3.
Bahr, D. B., E. W. . Hutton, J. P. . Syvitski, and L. F. Pratson (2001), Exponential
approximations to compacted sediment porosity profiles, Computers &
geosciences, 27(6), 691–700.
Banerjee, D. (2001), Supralinearity and sensitivity changes in optically stimulated
luminescence of annealed quartz, Radiation measurements, 33(1), 47–57.
Berger, A., and M. F. Loutre (1991), Insolation values for the climate of the last 10
million years, Quaternary Science Reviews, 10(4), 297–317.
Berner, R. A. (1971), Principles of Chemical Sedimentology, McGraw-Hill Inc.,US.
Blaauw, M. (2010), Methods and code for “classical” age-modelling of radiocarbon
sequences, Quaternary Geochronology, 5(5), 512–518.
Blome, M., A. Cohen, C. Tryon, A. Brooks, and J. Russell (in press), The environmental
context for the origins of modern human diversity: a synthesis of regional
variability in African climate 150,000-30,000 years ago, Journal of Human
Evolution.
Broccoli, A. J., K. A. Dahl, and R. J. Stouffer (2006), Response of the ITCZ to Northern
Hemisphere cooling, Geophys. Res. Lett, 33(10.1029).
Brooks, K., C. A. Scholz, J. W. King, J. Peck, J. T. Overpeck, J. M. Russell, and P. Y. .
Amoako (2005), Late-Quaternary lowstands of lake Bosumtwi, Ghana: evidence
from high-resolution seismic-reflection and sediment-core data, Palaeogeography,
Palaeoclimatology, Palaeoecology, 216(3), 235–249.
Brown, E. T., T. C. Johnson, C. A. Scholz, A. S. Cohen, and J. W. King (2007), Abrupt
change in tropical African climate linked to the bipolar seesaw over the past
55,000 years, Geophysical Research Letters, 34(20), L20702.
81
Bush, M. B., P. E. De Oliveira, P. A. Colinvaux, M. C. Miller, and J. E. Moreno (2004),
Amazonian paleoecological histories: one hill, three watersheds,
Palaeogeography, Palaeoclimatology, Palaeoecology, 214(4), 359–393.
Camberlin, P., S. Janicot, and I. Poccard (2001), Seasonality and atmospheric dynamics
of the teleconnection between African rainfall and tropical sea-surface
temperature: Atlantic vs. ENSO, International Journal of Climatology, 21(8), 973–
1005.
Carothers, W. W., L. H. Adami, and R. J. Rosenbauer (1988), Experimental oxygen
isotope fractionation between siderite-water and phosphoric acid liberated CO2siderite, Geochimica et Cosmochimica Acta, 52(10), 2445–2450.
Castañeda, I. S., S. Mulitza, E. Schefuss, R. A. . dos Santos, J. S. . Damsté, and S.
Schouten (2009), Wet phases in the Sahara/Sahel region and human migration
patterns in North Africa, Proceedings of the National Academy of Sciences,
106(48), 20159–20163.
Chacko, T., and P. Deines (2008), Theoretical calculation of oxygen isotope fractionation
factors in carbonate systems, Geochimica et Cosmochimica Acta, 72(15), 3642–
3660.
Chang, C.-P., Y. Ding, and N.-C. Lau (2011), The Global Monsoon System: Research and
Forecast, World Scientific.
Cheng, H., R. L. Edwards, W. S. Broecker, G. H. Denton, X. Kong, Y. Wang, R. Zhang,
and X. Wang (2009), Ice age terminations, Science, 326(5950), 248–252.
Cohen, A. S., J. R. Stone, K. R. M. Beuning, L. E. Park, P. N. Reinthal, D. Dettman, C. A.
Scholz, T. C. Johnson, J. W. King, and M. R. Talbot (2007), Ecological
consequences of early Late Pleistocene megadroughts in tropical Africa,
Proceedings of the National Academy of Sciences, 104(42), 16422.
Collins, J. A. et al. (2010), Interhemispheric symmetry of the tropical African rainbelt
over the past 23,000 years, Nature Geoscience, 4(1), 42–45.
Cook, K. H., and E. K. Vizy (2006), Coupled model simulations of the West African
monsoon system: Twentieth-and twenty-first-century simulations, Journal of
climate, 19(15), 3681–3703.
Cruz, F. W., S. J. Burns, I. Karmann, W. D. Sharp, M. Vuille, A. O. Cardoso, J. A. Ferrari,
P. L. Silva Dias, and O. Viana (2005), Insolation-driven changes in atmospheric
circulation over the past 116, 000 years in subtropical Brazil, Nature, 434(7029),
82
63–66.
Cruz, F. W., M. Vuille, S. J. Burns, X. Wang, H. Cheng, M. Werner, R. Lawrence
Edwards, I. Karmann, A. S. Auler, and H. Nguyen (2009), Orbitally driven east–
west antiphasing of South American precipitation, Nature Geoscience, 2(3), 210–
214, doi:10.1038/ngeo444.
deMenocal, P. B., and others (1995), Plio-pleistocene African climate, Science,
270(5233), 53–59.
deMenocal, P. B., and D. Rind (1993), Sensitivity of Asian and African Climate to
Variations in Seasonal Insolation, Glacial Ice Cover, Sea Surface Temperature,and
Asian Orography, Journal of Geophysical Research, 98(D4), PP. 7265–7287,
doi:199310.1029/92JD02924.
deMenocal, P. B., W. F. Ruddiman, and E. M. Pokras (1993), Influences of high-and lowlatitude process on African terrestrial climate to variations in seasonal insolation,
glacial ice cover, sea-surface temperature and Asian orography, Journal of
Geophysical Research, 98, 7265–7287.
Dezfuli, A. K., and S. E. Nicholson (2011), A note on long-term variations of the African
easterly jet, International Journal of Climatology, 31(13), 2049–2054.
Drake, N. A., R. M. Blench, S. J. Armitage, C. S. Bristow, and K. H. White (2011),
Ancient watercourses and biogeography of the Sahara explain the peopling of the
desert, Proceedings of the National Academy of Sciences, 108(2), 458–462.
EPICA (2004), Eight glacial cycles from an Antarctic ice core, Nature, 429(6992), 623–
628.
Fontaine, B., and S. Janicot (1996), Sea surface temperature fields associated with West
African rainfall anomaly types, Journal of climate, 9(11), 2935–2940.
Fontaine, B., S. Trzaska, and S. Janicot (1998), Evolution of the relationship between
near global and Atlantic SST modes and the rainy season in West Africa:
statistical analyses and sensitivity experiments, Climate dynamics, 14(5), 353–
368.
Forman, S., J. Pierson, J. Gómez, J. Brigham-Grette, N. Nowaczyk, and M. Melles
(2007), Luminescence geochronology for sediments from Lake El’gygytgyn,
northeast Siberia, Russia: constraining the timing of paleoenvironmental events
for the past 200 ka, Journal of Paleolimnology, 37(1), 77–88.
Frankenberg, C., J. F. Meirink, M. van Weele, U. Platt, and T. Wagner (2005), Assessing
83
methane emissions from global space-borne observations, Science, 308, 1010–
1014.
Ghil, M. et al. (2002), Advanced spectral methods for climatic time series, Rev. Geophys,
40(1), 1003.
Grodsky, S. A., J. A. Carton, and S. Nigam (2003), Near surface westerly wind jet in the
Atlantic ITCZ, Geophys. Res. Lett, 30(19), 2009.
Hanselman, J. A., W. D. Gosling, G. M. Paduano, and M. B. Bush (2005), Contrasting
pollen histories of MIS 5e and the Holocene from Lake Titicaca (Bolivia/Peru),
Journal of Quaternary Science, 20(7-8), 663–670.
He, C., H. Breuning-Madsen, and T. W. Awadzi (2007), Mineralogy of dust deposited
during the Harmattan season in Ghana, Danish Journal of Geography, 107.
Heil, C., J. King, J. Peck, T. Shanahan, N. McKay, J. Overpeck, and C. Scholz, (in prep.)
Geochronology of the Lake Bosumtwi sediment sequence for the past million
years, Quaternary Science Reviews.
Hoerling, M., and A. Kumar (2003), The perfect ocean for drought, Nature, 299, 691–
694.
Horng, C. S., M. Y. Lee, H. Palike, K. Y. Wei, W. T. Liang, Y. Iizuka, and M. Torii (2002),
Astronomically calibrated ages for geomagnetic reversals within the Matuyama
chron, Earth Planets and Space, 54(6), 679–690.
Hughen, K. A., T. I. Eglinton, L. Xu, and M. Makou (2004), Abrupt tropical vegetation
response to rapid climate changes, Science, 304, 1955–1959.
Huybers, P., and C. Wunsch (2004), A depth-derived Pleistocene age model: Uncertainty
estimates, sedimentation variability, and nonlinear climate change,
Paleoceanography, 19(1), PA1028.
Imbrie, J., J. D. Hays, D. G. Martinson, A. McIntyre, A. C. Mix, J. J. Morley, N. G.
Pisias, W. L. Prell, and N. J. Shackleton (1984), The orbital theory of Pleistocene
climate: Support from a revised chronology of the marine delta 18O record,
Milankovitch and climate: Understanding the response to astronomical forcing, p.
269.
Itambi, A. C., T. Von Dobeneck, S. Mulitza, T. Bickert, and D. Heslop (2009), Millennialscale northwest African droughts related to Heinrich events and DansgaardOeschger cycles: Evidence in marine sediments from offshore Senegal,
Paleoceanography, 24(1), PA1205.
84
Janicot, S., S. Trzaska, and I. Poccard (2001a), Summer Sahel-ENSO teleconnection and
decadal time scale SST variations, Climate Dynamics, 18(3), 303–320.
Janicot, S., S. Trzaska, and I. Poccard (2001b), Summer Sahel-ENSO teleconnection and
decadal time scale SST variations, Climate Dynamics, 18(3), 303–320.
Johnson, T. C., E. T. Brown, and J. Shi (2011), Biogenic silica deposition in Lake
Malawi, East Africa over the past 150,000 years, Palaeogeography,
Palaeoclimatology, Palaeoecology, 303(1-4), 103–109.
Jones, G. S., J. M. Gregory, P. A. Stott, S. F. . Tett, and R. B. Thorpe (2005), An AOGCM
simulation of the climate response to a volcanic super-eruption, Climate
Dynamics, 25(7), 725–738.
Jourdan, F., P. R. Renne, and W. U. Reimold (2009), An appraisal of the ages of terrestrial
impact structures, Earth and Planetary Science Letters, 286(1), 1–13.
Kanner, L. C., S. J. Burns, H. Cheng, and R. L. Edwards (2012), High-Latitude Forcing
of the South American Summer Monsoon During the Last Glacial, Science,
335(6068), 570–573.
Kido, Y., T. Koshikawa, and R. Tada (2006), Rapid and quantitative major element
analysis method for wet fine-grained sediments using an XRF microscanner,
Marine geology, 229(3), 209–225.
Kim, S., and J. Oneil (1997), Equilibrium and nonequilibrium oxygen isotope effects in
synthetic carbonates, Geochimica et Cosmochimica Acta, 61, 3461–3475.
Koeberl, C. et al. (2007), An international and multidisciplinary drilling project into a
young complex impact structure: The 2004 ICDP Bosumtwi Crater Drilling
Project—An overview, Meteoritics & Planetary Science, 42(4-5), 483–511.
Koeberl, C., R. Bottomley, B. P. Glass, and D. Storzer (1997), Geochemistry and age of
Ivory Coast tektites and microtektites, Geochimica et Cosmochimica Acta, 61(8),
1745–1772.
Koeberl, C., and B. Milkereit (2007), Continental Drilling and the Study of Impact
Craters and Processes-an ICDP Perspective, Continental Scientific Drilling (eds.
Harms, U., Koeberl, C., and Zoback, MD), Springer, Heidelberg, 95–161.
Koeberl, C., W. U. Reimold, J. D. Blum, and C. P. Chamberlain (1998), Petrology and
geochemistry of target rocks from the Bosumtwi impact structure, Ghana, and
comparison with Ivory Coast tektites, Geochimica et Cosmochimica Acta, 62(12),
85
2179–2196.
Lambert, F., B. Delmonte, J. R. Petit, M. Bigler, P. R. Kaufmann, M. A. Hutterli, T. F.
Stocker, U. Ruth, J. P. Steffensen, and V. Maggi (2008), Dust-climate couplings
over the past 800,000 years from the EPICA Dome C ice core, Nature, 452(7187),
616–619.
Lamb, P. J. (1978), Large-scale tropical Atlantic surface circulation patterns associated
with Subsaharan weather anomalies, Tellus, 30, 240–251.
Larrasoaña, J. C., A. P. Roberts, E. J. Rohling, M. Winklhofer, and R. Wehausen (2003),
Three million years of monsoon variability over the northern Sahara, Climate
Dynamics, 21(7), 689–698.
Lea, D. W., D. K. Pak, L. C. Peterson, and K. A. Hughen (2003), Synchroneity of tropical
and high-latitude Atlantic temperatures over the last glacial termination, Science,
301(5638), 1361–1364.
Lelieveld, J., P. J. Crutzen, and F. J. Dentener (1998), Changing concentration, lifetime
and climate forcing of atmospheric methane, Tellus B, 50(2), 128–150.
Lopes dos Santos, R. A., M. Prange, I. S. Castañeda, E. Schefuss, S. Mulitza, M. Schulz,
E. M. Niedermeyer, J. S. Sinninghe Damsté, and S. Schouten (2010), Glacial–
interglacial variability in Atlantic meridional overturning circulation and
thermocline adjustments in the tropical North Atlantic, Earth and Planetary
Science Letters, 300(3), 407–414.
Loulergue, L., A. Schilt, R. Spahni, V. Masson-Delmotte, T. Blunier, B. Lemieux, J. M.
Barnola, D. Raynaud, T. F. Stocker, and J. Chappellaz (2008), Orbital and
millennial-scale features of atmospheric CH4 over the past 800,000 years, Nature,
453(7193), 383–386.
Lu, J., C. Deser, and T. Reichler (2009), Cause of the widening of the tropical belt since
1958, Geophysical Research Letters, 36, 5 PP., doi:200910.1029/2008GL036076.
Maley, J. (1987), Fragmentation de la forêt dense humide africaine et extension des
biotopes montagnards au Quaternaire récent: nouvelles données polliniques et
chronologiques, implications paléoclimatiques et biogéographiques (1987).
Palaeoecology of Africa and the surrounding islands.
Maley, J. (1989), Late Quaternary climatic changes in the African rain forest: forest
refugia and the major role of sea surface temperature variations (1989).
Paleoclimatology and paleometeorology: modern and past patterns of global
atmospheric transport.
86
Maley, J. (1991), The African rain forest vegetation and palaeoenvironments during late
Quaternary, Climatic change, 19(1), 79–98.
Maley, J. (1997), Middle to Late Holocene changes in tropical Africa and other
continents: paleomonsoon and sea surface temperature variations, Nato ASI Series
I: Global Environmental Change, 49, 611–640.
Maley, J. (2002), A catastrophic destruction of African forests about 2,500 years ago still
exerts a major influence on present vegetation formations, IDS bulletin, 33(1),
13–30.
Maley, J., and D. A. Livingstone (1983), Extension d’un élément montagnard dans le sud
du Ghana (Afrique de l’Ouest) au Pléistocène supérieur et à l’Holocène inférieur:
premières données polliniques, CR Acad. Sci. Paris, Sér, 2(196), 1287–1292.
Masson-Delmotte, V., J. Jouzel, A. Landais, M. Stievenard, S. J. Johnsen, J. W. C. White,
M. Werner, A. Sveinbjornsdottir, and K. Fuhrer (2005), GRIP deuterium excess
reveals rapid and orbital-scale changes in Greenland moisture origin, Science,
309(5731), 118–121.
McManus, J. F., R. Francois, J. M. Gherardi, L. D. Keigwin, and S. Brown-Leger (2004),
Collapse and rapid resumption of Atlantic meridional circulation linked to
deglacial climate changes, Nature, 428(6985), 834–837.
Miller, G. H., F. Wendorf, R. Ernst, R. Schild, A. E. Close, I. Friedman, and H. P.
Schwarcz (1991), Dating lacustrine episodes in the eastern Sahara by the
epimerization of isoleucine in ostrich eggshells, Palaeogeography,
Palaeoclimatology, Palaeoecology, 84(1-4), 175–189.
Moernaut, J., D. Verschuren, F. Charlet, I. Kristen, M. Fagot, and M. De Batist (2010),
The seismic-stratigraphic record of lake-level fluctuations in Lake Challa:
Hydrological stability and change in equatorial East Africa over the last 140ákyr,
Earth and Planetary Science Letters, 290(1-2), 214–223.
Moreno, A. (2012), A Multiproxy Paleoclimate Reconstruction over the Last 250 kyr
from Marine Sediments: The Northwest African Margin and the Western
Mediterranean Sea, in: Hublin J.-J. and McPherron S.P. (Eds.), Modern Origins,
Springer Netherlands, pp. 3–17.
Moreno, A., I. Cacho, M. Canals, M. A. Prins, M.-F. Sánchez-Goñi, J. O. Grimalt, and G.
J. Weltje (2002), Saharan Dust Transport and High-Latitude Glacial Climatic
Variability: The Alboran Sea Record, Quaternary Research, 58(3), 318–328,
doi:10.1006/qres.2002.2383.
87
Moron, V. (1994), Guinean and Sahelian rainfall anomaly indices at annual and monthly
scales (1933–1990), International journal of climatology, 14(3), 325–341.
Moron, V., A. Navarra, M. N. Ward, C. K. Folland, P. Friederichs, K. Maynard, and J.
Polcher (2001), Analysing and combining atmospheric general circulation model
simulations forced by prescribed SST: northern extratropical response, Annals of
Geophysics, 44(4), doi:10.4401/ag-3565.
Mulitza, S., M. Prange, J. B. Stuut, M. Zabel, T. von Dobeneck, A. C. Itambi, J. Nizou,
M. Schulz, and G. Wefer (2008), Sahel megadroughts triggered by glacial
slowdowns of Atlantic meridional overturning, Paleoceanography, 23(4), PA4206.
Nicholson, S. E. (2001), Climatic and environmental change in Africa during the last two
centuries, Climate Research, 17(2), 123–144.
Nicholson, S. E. (2008), The intensity, location and structure of the tropical rainbelt over
west Africa as factors in interannual variability, International Journal of
Climatology, 28(13).
Nicholson, S. E. (2009a), On the factors modulating the intensity of the tropical rainbelt
over West Africa, International Journal of Climatology, 29(5), 673–689,
doi:10.1002/joc.1702.
Nicholson, S. E. (2009b), A revised picture of the structure of the “monsoon” and land
ITCZ over West Africa, Climate Dynamics, 29.
Nicholson, S. E., and J. P. Grist (2001), A conceptual model for understanding rainfall
variability in the West African Sahel on interannual and interdecadal timescales,
International Journal of Climatology, 21(14), 1733–1757.
Nicholson, S. E., and P. J. Webster (2007), A physical basis for the interannual variability
of rainfall in the Sahel, Quarterly Journal of the Royal Meteorological Society,
133(629).
Opoku-Ankomah, Y., and I. Cordery (1994), Atlantic Sea Surface Temperatures and
Rainfall Variability in Ghana., Journal of Climate, 7, 551–558.
Oppenheimer, C. (2002), Limited global change due to the largest known Quaternary
eruption, Toba ~74 kyr BP?, Quaternary Science Reviews, 21(14-15), 1593–1609.
Overpeck, J. T., and J. E. Cole (2006), Abrupt change in Earth’s climate system, Annu.
Rev. Environ. Resour., 31, 1–31.
88
Peck, J. A., R. R. Green, T. Shanahan, J. W. King, J. T. Overpeck, and C. A. Scholz
(2004), A magnetic mineral record of Late Quaternary tropical climate variability
from Lake Bosumtwi, Ghana, Palaeogeography, Palaeoclimatology,
Palaeoecology, 215(1-2), 37–57.
Peck, J., N. McKay, T. Shanahan, C. Heil, J. Overpeck, J. King, and C. Scholz (in prep.),
A million-year history of the West African Monsoon, in preparation for Science.
Peterson, L. C., G. H. Haug, K. A. Hughen, and U. Rohl (2000), Rapid changes in the
hydrologic cycle of the tropical Atlantic during the last glacial, Science, 290,
1947–1951.
Pu, B., and K. H. Cook (2010), Dynamics of the West African westerly jet, Journal of
Climate, 23(23), 6263–6276.
Pu, B., and K. H. Cook (2012), Role of the West African Westerly Jet in Sahel Rainfall
Variations, Journal of Climate, 25(8), 2880–2896.
Rampino, M. R., and S. Self (1992), Volcanic winter and accelerated glaciation following
the Toba super-eruption, Nature, 359(6390), 50–52.
Rosenbaum, J., and S. M. F. Sheppard (1986), An isotopic study of siderites, dolomites
and ankerites at high temperatures, Geochimica et Cosmochimica Acta, 50(6),
1147–1150.
Rowell, D. P. (2001), Teleconnections between the tropical Pacific and the Sahel,
Quarterly Journal of the Royal Meteorological Society, 127(575).
Rowell, D. P. (2003), The impact of Mediterranean SSTs on the Sahelian rainfall season,
Journal of Climate, 16(5), 849–862.
Ruddiman, W. F., and M. E. Raymo (2003), A methane-based time scale for Vostok ice,
Quaternary Science Reviews, 22(2-4), 141–155.
Russell, J., M. R. Talbot, and B. J. Haskell (2003), Mid-holocene climate change in Lake
Bosumtwi, Ghana, Quaternary Research, 60(2), 133–141.
Schneider, R. R., B. Price, P. J. Mtiller, D. Kroon, and I. Alexanderz (1997), Monsoon
related variations in Zaire (Congo) sediment load and influence of fluvial silicate
supply on marine productivity in the east equatorial Atlantic during the,
Paleoceanography, 12(3), 463–481.
Scholz, C. A., T. C. Johnson, A. S. Cohen, J. W. King, J. A. Peck, J. T. Overpeck, M. R.
Talbot, E. T. Brown, L. Kalindekafe, and P. Y. O. Amoako (2007), East African
89
megadroughts between 135 and 75 thousand years ago and bearing on earlymodern human origins, Proceedings of the National Academy of Sciences,
104(42), 16416.
Scholz, C. A., T. Karp, K. M. Brooks, B. Milkereit, P. Y. O. Amoako, and J. A. Arko
(2002), Pronounced central uplift identified in the Bosumtwi impact structure,
Ghana, using multichannel seismic reflection data, Geology, 30(10), 939–942.
Severinghaus, J. P., and E. J. Brook (1999), Abrupt Climate Change at the End of the Last
Glacial Period Inferred from Trapped Air in Polar Ice, Science, 286(5441), 930–
934, doi:10.1126/science.286.5441.930.
Shanahan, T., J. Beck, J. Overpeck, N. McKay, and J. Pigati (Submitted), Late Quaternary
sedimentological and climate changes at Lake Bosumtwi Ghana: new constraints
from laminae analysis and radiocarbon age modeling, Quaternary Science
Reviews.
Shanahan, T. M. (2006), West African monsoon variability from a high-resolution
paleolimnologic record (Lake Bosumtwi, Ghana), University of Arizona.
Shanahan, T. M., J. T. Overpeck, J. W. Beck, C. W. Wheeler, J. A. Peck, J. W. King, and
C. A. Scholz (2008), The formation of biogeochemical laminations in Lake
Bosumtwi, Ghana, and their usefulness as indicators of past environmental
changes, Journal of Paleolimnology, 40(1), 339–355.
Shanahan, T. M., J. T. Overpeck, W. E. Sharp, C. A. Scholz, and J. A. Arko (2007),
Simulating the response of a closed-basin lake to recent climate changes in
tropical West Africa (Lake Bosumtwi, Ghana), Hydrological Processes, 21(13).
Shanahan, T. M., J. T. Overpeck, C. W. Wheeler, J. W. Beck, J. S. Pigati, M. R. Talbot, C.
A. Scholz, J. Peck, and J. W. King (2006), Paleoclimatic variations in West Africa
from a record of late Pleistocene and Holocene lake level stands of Lake
Bosumtwi, Ghana, Palaeogeography, Palaeoclimatology, Palaeoecology, 242(3),
287–302.
Shanahan, T., J. Overpeck, K. Anchukaitis, J. Beck, J. Cole, D. Dettman, J. Peck, C.
Scholz, and J. King (2009), Atlantic forcing of persistent drought in West Africa,
Science, 324(5925), 377.
Shanahan, T., J. Peck, S. Forman, C. Heil, N. McKay, J. Overpeck, J. King, and C. Scholz
(in prep), Geochronology of the Lake Bosumtwi sediment sequence for the past
200,000 years, Quaternary Science Reviews.
das Sharma, S., D. J. Patil, and K. Gopalan (2002), Temperature dependence of oxygen
90
isotope fractionation of CO2 from magnesite-phosphoric acid reaction,
Geochimica et cosmochimica acta, 66(4), 589–593.
Sheldon, N. D., and G. J. Retallack (2001), Equation for compaction of paleosols due to
burial, Geology, 29(3), 247.
Singarayer, J. S., P. J. Valdes, P. Friedlingstein, S. Nelson, and D. J. Beerling (2011), Late
Holocene methane rise caused by orbitally controlled increase in tropical sources,
Nature, 470(7332), 82–85.
Stager, J. C., D. B. Ryves, B. M. Chase, and F. S. . Pausata (2011), Catastrophic drought
in the Afro-Asian monsoon region during Heinrich event 1, Science, 331(6022),
1299.
Stone, J. R., K. S. Westover, and A. S. Cohen (2011), Late Pleistocene paleohydrography
and diatom paleoecology of the central basin of Lake Malawi, Africa,
Palaeogeography, Palaeoclimatology, Palaeoecology, 303(1-4), 51–70.
Szabo, B., C. Haynes, and T. A. Maxwell (1995), Ages of Quaternary pluvial episodes
determined by uranium-series and radiocarbon dating of lacustrine deposits of
Eastern Sahara, Palaeogeography, Palaeoclimatology, Palaeoecology, 113(2),
227–242.
Talbot, M. R., and G. Delibrias (1977), Holocene variations in the level of Lake
Bosumtwi, Ghana, Nature, 268(5622), 722–724.
Talbot, M. R., and G. Delibrias (1980), A new late Pleistocene-Holocene water-level
curve for Lake Bosumtwi, Ghana, Earth and Planetary Science Letters, 47(3),
336–344.
Talbot, M. R., and T. Johannessen (1992), A high resolution palaeoclimatic record for the
last 27,500 years in tropical West Africa from the carbon and nitrogen isotopic
composition of lacustrine organic matter, Earth and Planetary Science Letters,
110(1-4), 23–37.
Talbot, M. R., and K. Kelts (1986), Primary and diagenetic carbonates in the anoxic
sediments of Lake Bosumtwi, Ghana, Geology, 14(11), 912–916.
Talbot, M. R., D. A. Livingstone, P. G. Palmer, J. Maley, J. M. Melack, G. Delibrias, and
S. Gulliksen (1984), Preliminary results from sediment cores from Lake
Bosumtwi, Ghana, Palaeoecology of Africa, 16, 173–192.
Tarutani, T., R. N. Clayton, and T. K. Mayeda (1969), The effect of polymorphism and
magnesium substitution on oxygen isotope fractionation between calcium
91
carbonate and water, Geochimica et Cosmochimica Acta, 33(8), 987–996,
doi:10.1016/0016-7037(69)90108-2.
Tiedemann, R., M. Sarnthein, and N. J. Shackleton (1994), Astronomic timescale for the
Pliocene Atlantic δ18O and dust flux records of Ocean Drilling Program Site 659,
Paleoceanography, 9(4), P. 619, doi:199410.1029/94PA00208.
Tierney, J. E., J. M. Russell, Y. Huang, J. S. S. Damste, E. C. Hopmans, and A. S. Cohen
(2008), Northern Hemisphere Controls on Tropical Southeast African Climate
During the Past 60,000 Years, Science, 322(5899), 252.
Tjallingii, R., M. Claussen, J. B. . Stuut, J. Fohlmeister, A. Jahn, T. Bickert, F. Lamy, and
U. Röhl (2008), Coherent high-and low-latitude control of the northwest African
hydrological balance, Nature Geoscience, 1(10), 670–675.
Tjallingii, R., U. Röhl, M. Kölling, and T. Bickert (2007), Influence of the water content
on X-ray fluorescence core-scanning measurements in soft marine sediments,
Geochemistry, Geophysics, Geosystems, 8(2), 1–12.
Toggweiler, J. R., and J. Russell (2008), Ocean circulation in a warming climate, Nature,
451(7176), 286–288, doi:10.1038/nature06590.
Turner, B. F., L. R. Gardner, and W. E. Sharp (1996a), The hydrology of Lake Bosumtwi,
a climate-sensitive lake in Ghana, West Africa, Journal of Hydrology, 183(3-4),
243–261.
Turner, B. F., L. R. Gardner, W. E. Sharp, and E. R. Blood (1996b), The geochemistry of
Lake Bosumtwi, a hydrologically closed basin in the humid zone of tropical
Ghana, Limnology and Oceanography, 1415–1424.
Vasconcelos, C., J. A. McKenzie, R. Warthmann, and S. M. Bernasconi (2005),
Calibration of the δ18O paleothermometer for dolomite precipitated in microbial
cultures and natural environments, Geology, 33(4), 317–320,
doi:10.1130/G20992.1.
Verschuren, D. et al. (2009), Half-precessional dynamics of monsoon rainfall near the
East African Equator, Nature, 462(7273), 637–641.
Visser, K., R. Thunell, L. Stott, and others (2003), Magnitude and timing of temperature
change in the Indo-Pacific warm pool during deglaciation, Nature, 421(6919),
152–155.
Wallace, J. M., and P. V. Hobbs (2006), Atmospheric science: an introductory survey,
Academic press.
92
Wang, X., A. S. Auler, R. L. Edwards, H. Cheng, E. Ito, Y. Wang, X. Kong, and M.
Solheid (2007), Millennial-scale precipitation changes in southern Brazil over the
past 90,000 years, Geophysical Research Letters, 34(23), L23701.
Wang, Y., H. Cheng, R. L. Edwards, X. Kong, X. Shao, S. Chen, J. Wu, X. Jiang, X.
Wang, and Z. An (2008), Millennial-and orbital-scale changes in the East Asian
monsoon over the past 224,000 years, Nature, 451(7182), 1090–1093.
Weldeab, S., D. W. Lea, R. R. Schneider, and N. Andersen (2007), 155,000 years of West
African monsoon and ocean thermal evolution, Science, 316(5829), 1303.
Wolff, E. W., J. Chappellaz, T. Blunier, S. O. Rasmussen, and A. Svensson (2010),
Millennial-scale variability during the last glacial: The ice core record, Quaternary
Science Reviews, 29(21–22), 2828–2838, doi:10.1016/j.quascirev.2009.10.013.
Wünnemann, B., K. Hartmann, N. Altmann, U. Hambach, H. J. Pachur, and H. Zhang
(2007), 22. Interglacial and glacial fingerprints from lake deposits in the Gobi
Desert, NW China, Developments in Quaternary Sciences, 7, 323–347.
Yarincik, K. M., R. W. Murray, T. W. Lyons, L. C. Peterson, and G. H. Haug (2000),
Oxygenation history of bottom waters in the Cariaco Basin, Venezuela, over the
past 578,000 years: results from redox-sensitive metals (Mo, V, Mn, and Fe),
Paleoceanography, 15(6), 593–604.
Zabel, M., and H. D. Schulz (2001), Importance of submarine landslides for non-steady
state conditions in pore water systems—Lower Zaire (Congo) deep-sea fan,
Marine Geology, 176(1), 87–99.
Zebiak, S. E. (1993), Air-sea interaction in the equatorial Atlantic region, Journal of
Climate, 6, 1567–1567.
Zielinski, G. A., P. A. Mayewski, L. D. Meeker, S. Whitlow, and M. S. Twickler (1996), A
110,000-yr record of explosive volcanism from the GISP2 (Greenland) ice core,
Quaternary Research, 45(2), 109–118.
93
Table A1. Elemental loadings of the first two principal components
of the XRF data
Element
Al
Ti
K
Si
Fe
Mn
Ca
Sr
Rb
Si
PC1
-0.39
-0.43
-0.44
-0.42
0.03
0.30
0.15
0.06
-0.40
0.12
PC2
0.03
0.11
0.00
0.03
0.51
0.35
-0.49
-0.50
-0.04
0.33
94
90
1°27' W
300
90
90
90
45
6°32' N
30 m
50 m
90
90
45
90
90
45
70 m
45
90
6°27' N
90
90
0
1
2
km
Figure A1. Site map of the Lake Bosumtwi crater, modified from Shanahan (2006). Solid
gray lines show bathymetry (10 m contour). Dashed line indicates crater rim. Inset:
location of Lake Bosumtwi in southern Ghana, West Africa.
4.8
10
4.4
8
4.0
6
3.6
4
3.2
2
2.8
monthly rainfall (cm)
Wind speed (m s-1)
95
0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Figure A2. 1961-2000 climatology for Kumasi, Ghana, 30 km from
Lake Bosumtwi. Grey bars indicate monthly precipitation totals (cm),
dashed line shows wind speed (m s-1). Modified from Shanahan (2006).
96
Figure A3. Multichannel seismic reflection profile showing draped reflectors of the lacustrine fill and
the brecciated bedrock of the central uplift (from Scholz et al., 2002). The 13 drill holes at the five drill
sites are shown as black bars. Drill hole 5B is shown as a solid line, along with the down core median
grain size profile at the site.
97
0
Age model ensemble envelope
Age model median estimate
EDC-dust-tuned Age Model
14C ages
OSL ages
U/Th ages
Geomagnetic reversal ages
Meteor impact age
RMCD turbidite−adjusted
50
100
150
200
0
200
400
0
600
Age (ka)
800
1000
RMCD turbidite−adjusted
10
20
30
40
50
60
70
80
0
50
100
150
200
Age (ka)
Figure A4. Upper: Age-depth relation and age models for all of core BOS04-5B. Error bars show
2-sigma uncertainty where it is larger than the size of the marker. Median age model and ensemble
range determined by the Classical Age Modelling approach of Blaauw (2010). Dashed box shows
depth-age range shown in lower plot.
260
285
310
335
360
385
410
435
460
485
510
535
98
BOS04−5B 10−1, 26 to 55 cm
Mean downcore x-ray density
X−ray intensity
500
1000
1500
2000
2500
560
585
610
635
66
X-radiograph
300
350
400
450
500
450
500
Core face image
300
350
400
depth (mm)
Figure A5. An example of a turbidite in Lake Bosumtwi sediments, and the analyses and tools
used to identify turbidites in the core. Upper: Mean x-radiograph intensitty of the sediment;
higher x-ray intensities correspond to lower density material. Middle: x-radiograph.
Bottom: core face image. Turbidite identified by red bar in all three plots.
600
650
600
650
99
A
mean x−ray intensity
Depth (mm)
2000
1190
1200
1200
1210
1210
1220
1220
1230
1230
Depth (mm)
1200
1190
mean x−ray
intensity
2000
Vert. var. = 3.4 x 104
Hor. var. = 9.3 x 102
log V:H ratio = 3.6
1200
mean x−ray intensity
1120
Depth (mm)
480
1240
480
490
490
500
500
510
510
520
520
mean x−ray
intensity
1240
Depth (mm)
B
Vert. var. = 6.8 x 102
Hor. var. = 3.4 x 102
log V:H ratio = 0.67
1120
Figure A6. Examples of the calculation of the vertical:horizontal density variance ratio for both
A) well-laminated and B) unlaminated sections of core BOS04-5B. In both A) and B), both core
face images and x-radiographs are shown for 5-cm sections core. To the right and below each
x-radiograph, horizontal and vertical intensity means are shown, respectively.
100
200
180
160
140
120
100
80
60
40
20
0
A
0
100
B
50
Bosumtwi Hc:Hcr
Dust/Moisture Proxy
ODP 663 Dust
Flux (g cm-2 kyr-1)
20
0.8 C
0.6
0.4
0.2
EDC Dust
Concentration ng/g
ODP 659 Dust
Flux (g cm-2 kyr-1)
Age (ka)
1000
D
200
180
160
140
120
100
80
Age (ka)
60
40
20
0
0
Figure A7. Records of dust flux over the past 200 kyr. A) Dust flux recorded in marine core
ODP 659 off the coast of West Africa (deMenocal, 1995). B) Dust flux recorded in marine core
ODP 663 in the Gulf of Guinea (deMenocal et al., 1993). C) Hc:Hcr a mineral-magnetic dust
proxy for Lake Bosumtwi (Peck et al., in prep.). D) Dust concentration in the EPICA Dome C
ice core (Lambert et al., 2008).
Age (ka)
800
700
600
500
400
300
200
100
0
20
0
100
0.8
0.6
0.4
0.2
1000
800
700
600
500
400
Age (ka)
300
200
100
0
0
EDC dust (ng/g)
Bosumtwi Hc:Hcr
Dust/Moisture Proxy
50
Dust flux (gcm-2 kyr-1 )
Dust flux (gcm-2 kyr-1 )
101
Figure A8. Records of dust flux over the past 800 kyr. A) Dust flux recorded in marine core
ODP 659 off the coast of West Africa (deMenocal, 1995). B) Dust flux recorded in marine core
ODP 663 in the Gulf of Guinea (deMenocal et al., 1993). C) Hc:Hcr a mineral-magnetic dust
proxy for Lake Bosumtwi (Peck et al., in prep.). D) Dust concentration in the EPICA Dome C
ice core (Lambert et al., 2008). Tie points used to tuned the Lake Bosumtwi to glacialinterglacial transitions on the EDC timescale are shown in red. Tie points are also estimated
for A) and B), but are not used in adjusting the chronology.
102
XRF-PC1
Untuned
2
10
XRF-PC1
Dust-tuned
2
10
270
1
Power
10
270
1
25 17
42 23 16
11 9
9 8
0
0
10
−1
10
10
10
110-90
10
110-90
−1
0
0.1
0.2
0
0.1
Mn
Untuned
4000
0.2
Mn
Dust-tuned
90
110-90
Power
3000
2000
34 16
14
11 10
22
26
1000
0.1
0.2
0
0
8
6
0.1
Ca
Untuned
6000
13
10
0
0
0.2
Ca
Dust-tuned
180
67
180
Power
16
41
4000
68
19
2000
34
26
0
0
13
11
25
8
0.1
Frequency (1 kyr-1)
0.2
0
0
17
12
8
7-6
0.1
0.2
Frequency (1 kyr-1 )
Figure A9. Multitaper-method power spectra for XRF-PC1, Mn and Ca, both before and
after tuning the age model to glacial-interglacial transitions in the EDC dust record (Lambert et
al., 2008).
103
Age (ka)
500
450
400
350
300
250
200
150
100
50
0
100
200
20
0
40
100
60
Si (cps)
0
4
200
0.5
300
1
Mn (cps)
3000
1.5
Fe (cps)
x 10
2000
2
1000
3000
0
2000
1000
S (cps)
60
0
40
20
500
450
400
350
300
250
Age (ka)
200
150
100
50
0
Figure A10a. Unadjusted XRF counts per second for Ti, Si, K, Fe, Mn, Ca, and S.
Ca (cps)
K (cps)
Ti (cps)
0
104
Age (ka)
Ti (cps)
0
500
450
400
350
300
250
200
150
100
50
0
50
100
0
20
40
200
5000
1500
10000
1000
15000
Fe (cps)
100
500
2000
0
1000
S (cps)
60
40
0
20
0
500
450
400
350
300
250
Age (ka)
200
150
100
50
0
Figure A10b. XRF counts per second for Ti, Si, K, Fe, Mn, Ca, and S after adjusting for water
content, down-core compaction and turbidites. See text for details.
Ca (cps)
Mn (cps)
K (cps)
0
Si (cps)
150
105
40%
30%
20%
50
Si
30
Water−contentadjusted Ti (cps)
10
Water−contentadjusted Si (cps)
Estimated reduction in
water content since
core splitting
Ti
200
150
100
0
20
40
60
80
100
Depth (RMCD)
120
140
Figure A11. The effect on Si and Ti XRF counts of uncertain drying of the
cores on the adjustment for attenuation and absorption of x-rays by water.
The median estimate of 30% was used in the study, see text for details.
106
0.6
Ca Sr
0.4
PC2
0.2
Rb
K
Al Si
Ti
0
−0.2
−0.4
Mn
S
Fe
−0.4
−0.2
0
PC1
0.2
0.4
0.6
Figure A12. Loadings of XRF data on XRF-PC1 and XRF-PC2.
XRF-PC1 is dominated by the abundance of terrigenous elements,
whereas PC2 tracks carbonate mineralogy.
107
A
B
C
D
Figure A13. Relation between XRF-PC1 and Mn (A) and Ca (C) showing the
effect of dilution of terrigenous material by carbonates. Black lines show
linear regressions used to adjust for dilution, and B) and D) show the relation
after adjustment.
108
Depth (RMCD)
0
50
100
150
2.5
2
1.5
1
4
0.5
3
0
2
1
0
Mn and Ca (cps)
3000
−1
−2
log V:H density variance ratio
Lamination index
3
2000
1000
0
0
50
100
Depth (RMCD)
150
Figure A14. Variations in the A) subjective and B) objective lamination
indices for the upper 160 m of core BOS04-5B. The objective index is
calculate as the ratio of vertical to horizontal density variance in the
x-radiographs. C) Mn (red) and Ca (gray), indicative of carbonates
in the sediment. Well-laminated sediments and most carbonates in
the sediment are indicative of anoxic bottom waters (Talbot and Kelts,
1986).
109
Turbidites (cm m−1)
100
80
60
40
20
0
0
50
100
150
200
Depth (RMCD)
250
300
Figure A15. Abundance of turbidites in core BOS04-5B.
Turbidite thicknesses summed over contiguous 1-meter
increments.
110
550
0.8
500
450
400
350
300
250
200
150
100
50
0
A
0.6
0.4
0.2
1000
B
0
100-kyr
(54% var)
-4
(-) 0.01
0
0
(+) -0.01
XRF-PC1
Lake Level
Eccentricity
C
400-kyr
(41% var)
(Axis inv.)
EDC Dust (ng/g)
Bosumtwi Hc:Hcr
Dust/Moisture Proxy
Age (ka)
D
5000
2000
600
Turbidite abundance
(cm kyr-1)
400
2000
0
50 E
0
550
500
450
400
350
300 250
Age (ka)
200
150
100
50
Mn and Ca (cps)
800
4
EPICA CH (ppbv)
4
0
Figure A16. Lake Bosumtwi paleohydrology indicators compared with orbital forcing and Antarctic
dust and global methane. A) Dust abundance in Lake Bosumtwi inferred from mineral magnetic
properties (Hc:Hcr; Peck et al., in prep.). B) Dust content in the EPICA Dome C (EDC) Antarctic
ice core (Lambert et al., 2008). C) The two primary frequency components of orbital eccentricity
along with Lake Bosumtwi XRF-PC1 (This study). D) Atmospheric methane concentration
(Loulergue et al., 2008) with Lake Bosumtwi Mn and Ca abundance (This study). Mn and Ca
are indicative of manganosiderite and Ca-carbonates, respectively. E) Total thickness of turbidites
summed over each 1000-yr interval in the record (This study). Interglacial periods shown in yellow.
1
A
2000
0.5
correlation
coefficient (r)
0
1000
500
400
300
200
Age (ka)
0.3 B
0.3 C
0.2
0.2
0.1
0.1
−100
0
turbidite lag
(cm)
100
−5000
100
0
turbidite lag
(yrs)
0
0
5000
Figure A17. A) Turbidite abundance and Mn over the past 530 ka.
The lagged correlations between turbidites and Mn with B) depth
and C) time. On average, turbidites lag Mn by 20-30 cm, or 1000
to 3000 yrs.
Mn cps
Turbidites (m m-1)
111
112
1000
800
600
400
200
0
A
100
50
Lamination Index
0
3
2
1
0
B
1000
800
600
400
Age (ka)
200
0
Figure A18. A) Turbidite abundance and B) the presence and
quality of laminations over the past 1.08 Ma. For the lamination
index 3=well laminated, 2=moderately laminated, 1=poorly
laminated, and 0=unlaminated.
Turbidites (cm kyr-1)
Age (ka)
−2
10° JJA insolation (W m )
10° JJA insolation (W m−2)
113
480 A
Timing of rainy
seasons
Jun
Sep
May
Oct
Apr
Nov
460
440
420
400
20
15
10
Age (ka)
5
0
480 B
Timing of rainy
seasons
Jun
Sep
May
Oct
Apr
Nov
460
440
420
400
225
220
215
Age (ka)
210
Figure A19. Boreal summer (JJA) insolation during intervals of A) low and B) high eccentricity.
Conceptual changes in the in timing of rainy seasons on the Guinea Coast, in response to
shifts in the mean, and estimated summer maximum positions are shown in red (qualitative scales).
When insolation is highest the rainy seasons shift away from the boreal solstice. This effect is more
dramatic when insolation is higher, limiting the duration of wet conditions at the lake.
Age (ka)
NGRIP δ18O
100
50
0
1.5
B
1
21°N
0
-1
C
12°N
2.5
3.5
-30
-28
-26
20
40
E
9°N
25
60
80
F
24
5
6.5°N
H
0
5
10°N July
Insolation (W m-2)
4
20
I
500
2-6°N
25
J
30
450
K
7
6
11°S-6°N
150
100
50
0
5
Gulf of Guinea
SSS (p.s.u.)
0
6
4-14°N
10
Congo drainage
weathering (Al:K)
Lake Bosumtwi
XRF-PC1
G
18
23
7
−4
Drier
δ Olw Niger drainage Obliquity
(‰) weathering (Al:K)
(°)
Terrigenous dust (%)
D
C3-C4 veg. balance
(C29 n-alkane δ13C)
−45
Terrestrial
Humidity Index
−40
10°N
Terrestrial moisture
114
75°N
A
−35
Eolian vs Fluvial
sediment (Ti:Al)
Wetter
150
Age (ka)
Figure A20. West African terrestrial moisture indicators and climatic forcing conditions over the past 150 kyr.
A) NGRIP δ18O on the GICC05ext timescale (NGRIP 2004, Wolff et al., 2010). B) Terrestrial humidity index
from core GeoB7920-2 (Tjallingii et al., 2008). C) Ti:Al ratio from core GeoB9527-5 (Itambi et al., 2009) D)
C29 n-alkane δ13C from core GeoB9528-3 (Castañeda et al., 2009). E) Terrigenous dust content from core
ODP 661 (Demenocal, 1995). F) Obliquity (Berger and Loutre, 1991). G) Al:K ratios from core GeoB4901-8
(Zabel et al., 2001). H) Lake Bosumtwi XRF-PC1 and δ18Olw (this study). I) Sea surface salinity inferred from
Ba/Ca from core MD03-2707 (Weldeab et al., 2007). J) 10°N July insolation (Berger and Loutre, 1991). K)
Al:K ratios from core GeoB1008-3 (Schneider et al., 1997). Vertical colored bars indicate intervals of high
eccentricity and precession, and the latitudinally varied terrestrial moisture response in each interval.
115
Age (ka)
−35
100
80
60
H7
A
40
H5.2
H6
H4
H5
20
H3
0
YD
H2 H1
−6
−45
−4
B
−2
O lake water (‰)
18
δ
Lake Bosumtwi
0
2
0
4
C
5
10
29
D
28
27
26
25
Gulf of Guinea SSS (psu)
XRF−PC1 (Terrigenous)
−40
22
24
24
23
26
Gulf of Guinea SST (°C)
18
NGRIP δ O (‰)
120
E
28
30
32
120
100
80
60
Age (ka)
40
20
0
Figure A21. Lake Bosumtwi hydrology over the past 130 kyr, compared with regional records and
hemispheric boundary conditions. A) δ18O of Greenland ice tracking North Atlantic temperature
(NGRIP, 2004; Wolff et al., 2010). B) Lake Bosumtwi terrigenous input, an indicator of lake level (This
study). C) δ18O of lake water inferred from authigenic carbonates in Lake Bosumtwi (this study). D)
Sea surface temperature and E) salinity in the Gulf of Guinea (Weldeab et al., 2007). Millennial-scale
cold events in Greenland shown in yellow, often coincident with labeled Heinrich events.
116
Age (ka)
AI-6
700
600
250
200
150
AI-5
100
AI-3
AI-4
A
50
0
AI-1
AI-2
500
400
300
−4
B
−2
July Insolation
15oN (W m-2)
500
−6
0
2
400
4000
4
C
2000
2000
0
−360
−380
D
−400
−420
−440
EDC δD (per mil)
Mn and Ca (cps)
XRF−PC1
EDC CH4 (ppbv)
300
800
−460
300
250
200
150
Age (ka)
100
50
0
Figure A22. The timing of extreme arid intervals during the past 300 ka. A) EDC methane
concentration (Lambert et al., 2008). B) Lake Bosumtwi XRF-PC1 lake level indicator (black; This
study) with July insolation at 15oN (red; Berger and Loutre, 1991). C) Mn (red) and Ca (gray)
intensities, indicating the presence of manganosiderite and calcium carbonates, respectively
(This study). D) EDC δD (Jouzel et al., 2007), indicative of Antarctic temperatures. Aridity
intervals (AI) shown in yellow bars.
117
APPENDIX B
A 12,000-YEAR-LONG, ANNUALLY-RESOLVED VARVE RECORD SPANNING
THE LAST INTERGLACIAL FROM LAKE BOSUMTWI, SOUTHERN GHANA
118
A 12,000-YEAR-LONG, ANNUALLY-RESOLVED VARVE RECORD SPANNING
THE LAST INTERGLACIAL FROM LAKE BOSUMTWI, SOUTHERN GHANA
Nicholas McKay1, Jonathan T. Overpeck1,2,3, Timothy M. Shanahan4, John A. Peck5,
Clifford W. Heil6, John W. King6, and Chris A. Scholz7
1
Department of Geosciences, University of Arizona
Department of Atmpospheric Science, University of Arizona
3
Institute of the Environment, University of Arizona
4
Jackson School of Geosciences, University of Texas at Austin
5
Department of Geology and Environmental Science, University of Akron
6
Graduate School of Oceanography, University of Rhode Island
7
Deptartment of Earth and Environmental Sciences, Syracuse University
2
119
Abstract
The impact of continued global warming on the likelihood of severe drought in
sub-Saharan West Africa remains uncertain, as climate models generally do not simulate
realistic climate dynamics in the region and have inconsistent projections for the future.
The Last Interglacial period (LIG), occurring between 128 and 116 thousand years ago, is
a partial analog for future warming because at its peak, global temperatures were slightly
higher, and this warming was accentuated in Northern Hemisphere terrestrial summer
temperatures. Here we present a new, annually-resolved, 12,100-year-long varve record
for the LIG from Lake Bosumtwi in southern Ghana (6.5°N, 1.4°W). The abundance of
terrigenous elements in the sediment, varve thickness, and the isotope geochemistry and
mineralogy of authigenic carbonates in the sediment are all sensitive to changes in lake
level, and record a dynamic history of hydrologic variability in the region. The LIG lake
highstand was lower and shorter-lived than the the prolonged highstand in the early
Holocene, and unlike the Holocene, the lake never overflowed during LIG. The overall
drier conditions during the Holocene are most likely driven by amplified precessional
forcing during the interval, resulting in a northward shift in the rainbelt. The LIG, like the
Holocene, had two distinct millennial-scale moist intervals, from 125 - 123 and 121 – 120
ka. In both the LIG and the Holocene, these peaks occurred during times of precessiondriven insolation maxima in July and October, corresponding to the two rainy seasons in
the modern climatology. This suggests that, at least during interglacials, prolonged wet
conditions occur at the lake when rainy season insolation is highest. Over the course of
the LIG, lake level generally tracked sea surface temperatures (SST) in Gulf of Guinea,
120
including an abrupt drop in lake level that lasted about 500 years ca. 118 ka,
corresponding to cooling in the Gulf of Guinea and much of the North Atlantic during the
interval. The timing and duration of the event are comparable to the Late Eemian Aridity
Pulse (LEAP) that is observed at several European sites, and has been interpreted to
result from abrupt cooling in the North Atlantic, and possibly a reduction of Atlantic
meridional overturning circulation. This scenario would likely result in drought in West
Africa, so the aridity ca. 118 ka is the first indication that the LEAP occurred in Africa as
well as Europe. The occurrence of quasiperiodic variability at multidecadal to centennial
timescales is consistent with the hypothesis that slowly varying changes in Atlantic SST
structure (e.g., the Atlantic Multidecadal Oscillation) drives long term hydrologic
variability in the region. The periodicities vary over the course of the record, further
implicating the role ocean circulation.
1. Introduction
Sub-Saharan West Africa is prone to decadal to century-scale droughts, which often
begin abruptly, and are difficult to predict. The recent “Sahel Drought” began suddenly
and unexpectedly in the late 1960's, resulting in widespread famine, mass emigration, and
political instability, and exemplifies the effects of abrupt climate change (Benson and
Clay, 1998). Virtually no instrumental precipitation records in the region were collected
before 1915 (Nicholson, 2001), and the prolonged drought spanning the end of the 20th
Century is well outside the variability observed prior to 1968, when the drought began.
For example, annual precipitation totals for the region for every year from 1970 to 1997
121
were below the 100-yr mean (Nicholson and Webster, 2007). The instrumental record is
too short to understand the full range of hydrologic variability in the region, especially
because of the occurrence of multidecadal droughts and pluvials in the pre-instrumental
record (Shanahan et al., 2009). This motivates the investigation of paleoclimate archives.
Unfortunately, terrestrial West Africa has among the poorest coverage of paleoclimate
data in the world; however, recent work on Lake Bosumtwi has revealed extreme
variability in hydrology over the past 2700 years, indicating that the “Sahel Drought” is
well within the range of past variability, and that the region has experienced more severe
and longer lasting droughts (Shanahan et al., 2009). The implications of this finding for
the influence of continued global warming on moisture availability in the region remains
unclear.
Climate model projections for precipitation and evaporation in sub-Saharan west
Africa vary widely from model to model (Cook and Vizy, 2006; Randall et al., 2007;
Christensen et al., 2007; Rodriguez-Fonseca et al., 2011). This is partly because the
region lies on the boundary between two zonal bands with well-understood responses to
global warming. Climate models consistently simulate increases in water vapor with
increases in temperature, following the Clausius-Clapeyron relation, which reduces the
poleward redistribution of moisture in the tropics, resulting in increases in equatorial
precipitation and decreases in subtropical moisture (Held and Soden, 2006). In addition,
continued tropospheric warming and stratospheric cooling, driven by both increases in
greenhouse gases and anthropogenic ozone depletion, results in a widening of the tropical
belt (Lu et al., 2009) and a poleward shift in the westerlies (Shindell and Schmidt, 2004;
122
Toggweiler and Russell, 2008), further drying the subtropics. The Sahel and Guinea
Coast regions are in the transition zone between increases in precipitation in the tropics
and decreases in the subtropics.
Projecting changes in hydrology in the region is further complicated by the
dynamics of the West African Monsoon (WAM). Unlike convective precipitation in other
equatorial and monsoonal regions, the West African monsoon is not driven by the
convergence of moist southwesterly monsoonal flow and the ITCZ, but rather, by the
complex interaction of upper-level circulation and jets with low level moisture transport
(See section 1.2) for more detail; Nicholson, 2009; Pu and Cook, 2010; 2012). This
complexity, and the importance of small-scale features, such as the West African Westerly
Jet, is probably part of the reason why GCMs struggle to simulate the West African
Monsoon, both over the 20th century and in the mid-Holocene (Cook and Vizy, 2006;
Randall et al., 2007; Christensen, 2007; Pu and Cook; 2012). Only 4 of the 18 AR4
models analyzed by Cook and Vizy (2006) demonstrated reasonably realistic simulations
of the WAM during the second half of the 20th century. The models that did well in the
20th century, each simulated substantially different responses for the 21st century. For the
A2 scenario, GFDL CM2.0 and CM2.1 (Held et al., 2005) showed slight drying in the
Guinea Coast region, and extreme drying in the Sahel, MIROC (medres model) simulated
wetter conditions in the Sahel and drier on the Guinea Coast, and MRI simulated small
increases in precipitation on the Guinea Coast, and a slight drying of the Sahel (Cook and
Vizy, 2006).
Given the uncertainty in climate model projections for the region, using
123
paleoclimate records to understand how hydrology in the region responds to changes in
forcing may provide insight to the region will respond to continued warming. The Last
Interglacial period (LIG; roughly equivalent to the Eemian, or Marine Isotope Stage 5e),
which occurred from about 128 to 118 ka, is an interesting period to study for several
reasons; first, it's the most recent analog to the Holocene, with generally similar
conditions, although differences in the orbital configuration resulted in altered insolation
forcing, especially seasonally (Berger and Loutre, 1991). It is often considered a partial
analog for future climate, because global mean temperatures were slightly warmer (02°C; Turney and Jones, 2010; McKay et al., 2011) and sea level was 6-8 m higher than
preindustrial conditions (Hearty et al., 2007; Kopp et al., 2009), although global CO2 (and
other GHG) concentrations were far below modern and projected future concentrations
(270-290 ppm), Luthi et al. (2008). The largest difference in forcing relative to the
Holocene during the LIG is the change in seasonal forcing due to the amplification of the
precession cycle by eccentricity (Berger and Loutre, 1991). This resulted in a >30 W m-2
increase in Northern Hemisphere insolation summer, and a corresponding decrease in
Southern Hemisphere winter insolation at the height of the interglacial ca. 124 ka. This
resulted in much warmer temperatures in the Northern Hemisphere, especially on land
during the summer (Turney and Jones, 2010; McKay et al., 2011). The northern
Hemisphere is consistently projected to warm faster in the future than the southern
Hemisphere in all seasons, and especially on land. More regionally, Weldeab et al.,
(2007) show that the SST in the Gulf of Guinea was generally warmer during the LIG
than the Holocene, which is also a consistent projection of GCMs for the 21st century
124
(Cook and Vizy, 2006). Altogether, reconstructing hydrologic variability in the region
during the LIG has the potential to provide insight into future climate in the region.
1.1 Site
Lake Bosumtwi is a 1.08 +/- Ma (Koeberl et al., 1997; Jourdan et al., 2009) meteoriteimpact crater lake located in southern Ghana (6°30' N, 1°25' W) approximately 150 km
north of the Gulf of Guinea (Figure B1). The lake has no outlet, is isolated from the
groundwater table, and the lake surface occupies about half of the basin, and
approximately 80% of the annual water input from rainfall directly on the surface (Turner
et al., 1996a). These features make lake level extremely sensitive to changes precipitation
and evaporation, and lake level responds to interannual and decadal-scale changes in
precipitation, rising more than 5 m during the wet interval from 1940-1970, and falling
more than 2 m during the “Sahel Drought” from 1970-2000 (Shanahan et al., 2007).
Sedimentation in the lake and the stable isotope geochemistry of carbonates precipitated
in the lake are sensitive to lake level (Talbot and Delibrias, 1977; Talbot and Kelts, 1986;
Peck et al., 2004; Shanahan, 2006; Shanahan et al., 2009). The lake is meromictic and
typically anoxic below 15 m depth (Turner et al., 1996b), allowing for the preservation of
finely laminated varves, which have accumulated over the past ca. 2700 yr (Shanahan et
al., 2008a, 2009). Finely laminated, presumably varved, sediment is common in intervals
throughout the 291-m sediment sequence deposited since the crater was formed, 1.08 Ma
(Peck et al., in prep.; Appendix A). This study focuses on a continuously laminated
interval between 57.4 and 60.7 meters below the present-day lake floor.
125
1.2 Modern Climate
Climate in the region is strongly seasonal, and most of the variability can be
attributed to the annual migration of the tropical rainbelt (Nicholson, 2009) (Figure B2).
Annual precipitation at the lake has a bimodal distribution, with peak rainfall occurring in
May-July and a smaller peak from September-October. Mean temperature at nearby
Kumasi varies from about 24°C in the boreal summer to 28°C in the winter (Shanahan et
al., 2008). The cooler temperatures in the summer are attributable to enhanced upwelling
in the Gulf of Guinea, while the warm winters are indicative of the influence of the
terrestrial air masses to the north. Wind in the region is also seasonal, and is controlled by
the position of the Intertropical Convergence Zone (ITCZ). In the boreal summer the
ITCZ is positioned north of the lake over the Sahel and Sahara, and the monsoonal
southwesterlies peak in August. In the winter the ITCZ is positioned near the Guinea
Coast, wind speed is reduced, and aerosol-rich northeasterly 'Harmattan' winds are
common. Dust supplied to the region by these winds has been shown to be a substantial
component of Ghanaian soils (He et al., 2007) and solute load in Lake Bosumtwi (Turner
et al., 1996b).
Interannual variability in precipitation in the region has two primary modes of
variability: a monopole and dipole mode. The monopole mode describes years when the
both the Guinea Coast and the Sahel are wetter or drier than average. This mode is
associated with the amount of low-level moisture transport (Pu and Cook, 2012) and the
strength of the Tropical Easterly Jet (TEJ; Nicholson 2008a,b), which, when stronger,
drives upper level divergence and promotes deep convection through the rain belt. The
126
strength of the TEJ may in turn be associated with the strength of the Asian Monsoon and
tropical Pacific SSTs (Chen and van Loon, 1987; Janicot et al., 2001; Nicholson, 2008b).
The dipole mode describes rainfall anomalies in the Sahel region (north of 10° N) are of
the opposite sign of those in the Guinea Coast region (south of 10° N) (Nicholson and
Webster, 2007; Nicholson, 2008a). The mode of variability, the sign of that mode, and the
amplitude of the anomalies are related to the strength, position and structure of the upper
level TEJ (Nicholson 2008a,b) and the mid-level African Easterly Jet (AEJ; Nicholson
and Webster, 2007; Nicholson, 2008a, Dezfuli and Nicholson, 2011). Whereas upper
atmospheric circulation drives the convective conditions, moisture is primarily derived
from low levels from the southerly monsoonal flow, and a distinct, low-level westerly jet
originating off the coast of West Africa near the Atlantic ITCZ, which was identified by
Grodsky et al. (2003). Recent work suggests that the West African Westerly Jet (WAWJ)
is critical in driving interannual and decadal variability in precipitation in the Sahel (Pu
and Cook, 2010; 2012). Variability in these atmospheric circulation patterns are related to
a number of factors, including sea surface temperatures (SST) in the tropical Pacific (e.g.,
Rowell 2001; Janicot et al., 2001) and eastern Mediterranean (Rowell, 2003), as well as
the strength of the Asian Monsoon (Chen and van Loon, 1987; Janicot et al., 2001;
Nicholson, 2008b). However, SST variability and especially the interhemispheric sea
level pressure (SLP) and SST gradients in the tropical Atlantic are most strongly linked to
interannual and decadal scale variability in the regional precipitation (e.g., Lamb, 1978;
Nicholson and Grist, 2001; Rowell, 2003; Nicholson and Webster, 2007; Nicholson,
2008a,b; Nicholson, 2009, Pu and Cook, 2012).
127
Interhemispheric SST structure has been implicated in past hydrologic changes in
the region as well. Shanahan et al., (2009) showed that lake level at Lake Bosumtwi was
tied to the Atlantic Multidecadal Oscillation (AMO) over the past 2700 years. Cold
intervals in the North Atlantic have been associated with extremely dry conditions in
much of West Africa, most notably during Heinrich 1 (Stager et al., 2011); although
similar conditions have been observed in both terrestrial and marine proxies throughout
the past 100 ka (Peck et al., 2004; Weldeab et al., 2007; Mulitza et al., 2008; Tjallingii et
al., 2008; Castañeda et al., 2009; Itambi et al.; 2009).
2. Methods
2.1 Core Drilling and sampling
In July and August of 2004, a total of 1.83 km of sediment was recovered from
five drill sites using the global lake drilling system (GLAD-800; Peck et al., 2005;
Koeberl et al., 2007). This study focuses on cores recovered from deep-water (74 m) site
5, drilled into the annular sediment moat around the impact-derived central uplift (Figure
B3). Two cores from that site (5B and 5C) capture the entire 296-m-long lacustrine
sediment sequence. After drilling, the cores were shipped to the University of Rhode
Island, where they were split, described, imaged, and analyzed for a wide variety of
physical parameters. U-channels were taken from the axis of each core to measure a suite
of mineral magnetic parameters. Core logging and physical sediment properties were
used to integrate the depth scales for the three drill cores from site 5 (BOS04-5A,B&C)
into a relative mean core depth (RMCD; pers. comm. C. Heil).
128
2.2 Sampling and laminae analysis
The U-channels from the core sections of cores 5B and 5C were subsampled, flash-frozen
using liquid nitrogen, and freeze dried at the University of Arizona. Once dehydrated, the
sediments were vacuum-embedded with epoxy resin (cf., Lotter and Lemcke, 1999). For
optical analysis, the embedded sediments were cut and polished down into thin sections
of 30-40 μm thickness, and then photographed at 25x magnification under crosspolarized light using a Canon EOS D20 SLR camera attached to an Olympus SZX9
binocular microscope. To create mosaic images without distortion, the photographs were
stitched together using Canon PhotoStitch software version 3.1 in “Document” mode.
The images were then spatially rectified in ArcGIS 10.0 (ESRI Software, Inc.) by
aligning the images with the 25-μm-resolution μXRF data, or a 100-μm-resolution glass
scale included in the photographs. Laminae were then counted and measured in ArcGIS
10.0.
2.3 XRF analysis
The relative abundance of major and minor elements in the sediments was
analyzed using two scanning X-ray fluorescence techniques. The embedded sediments
from core 5B were analyzed at 25-μm resolution, under vacuum, using the EDAX Eagle
III scanning XRF at the University of Arizona, using a Mo X-ray source and 16 sec count
times. The short count times used on the embedded sediments were necessary to allow for
high-resolution measurements of elemental variability within small lamina present in the
the sediments, and to support laminae counting. After the varves were identified and
129
counted, two methods were used to calculate annual averages. In the first approach, the
raw XRF spectra for each year were summed and reanalyzed in the EDAX software. This
approach allows for better defined spectra (with effectively longer count times), but was
too inefficient to use for the Monte Carlo ensembles (see section 2.4). For the ensemble
members, the XRF counts for each point were simply averaged. The latter approach result
is less precise than the former, however comparison of the two methods suggests that the
large changes in XRF intensities throughout the record are well-captured by either
approach. Nevertheless, we hope to improve the efficiency of the first approach and fully
adopt this methodology in future research, as this two-step procedure would allow the
XRF data to be collected and used to aid laminae counting at extremely high resolution,
and then subsequently integrated into yearly averages from very well-defined spectra.
Over the course of the analyses, a decrease in count rates were observed in all elements,
possibly due to a weakening of the X-ray source. To account for this, Cl counts on pure
resin at the beginning, end, or in gaps in the sediment in each puck were used to
standardize the count rates. The standardized counts compare well with lower resolution
XRF data collected using the Itrax scanning XRF at the University of Minnesota, Duluth,
justifying the normalization (Figure B4; Appendix A).
2.4 Development of the varve chronology
The fine laminations found throughout most of the Lake Bosumtwi record have
been carefully studied in several intervals. The lamina comprising the top 1.5 m of the
sediment sequence are annual layers, or varves, as shown by their correspondence with
130
210
Pb, and 14C “bomb-pulse” ages over recent decades and centuries, and 14C ages over the
past 2700 yrs (Shanahan et al., 2007, 2009). A longer interval of laminations deposited
between 12 and 29 ka closely correspond with the nearby 14C dates, confirming that the
laminations deposited during this interval are also varves (Shanahan et al., in prep). The
internal structure of the varves through these intervals varies considerably (Shanahan et
al., 2007). Varves formed during the late Holocene are comprised of a clastic unit, which
most likely washes in during the rainy season, and organic matter and calcite-rich unit
which forms during the productive season (Shanahan et al., 2007). In intervals deeper in
the core, the calcite is often absent or replaced by another carbonate mineral (i.e.,
dolomite, aragonite, siderite, or ankerite), and the organic layer typically appears in thin
section as a thin, opaque layer. Nevertheless, the minerogenic (rainy season) component
and the organic-rich (productive season) component, appear to be a robust indicator of
the annual cycle across a wide range of environmental and climatic conditions.
The laminations studied here also consist of couplets of alternating minerogenic
and organic-rich units, among other components (including several carbonate minerals)
which occur in sections of the study interval. The interval is well beyond the limits of
radiocarbon dating, and there are no independent geochronological approaches that can
be used to verify my interpretation of the annual nature of the laminations in this interval.
Nevertheless, because the layers have a similar structure as the Holocene and late
Pleistocene varves, and because their components are consistent with an annual cycle of
sedimentation, it is most likely that they represent annual layers. In this study, the top of
each varve was identified by the opaque, organic-rich unit deposited each productive
131
season. For some laminae, it is unclear whether or not they represent a year, or are
subannual depositional events. For these layers, we followed the approach Rasmussen et
al., (2006) used in layer counting the NGRIP ice core, and recorded these layers 0.5 +/0.5 years. These layers were later used to help quantify the uncertainty in the varve
chronology, varve thickness, and XRF records.
The entire LIG section (57.3 to 60.7 relative mean composite depth (RMCD))
from both cores 5B and 5C was thin sectioned, and the varves were counted and
measured twice for each core, resulting in nearly complete replication of the varve
sequence, and for most intervals, four estimates of varve counts and thickness. To allow
for comparison and evaluation of counting uncertainty between cores, 379 marker layers,
primarily turbidites, and units with distinct mineralogies (iron-sulfides and carbonates),
were identified in cores 5B and 5C. The composite varve chronology was developed by
creating an ensemble of varve counts and thicknesses between each set of marker layers,
moving progressively up the sediment section. For most intervals, this resulted in four
ensemble members for each interval. In sections where there were gaps in the sediment in
one of the cores, only the two estimates from the continuous core were used. For each of
the four varve count and thickness estimates, hundreds of additional ensemble members
were generated stochastically by allowing the uncertain years identified during counting
and measurement (i.e., 0.5 +/- 0.5 years) to be either a separate year, or part of the
adjacent lamination. Following this approach, 1000 ensemble members for varve count
and thickness were created for the gap between each set of marker layers. These
ensembles allow quantification of the uncertainty associated with identifying, and
132
measuring the varves throughout the record and estimation of the chronological
uncertainty between marker layers. By repeating analyses on individual ensemble
members, we can investigate how counting and identification uncertainty influence the
varve thickness and XRF records. It is important to note, however, that this approach
represents only one component of the chronological uncertainty in the record. The other
component, the veracity of my interpretation of annual cycles in the sediment cannot be
independently determined, and is not included in the uncertainty measurements.
Because we have identified synchronous marker layers in cores BOS04-5B and
BOS04-5C, two different approaches for creating XRF and varve thickness ensembles are
used. In the first approach, marker layers are forced to occur in the same year in each
ensemble, so the the gap between marker layers also remains constant. This approach is
preferable for examining the records in time space, since it enforces the observed
stratigraphic order in the record, not allowing for varves below a marker layers in one
core to be integrated with varves above the same layer in the other core. These
ensembles, however, alter the serial nature and autocorrelation structure of the records.
Therefore, a second set of ensembles, that disregard the marker layers, is used for spectral
analysis of the records.
2.5 Geochronology
Although the varve chronology for the interval is well constrained, the absolute
age of any given varve is unknown; it is a “floating” chronology. For the purposes of
comparing with other regional and global records, and understanding the intervals
133
response to orbital forcing, it is worthwhile to try to “pin” the “floating” chronology as
accurately as possible. The uppermost sediments of the Lake Bosumtwi sediment
sequence are well-dated by 14C ages and varve counting (Shanahan et al., 2009; Shanahan
et al., in review), however, beyond the limits of radiocarbon dating the chronology
becomes less certain. Deeper in the core, optically stimulated luminescence (OSL) ages,
and U/Th ages on authigenic carbonates provide further age control, however, the
analytical uncertainty and lack of agreement between ages make identification of MIS 5e
in the sediment sequence ambiguous (Figure B5). In Appendix A, the Lake Bosumtwi
sediment sequence was roughly synchronized with the EPICA Dome C (EDC) timescale
by matching abrupt changes in the Lake Bosumtwi and EDC dust records (Lambert et al.,
2008; Peck et al., in prep) that coincide with glacial-interglacial transitions. The two
nearest tie points relevant to the LIG are the transition from MIS 6 to MIS 5e (133 ka on
the EDC timescale) and the onset of abundant dust deposition associated with MIS 4 (70
ka on the EDC timescale). This tuned chronology includes its own assumptions, and does
not allow for evaluation of leads and lags in the onset and termination of the period
(Appendix A). However, it does increase our confidence that the varve interval was
deposited during the LIG. Over the past 140 ka, the relative abundance of terrigenous
sediment in the Lake Bosumtwi record generally covaries with nearby SST and SSS
records (Weldeab et al., 2007), and to a lesser extent the NGRIP δ18O record (NGRIP,
2004; Figure B6). A distinct cold event in the NGRIP record, Greenland stadial 25 (GS25), followed by an abrupt shift into Greenland interstadial 24 (GI-24; ~110 ka) is clearly
recognizable in SST and SSS records in the Gulf of Guinea, and likely corresponds to a
134
dry-to-wet transition in Lake Bosumtwi, apparent in the abundance of terrigenous
material in the sediment at about 109 ka on the timescale from Appendix A. Assuming
this excursion in Lake Bosumtwi hydrology is correlative with GS-25 and GI-24 implies
a 1 kyr offset in the Lake Bosumtwi timescale (Appendix A; Shanahan et al., in prep)
relative to the NGRIP GICC05ext timescale (Wolff et al., 2010). Therefore, we use this
offset to pin the “floating” varve chronology to an absolute age; suggesting that the 12.1
kyr-long varve record spans the interval from 128.6 to 116.5 ka. It is very difficult to
assess the uncertainty in these absolute ages. The assignment relies on the several
assumptions, including that the large scale climate changes are synchronous between
Greenland and West Africa. Although this hypothesis is well supported in the literature,
there are few if any records for this time from West Africa that have chronologies that are
completely independent from the NGRIP chronology. Another assumption is that the 1
kyr offset between the Lake Bosumtwi and NGRIP records apparent 110 kyr remains
constant back to 116.5 ka. This assumption could be avoided by pinning the record to an
event within the varve chronology, however, there are not any unambiguously correlative
events in the varved interval. By tying the floating varve chronology to a single event we
avoid potential biases associated with “wiggle-matching” the varve chronology to nearby
high-resolution records.
There are several other lines of evidence supporting my hypothesis that the study
interval corresponds to the LIG. Dust proxies for Lake Bosumtwi generally share the
main features of both nearby and high latitude dust records, particularly the difference
between glacials and interglacial periods (Peck et al., in prep.; Appendix A). The study
135
interval comprises a period of low dust content immediately following high dust
deposition at the end of MIS 6. The 3.4-m-long varved interval studied here has relatively
high organic matter and manganosiderite content, both of which are indicative of
relatively high lake level (Appendix A). Furthermore, the interval contains continuous,
sub-mm laminations indicating the persistence of anoxic bottom waters, another indicator
of wet or warm conditions at the lake. The overwhelming majority of terrestrial moisture
sensitive records for West Africa indicate that interglacials are generally wet with respect
to glacial periods (e.g., deMenocal et al., 1993; deMenocal, 1995; Weldeab et al., 2007;
Mulitza et al., 2008; Tjallingii et al., 2008; Itambi et al.; 2009), and this was clearly the
case at Lake Bosumtwi for the Holocene relative to the late Pleistocene (Talbot and
Johannessen, 1992; Peck et al., 2004). Therefore, it is likely that lake level was higher
during MIS 5e than MIS 6 or MIS 5d.
Taken together, the general agreement of lake level indicators at Lake Bosumtwi
and global-scale climate phenomena during MIS 5, and the occurrence of a high-moisture
balance interval immediately following dust-rich MIS 6 makes the study interval (57.3 to
60.7 RMCD) the most likely candidate for the LIG at the lake. This is consistent with the
nearby U/Th and OSL ages (Figure B5; Shanahan et al., in prep.). Finally, the Blake
geomagnetic excursion, which occurred during the LIG (123 +/- 3 ka; Lund et al., 2006),
has been identified in this interval (Heil et al., in prep.).
2.6 Carbonate isotope geochemistry
Potential samples for carbonate mass spectrometry were identified from thin
136
section analysis and the XRF intensities of Ca (for calcite, dolomite and ankerite) or Mn
(for manganosiderites). A 300 μm drill bit was then used to sample the carbonates from
the resin-embedded sediments. Once cured, the epoxy resin does not react with the
phosphoric acid used in isotopic analysis; powdered resin added to calcite standards
NBS-19 and NBS-18 resulted in no offset from pure carbonates. The powdered
carbonates were heated under vacuum at 100°C for several hours to remove volatile
organic compounds prior to measurement. The δ18O and δ13C of these carbonates were
measured using an automated carbonate preparation device (KIEL-III) coupled to a gasratio mass spectrometer (Finnigan MAT 252). Powdered samples were reacted with
dehydrated phosphoric acid under vacuum at 70°C. The isotope ratio measurement is
calibrated based on repeated measurements of NBS-19 and NBS-18 and precision is ±
0.10 ‰ for δ18O and ±0.08‰ for δ13C (1 sigma).
3. Results
3.1 Generalized Lithology and Laminae structure
A generalized lithology is shown for the study interval in Figure B7. The studied
interval is a 3.4-m-long (57.4 to 60.7 m RMCD), continuously laminated with variable
lithologies. The interval is bounded above and below by unlaminated sediments,
immediately below the study interval is 2-cm-thick carbonate-rich layer, underlain by
massive, dense clays and silts with little internal structure.
137
3.1.1 Unit 1a (60.3 – 60.7 m RMCD)
The base of the study interval is composed of dark-brown finely laminated (0.1 to
0.5 mm) muds and opaque organic layers with abundant well-rounded quartz grains,
varying from 10 to 50 μm in diameter, presumably of eolian origin. Moving up in the
section, the well-rounded grains gradually become finer and sparser, and are rare above
60.4 m RMCD. Calcite-rich sublaminae are common in this interval, but are not found
within every lamination.
3.1.2 Unit 1b (60.2 – 60.3 m RMCD)
In unit 1b, the laminae remain finely laminated (0.1 to 0.4 mm), but carbonaterich laminae are rare. The laminae are clearly defined as alternating tan, minerogenic
layers and thinner, opaque, organic-rich layers. The minerogenic layers often have a
distinctly reddish color, associated with fine-grained iron sulfide minerals.
3.1.3 Unit 1c (59.8 – 60.2 m RMCD)
The base of unit 1c is composed of very finely laminated (0.05 to 0.15 mm)
terrigenous-calcite-organic triplets that gradually thicken upwards and into thick (0.4 to
1.0 mm) terrigenous layers divided by poorly defined opaque organic units. Through this
interval, distinct iron-sulfide-rich bands become common, and carbonate layers are rare
in the thickest lamina. Above, 59.9 m RMCD, the laminations thin somewhat (0.3 to 0.6
mm), and thin calcite layers are common in most laminae, and thick, red iron sulfide
layers are found occasionally.
138
3.1.4 Unit 1d (59.6 – 59.8 m RMCD)
Here, finely laminated (0.1 to 0.4 mm) alternating minerogenic terrigenous layers
alternate with opaque organic layers, with light tan calcite layers in each packet. Distinct,
bright red discontinuous layers of coarse-grained iron sulfide occur sporadically, once
every several layers. The coarseness of morphology of the layers suggest that these
sulfide layers were formed post-depositionally, however, they remain confined to
individual lamina and are traceable between cores, suggesting that they form in the
shallow subsurface in response to bottom water geochemistry.
3.1.5 Unit 2a (59.1 – 59.6 m RMCD)
Unit 1 grades into Unit 2, where the laminae become extremely fine (0.05 to 0.1
mm), and calcite layers are rare. Bright red iron-sulfide-rich layers are common, but are
thinner and more continuous than their counterparts in Unit 1d. The laminae are still
primarily formed of alternating minerogenic and organic units, but the minerogenic units
become very thin, and the overall composition of the sediment is more organic-rich.
Throughout the interval, there are distinctive, relatively thick (0.1 to 1 mm) orange
manganosiderite units, and above 59.5 RMCD, thick (up to 7 cm) turbidites are common.
The turbidites are primarily minerogenic, and coarsen upwards, terminating in lightcolored fine-grained caps. In places, the turbidites occur in packets, with no or few
laminae between turbidites. Elsewhere in the unit, however, turbidites of various
thicknesses are separated by hundreds of fine lamina. There is no evidence of erosive
contacts at the base of any of the turbidites, although some of the thicker turbidites
139
include packets of laminae reworked from other, presumably shallower areas of the lake.
3.1.6 Unit 2b (58.7 – 59.1 m RMCD)
Above the extremely fine laminations of unit 2a, the lamina thicken (0.1 to 0.4
mm) and the grain sizes of the terrigenous layers become visibly coarser, composed
primarily of silt-size material. The unit is less organic-rich than unit 2a, but contains more
dark-colored organic material than unit 1. The laminations in this interval often contain a
thin reddish-brown layer in the middle of the clastic terrigenous layer. The layers are rich
in Fe, but contain no sulfur, and may be associated with redox geochemistry
accompanying an annual overturning of lake during this interval. Large, (1-3 mm)
individually occurring gypsum crystals that cut across nearby lamina are found in this
interval. The crystals are clearly diagenetic, and may be associated with increased amount
of sedimentary organic-material in the interval, although the origin of the crystals and
their significance remains uncertain.
3.1.7 Unit 2c (58.6 - 58.7 m RMCD)
In unit 2c, the fine laminations (0.1 to 0.3 mm) are predominantly composed of a
minerogenic clastic unit interbedded with a thin, opaque organic unit. The sediments here
contain abundant, well-round silt-sized quartz, similar to the base of unit 1a, but
somewhat finer grained and more densely packed. Dolomite, present in both distinct
yellow bands, and mixed in with the terrigenous layer is also present in this interval.
140
3.1.8 Unit 2d (58.4 - 58.6 m RMCD)
This unit is very finely laminated (0.05 to 0.15 mm), and the grain size of the
terrigenous clastic unit is finer than in units 2c and 2b, with rare well-rounded quartz
grains. The laminations are primarily composed of distinct minerogenic and opaque
organic couplets, although reddish, iron-sulfide-rich layers are common. As in Unit 2b,
large diagenetic (1-3 mm) gypsum crystals cut across laminations in places.
3.1.9 Unit 3a (58.1 - 58.4 m RMCD)
Dark brown muds alternating with thin opaque organic layers form the
laminations in this interval, which gradually thicken upwards (from about 0.2 mm to 0.3
mm). Well rounded, presumably eolian, quartz grains are abundant through this interval.
Yellow-tan carbonates of variable compositions between dolomite and ankerite occur in
places through the interval, as both fine grained crystals interbedded with terrigenous
clastic sediments, and as larger (100 μm to 1 mm) cyrstals that form discontinuously
across lamina and appear to have formed post-depositionally.
3.1.10 Unit 3b (57.4 - 58.1 m RMCD)
Tan, terrigenous, minerogenic sediment dominates this interval, alternating with
discontinuous thin opaque organic-rich layers to define laminations that thicken upward
(from 0.4 mm near the base to about 1.5 mm near the top). Distinct red iron-sulfide-rich
layers commonly form distinct bands that are traceable across cores. As in Unit 3a,
dolomitic to ankeritic carbonates are found as both distinct yellow-tan units with large
141
crystals, and as fine-grained minerals within the terrigenous units. Near the top of the
section the laminations continue to thicken and the opaque organic layer becomes less
distinct until the laminations are no longer clearly recognizable, marking the top of the
study interval.
3.2 Carbonate isotope geochemistry and mineralogy
The δ18O and δ13C of the carbonates (δ18Ocarb, δ13Ccarb) show extreme variability
across through the 12,000 yr record, varying from 0.7 to 12.5‰ and 3.0 to 23.5‰,
respectively (Figure B8). Generally, δ18Ocarb and δ13Ccarb covary, consistent with previous
studies at the lake (Figure B8; Talbot and Kelts, 1986). Furthermore, the mineralogy and
mechanism of formation of the carbonates also have a strong control over their carbon
and oxygen isotopic composition. Talbot and Kelts (1986) identified two primary modes
of formation for the carbonates in Lake Bosumtwi, 1) near the lake surface driven by CO2
drawdown during the productive season, and 2) near or just below the sediment-water
interface driven by methanogenesis. The latter mode of formation results in very high
δ13C values (7 – 27‰) because the methane produced is extremely depleted (-60‰) in
13
C, causing the remaining carbon is isotopically enriched. Methanogenesis does not
affect the oxygen isotopic composition of the carbonates, and the oxygen isotopes of
carbonates formed by either mechanism should track that of the surface or deep water
(Talbot and Kelts, 1986). Because the lake is small, has a short residence time (~30 yrs ;
Turner et al., 1996a), and is relatively well-mixed (Turner et al., 1996b), the carbonates
formed in the bottom waters probably represent a long-term average of the lake water
142
δ18O (δ18Olw), whereas those formed near the surface may reflect a shorter-term
“snapshot” of δ18Olw. On decadal and longer timescales, however, the two are likely
comparable.
Changes in carbonate mineralogy complicate the δ18Ocarb record, because the
water-carbonate oxygen isotope fractionation varies considerably between minerals (cf.,
Chacko and Deines, 2008). At least four species of carbonate are found in the study
interval. The minerals were identified using several methods. First, the manganosiderites,
found between 59.1 and 59.6 m RMCD, are readily apparent in thin section, because of
their bright orange color, and in the XRF Mn intensities, which are very high (Figure B9).
Calcite is identifiable in thin section, as distinct white layers, with sharp Ca XRF peaks
between 59.6 and 60.6 m RMCD, and 58.8 and 59.1 m RMCD. The transition from
calcite to dolomite at 59.1 m RMCD occurs abruptly, in less than 4 cm. Between 58.4 and
59.1 m RMCD, dolomite is visually distinct in thin section, forming thicker layers and is
yellow in thin section. The shift in mineralogy across this transition was confirmed using
Raman spectroscopy (Rutt and Nicola, 1974; Herman et al., 1987). It is likely that the
calcite near the transition to dolomite is relatively rich in Mg, however the inclusion of
small amounts Mg into the calcite structure has a minimal impact on the water-carbonate
fractionation (~0.06‰/1% Mg at 25°C; Tauritani et al., 1969). Talbot and Kelts (1986)
report up to 18% mol MgCO3 in Lake Bosumtwi calcites, that varied significantly, even
within individual laminations. This is consistent with the results of Abebe (2010), who
found that the molar % of MgCO3 varies from 7.9 to 19.2 throughout drill core 5B. This
would correspond to a maximum offset of ~1‰, which would have a minor impact on
143
major trends in δ18Ocarb. Therefore, we ignore the amount of Mg in the calcites; however,
ideally these data would be included in the analysis.
Finally, both dolomite and ankerite [Ca(Fe,Mn,Mg)(CO3)2], occur between 57.4
and 58.4 m RMCD. Much of the carbonate in this section of the core is fine grained and
intermixed with terrigenous sediment, and is only identifiable because of the weak Ca
(and sometimes Mn) peak in the XRF data. It is difficult to obtain well-defined Raman
spectra on the fine grained material, although ankerite is observed in X-ray diffraction
analyses on samples from 57.70 and 57.84 m RMCD (Abebe, 2010). The occurrence of
ankerite in the sediment complicates the interpretation of the δ18Ocarb through this interval
for several reasons. First, the oxygen isotopic fractionation from water to ankerite has not
been empirically determined, although a theoretical estimate suggests that it should be
similar to calcite, and substantially lower than dolomite at relevant temperatures. This
implies that the δ18O of an ankerite would be ~4‰ lower than a dolomite formed from the
same water (Chacko and Deines, 2008). Secondly, the amount of Fe and Mn as opposed
to Mg in the carbonates most likely falls on a continuum between the dolomite and
ankerite end-member compositions. In fact, pure ankerite [CaFe(CO3)2] isn't found in
nature, (Davidson et al., 1993). Again, the impact of substitution of Fe and Mn for Mg in
the dolomite structure on oxygen isotopic fractionation has not been measured
empirically, but theoretically, replacing Fe with Mg should increase the fractionation,
while replacing Fe with Mn should decrease it (Chacko and Deines, 2008). The elemental
composition of the dolomites and ankerites in this interval has not been carefully
determined, however, comparison of the relative intensities of the Ca, Mn, and Fe XRF
144
peaks allows further investigation of the relationship between mineralogy and δ18Ocarb.
Interestingly, Ca/Mn and Ca/Fe ratios generally correspond to δ18Ocarb measurements in
this interval (Figure B10). Generally, lower δ18Ocarb values are associated with lower
Ca/Mn and Ca/Fe ratios, probably corresponding to more ankeritic carbonates, whereas
samples with high Ca/Mn and Ca/Fe ratios are more enriched in 18O. Accounting for the
~4‰ offset between dolomite and ankerite would reduce the variability in δ18O in this
interval, and suggest that generally, the apparently more ankeritic samples would have
been formed in lake water with higher δ18O. This is consistent with the observation that
over the history of the lake, ankerite is only found in intervals that other indicators
suggest are uncommonly dry (Abebe, 2010).
To compensate for the effect of mineralogy on carbonate δ18O, for the ankeritefree interval between 58.4 and 60.7 m RMCD, we used the fractionation equations for
calcite (Kim and O'Neill, 1997), siderite (Carothers et al., 1988), and dolomite
(Vasconcelos et al., 2005), to calculate the δ18Olw at the time of formation. Because the
water-carbonate oxygen isotope fractionation is temperature dependent, changes in lake
water temperature could complicate the δ18Olw reconstruction. To account for this, we
estimate changes in δ18Olw over a 6°C range (23°C to 29°C). This is similar to the modern
seasonal range in temperature, and annual mean temperatures are unlikely to have varied
by 6°C over the course of the record. Furthermore, a period of lake level decline
(enriching δ18Olw) would most likely be associated with warmer temperatures, which
would have the opposite effect on δ18Ocarb. Ultimately, the inferred variability in δ18Olw is
substantially larger than even a large temperature effect can explain (Figure B11),
145
reinforcing the assumption of past workers that carbonate δ18O records the composition
of the lake water (Talbot and Kelts, 1986; Shanahan et al., 2009).
3.3 Varve thickness
Varve thickness varies over several orders of magnitude over the study interval,
from 0.03 to 70 mm. The thickest layers correspond to turbidites, and although they occur
rarely, they make up a substantial portion of the sediment sequence. The turbidites are
typically found in very finely laminated intervals, and represent sedimentation rates
several orders of magnitude greater than nearby years. The extreme variability in varve
thickness illustrates the non-linear relationship between hydrology and sedimentation at
the core site. Varve thickness is controlled by several factors. On short timescales (i.e., <
30 years), varve thickness is probably dominated by changes in precipitation. During the
instrumental record, varve thickness tracks precipitation well; more sediment is deposited
at the core site during rainier years (Shanahan et al, 2008). On long timescales (> 100
years), this relationship is dominated by other factors. First, because of the small size of
the drainage basin relative to the lake, changes in lake level have a large impact on the
area available for sediment supply. For example, the drop in lake level from its early
Holocene maximum to modern level corresponds to a 160% increase in the size of the
non-lake drainage area (Turner et al., 1996a, Shanahan et al., 2006). Additionally, a
decrease in lake level means that there is a reduced lake floor area over which sediment is
distributed, and the core site, which is near the center of the lake, is nearer the shore.
Both of these factors would also increase sedimentation rates at the core site, even
146
without a change in the volume of sediment entering the lake. Finally, drops in lake level
may be associated with changes in the vegetation structure of the lake basin, varying
between rainforest during wet intervals, and savannah and grassland flora during drier
intervals. This along, with lake previously-deposited lake silts could increase the supply
of available sediment, although this feedback is less certain. Altogether, the morphology
of the lake basin is expected to drive non-linear increases in sedimentation rates during
lowstands, and greatly reduced sedimentation during highstands.
3.4 XRF data
The relative abundance of Al, Si, K, and Ti, as indicated by changes in the
intensities of their XRF peaks, generally tracks varve thickness (Figures B9 and B11).
The primary source of these elements in the lake sediments is terrigenous material that
washes in from the drainage basin during the rain season, and every varve contains a
layer rich in these elements (Shanahan et al., 2008; 2009). Because they are associated
with minerals washing into the lake, as opposed to forming in the lake, we refer to these
elements as “terrigenous elements”. The terrigenous elements are washed in from the lake
basin, and their relative abundance is strongly influenced by the size of the drainage basin
and the proximity of the core site to the shoreline, both of which are primarily functions
of lake level. Therefore, the terrigenous elements are expected to vary inversely with lake
level; most abundant when the lake is low, and least when it is high. Changes in dust flux
would amplify this effect, further increasing the abundance of terrigenous material when
lake level is low. The abundance of terrigenous elements is a robust lake level indicator
147
for the late Holocene, covarying with carbonate oxygen isotope values, geomorphic lake
level indicators, and a variety of other forms evidence (Shanahan et al., 2009). In this
study, we focus on the intensities of Ti as an indicator of terrigenous content, because it
forms relative low-noise peaks even at short count times (unlike Al and Si), and is not
complicated by changes in weathering and dust input (unlike Si and K; see section).
Fe is also a terrigenous element, and is an important component of the sediment
that washes into the lake. However, Fe is also present in minerals precipitated in the lake,
or post-depositionally in the sediment, primarily manganosiderites and various species of
iron-sulfide minerals (Shanahan 2006; Shanahan et al., 2008; 2009). This complicates the
interpretation of the Fe intensities, and we consider them separately from the other
terrigenous elements. Mn in the sediment record is present in several minerals, but is
predominately associated with manganosiderites. Mn is also associated with iron-sulfide
minerals and ankerite, but the intensity of the Mn peaks associated with these minerals is
much lower (<100 cps) than with the manganosiderites (>1000 cps). Therefore, we
consider Mn to be a robust indicator of the presence of manganosiderite.
Ca XRF intensities primarily indicate the presence, and relative abundance of
various species of calcium carbonates in the sediments (Shanahan 2006; Shanahan et al.,
2008; 2009). There is very little Ca in the bedrock of the drainage, which is
predominantly schist with small outcrops of granite (Koeberl et al., 1998), and the low
intensities of Ca in the terrigenous layers of the sediment are consistent with this
observation. The Ca that is incorporated in authigenic carbonates in the sediments is most
likely delivered to the lake as dust. Dust deposited by northerly Harmattan winds is an
148
important component of Ghanaian soils (He et al., 2007), and is the likely origin of much
of the dissolved Na in Lake Bosumtwi (Turner et al., 1996b). During the early part of the
LIG, Ca forms distinct peaks associated with the annual, or semi-annual precipitation of
calcium carbonates in the lake, after about 7000 varve years, however, the peaks become
broader and smaller, reflecting the change from distinct carbonate layers to diffuse fine
grained carbonates that are partially intermixed with terrigenous sediment and probably
formed post-depositionally. Finally, the relative abundance of S in the sediment shows
pronounced variability, and is present in three distinct components of the sediment. For
most of the sediment, S XRF intensities are fairly low, tracking the abundance of organic
matter in the sediment. Higher S intensities are associated with various the various ironsulfide minerals (pyrrhotite), and less frequently, large (1-3 mm) crystals of diagenetic
gypsum found in units 2b and 2d.
4. Discussion
4.1 Hydrology during the interglacial
Varve thickness, the abundance of terrigenous sediment as indicated by Ti, and
δ18Olw all show the same main patterns over the 12 kyr-long interval and indicate a
consistent, dynamic history of hydrologic variability during the Last Interglacial (Figure
B11). Whereas there are a few varve records for the interval (e.g., Sirocko et al., 2005;
Brauer et al., 2007; Seelos and Sirocko, 2007), the overwhelming majority of the
paleoclimatic and paleoenvironmental work done on the LIG is comprised of
paleoecological records from Europe (e.g., Woillard 1978; Klotz et al., 2004; Sirocko et
149
al., 2005; Kühl et al., 2007; Brewer et al., 2008). This work has identified three primary
phases of climate during the interval (c.f., Brewer et al., 2009), and although there is little
a priori reason to assume that West African climate and hydrology experienced a similar
history, both regions are strongly influenced by North Atlantic SST, and the framework
provides a convenient way to examine climate as it evolved through the the interglacial.
4.1.1 Early LIG (Varve years 1-5600; 128.6 – 123 ka)
The beginning of the early LIG interval shows gradually decreasing varve
thickness and terrigenous sediment abundance, corresponding with decreasing δ18Olw, all
of which suggest a gradual increase in lake level. This corresponds to a gradual decrease
in the abundance of well-rounded eolian quartz grains in the sediment (Figure B7),
suggesting that the increase in lake level was accompanied by some combination of
decreasing local dust supply, driven by increasing moisture in the region, and less intense,
or less frequent, northerly winds. Such a decrease in northerly winds may be associated
with a northward shift in the boreal winter position of the ITCZ. All of this is consistent
with a gradual strengthening of the West African Monsoon during the early part of the
interglacial.
By 127 to 126.3 ka, varve thicknesses were generally very thin, although the Ti
and the δ18Olw data are more indicative of intermediate lake level. Beginning 126.2 ka,
varve thicknesses dramatically increase from less than 0.1 mm to more than 1 mm by
125.8 ka, before decreasing to thin varves (<0.1 mm) by 125 ka. This feature is among
the most distinctive throughout the LIG varve record, and is generally supported by a
150
corresponding increase in inferred δ18Olw. Interestingly, this feature is not apparent in the
average abundance of terrigenous material, which remains near the long term average
during this interval. This means that although the flux of terrigenous material into the
lake certainly increased during this interval (since flux is proportional to the product of
average terrigenous input and sedimentation rate), the relative proportions of terrigenous
material did not increase, which is typical of lower lake levels (Figure B11). The
significance of this is not yet clear.
Following the return to thin varves, all three proxies suggest an extended period
of high lake level from 125 to 123 ka. This interval is characterized by organic-rich
sediment, with abundant manganosiderite layers and occasional, thick turbidites. The
presence of turbidites during intervals of high lake level is consistent with observations
over longer timescales; turbidites are frequent during relatively wet intervals throughout
at least the past 530 kyr (Appendix A). Most likely, the turbidites formed during brief,
and perhaps relatively small, decreases in lake level that exposed deltas formed during
higher lake level. These landforms would be unsupported and poorly consolidated, and
would be particularly susceptible to mass wasting events. Taken together, the
combination of thin varves, low Ti content, low δ18Olw and the presence of abundant
manganosiderite and turbidites identify this interval as the longest period of sustained
lake level highstand in the LIG.
4.1.2 Mid-LIG (Varve years 5600-9000; 123 – 119.6 ka)
Lake level began to decrease following the sustained highstand from 125-123 ka,
151
with Ti and varve thickness increasing to intermediate levels. δ18Olw increased to the
highest early LIG values by 122.3 ka, after which varve thickness and δ18Olw both
decrease suggesting a return to somewhat higher lake level by 121.8 ka. Following this
varve thickness and δ18Olw rapidly increase to local maxima around 121.1 ka. This
transition is accompanied by an abrupt change in carbonate mineralogy from calcite
(probably Mg-rich calcite) to dolomite, and calcite is not found in the rest of the LIG
interval. All three lake level indicators suggest a return to relatively higher lake level
following 121 ka, with lake level remaining relatively high for the rest of the mid-LIG,
until about 119.6 ka.
4.1.3 Late LIG (Varve years 9000-12100; 119.6 – 116.5 ka)
The beginning of the late LIG interval corresponds to an abrupt increase in varve
thickness, and the onset of ankeritic carbonates in the sediment sequence. The carbonates
are not pure ankerite, which doesn’t exist in nature, but most likely a represent carbonates
of variable composition between the dolomite and ankerite end members, where the Mg
of the dolomite is progressively substituted with Fe or Mn. Water-ankerite oxygen isotope
fractionation factors have not been measured empirically, and the effect of partial
substitution of Fe or Mn for Mg remains uncertain, although theoretical arguments
suggest that increasing Fe and Mn should decrease the fractionation factor. Interestingly,
the oxygen isotopes of the carbonates in this interval closely track both varve thickness
and Ti, with thicker, more terrigenous varves corresponding to lower δ18Ocarb values,
opposite of what is observed throughout the rest of the study interval (Figure B12). This
152
is consistent with the hypothesis that δ18Ocarb in this interval tracks the relative amount of
Fe and Mn in the carbonate, and that carbonates with lower δ18Ocarb values are more
ankeritic, and deposited during intervals of very low lake level (Abebe, 2010). This
interpretation is consistent with varve thickness and Ti during these intervals.
Mineralogical complications prevent calculation of the δ18Olw during this interval,
however the inference that more ankeritic samples have lower δ18Ocarb and dolomitic
samples have higher δ18Ocarb, due to changes in water-carbonate fractionation factors,
suggests that δ18Olw was on the order of 9 to 11 per mil during the interval, the highest
values in the LIG varve record. This suggests a highly evaporated water mass, and is
consistent with the other indicators of very low lake level during at this time (Figure
B12).
Varve thickness and Ti intensities both show gradual increases through this
interval, suggesting a slow decrease in lake level, interrupted by two multicentury-scale
lowstands from 118 to 117.5 ka and 117.2 to 116.9 ka (Figure B12). The end of the
interval is defined by increasingly poorly-defined, terrigenous-rich varves with disparate
opaque organic layers, probably indicative of seasonally oxic conditions at the lake floor,
consistent with a continued drop in lake level, until the varves are no longer recognizable.
4.2 Comparison between the LIG and the Holocene
The most pronounced difference between the LIG and Holocene intervals of the
Lake Bosumtwi sediment sequence (Figure B13) is the lack of an equivalent to the thick,
organic-rich unit deposited during the early Holocene (often referred to as the “sapropel”;
153
Talbot et al., 1984). Geomorphic evidence and and exposed shorelines indicate that the
lake overflowed frequently during the early Holocene, from ca. 11 to 9 ka, and
maintained very high lake levels until 3.2 ka (Shanahan et al., 2006). This interval of high
lake level is distinct in the sediment sequence; characterized by the lack of varves, greatly
reduced precipitation of carbonates, and a dramatic change in the aquatic biology, as
indicated by δ15N (Talbot and Johannessen, 1992). There is no LIG equivalent of the
Holocene highstand sediment (“sapropel”). The organic-rich interval from 59.7 to 59.1 m
RMCD, corresponding to 124.9 to 122.8 ka (700-5800 varve years), is the most similar,
but the laminations are still apparent, although very thin (often <50 μm). The presence of
a very thin terrigenous component in these varves suggest a much smaller, but still active,
terrestrial watershed, consistent with a high, but not overflowing lake. Manganosiderite is
much more abundant in the LIG interval than in the Holocene highstand sediment,
another indication that the lake did not overflow during the LIG. Dissolved Fe and Mn
were likely abundant in the deep anoxic water masses during both intervals, however,
closed-lake conditions throughout the LIG would have kept the alkalinity and
concentration of HCO3 high, resulting in more carbonate precipitation. The final
indication that the wettest part of the LIG was not overflowing is the inferred δ18Olw from
the carbonates in the interval. There are no δ18Olw measurements from the early Holocene,
but the lowest δ18Olw values in the LIG (2.5 to 3.5 ‰) are comparable to similar values
derived from calcites precipitated between 2.7 and 1.7 ka, after the lake had fallen
dramatically from the early Holocene highstand (Shanahan et al., 2009).
154
4.2.1 Relationship with solar insolation
Unlike what has been observed in parts of the Asian (Wang et al., 2001; Wang et
al., 2008) and South American (Cruz et al., 2005; Cruz et al., 2009) Monsoon regions, the
West African Monsoon, at least in region of Lake Bosumtwi, does not vary in phase with
changes in summer insolation (Appendix A; Shanahan, 2006; Peck et al., in prep). During
the Holocene lake level remained very high for several thousand years after summer
insolation began to decrease, not falling until insolation was nearing its minima 2.7 ka
(Figure B13) . Lake level also appears to have lagged insolation during the LIG as well,
reaching its maximum highstand 2 – 4 ka after the summer insolation maxima 127 ka
(Figure B13). Lake level does appear to have decreased along with decreasing insolation
between 123 and 121 ka, before recovering to higher lake level between 121 and 120 ka
despite continued decrease in insolation. Around 120 ka, increases in varve thickness and
the appearance of ankerite in the sediment suggest an abrupt decrease in lake level,
although this is not well-recorded in Ti intensities (Figure B11). This occurred at about
the same phase of the precession cycle (2-3 ka before the northern Hemisphere summer
insolation minima) as the abrupt fall during the late Holocene. This may suggest a
nonlinear response of the West African Monsoon to gradual decreases in summer
insolation, although the uncertainty in the chronology during the LIG interval makes it
impossible to compare the phasings precisely.
Interestingly, both interglacials had two millennial-scale peaks in lake level, at 109 ka and 4-3 ka in the Holocene and from 124-123 and 121-120 ka in the LIG. These
peaks both occurred at around the same phasing within the precession cycle during both
155
interglacials, the first peak corresponding to maximum 10N insolation in July, the second
peak corresponding to maximum insolation in October (Figure B13). Sub-precessional
periodicities in precipitation are predicted and observed for equatorial regions that
experience double maximums in rainfall associated with the twice-yearly passing of the
tropical rainbelt and ITCZ (Short et al., 1991; Trauth et al, 2003; Berger et al., 2006;
Verschuren et al., 2009). Despite this, the climate dynamics associated with these subprecessional peaks and troughs remain somewhat uncertain, and most likely vary
regionally. Trauth et al. (2003) suggested that lake level high stands at equatorial Lake
Naivasha coincided with March and September local insolation maxima, corresponding
to the timing of the two rainy seasons in the modern climatology. This would imply that
higher insolation was driving enhanced convection and onshore moisture transport, and
thus, an intensification of the East African Monsoon during whatever season has higher
than normal solar insolation. Verschuren et al. (2009), came to a contradictory conclusion
based on a record from nearby Lake Challa, where subprecessional lake level highstands
correspond to insolation maxima near the solstices, and lowstands when insolation
maxima occurred near the equinoxes. The authors suggested that onshore moisture
transport was highest when either the Northern or the the Southern Hemispheric
components of the East African Monsoon was strongest, and that the very low insolation
during the other rainy season when March or September insolation may have contributed
to aridity during those intervals.
Lake Bosumtwi also experiences two distinct rainy seasons, centered on June and
October (Figure B2). The two millennial-scale highstands in each interglacial correspond
156
to the periods of maximum insolation during the two rainy seasons (Figure B13).
Furthermore, in both interglacials, the first of the two highstand intervals were longer and
are associated with higher lake level than the second highstand. The same pattern is also
observed in modern precipitation climatology; the first rainy season is both longer and
provides more precipitation than the second rainy season (Figure B13). It is reasonable to
assume that an intensification of a longer and wetter rainy season would result in a longer
and wetter highstand, although how the relative amplitudes and durations of the rainy
seasons may have varied in the past is uncertain.
4.3 Relationships with sea surface temperature records
There are very few records of sea surface temperature in the region for the LIG,
and only a Mg/Ca-based SST record from the Gulf of Guinea (Weldeab et al., 2007) is
sufficiently well-resolved to allow comparison of the variability within the interval. The
floating varve chronology is pinned to this record based on an abrupt event 110 ka, so
although some uncertainty remains in the absolute timing of the records, relative changes
between the two should be comparable. Generally speaking, varve thickness, Ti
intensities and carbonate mineralogy and isotope geochemistry covary with SSTs, with
warmer SSTs co-occurring with periods of higher lake level (Figure B11). The SST
record shows a gradual cooling trend after 127 ka that is consistent with gradual drying in
the Lake Bosumtwi region, as shown by increases in varve thickness and δ18Olw, but this
is not reflected in the abundance of Ti, possibly due to dilution by authigenic minerals
during the interval. A local minimum in SSTs occurred between 122 and 120 ka,
157
coincident with the drier conditions in between the two LIG highstands. This decrease in
SST is consistent with a northward shifted maximum position of the ITCZ, which would
drive increased upwelling off the Guinean Coast and potentially drier conditions at Lake
Bosumtwi.
During the late Holocene, variability in North Atlantic SSTs played a prominent
role in driving hydrologic variability at Lake Bosumtwi (Shanahan et al., 2009). High
resolution SST data for the LIG from the North Atlantic are also sparse, however, SST
records from the western Mediterranean and Iberian Margin regions (Martrat et al., 2004;
2007) and from the eastern subpolar and western North Atlantic (Lehman et al., 2002;
Oppo et al., 2006), provide some context for the interval (Figure B14). Both regions
experienced rapid warming between 128 and 127 ka, followed by generally warm
temperatures until about 122 ka before gradually cooling through the rest of the interval
(Figure B14). The initial warming apparent in the SST data from all three regions
corresponds to a period of low varve thickness, and perhaps wet conditions at Lake
Bosumtwi, but gaps in the XRF and δ18Olw data during this interval prevent further
conclusions. Interestingly, SSTs in both the Gulf of Guinea and Iberian Peninsula region
suggest cooler conditions from about 127 to 125.5 ka, at about the same time as a
pronounced, gradual increase and decrease in varve thickness, that is partly corroborated
by δ18Olw data. The low resolution and apparent noise in the North Atlantic records make
it impossible to evaluate potential linkages on shorter timescales.
158
4.4 An multi-century aridity event 118 ka
Immediately after 118 ka, Ti intensity and varve thickness both rapidly increase in
less than 100 years (Figure B12), corresponding with an abrupt decrease in SSTs in the
Gulf of Guinea (Figure B11). Carbonate mineralogy and geochemistry also shows abrupt
changes a this interval, with an abrupt decrease in δ18Ocarb most likely associated with a
transition towards more ankeritic carbonate. The hypothesis that lower δ18Ocarb values is
consistent with differences between dolomite and ankerite fractionation factors (Chacko
and Deines, 2008), but has yet to be tested for this interval. Ankerite is present in the two
XRD samples measured in the interval, corresponding to ca. 117.3 and 117.1 ka (Abebe,
2010). Throughout the 1.08 Myr record, ankerite is only rarely found in Lake Bosumtwi
sediments, and is consistently associated with extreme aridity (Abebe, 2010; Peck et al.,
in prep.). The arid interval lasted about 400-500 years, before recovering somewhat
around 117.5 ka (Figure B12). Afterward, varve thicknesses gradually increase towards
the top of the LIG section, and Ti counts remain high. δ18Ocarb also increases, although
whether this is due to continued drying of the lake driving an increase in δ18Ocarb, or a
shift in mineralogy toward more dolomitic carbonates is unclear.
The arid interval is consistent in timing and duration with the Late Eemian Aridity
Pulse (LEAP), which has been observed in some European records, most notably those
from the Eifel Maar complex (Figure B12; Sirocko et al., 2005; Seelos and Sirocko,
2007). In northern Germany, LEAP corresponds to about 500 years of greatly increased
loess deposition, suggesting much colder, and more arid conditions in the region at the
time, although the precise timing and duration of the event vary from site to site (Sirocko
159
and Seelos, 2007). LEAP most likely corresponds to the North Atlantic cooling event
C26, which is observed in some, but not all SST records from the North Atlantic, and
may be associated with a small peak in ice rafted debris at the time (Lehman et al., 2002;
Sirocko et al., 2005; Oppo et al., 2006). Sirocko et al., (2005) hypothesize that LEAP, and
C26, resulted from a slowdown of AMOC, or at least a southward shift of the northern
limit of Atlantic heat transport. A reduction in AMOC, and a cooling of North Atlantic
would likely result in aridity in West Africa (Street-Perrott and Perrot, 1990; Chang et al.,
2008; Mulitza et al., 2008; Shanahan et al., 2009), so it is reasonable to assume that the
aridity observed near the end of the LIG varve record may be the West African expression
of LEAP, although uncertainty in the ages of the records make it impossible to link the
two records conclusively.
4.5 Drought frequency during the LIG
Both varve thickness and Ti counts (Figure B11), as well as other XRF counts
from other terrigenous elements (Figure B9), show pronounced decadal to century scale
variability during the LIG. To investigate the timescales of variability in the records, as
well as the influence of uncertainty in the varve chronologies, spectral estimates were
calculated using the multi-taper method (MTM; Mann and Lees, 1996) for all of the
ensemble members of the varve thickness, and Ti count records (Figure B15). Confidence
limits were calculated for each ensemble member using the “Robust” noise estimation
method of Mann and Lees (1996), the median 90 and 95% confidence limits are shown
(Figure B15), but there is little variability in the confidence levels between ensemble
160
members.
The Monte Carlo MTM spectra (MC-MTM) for both Ti and varve thickness
(Figure B15) show distinct changes in slope around frequencies of 1/100 years. This
“kink” in the spectra is a common feature of paleotemperature records, and may reflect
the midpoint between the predominant influence of the influence of the annual cycle at
higher frequencies, and Milankovitch cycles at lower frequencies (Huybers and Curry,
2006). Although Huybers and Curry restricted their analysis to temperature records, it is
not surprising to find a similar feature at Lake Bosumtwi, since hydrology at the lake is
greatly influenced by annual and orbital cycles.
The median MC-MTM spectra for both varve thickness and Ti do not have any
significant peaks at frequencies higher than 1/50 yrs, although there are significant peaks
in each of the individual ensemble members. On the other hand, whereas individual
ensemble members may have a grouping of individual significant peaks, the median
estimate shows broader bands of relatively higher variance, such as the peaks centered on
periodicities of 19, 16, and 14 years in Ti, and the peaks at 28 and 20 years in varve
thickness (Figure B15). The diffusion of high frequency peaks is a predictable result of
incorporating age model uncertainty, both because of the relatively greater influence of
small changes in the age model, and because each higher frequency spectral estimate
corresponds to narrower window of time. The broad bands of power at multidecadal
frequencies are not significant, but most likely reflect real modes variability in the record,
that are either obscured by age model uncertainty, or are actually non-periodic and really
do occur over a range of frequencies. To investigate the latter hypothesis, we calculated
161
evolutive MC-MTM spectra for Ti and varve thickness to analyze changes in periodicities
over the course of the varve sequence.
The evolutive MC-MTM (EvMC-MTM) spectra were calculated using the same
method as the MC-MTM spectra, but over nine, partially overlapping, 2000-yr-windows
for each series, with the exception of the first window for Ti, which ends at the start of
the gap in XRF data coverage. Both series show concentrations of multidecadal to
centennial-scale variance, including many that are significant at the 95% confidence level
(Figures B16 and B17). In both EvMC-MTM spectra, however, the periodicities vary
over the course of the interglacial. For example, significant peaks, or nearly significant
peaks are found at periodicities between 37 and 33 years, and 77 and 67 years in Ti, and
32-27, 61-53, and 83-71 years in varve thickness, but the periodicities and relative
amplitudes of the peaks vary greatly between windows. This suggests the presence of a
non-periodic, but preferred range of timescales of hydrological variability at the lake.
Alternatively, bias in the varve chronology could also result in such variability, distorting
a truly periodic signal. However, the same phenomenon is observed at lower frequencies,
which should be less susceptible to age bias. In the Ti EvMC-MTM spectra, peaks with
periodicities from 120-95, and 220 to 170 years, and peaks from 180 to 130 years in the
log varve thickness record show similar changes in frequency and relative amplitudes
(Figures B16 and B17).
Multichannel Singular Spectrum Analysis (M-SSA) was performed on the Ti and
log varve thickness datasets, such that each input was a timeseries ensemble member
(Ghil et al., 2002). This non-parametric approach identifies the primary oscillatory modes
162
of variability in the time series that are robust to uncertainty in the varve chronologies.
Because we are particularly interested in multidecadal to centennial (M2C) variability, we
detrended the series with a spline before performing the analysis to remove millennialscale variability, The M-SSA was performed with an embedding window of 400 yrs, and
with 50 ensemble members. The full set of ensemble members could not be used because
the procedure is extremely computationally expensive, even with high-performance
computing capabilities. Generally, the results support the MC-MTM and EvMC-MTM
analyses, indicating the predominance of M2C variability, especially periodicities around
110 and 275 yrs (Table B1). The M-SSA analyses on the ensembles without marker
layers suggested more variance concentrated at high periodicities, likely due to increased
variability between ensemble members in that set. For both types of ensembles, some of
the predominance of lower frequencies likely reflects chronological uncertainty, which
has a greater impact on higher frequency variability. That said, the Ti ensemble with
marker layers has several important reconstructed components with multidecadal-scale
variability, including those with periodicities of ~70, 55 and 24 years, consistent with the
EvMC-MTM analysis for Ti (Figure B16).
The finding that Ti abundance and log varve thickness show pronounced M2C
variability, but over a range of frequencies, is consistent with the finding that changes in
Atlantic sea surface temperature structure controlled by the AMO drove M2C drought
variability during the late Holocene (Shanahan, 2006). The origin of the AMO remains a
topic of debate; with hypotheses ranging from a global-scale series of ocean-atmospheresea ice teleconnections (Dima and Lohman, 2006), to internal variability in Atlantic
163
meridional overturning circulation (AMOC) (Knight et al., 2005), to the possibility that
the AMO in the instrumental record is an artifact of the 20th century warming trend and
two abrupt cooling events (Thompson et al., 2010), although this last hypothesis seems
unlikely, given that there is evidence of AMO variability for at least the past several
hundred years (Gray et al., 2004; Shanahan et al., 2009). The AMO is typically simulated
in state-of-the-art climate models, but the underlying mechanism, as well as the
timescales of variance vary from model to model (Danabasoglu, 2008). In CCSM4, AMO
variability is associated with changes in AMOC and covers a broad low frequency
spectrum (50-200 yr periodicities) (Danabasoglu et al., in press). The variability in
CCSM4 is associated with deep-water formation in the Labrador Sea, and may be
associated with the NAO, however the origin of the lower frequency time scale remains
unclear (Danabasoglu et al., in press). Interestingly, in CCSM4, a externally-forced (with
solar and volcanic variability) last millennium simulation shows substantially more
multidecadal-scale variability in the AMO than the long control run (Landrum et al.,
submitted). The finding that Lake Bosumtwi hydrology varied with periods around 30,
50, 70 and 150 yrs during the LIG suggests that AMO, and AMOC variability during the
Last Interglacial may have been similar to the late Holocene, despite the substantial
differences in seasonal insolation forcing. That said, the lack of a well-defined timescale
of variability and uncertainty in the underlying mechanism prevents further analysis.
4.5.1 Influence of solar variability
The similarities between paleorecords of tropical and subtropical hydrology and
164
indicators of past solar variability (e.g., 14C, 10Be), have motivated the hypothesis that
M2C variability in solar insolation drives changes in these regions (e.g., deMenocal,
2001; Hodell et al., 2001; Cook et al., 2004, Shindell et al., 2006, Meehl et al., 2009)
although recent syntheses of global climate variability and cold events during the
Holocene concluded that long-term solar-variability was not consistently associated with
climate on global scale (Wanner et al., 2008; 2011). Modelling studies have suggested
that certain regions may be more vulnerable than others to solar variability, especially the
hydrology of the tropics, due both to direct heating of the land surface and changes the
ozone chemistry response to altered ultraviolet radiation (Shindell et al., 1999; Shindell
and Schmidt, 2004). Furthermore, tropical Atlantic SSTs and AMOC respond to
simulated solar variability as well (Goosse and Renssen, 2004; Weber et al., 2004;
Cubasch et al., 2005). Altogether, this indicates that Lake Bosumtwi should be
particularly well situated to record the climatic influence of solar variability. The
Holocene record from Lake Bosumtwi offered limited support for the hypothesis that
solar variability modulates hydrology in West Africa. Shanahan (2006) found coherent
variability between terrigenous sediment abundance and reconstructed atmospheric 14C
concentrations at several frequencies, including inferred solar periodicities of 120 and
220 yrs. During the LIG, however, varve thickness and Ti abundance do not show
significant variability at these periodicities. As discussed above, M2C variability during
the LIG varied in periodicity and amplitude throughout the interglacial. It remains
uncertain whether or not solar variability is periodic, or whether solar variability during
the LIG occurred at the same periodicities as observed during the Holocene. It is clear,
165
however, that there is no noticeable concentration of spectral power at known modern
solar periodicities, that is, the 11-year sunspot cycle, the 87-year Gleissberg cycle, or the
210-year DeVries cycle (Figures B16 and B17; Braun et al., 2005). Ultimately, the role of
solar variability during the LIG remains unclear.
5. Conclusions and summary
5.1 Primary findings

In this study we presented a new, 12,100 year varve record from Lake
Bosumtwi that spans an interval that represents the Last Interglacial
period. Radiometric and cosmogenic age control points from within or
near the study interval do not have sufficient precision to accurately “pin”
the floating varve chronology, however, correlation between global and
regional dust and SST records suggest that the varve section spans the
interval from 128.6 to 116.5 ka. This is consistent nearby U/Th and OSL
ages, as well the Blake geomagnetic excursion (123 +/- 3 ka; Lund et al.,
2006) found within the varved interval (Heil et al., in prep).

A new approach was used to account for uncertainty in the varve count age
model: Following the approach of Rasmussen et al., (2006) in layer
counting the NGRIP ice core, ambiguous (possibly annual, possibly subannual) layers were recorded as 0.5 +/- 0.5 years. Varve count, thickness,
and XRF ensembles were created by using the four separate varve counts
(two on each core) and allowing the uncertain years to vary stochastically
166
between marker layers. For spectral analysis, a second set of ensembles
were created that allowed the gaps between marker layers to vary as well.
This approach allows quantification of the effect of counting uncertainty
on the resulting record and analyses.

Varve thickness and XRF titanium intensities generally agree with the δ18O
of lake water (δ18Olw), which was inferred from carbonate isotope
geochemistry and mineralogy over a range of temperatures. All three of
these parameters are sensitive lake level, and vary substantially throughout
the interglacial.

The maximum LIG highstand was both smaller and shorter-lived than the
the prolonged highstand in the early Holocene, and unlike the Holocene,
the lake level indicators, including the continuous presence of varves
through the interval, suggests that the lake never overflowed during LIG.

The LIG, like the Holocene, had two distinct millennial-scale moist
intervals, from 125 - 123 and 121 – 120 ka. In both the LIG and the
Holocene, these peaks occurred during times of precession-driven
insolation maxima in July and October, corresponding to the two rainy
seasons in the modern climatology. This suggests that, at least during
interglacials, prolonged wet conditions occur at the lake when rainy season
insolation is highest. Furthermore, in both interglacials, the larger
highstand occurred during the July insolation maxima, which is the more
substantial of the two rainy seasons.
167

Over the course of the LIG, lake level generally tracks sea surface
temperatures (SST) in Gulf of Guinea, including an abrupt drop in lake
level that lasted about 500 years ca. 118 ka, corresponding to cooling in
the Gulf of Guinea and much of the North Atlantic during the C26 cooling
event. The timing and duration of the event are comparable to the Late
Eemian Aridity Pulse (LEAP; Sirocko et al.,, 2005) that is observed in
several sites in Europe, most notably several lakes in the Eifel Maar
complex (Sirocko et al., 2005; Seelos and Sirocko, 2007). The event is
interpreted to have resulted from abrupt cooling in the North Atlantic, and
possibly a reduction of Atlantic meridional overturning circulation. These
two phenomena are known related to West African aridity over a range of
timescales, so it seems likely that the 500-yr-long arid interval around 118
ka is responding to the same forcing.

The occurrence of quasiperiodic variability at multidecadal to centennial
timescales is consistent with the hypothesis that slow-varying changes in
Atlantic SST structure (e.g., the Atlantic Multidecadal Oscillation) drives
long term hydrologic variability in the region. The periodicities vary over
the course of the record, further implicating the role ocean circulation.
This does not support the hypothesis that solar variability plays an
important, direct role in regional hydrology, especially because there is no
concentration of spectral power at well known periodicities of solar
variability.
168
5.2 Implications for future hydrology in the region
If the Last Interglacial holds as a partial analog for future warming in the region,
the varve record for the interval has several important implications for the future in the
region. One robust comparison between future projections and the LIG is the increase in
Northern Hemisphere summer heating, especially on land, relative to the rest of the
world. This asymmetric heating affects the hemispheric heat balance during the monsoon
season and would likely effect the timing and intensity of monsoon rains throughout West
Africa. When this occurred during the Last Interglacial, Lake Bosumtwi, and probably
most of the Guinea Coast region, was substantially drier, on average, than during the
Holocene. Most evidence from the Sahel (and Sahara) during the interval suggests wetterthan-Holocene conditions during the LIG (Drake et al., 2011). This scenario is precisely
what is predicted for the 21st century by the AR4 MIROC model, one of the four AR4
models which simulated the WAM well.
Whereas the mean state was drier in the Guinea Coast region, drought variability
during the Eemian was similar to the late Holocene, dominated by multidecadal to
centennial scale droughts and pluvials. Further extending the LIG analogy to the future
would suggest similar variability throughout coming centuries superimposed on generally
drier conditions on the Guinea Coast, and perhaps generally wetter conditions in the
Sahel. That said, drought variability is most likely linked to variability in Atlantic SST
and circulation, and the impact of future warming on the AMO and AMOC remains
uncertain. The results presented here do imply that variability in the system was robust to
the changes in seasonal forcing during the LIG.
169
Finally, the occurrence of an abrupt aridity event at the end of the interglacial, that
lasted about 500 years, and may be correlative with widespread aridity in Europe,
highlights the capacity of the system to rapidly fall into severe, prolonged drought. This
drought appears to have been far more severe than the “Sahel Drought”, and occurred
during an interglacial. The exact cause and the likelihood of such an event remain
uncertain, the fact that such widespread drought can occur during conditions similar to
today, justifies further investigation into the underlying mechanisms and the potential of
such an event in the future.
6. References
Abebe, N. T. (2010), Paleohydrology of West Africa Using Carbonate, Detrital and
Diagenetic Minerals of Lake Bosumtwi, Ghana, M.S., University of Akron.
Benson, C., and E. J. Clay (1998), The impact of drought on sub-Saharan African
economies: a preliminary examination, World Bank Publications.
Berger, A., and M. F. Loutre (1991), Insolation values for the climate of the last 10
million years, Quaternary Science Reviews, 10(4), 297–317.
Bergner, A. G. N., M. H. Trauth, and B. Bookhagen (2003), Paleoprecipitation estimates
for the Lake Naivasha basin (Kenya) during the last 175 ky using a lake-balance
model, Global and Planetary Change, 36(1-2), 117–136.
Brauer, A., J. R. . Allen, J. Mingram, P. Dulski, S. Wulf, and B. Huntley (2007), Evidence
for Last Interglacial chronology and environmental change from Southern Europe,
Proceedings of the National Academy of Sciences, 104(2), 450.
Braun, H., M. Christl, S. Rahmstorf, A. Ganopolski, A. Mangini, C. Kubatzki, K. Roth,
and B. Kromer (2005), Possible solar origin of the 1,470-year glacial climate
cycle demonstrated in a coupled model, Nature, 438(7065), 208–211.
Brewer, S., J. Guiot, M. F. Sánchez-Goņi, and S. Klotz (2008), The climate in Europe
during the Eemian: a multi-method approach using pollen data, Quaternary
170
Science Reviews, 27(25-26), 2303–2315.
Carothers, W. W., L. H. Adami, and R. J. Rosenbauer (1988), Experimental oxygen
isotope fractionation between siderite-water and phosphoric acid liberated CO2siderite, Geochimica et Cosmochimica Acta, 52(10), 2445–2450.
Castañeda, I. S., S. Mulitza, E. Schefuss, R. A. . dos Santos, J. S. . Damsté, and S.
Schouten (2009), Wet phases in the Sahara/Sahel region and human migration
patterns in North Africa, Proceedings of the National Academy of Sciences,
106(48), 20159–20163.
Chacko, T., and P. Deines (2008), Theoretical calculation of oxygen isotope fractionation
factors in carbonate systems, Geochimica et Cosmochimica Acta, 72(15), 3642–
3660.
Chang, P., R. Zhang, W. Hazeleger, C. Wen, X. Wan, L. Ji, R. J. Haarsma, W.-P.
Breugem, and H. Seidel (2008), Oceanic link between abrupt changes in the
North Atlantic Ocean and the African monsoon, Nature Geoscience, 1(7), 444–
448, doi:10.1038/ngeo218.
Chen, T. C., and H. van Loon (1987), Interannual variation of the tropical easterly jet,
Monthly Weather Review, 115(8), 1739–1759.
Christensen, J. H. et al. (2007), Regional climate projections, Climate Change, 2007: The
Physical Science Basis. Contribution of Working group I to the Fourth
Assessment Report of the Intergovernmental Panel on Climate Change,
University Press, Cambridge, Chapter 11, 847–940.
Cook, E. R., C. A. Woodhouse, C. M. Eakin, D. M. Meko, and D. W. Stahle (2004),
Long-term aridity changes in the western United States, Science, 306(5698),
1015–1018.
Cook, K. H., and E. K. Vizy (2006), Coupled model simulations of the West African
monsoon system: Twentieth-and twenty-first-century simulations, Journal of
climate, 19(15), 3681–3703.
Cubasch, U., G. Bürger, I. Fast, T. Spangehl, and S. Wagner (2005), The direct solar
influence on climate: modeling the lower atmosphere, Memorie della Societa
Astronomica Italiana, 76, 810.
Danabasoglu, G. (2008), On Multidecadal Variability of the Atlantic Meridional
Overturning Circulation in the Community Climate System Model Version 3,
Journal of Climate, 21(21), 5524–5544, doi:10.1175/2008JCLI2019.1.
171
Danabasoglu, G., Yeager, S.G., Kwon, Y.-O., Tribbia, J. J., Phillips, A. S., and Hurrell, J.
W. (n.d.), Variability of the Atlantic Meridional Overturning Circulation in
CCSM4., Journal of Climate.
Davidson, P. M., G. H. Symmes, B. A. Cohen, R. J. Reeder, and D. H. Lindsley (1993),
Synthesis of the new compound CaFe (CO3) 2 and experimental constraints on
the (Ca, Fe) CO3 join, Geochimica et Cosmochimica Acta, 57(23-24), 5105–
5109.
deMenocal, P. B. (2001), Cultural Responses to Climate Change During the Late
Holocene, Science, 292(5517), 667–673, doi:10.1126/science.1059287.
deMenocal, P. B., and others (1995), Plio-pleistocene African climate, Science,
270(5233), 53–59.
deMenocal, P. B., W. F. Ruddiman, and E. M. Pokras (1993), Influences of high-and lowlatitude process on African terrestrial climate to variations in seasonal insolation,
glacial ice cover, sea-surface temperature and Asian orography, Journal of
Geophysical Research, 98, 7265–7287.
Dezfuli, A. K., and S. E. Nicholson (2011), A note on long-term variations of the African
easterly jet, International Journal of Climatology, 31(13), 2049–2054.
Dima, M., and G. Lohmann (2007), A hemispheric mechanism for the Atlantic
Multidecadal Oscillation, Journal of Climate, 20(11), 2706–2719.
Drake, N. A., R. M. Blench, S. J. Armitage, C. S. Bristow, and K. H. White (2011),
Ancient watercourses and biogeography of the Sahara explain the peopling of the
desert, Proceedings of the National Academy of Sciences, 108(2), 458–462.
Ghil, M. et al. (2002), Advanced spectral methods for climatic time series, Rev. Geophys,
40(1), 1003.
Goosse, H., and H. Renssen (2004), Exciting natural modes of variability by solar and
volcanic forcing: idealized and realistic experiments, Climate Dynamics, 23(2),
153–163.
Gray, S. T., L. J. Graumlich, J. L. Betancourt, and G. T. Pederson (2004), A tree-ring
based reconstruction of the Atlantic Multidecadal Oscillation since 1567 AD,
Geophysical Research Letters, 31(12), L12205.
Grodsky, S. A., J. A. Carton, and S. Nigam (2003), Near surface westerly wind jet in the
Atlantic ITCZ, Geophys. Res. Lett, 30(19), 2009.
172
Hearty, P. J., J. T. Hollin, A. C. Neumann, M. J. O’Leary, and M. McCulloch (2007),
Global sea-level fluctuations during the Last Interglaciation (MIS 5e), Quaternary
Science Reviews, 26(17-18), 2090–2112.
He, C., H. Breuning-Madsen, and T. W. Awadzi (2007), Mineralogy of dust deposited
during the Harmattan season in Ghana, Danish Journal of Geography, 107.
Heil, C., J. King, J. Peck, T. Shanahan, N. McKay, J. Overpeck, and C. Scholz, (in prep)
Geochronology of the Lake Bosumtwi sediment sequence for the past million
years, In preparation for Quaternary Science Reviews.
Held, I., T. Delworth, J. Lu, K. Findell, and T. Knutson (2005), Simulation of Sahel
drought in the 20th and 21st centuries, Proceedings of the National Academy of
Sciences of the United States of America, 102(50), 17891.
Held, I. M., and B. J. Soden (2006), Robust responses of the hydrological cycle to global
warming, Journal of Climate, 19(21), 5686–5699.
Herman, R. G., C. E. Bogdan, A. J. Sommer, and D. R. Simpson (1987), Discrimination
among carbonate minerals by Raman spectroscopy using the laser microprobe,
Applied spectroscopy, 41(3), 437–440.
Hodell, D. A., M. Brenner, J. H. Curtis, and T. Guilderson (2001), Solar forcing of
drought frequency in the Maya lowlands, Science, 292(5520), 1367–1370.
Huybers, P., and W. Curry (2006), Links between annual, Milankovitch and continuum
temperature variability, Nature, 441(7091), 329–332.
Itambi, A. C., T. Von Dobeneck, S. Mulitza, T. Bickert, and D. Heslop (2009), Millennialscale northwest African droughts related to Heinrich events and DansgaardOeschger cycles: Evidence in marine sediments from offshore Senegal,
Paleoceanography, 24(1), PA1205.
Janicot, S., S. Trzaska, and I. Poccard (2001), Summer Sahel-ENSO teleconnection and
decadal time scale SST variations, Climate Dynamics, 18(3), 303–320.
Jourdan, F., P. R. Renne, and W. U. Reimold (2009), An appraisal of the ages of terrestrial
impact structures, Earth and Planetary Science Letters, 286(1), 1–13.
Kim, S., and J. Oneil (1997), Equilibrium and nonequilibrium oxygen isotope effects in
synthetic carbonates, Geochimica et Cosmochimica Acta, 61, 3461–3475.
Klotz, S., U. Müller, V. Mosbrugger, J.-L. de Beaulieu, and M. Reille (2004), Eemian to
early Würmian climate dynamics: history and pattern of changes in Central
173
Europe, Palaeogeography, Palaeoclimatology, Palaeoecology, 211(1–2), 107–126,
doi:10.1016/j.palaeo.2004.04.009.
Knight, J. R., R. J. Allan, C. K. Folland, M. Vellinga, and M. E. Mann (2005), A signature
of persistent natural thermohaline circulation cycles in observed climate,
Geophys. Res. Lett, 32(L20708).
Koeberl, C. et al. (2007), An international and multidisciplinary drilling project into a
young complex impact structure: The 2004 ICDP Bosumtwi Crater Drilling
Project—An overview, Meteoritics & Planetary Science, 42(4-5), 483–511.
Koeberl, C., R. Bottomley, B. P. Glass, and D. Storzer (1997), Geochemistry and age of
Ivory Coast tektites and microtektites, Geochimica et Cosmochimica Acta, 61(8),
1745–1772.
Koeberl, C., W. U. Reimold, J. D. Blum, and C. P. Chamberlain (1998), Petrology and
geochemistry of target rocks from the Bosumtwi impact structure, Ghana, and
comparison with Ivory Coast tektites, Geochimica et Cosmochimica Acta, 62(12),
2179–2196.
Kopp, R. E., F. J. Simons, J. X. Mitrovica, A. C. Maloof, and M. Oppenheimer (2009),
Probabilistic assessment of sea level during the Last Interglacial stage, Nature,
462(7275), 863–867.
Kühl, N., T. Litt, C. Schölzel, and A. Hense (2007), Eemian and Early Weichselian
temperature and precipitation variability in northern Germany, Quaternary
Science Reviews, 26(25-28), 3311–3317.
Lambert, F., B. Delmonte, J. R. Petit, M. Bigler, P. R. Kaufmann, M. A. Hutterli, T. F.
Stocker, U. Ruth, J. P. Steffensen, and V. Maggi (2008), Dust-climate couplings
over the past 800,000 years from the EPICA Dome C ice core, Nature, 452(7187),
616–619.
Lamb, P. J. (1978), Large-scale tropical Atlantic surface circulation patterns associated
with Subsaharan weather anomalies, Tellus, 30, 240–251.
Landrum, L., Otto-Bliesner, B.L., Conley, A., Lawrence., P., Rosenbloom, N., and Teng,
H. (in review), Last Millennium Climate and Its Variability in CCSM4., Journal of
Climate.
Lotter, A. F., and G. Lemcke (1999), Methods for preparing and counting biochemical
varves, Boreas, 28(2), 243–252.
Lu, J., C. Deser, and T. Reichler (2009), Cause of the widening of the tropical belt since
174
1958, Geophysical Research Letters, 36, 5 PP., doi:200910.1029/2008GL036076.
Lund, S., J. S. Stoner, J. E. T. Channell, and G. Acton (2006), A summary of Brunhes
paleomagnetic field variability recorded in Ocean Drilling Program cores, Physics
of the earth and planetary interiors, 156(3-4), 194–204.
Lüthi, D. et al. (2008), High-resolution carbon dioxide concentration record 650,000–
800,000 years before present, Nature, 453(7193), 379–382.
Mann, M. E., and J. M. Lees (1996), Robust estimation of background noise and signal
detection in climatic time series, Climatic Change, 33(3), 409–445.
Martrat, B., J. O. Grimalt, C. Lopez-Martinez, I. Cacho, F. J. Sierro, J. A. Flores, R. Zahn,
M. Canals, J. H. Curtis, and D. A. Hodell (2004), Abrupt temperature changes in
the western Mediterranean over the past 250,000 years, Science, 306(5702), 1762.
Martrat, B., J. O. Grimalt, N. J. Shackleton, L. de Abreu, M. A. Hutterli, and T. F. Stocker
(2007), Four climate cycles of recurring deep and surface water destabilizations
on the Iberian margin, Science, 317(5837), 502.
McKay, N. P., J. T. Overpeck, and B. L. Otto-Bliesner (2011), The role of ocean thermal
expansion in Last Interglacial sea level rise, Geophysical Research Letters,
38(14), L14605.
Meehl, G. A., J. M. Arblaster, K. Matthes, F. Sassi, and H. van Loon (2009), Amplifying
the Pacific climate system response to a small 11-year solar cycle forcing,
Science, 325(5944), 1114–1118.
Mulitza, S., M. Prange, J. B. Stuut, M. Zabel, T. von Dobeneck, A. C. Itambi, J. Nizou,
M. Schulz, and G. Wefer (2008), Sahel megadroughts triggered by glacial
slowdowns of Atlantic meridional overturning, Paleoceanography, 23(4), PA4206.
Nicholson, S. E. (2001), Climatic and environmental change in Africa during the last two
centuries, Climate Research, 17(2), 123–144.
Nicholson, S. E. (2008), The intensity, location and structure of the tropical rainbelt over
west Africa as factors in interannual variability, International Journal of
Climatology, 28(13).
Nicholson, S. E. (2009a), A revised picture of the structure of the “monsoon” and land
ITCZ over West Africa, Climate Dynamics, 29, doi:10.1007/s00382-008-0514-3.
Nicholson, S. E. (2009b), On the factors modulating the intensity of the tropical rainbelt
over West Africa, International Journal of Climatology.
175
Nicholson, S. E., and J. P. Grist (2001), A conceptual model for understanding rainfall
variability in the West African Sahel on interannual and interdecadal timescales,
International Journal of Climatology, 21(14), 1733–1757.
Nicholson, S. E., and P. J. Webster (2007), A physical basis for the interannual variability
of rainfall in the Sahel, Quarterly Journal of the Royal Meteorological Society,
133(629).
Oppo, D. W., J. F. McManus, and J. L. Cullen (2006), Evolution and demise of the Last
Interglacial warmth in the subpolar North Atlantic, Quaternary Science Reviews,
25(23-24), 3268–3277.
Peck, J. A., R. R. Green, T. Shanahan, J. W. King, J. T. Overpeck, and C. A. Scholz
(2004), A magnetic mineral record of Late Quaternary tropical climate variability
from Lake Bosumtwi, Ghana, Palaeogeography, Palaeoclimatology,
Palaeoecology, 215(1-2), 37–57.
Peck, J., C. Koeberl, J. King, B. Milkereit, J. Overpeck, and C. A. Scholz (2005), The
Lake Bosumtwi drilling project: Initial Report, Geological Society of America,
Limnogeology Division Newsletter, 2(2), 3–7.
Peck, J., N. McKay, T. Shanahan, C. Heil, J. Overpeck, J. King, and C. Scholz (in prep),
A million-year history of the West African Monsoon, in preparation for Science.
Pu, B., and K. H. Cook (2010), Dynamics of the West African westerly jet, Journal of
Climate, 23(23), 6263–6276.
Pu, B., and K. H. Cook (2012), Role of the West African Westerly Jet in Sahel Rainfall
Variations, Journal of Climate, 25(8), 2880–2896, doi:10.1175/JCLI-D-1100394.1.
Randall, D. A. et al. (2007), Climate models and their evaluation, Climate Change, 2007:
The Physical Science Basis. Contribution of Working group I to the Fourth
Assessment Report of the Intergovernmental Panel on Climate Change,
University Press, Cambridge, Chapter 8, 589-662.
Rasmussen, S. O. et al. (2006), A new Greenland ice core chronology for the last glacial
termination, Journal of geophysical research, 111(D6), D06102.
Rodríguez-Fonseca, B. et al. (2011), Interannual and decadal SST-forced responses of the
West African monsoon, Atmospheric Science Letters, 12(1), 67–74.
Rowell, D. P. (2003), The impact of Mediterranean SSTs on the Sahelian rainfall season,
176
Journal of Climate, 16(5), 849–862.
Rutt, H. N., and J. H. Nicola (1974), Raman spectra of carbonates of calcite structure,
Journal of Physics C: Solid State Physics, 7, 4522.
Seelos, K., and F. Sirocko (2007), 14. Abrupt cooling events at the very end of the Last
Interglacial, Developments in Quaternary Sciences, 7, 207–229.
Shanahan, T., J. Beck, J. Overpeck, N. McKay, and J. Pigati, (submitted) Late Quaternary
sedimentological and climate changes at Lake Bosumtwi Ghana: new constraints
from laminae analysis and radiocarbon age modeling, Quaternary Science
Reviews.
Shanahan, T. M. (2006), West African monsoon variability from a high-resolution
paleolimnologic record (Lake Bosumtwi, Ghana), University of Arizona.
Shanahan, T. M., J. T. Overpeck, J. W. Beck, C. W. Wheeler, J. A. Peck, J. W. King, and
C. A. Scholz (2008a), The formation of biogeochemical laminations in Lake
Bosumtwi, Ghana, and their usefulness as indicators of past environmental
changes, Journal of Paleolimnology, 40(1), 339–355.
Shanahan, T. M., J. T. Overpeck, J. B. Hubeny, J. King, F. S. Hu, K. Hughen, G. Miller,
and J. Black (2008b), Scanning micro-X-ray fluorescence elemental mapping: A
new tool for the study of laminated sediment records, Geochemistry, Geophysics,
Geosystems, 9(2), Q02016.
Shanahan, T. M., J. T. Overpeck, W. E. Sharp, C. A. Scholz, and J. A. Arko (2007),
Simulating the response of a closed-basin lake to recent climate changes in
tropical West Africa (Lake Bosumtwi, Ghana), Hydrological Processes, 21(13).
Shanahan, T. M., J. T. Overpeck, C. W. Wheeler, J. W. Beck, J. S. Pigati, M. R. Talbot, C.
A. Scholz, J. Peck, and J. W. King (2006), Paleoclimatic variations in West Africa
from a record of late Pleistocene and Holocene lake level stands of Lake
Bosumtwi, Ghana, Palaeogeography, Palaeoclimatology, Palaeoecology, 242(3),
287–302.
Shanahan, T., J. Overpeck, K. Anchukaitis, J. Beck, J. Cole, D. Dettman, J. Peck, C.
Scholz, and J. King (2009), Atlantic forcing of persistent drought in West Africa,
Science, 324(5925), 377.
Shanahan, T., J. Peck, S. Forman, C. Heil, N. McKay, J. Overpeck, J. King, and C. Scholz
(in prep), Geochronology of the Lake Bosumtwi sediment sequence for the past
200,000 years, in preparation for Quaternary Science Reviews.
177
Shindell, D. T., G. Faluvegi, R. L. Miller, G. A. Schmidt, J. E. Hansen, and S. Sun (2006),
Solar and anthropogenic forcing of tropical hydrology, Geophys. Res. Lett,
33(24), L24706.
Shindell, D. T., R. L. Miller, G. A. Schmidt, and L. Pandolfo (1999), Simulation of recent
northern winter climate trends by greenhouse-gas forcing, Nature, 399(6735),
452–455.
Shindell, D. T., and G. A. Schmidt (2004a), Southern Hemisphere climate response to
ozone changes and greenhouse gas increases, Geophysical Research Letters, 31, 4
PP., doi:200410.1029/2004GL020724.
Shindell, D. T., and G. A. Schmidt (2004b), Southern Hemisphere climate response to
ozone changes and greenhouse gas increases, Geophysical Research Letters,
31(18), L18209.
Short, D. A., J. G. Mengel, T. J. Crowley, W. T. Hyde, and G. R. North (1991), Filtering
of Milankovitch cycles by Earth’s geography, Quaternary Research, 35(2), 157–
173.
Sirocko, F., K. Seelos, K. Schaber, B. Rein, F. Dreher, M. Diehl, R. Lehne, K. Jäger, M.
Krbetschek, and D. Degering (2005), A late Eemian aridity pulse in central
Europe during the last glacial inception, Nature, 436(7052), 833–836.
Stager, J. C., D. B. Ryves, B. M. Chase, and F. S. . Pausata (2011), Catastrophic drought
in the Afro-Asian monsoon region during Heinrich event 1, Science, 331(6022),
1299.
Street-Perrott, F. A., and R. A. Perrott (1990), Abrupt climate fluctuations in the tropics:
the influence of Atlantic Ocean circulation, Nature, 343(6259), 607–612,
doi:10.1038/343607a0.
Talbot, M. R., and G. Delibrias (1977), Holocene variations in the level of Lake
Bosumtwi, Ghana, Nature, 268(5622), 722–724.
Talbot, M. R., and T. Johannessen (1992), A high resolution palaeoclimatic record for the
last 27,500 years in tropical West Africa from the carbon and nitrogen isotopic
composition of lacustrine organic matter, Earth and Planetary Science Letters,
110(1-4), 23–37.
Talbot, M. R., and K. Kelts (1986), Primary and diagenetic carbonates in the anoxic
sediments of Lake Bosumtwi, Ghana, Geology, 14(11), 912–916.
Talbot, M. R., D. A. Livingstone, P. G. Palmer, J. Maley, J. M. Melack, G. Delibrias, and
178
S. Gulliksen (1984), Preliminary results from sediment cores from Lake
Bosumtwi, Ghana, Palaeoecology of Africa, 16, 173–192.
Tarutani, T., R. N. Clayton, and T. K. Mayeda (1969), The effect of polymorphism and
magnesium substitution on oxygen isotope fractionation between calcium
carbonate and water, Geochimica et Cosmochimica Acta, 33(8), 987–996,
doi:10.1016/0016-7037(69)90108-2.
Thompson, D. W. J., J. M. Wallace, J. J. Kennedy, and P. D. Jones (2010), An abrupt drop
in Northern Hemisphere sea surface temperature around 1970, Nature, 467(7314),
444–447.
Tjallingii, R., M. Claussen, J. B. . Stuut, J. Fohlmeister, A. Jahn, T. Bickert, F. Lamy, and
U. Röhl (2008), Coherent high-and low-latitude control of the northwest African
hydrological balance, Nature Geoscience, 1(10), 670–675.
Toggweiler, J. R., and J. Russell (2008), Ocean circulation in a warming climate, Nature,
451(7176), 286–288, doi:10.1038/nature06590.
Trauth, M. H., A. L. Deino, A. G. . Bergner, and M. R. Strecker (2003), East African
climate change and orbital forcing during the last 175 kyr BP, Earth and Planetary
Science Letters, 206(3), 297–313.
Turner, B. F., L. R. Gardner, and W. E. Sharp (1996a), The hydrology of Lake Bosumtwi,
a climate-sensitive lake in Ghana, West Africa, Journal of Hydrology, 183(3-4),
243–261.
Turner, B. F., L. R. Gardner, W. E. Sharp, and E. R. Blood (1996b), The geochemistry of
Lake Bosumtwi, a hydrologically closed basin in the humid zone of tropical
Ghana, Limnology and Oceanography, 1415–1424.
Turney, C. S. ., and R. T. Jones (2010), Does the Agulhas Current amplify global
temperatures during super-interglacials?, Journal of Quaternary Science, 25(6),
839–843.
Vasconcelos, C., J. A. McKenzie, R. Warthmann, and S. M. Bernasconi (2005),
Calibration of the δ18O paleothermometer for dolomite precipitated in microbial
cultures and natural environments, Geology, 33(4), 317–320,
doi:10.1130/G20992.1.
Verschuren, D. et al. (2009), Half-precessional dynamics of monsoon rainfall near the
East African Equator, Nature, 462(7273), 637–641.
Wanner, H. et al. (2008), Mid-to Late Holocene climate change: an overview, Quaternary
179
Science Reviews, 27(19-20), 1791–1828.
Wanner, H., O. Solomina, M. Grosjean, S. P. Ritz, and M. Jetel (2011), Structure and
origin of Holocene cold events, Quaternary Science Reviews, 30(21–22), 3109–
3123, doi:10.1016/j.quascirev.2011.07.010.
Weber, S. L., T. J. Crowley, and G. der Schrier (2004), Solar irradiance forcing of
centennial climate variability during the Holocene, Climate Dynamics, 22(5),
539–553.
Weldeab, S., D. W. Lea, R. R. Schneider, and N. Andersen (2007), 155,000 years of West
African monsoon and ocean thermal evolution, Science, 316(5829), 1303.
Woillard, G. M. (1978), Grande Pile peat bog: a continuous pollen record for the last
140,000 years, Quaternary Research, 9(1), 1–21.
Wolff, E. W., J. Chappellaz, T. Blunier, S. O. Rasmussen, and A. Svensson (2010),
Millennial-scale variability during the last glacial: The ice core record, Quaternary
Science Reviews, 29(21–22), 2828–2838, doi:10.1016/j.quascirev.2009.10.013.
180
Table B1. Multi-channel singular spectrum analysis of the detrended Ti and log
varve thickness time series, for ensembles with and without marker layers (ML).
Ti (ML-ensemble)
Ti (noML-ensemble)
Reconstructed
Variance
Reconstructed
Variance
Component Period (yrs) exlained (%) Component Period (yrs) exlained (%)
1+2
650-750
12
1+2
600-800
20
3+4
240-270
6
3+4
270-330
6
5+6
150-180
3
5+6
230-250
5
7
90
1
7 – 10
160-190
5
8
70
1
11+12
125-145
2
9+10
60
2
13+14
100-110
1.3
11+12
110-120
2
15+16
230-250
1.3
13+14
54-55
1.6
15+16
62-63
1.4
17+18
24
1.4
Log of varve thickness (ML-ensemble)
Log of varve thickness (noML-ensemble)
Reconstructed
Variance
Reconstructed
Variance
Component Period (yrs) exlained (%) Component Period (yrs) exlained (%)
1
1200
29
1
1300
38
2
660
14
2
750
15
3
340
5
3
400
4
4
260
3
4
570
3
5
190
2
5
430
3
6
160
1.5
6–8
250-280
4
7
130
1
9+10
230
2
8
112
0.9
11+12
155
1.5
9
100
0.8
13-17
190-220
3
10+11
80-90
1.2
12+13
55-60
1.1
14+15
40-55
1
181
90
1°27' W
300
90
90
90
45
6°32' N
30 m
50 m
90
90
45
90
90
45
70 m
45
90
6°27' N
90
90
0
1
2
km
Figure B1. Site map of the Lake Bosumtwi crater, modified from Shanahan (2006).
Solid gray lines show bathymetry (10 m contour). Dashed line indicates crater rim.
Inset: location of Lake Bosumtwi in southern Ghana, West Africa.
4.8
10
4.4
8
4.0
6
3.6
4
3.2
2
2.8
monthly rainfall (cm)
Wind speed (m s-1)
182
0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Figure B2. 1961-2000 climatology for Kumasi, Ghana, 30 km from
Lake Bosumtwi. Grey bars indicate monthly precipitation totals (cm),
dashed line shows wind speed (m s-1). Modified from Shanahan (2006).
183
Figure B3. Multichannel seismic reflection profile showing draped reflectors of the lacustrine fill and
the brecciated bedrock of the central uplift (from Scholz et al., 2002). The 13 drill holes at the five drill
sites are shown as black bars. Drill hole 5B is shown as a solid line, along with the down core median
grain size profile at the site.
Ti intensity (cps)
184
200
Itrax XRF
μXRF
150
100
50
0
2000
4000
6000 8000
Varve Year
10000 12000
Figure B4. Comparison between Itrax and μXRF data for the
study interval. The very high values between varve year 4000
and 6000 correspond to turbidites that are averaged into one
year in the varve record.
185
Age model ensemble envelope
Age model median estimate
EDC-dust-tuned Age Model
Blake Excursion (Top and Bottom)
OSL ages
U/Th ages
30
Depth (RMCD turbidite−adjusted)
35
40
Interpreted LIG Interval
(varve section)
45
50
55
60
65
70
50
100
150
Age (ka)
Figure B5. Age control for Lake Bosumtwi core 5B between 30 and 75 RMCD. Error bars on OSL
and U/Th ages show 2 sigma uncertainty ranges. The yellow band corresponds to the of the
floating chronology after "pinning" to the NGRIP (Wolff et al., 2010) chronology, see text for details.
186
Age (ka)
−35
A
100
80
60
H7
GS25
40
H5.2
H6
H4
H5
20
H3
0
YD
H2 H1
−6
−45
1 kyr
−4
B
−2
O lake water (‰)
18
δ
Lake Bosumtwi
0
2
0
4
C
5
10
29
D
28
27
26
25
Gulf of Guinea SSS (psu)
XRF−PC1 (Terrigenous)
−40
22
24
24
23
26
Gulf of Guinea SST (°C)
18
NGRIP δ O (‰)
120
E
28
30
32
120
100
80
60
Age (ka)
40
20
0
Figure B6. A) δ18O of Greenland ice tracking North Atlantic temperature (NGRIP, 2004; Wolff et al.,
2010). B) Lake Bosumtwi terrigenous input, an indicator of lake level (Appendix A). C) δ 18O of lake
water inferred from authigenic carbonates in Lake Bosumtwi (this study). D) Sea surface temperature
and E) salinity in the Gulf of Guinea (Weldeab et al., 2007).The broad agreement between the series,
and especially the abrupt event ca. 110 ka corresponding to GS25 and GI24, suggest that the MIS 5
section of the Bosumtwi chronology is too young by about 1 ka. This suggests that the end of the
varved interval (red circle) occurred ca. 116.5 ka.
187
Calcite
Siderite
Dolomite
Dolomite/Ankerite
Unit
Varve
year
11000
58
3a
10000
2d
9000
8000
2c
Depth (RMCD)
Core Description
Laminations thicken upwards, minerogenic muds
with diffuse dolomite layers, diagenetic FeS layers,
and occasional well-rounded quartz grains
3b
59
Fine-grained FeS
12000
57.5
58.5
Mineralogy
Diagenetic FeS
Diagenetic Gypsum
2b
Dark brown muds with abundant well-rounded quartz
grains and diagenetic dolomite
7000
Finely laminated organic-rich mud, with packets of
dolomite and occasional large (1 - 3 mm) diagenetic
gypsum crsytals
Coarsely laminated mineral-rich muds with abundant
medium to coarse silt-sized quartz grains
6000
Organic-rich silts with abundant reddish, FeS-rich layers,
with occasionaly calcites and diagenetic gypsum
5000
2a
Very finely-laminated dark brown, organic rich muds
interbedded with large coarse-grained turbidites, and
manganosiderite
59.5
4000
Well-laminated with occasional diagenetic FeS layers
and well-defined calcite layers.
1d
3000
60
1c
Thick, poorly defined minerogenic laminae.
1b
Well-laminated tan terrigenous layers alternating with
calcite and organic material, with more reddish, finegrained FeS-rich lamina at the base
2000
1000
60.5
1a
Dark brown finely-laminated with occasional turbidites, and
abundant well-rounded quartz grains at its base
1
Figure B7. Generalized lithology, mineralogy and lamination structure of the study interval.
188
25
20
δ13Cpdb
15
10
Calcite
Diagenetic Calcite
Manganosiderite
Dolomite
Dolomite/Ankerite
5
0
0
2
4
6
8
10
12
14
18
δ Opdb
Figure B8. Relationship between δ18O, δ13C , and carbonate mineralogy.
189
Varve Years
4000
6000
8000
10000
12000
−4
−2
50
Ti (cps)
0
100
20
40
60
80
100
120
140
10
20
200
4
2
x 10
1.5
K (cps)
100
30
300
1
10000
0.5
Mn (cps)
Fe (cps)
Si (cps)
2
150
Al (cps)
Log varve
thickness
2000
5000
0
60
0
40
S (cps)
Ca (cps)
2000
20
2000
4000
6000
Varve Years
8000
10000
12000
Figure B9. 10-yr means of varve thickness and XRF ensemble members for Ti, Si, Al, K,
Fe, Mn, Ca and S. The 5 to 95% annual ensemble envelopes are shown in light gray. Note
that the axes for log varve thickness, Ti, Si, Al, and K are inverted.
190
Ca/Mn ratio
30
12
25
20
10
15
8
10
6
10500
9500
10000
Varve year
10
0.4
8
9000
13
C
12
11
11
δ Ocarb
10
18
18
δ Ocarb
0.6
12
9
8
7
12
0.8
0.2
5
9000
13
14
B
carb
1
δ18O
14
A
Ca/Fe ratio
35
9500
10000
Varve year
6
10500
D
10
9
8
10
15
20
25
Ca/Mn ratio
30
35
7
0.2
0.4
0.6
0.8
Ca/Fe ratio
1
Figure B10. XRF Ca/Mn (A) and Ca/Fe (B) ratios plotted along with the corresponding δ 18O values
in the ankeritic interval of the record. (C) and (D) show the same data as (A) and (B), respectively,
plotted against each other. High Ca/(Mn or Fe) ratios likely corresponds to more dolomitic carbonate,
which typically is more enriched in δ18O, consistent oxygen isotope fractionation theory. Samples
with lower elemental ratios and δ18O values are probably more ankeritic.
191
128
126
124
122
120
118
116
50
100
10
150
-1
Gulf of Guinea SST (°C)
10
0
10
28
Turbidites
Varve thickness
(mm)
Lake Level
5
Variable ankerite content,
18
likely range of lake δ O
Low
Ti intensity (cps)
0
18
Lake water δ O
High
Age (ka)
27
26
128
126
124
Age (ka)
122
120
118
116
Figure B11. Lake level indicators for Lake Bosumtwi last interglacial varve record, along with Gulf of
Guinea sea surface temperatures (Weldeab et al., 2007). The blue band corresponds to the range
of possible lake water δ18O, given the carbonate δ18O, mineralogy, and a range of temperature.
The rectangular band corresponds to an interval of variably dolomitic to ankeritic carbonate, which
cannot be properly accounted for. However, the band corresponds to the best estimate of mean lake
water δ18O for the interval. See text for details. Sharp, single year peaks in varve thickness are
turbidites.
192
Age (ka)
119
118
117
13
Dolomite
120
11
18
80
Ankerite
9
Lake Level
δ OCarb
Ti (cps)
High
40
120
−1
Low
10
0
10
Varve thickness (mm)
7
Loess content (%)
0
20
40
HL2
Ei2
RS1
60
120
119
118
117
Age (ka)
Figure B12. Lake Bosumtwi Ti abundance, varve thickness and carbonate δ18O between 120
and 116.5 ka, along with the loess content of lake sediments from the Eifel Maar Complex
(Sirocko et al., 2005; Seelos and Sirocko, 2007. Eifel Maar lakes West of Hoer List (HL2),
Eigelback (Ei2), and Rederstall (RS1), record abrupt increases in less during the Late Eemian
Aridity Pulse (LEAP), ca. 118 ka. The yellow bar indicates a ~500 yr interval of extremely dry
conditions at Lake Bosumtwi that is consistent in timing with the LEAP.
128
126
124
122
Age (ka)
120
400
118
Lake
overflowing
470
460
40
450
80
440
430
120
10
8
6
Age (ka)
4
2
0
420
440
420
400
420
400
380
Figure B13. Comparison of Lake Bosumtwi lake level and 10N rainy season insolation for July
(solid red) and October, for the last interglacial (top) and the Holocene (bottom). Inferred
lake water δ18O, and Ti intensity as in figure B11. Geomorphic evidence indicates that the
lake was overflowing between 11 and 9 ka, despite somewhat higher terrigenous content
in the sediment (Shanahan, 2006; Shanahan et al., 2006).
October insolation (W m−2)
420
150
10
Ti (cps)
Low
440
440
October insolation (W m−2)
8
100
July insolation (W m−2)
6
460
460
July insolation (W m−2)
4
480
50
Ti (cps)
Inferred water δ18O
2
Lake Level
High
lw
193
194
128
126
124
122
120
118
116
50
5
100
150
0
North Atlantic
(ODP 980)
JJA SST (oC)
10
10
5
25
0
20
23
10
21
28
19
27
26
128
126
124
122
Age (ka)
120
118
Varve thickness
(mm)
-1
10
Western Mediterranean
and Iberian Margin SST (oC)
10
Variable ankerite content,
18
likely range of lake δ O
Western North Atlantic
(MD95-2036)
JJA SST (oC)
Ti intensity (cps)
0
Gulf of Guinea
SST (oC)
Lake water δ18O
Age (ka)
116
Figure B14. Same as figure B11, but with additional high resolution SST records included, from the
western Mediterranean and Iberian Margin regions (Martrat et al., 2004; 2007), the eastern
subpolar North Atlantic (Oppo et al., 2006), the western North Atlantic (Lehman et al., 2002)
as well as Gulf of Guinea SSTs (Weldeab et al., 2007).
195
A
4
10
269
112
73
3
10
Power
59 42
34 28
19
17
95%
90%
14
2
10
0
0.02
0.04
0.06
0.08
0.1
0.12
Frequency (1/yr)
B
1
10
279
154
Power
0
10
104
78 59
46
27
95%
90%
20
−1
10
0
0.02
0.04
0.06
0.08
0.1
0.12
Frequency (1/yr)
Figure B15. Monte carlo MTM spectra for (A) Ti intensities and
(B) the log of varve thickness. For both, the black line indicates
the median power spectrum for all ensembles, and the gray
envelope shows the 5 to 95% range of the spectra. Red dashed
lines show the median 90 and 95% confidence levels The
periodicites of pronounced peaks are labeled, rounded to the
nearest years.
196
Ti intensity (cps)
100
200
Frequency (1/yr)
0.02
0.03
0.04
0.01
400 120
12000
150
10000
110
44
77
7765
Varve Year
120
71
12000
23
10000
49
37 33
140 100
74
250
23
33
48
130 95
77
8000
36 32
0.05
40
50
33
34
26
23
28
8000
22
Varve Year
0
29
500
6000
95 74
220
140 91 69
4000
6000
53
56
38 33 30
26 24
43
4000
25 23
220
120
71
57
35
27
21
2000
2000
180 110
67 56
0
100
Ti intensity (cps)
200
0.01
43
0.02
34
28
0.03
26
0.04
23
0.05
Frequency (1/yr)
Figure B16. Evolutive spectra for Ti intensity. Left panel: median (black) and 95% confidence
envelope (gray) of full ensemble (no marker layers) used to calculate spectra. Right panel: MC-MTM
spectra, as in Figure B15, for partially overlapping 2000 year windows. Vertical position corresponds
to mid pointof each window. Dot-dashed vertical blue lines corresponds to well-known frequencies of
solar variability. Periodicities of pronounced peaks are labeled on each spectrum.
197
log varve thickness
−3 −2 −1
0
1
12000
Frequency (1/yr)
0.02
0.03
0.04
0.01
400
140 110
83
0.05
12000
59
47 42
180
10000
110 74
57
10000
32
330
170 120
250
330
130
47
33
28
25
8000
83
49
130 83
27
53
6000
23
27
21
6000
Varve Year
Varve Year
8000
74 59
290
80
61
45
27
36
21
4000
4000
290
130
61
28
45
140
110 77
2000
50
32
20
28
2000
170
110 80
−3 −2 −1
0
1
log varve thickness
0.01
59 49 43
31 29
0.02
0.03
0.04
Frequency (1/yr)
23
0.05
Figure B17. Evolutive spectra for log varve thickness. Left panel: median (black) and 95% confidence
envelope (gray) of full ensemble (no marker layers) used to calculate spectra. Right panel: MC-MTM
spectra, as in Figure B15, for partially overlapping 2000 year windows. Vertical position corresponds
to mid pointof each window. Dot-dashed vertical blue lines corresponds to well-known frequencies of
solar variability. Periodicities of pronounced peaks are labeled on each spectrum.
198
APPENDIX C
THE ROLE OF OCEAN THERMAL EXPANSION IN
LAST INTERGLACIAL SEA LEVEL RISE
Published in the professional journal: Geophysical Research Letters
Copyright 2011 American Geophysical Union
Reproduced by permission of American Geophysical Union.
McKay, N. P., J. T. Overpeck, and B. L. Otto-Bliesner, The role of ocean thermal
expansion in Last Interglacial sea level rise, Geophysical Research Letters, 38(14),
L14605, 2011. Copyright 2011 American Geophysical Union.
199
The Role of Ocean Thermal Expansion in Last Interglacial Sea Level Rise
1*
Nicholas P. McKay , Jonathan T. Overpeck
1,2,3
, Bette L. Otto-Bliesner
4
1. Department of Geosciences, University of Arizona, Tucson, AZ 85721, USA
2. Institute of the Environment, University of Arizona, 845 North Park Avenue, Suite 532,
Tucson, AZ 85719, USA
3. Department of Atmospheric Science, University of Arizona, Tucson, AZ 85721, USA
4. National Center for Atmospheric Research, Post Office Box 3000, Boulder CO 80307
USA
*Corresponding author: nmckay@email.arizona.edu
200
Abstract
A compilation of paleoceanographic data and a coupled atmosphere-ocean climate
model were used to examine global ocean surface temperatures of the Last Interglacial
(LIG) period, and to produce the first quantitative estimate of the role that ocean thermal
expansion likely played in driving sea level rise above present day during the LIG. Our
analysis of the paleoclimatic data suggests a peak LIG global sea surface temperature
(SST) warming of 0.7±0.6°C compared to the late Holocene. Our LIG climate model
simulation suggests a slight cooling of global average SST relative to preindustrial
conditions (ΔSST = -0.4°C), with a reduction in atmospheric water vapor in the Southern
Hemisphere driven by a northward shift of the Intertropical Convergence Zone, and
substantially reduced seasonality in the Southern Hemisphere. Taken together, the model
and paleoceanographic data imply a minimal contribution of ocean thermal expansion to
LIG sea level rise above present day. Uncertainty remains, but it seems unlikely that
thermosteric sea level rise exceeded 0.4±0.3 m during the LIG. This constraint, along
with estimates of the sea level contributions from the Greenland Ice Sheet, glaciers and
ice caps, implies that 4.1 to 5.8 m of sea level rise during the Last Interglacial period was
derived from the Antarctic Ice Sheet. These results reemphasize the concern that both the
Antarctic and Greenland Ice Sheets may be more sensitive to temperature than widely
thought.
Introduction
Sea-level rise is one of the major socio-economic hazards associated with global
201
warming, and a better understanding the mechanisms that underly sea-level rise is a
prerequisite to accurate projections of global and regional sea-level rise. Despite this,
variability in the different components of sea level rise (i.e., ocean thermal expansion,
melting of glaciers, and wasting of the Greenland and Antarctic Ice Sheets) is poorly
understood, especially with respect to the future. The Fourth Assessment Report of the
Intergovernmental Panel on Climate Change (IPCC), which explicitly excluded rapid ice
flow dynamics, projected that ocean thermal expansion would make up 55 to 70% of the
sea level rise over the 21st century [Meehl et al., 2007], whereas the empirical model of
Vermeer and Rahmstorf [2009] projects a much smaller proportion, between 20 and 30%,
although this result is primarily driven by a larger contribution of ice melt. On longer
timescales, the equilibrium response of ocean thermal expansion to warming has been
estimated as 0.2 to 0.6 m °C-1 [Meehl et al. 2007], but the relative contributions of ice
sheet melt and thermal expansion during a millennial-scale highstand remains unclear.
One approach to address this uncertainty is to study past sea-level changes. The Last
Interglacial period (LIG) is the most recent warm interval with substantially higher-thanmodern global sea level. During the LIG, from ca. 130 to 120 ka, sea level reached at
least 6 m above present levels [Hearty et al., 2007; Kopp et al., 2009]. The majority of
the sea level rise originated from melting of the Greenland Ice Sheet (GIS) and the
Antarctic Ice Sheets [Otto-Bliesner et al., 2006; Overpeck et al., 2006; Kopp et al., 2009;
Clark and Huybers, 2009], but the role of thermal expansion has not been carefully
examined. Here, we compile the available paleoceanographic records and examine global
climate model simulations to better constrain the amount of thermal expansion during the
202
LIG.
Methods
Paleoclimate data
We compiled a dataset of 76 published sea surface temperature (SST) records that
met several criteria. Only quantitative SST records that included both the LIG and the
Holocene were included so that ΔSST values (warmest LIG – late Holocene) could be
calculated internally for each record. We restricted our analyses to records that had an
average temporal resolution of 3 kyr or better during both LIG and the Holocene.
Records were obtained through the NOAA Paleoclimatology World Data Center
(www.ngdc.noaa.gov/paleo/paleo.html), the Pangaea database (www.pangaea.de), and
from individual site reports and papers (Supplemental Table 1). The sea surface
temperatures (SSTs) were determined using Mg/Ca ratios in foraminifera, alkenone
unsaturation ratios (i.e., Uk37), and faunal assemblage transfer functions (for radiolaria,
foraminifera, diatoms and coccoliths), and were interpreted to reconstruct annual, austral
summer, and boreal summer sea surface temperatures. Only records with published age
models were included; however, there is substantial uncertainty between age estimates at
different sites. For this study we chose to determine a maximum estimate of ocean
warming during a sustained sea level highstand, so the average SST of a 5 kyr period
centered on the warmest temperature between 135 and 118 ka was calculated for each
record. ΔSST values were determined by subtracting the average SST of the late
Holocene (5 to 0 ka) from the 5 kyr LIG average. The data set was supplemented by 94
203
LIG SST estimates from the CLIMAP project [CLIMAP Project Members, 1984]. For the
CLIMAP data, ΔSST values were determined as the difference between LIG temperatures
and core top temperature estimates at each site. Global mean SST anomalies (ΔSST)
were calculated by averaging anomalies in 10°x10° boxes, then determining zonal
averages, which were finally averaged after weighting each zonal average by the area of
ocean for each latitudinal band.
To complement our data synthesis, we performed the same analyses on the LIG
SST dataset assembled by Turney and Jones [2010]. The Turney and Jones [2010] dataset
differs from our synthesis in several regards: 1) only data that were interpreted to
reconstruct annual mean temperatures were included, 2) the timing of LIG mean SST
estimates was determined by corresponding marine δ18O records and 3) ΔSST values
were calculated as the difference between LIG SSTs and instrumental SST climatology.
Many of the same records went into both LIG SST syntheses, but analyzing both datasets
allows us to evaluate the sensitivity of our results to markedly different data treatment
approaches.
Global Climate Model Simulations
Climate simulations were conducted using a global, coupled ocean-atmosphereland-sea ice general circulation model (Community Climate System Model [CCSM],
Version 3) [Collins et al., 2006]. The atmospheric model has ~1.4° latitude-longitude
resolution (T85) with 26 levels, and the ocean model has ~1° resolution and 40 levels.
The preindustrial 1870 AD control simulation includes the appropriate forcing conditions,
204
including trace gas concentrations (CO2: 289 ppmv; CH4 901 ppbv), solar constant (1365
W/m2), and orbital characteristics (obliquity: 23.44°, perihelion: 3 January, and
eccentricity: 0.0167). The preindustrial control simulation was run for 550 model years,
and climatologies were calculated using model years 530 to 549. The LIG simulation
included forcing conditions appropriate for 125 ka; obliquity was 23.80°, perihelion was
23 July, and eccentricity was 0.0400 [Berger et al., 1991]. The trace gas concentrations
were estimated from ice core data (CO2: 273 ppmv; CH4: 642 ppbv) [Petit et al., 1999].
The solar constant was set to the model present-day value of 1367 W/m2. Vegetation and
land ice coverage were prescribed at their present-day distributions for both the
preindustrial and LIG simulations. The LIG simulation was run for 200 model years, and
climatologies were calculated using model years 180 to 199. CCSM3 is known to have
regional SST biases, but is very well-suited for simulating global mean SST [Collins et
al., 2006], which is the focus of this study. Global SST anomalies were calculated by
zonal averaging and then calculating an area-weighted global mean.
Paleoceanographic data synthesis
LIG-Holocene SST anomalies varied regionally, and importantly, were not
uniformly warmer during the LIG (Figure C1). Furthermore, our data synthesis shows the
same primary patterns as the synthesis of Turney and Jones [2010], suggesting that the
primary results results are robust to the choices of averaging and Holocene reference
period. Records from the high latitudes of the Northern Hemisphere (>30°N) were
consistently warmer during the LIG. This is consistent with the dramatic increase in
205
summer insolation (~12% above preindustrial), and extensive evidence for much warmer
(4-5 °C) conditions in the Arctic during the interval [CAPE Project Members, 2006].
South of ~30°N, the anomalies are regionally variable (Figure C1). The Caribbean Sea
and the tropical Atlantic Oceans appear to have been generally cooler during the LIG than
the late Holocene (and late 20th century). The eastern equatorial Pacific Ocean shows both
positive and negative anomalies, as does the rest of the Pacific ocean. The western Indian
Ocean appears to have been slightly warmer, and the central Indian Ocean somewhat
cooler, but the data coverage in both the Pacific and Indian Ocean is poor. The
southeastern Atlantic Ocean, off the west coast of South Africa, was consistently warmer.
The Southern Ocean changes are mixed, apparently cooler west of South America,
somewhat warmer in the Atlantic sector and near New Zealand, and mixed in the Indian
sector.
The regional variability is interesting, and warrants further investigation, however
interpreting the patterns in terms of modes of climatic and ocean variability is
confounded by chronological errors, resolution differences and poor data coverage.
Consequently, we chose to focus on the primary, global pattern: the warmer temperatures
between 30°N and 70°N, and equivocal anomalies further south (Figure C1). The oceanarea-weighted global average SST anomaly is 0.7±0.6°C for our data synthesis, and
0.7°C for that of Turney and Jones [2010]. Interestingly, these global ΔSST estimates are
lower than the global land and ocean temperature anomaly (1.5±0.1°C) calculated by
Turney and Jones [2010]. This discrepancy may be due to the predominance of terrestrial
records from the Northern Hemisphere that are particularly sensitive to summer
206
temperatures in their global synthesis.
The error calculated for global ΔSST incorporates errors in the SST proxies,
which typically range from 1 to 2°C, and the error associated with estimating global
ΔSST from limited spatial coverage (Auxiliary Material). Nevertheless, this estimate
does not capture all of uncertainty in global ΔSST. Because we calculated a maximum
estimate for ΔSST, we excluded chronological errors, although differences in temporal
resolution between sites contributes additional uncertainty. Furthermore, each of the SST
proxies comes with its own set of errors and biases. A particular concern is that all three
of the primary SST proxies in our database (faunal assemblages, Mg/Ca, Uk37) are known
to be sensitive to changes in seasonality [Anand et al., 2003; Morey et al., 2005; Lorenz
et al., 2006], and each proxy may exhibit different responses to changes in seasonality,
even at the same location [e.g., Weldeab et al., 2007; Saher et al., 2009]. Given the
extreme differences in seasonal insolation forcing during the LIG relative to the late
Holocene, changes in the timing and distribution of the productive seasons likely biased
the SST estimates.
To evaluate some of the potential biases in our analysis, we subsampled our
database by proxy type and seasonality (Auxiliary Material). Globally, ΔSST for the Uk37
and Mg/Ca proxies was about 1.5°C higher than the faunal assemblage proxies. Some of
this offset is likely due to lower sample density and different spatial coverage of the Uk37
and Mg/Ca proxies, which are commonly located near coasts in upwelling regions.
Regionally, ΔSST appears generally consistent between proxies, with some exceptions
(Figure C1a). Subdividing seasonally, ΔSST in boreal summer (JJA) records was slightly
207
higher (0.2°C) than in austral summer (DJF) records, consistent with the change in
insolation forcing.
Global Climate Model Simulations
Like the paleoceanographic data, the model simulations for 125 ka show
substantial warming north of 40°N, and similar or slightly cooler SSTs south of 30°N
(Figure C2a-b), similar to previously published simulations for this time period [Montoya
et al., 2000; Kaspar and Cubasch, 2007]. The ocean-area-weighted global average ocean
temperature difference between the 125 ka simulation and the preindustrial control is
-0.4°C for both the surface temperatures and the top 200 m. The result of a cooler average
ocean surface in the 125 ka simulation is surprising given that the annual insolation
anomalies are positive globally (Figure C2c). This result merits a discussion of the
climate dynamics simulated in the model that contribute to the cooling in the Southern
Hemisphere.
The most significant difference between the forcings for the LIG simulation and
the preindustrial control are the different orbital parameters, and among those, the date of
perihelion (or phase in the precession cycle) is most different. In the LIG simulation,
perihelion occurs during the boreal summer, as opposed to the preindustrial control, when
aphelion occurs during the boreal summer. The result is that relative to the preindustrial
simulation, the Northern Hemisphere should experience much greater seasonality
(warmer summers and colder winters), while the Southern Hemisphere should have
colder summers and warmer winters (Figure C2c).
208
The Southern Hemisphere cooling in the model is associated with a decrease in
longwave radiative forcing (Figure C2d), which is a function of decreased water vapor
concentrations in the southern hemisphere (Figure C2e). Annually averaged, water vapor
content was consistently lower in the LIG simulation than the preindustrial control
throughout most of the Southern Hemisphere and over the Pacific Ocean, and
substantially higher over the Northern Hemisphere monsoon regions, and the high
Northern latitudes (Figure C2e). There appear to be two global scale mechanisms
responsible for the hemispheric shift in water vapor.
First, the large increase in summer insolation in the Northern Hemisphere results
in a strengthening of the Asian, African and North American Monsoons in the model,
along with a northward shift of the Intertropical Convergence Zone (ITCZ) (Figure C2e).
The strengthened and northward-shifted monsoon systems pull more moisture further
across the equator into the Northern Hemisphere, focusing precipitation in the monsoons
while effectively drying the southern tropics. The effect of this northward shift on the
Earth's energy budget is apparent in the changes in outgoing longwave radiation (OLR;
Figure C2f), which is substantially reduced over the Northern Hemisphere monsoon
regions, and increased over the Southern Hemisphere monsoon systems (e.g., South
America, equatorial and southern Africa, Australia), effectively cooling the tropical
Southern Hemisphere.
The second mechanism is associated with the opposing changes in seasonality in
each hemisphere. Due to the nonlinearity in the capacity of air to hold water vapor as a
function of temperature (the Clausius-Clapeyron relation), the large decrease in insolation
209
during the Southern Hemisphere summer and fall is not compensated, in terms of specific
humidity, by an equivalent increase in winter and spring insolation. This effect should be
most important at higher latitudes, and the increase in Northern Hemisphere specific
humidity is consistent with this hypothesis (Figure C2e). The impact is not immediately
apparent in OLR (Figure C2f). At the high southern latitudes, both downwelling radiation
at the surface (Figure C2d) and OLR (Figure C2f) are decreased. This is due to decreased
absorption and attenuation of longwave radiation in the atmosphere, and is a function of
both decreased specific humidity and cooler surface temperatures decreasing the amount
of outgoing longwave radiation produced at the surface. The opposite scenario is apparent
at the high northern latitudes.
These two mechanisms provide a plausible explanation for the cooling over most
of the world's ocean. The results do not appear to be specific to the CCSM3 model.
Simulations with other climate models show cooler temperatures in the Southern
Hemisphere, and near-zero or negative annual SST anomalies relative to preindustrial
controls [e.g., Montoya et al., 2000; Kaspar and Cubasch, 2007]. Furthermore, an
additional simulation using CCSM3 for the period 130 ka yields a similar cooling in the
Southern Hemisphere, despite regional differences in SST, suggesting that our result is
not specific to only this interval of the LIG (Auxiliary Material). This result calls into
question the belief that the LIG was substantially warmer globally [e.g., LIGA Members,
1991; Clark and Huybers, 2009; Turney and Jones, 2010; Masson-Delmotte et al., 2010].
Much uncertainty remains in model simulations; but it is possible that the predominance
of terrestrial, Northern Hemisphere, summer-sensitive temperature proxies may have
210
biased our understanding of global temperature anomalies during the interval.
The thermosteric component of LIG sea level rise
The amount of steric sea level rise can be determined by calculating the specific
volume of the ocean, which requires integrating the temperature and salinity structure of
the ocean. This is possible for the model simulations, but not for the paleoceanographic
data, so other approaches must be utilized. A simple empirical approach is to estimate a
thermal expansion sensitivity (i.e., cm/°C). This can be achieved with instrumental data;
the IPCC [Bindoff et al., 2007] concluded that the top 700 m of the ocean warmed 0.1°C
from 1961-2003, and that thermal expansion of the ocean was about 1.3 cm over the
same interval, resulting in a sensitivity of ~13 cm/°C. To determine a maximum estimate,
we assumed our average ΔSST of 0.7±0.6°C is representative of the top 700 m, resulting
in 9±8 cm of thermosteric sea level rise. Alternatively, we estimate the thermal
expansion using the Thermodynamic Equation of Seawater 2010 (TEOS-10) to calculate
the change in the specific volume of the top 700 m of the ocean due to a 0.7±0.6°C
warming, while holding the salinity constant, and neglecting changes in ocean area. This
approach results in ~12±10 cm of thermosteric sea level rise. It is possible that sustained,
warmer-than-modern conditions resulted in warming below 700 m in the oceans. If the
average warming extended to 2000 m, the thermal expansion of the ocean would have
been about 35±30 cm, consistent with the equilibrium ocean-temperature thermal
expansion sensitivity observed in long climate model simulations (0.2 to 0.6 m °C-1)
[Meehl et al. 2007] .
211
For the model simulation, the whole-ocean global average steric sea level change
was -18 cm, primarily due to cooler ocean temperatures in the Southern Hemisphere.
Because the model simulations are relatively short, the deep ocean was not equilibrated.
This introduces additional uncertainty in our estimate of steric sea level change; however
the volume-integrated ocean temperature trends are the same in both the LIG and
preindustrial simulations (-0.12°C/century), suggesting that the steric sea level change
would be comparable after equilibration.
Altogether, it is clear that ocean thermal expansion during the LIG was a small
component of the maximum LIG sea level highstand. A conservative estimate from the
available paleoclimatic data is 0.4±0.3 m. The climate model simulations suggest that the
thermosteric component may have been smaller or even negative 125 ka, near the time of
the maximum highstand. This has several important implications. First the high-end
estimate of sea level exceedance (33% probability that sea level exceeded 9.4 m during
the LIG) by Kopp et al. [2009] is probably too high, because the stochastic thermosteric
component in their model was unrealistically large (mean = 0 m, 1σ = 2 m). Using a
more realistic thermosteric component should reduce the variance of the distribution of
sea level histories, resulting in tighter error estimates and exceedance levels that are
nearer to the median.
Secondly, our results provide further constraints on the relative contributions to
sea level rise during the Last Interglacial. The contribution from the GIS was likely 2.2 –
3.4 m [Otto-Bliesner et al., 2006], or even less [Oerlemans et al., 2006]. The maximum
possible contribution from mountain glaciers and ice caps is 0.6±0.1 m [Radić and Hock,
212
2010], and our conservative estimate of maximum thermal expansion during the LIG
(0.4±0.3 m). These data, combined with the median projection (50% exceedance) of
maximum LIG sea level rise (8.5 m) [Kopp et al., 2009] imply that the Antarctic Ice
Sheet (AIS), most likely the West Antarctic Ice Sheet (WAIS), contributed at least
4.1±0.3 m. Assuming a low-end contribution from the GIS (2.2 m), only glaciers and ice
caps from the northern Hemisphere (0.4±0.1 m) and our low-end estimate for thermal
expansion (0.1±0.1 m), the maximum contribution from Antarctica is 5.8±0.1 m.
It remains unclear why so much more ice (4.1 to 5.8 m sea level equivalent) was
lost from Antarctica during the LIG than the Holocene. Antarctic ice cores all suggest
warmer-than-modern annual temperatures for East Antarctica [cf., Petit et al., 1999;
EPICA, 2004; Kawamura et al., 2007], and recent evidence suggests that the warming
anomaly may have been larger (~6°C warmer than the Holocene) than previously
estimated [Sime et al., 2009]. This stands in contrast to the cooling simulated by our LIG
simulation (Figure C2a-b), and is a consistent frustration of model-paleodata comparisons
[Masson-Delmotte et al., 2010]. Furthermore, melt season solar insolation was
substantially lower than present-day (Figure C2c). A recent study by Huybers and
Denton [2008] suggested that Antarctic temperatures are primarily controlled by the
duration of summer, which was very long during this interval, rather than the intensity of
solar insolation (like the Northern Hemisphere), although this mechanism does not drive
Antarctic temperatures in our LIG simulation. It has also been hypothesized that poorly
simulated climatic feedbacks and changes in ocean circulation may be responsible for the
mismatch [Overpeck et al., 2006; Masson-Delmotte et al., 2010], a hypothesis that
213
implies substantial vulnerability of the AIS in the future [Yin et al., 2011]. Finally, it is
possible that much of the Antarctic contribution was derived during late deglaciation or
early LIG, when melt-season insolation was much higher [Overpeck et al., 2006]. This
possibility is consistent with the observation that substantial downwasting of the WAIS is
necessary to simulate the high temperatures inferred from East Antarctic ice cores during
the LIG in climate models [Holden et al., 2010].
Conclusion
The available paleoceanographic records and our LIG GCM simulation suggest
that global SSTs were not dramatically warmer than preindustrial conditions (paleodata =
0.7±0.6°C warmer, model = 0.4°C cooler). Taken together, the model and paleodata
imply a minimal (-0.2 to 0.4 m) contribution of thermal expansion to LIG sea level rise.
This constraint, along with estimates of the sea level contributions from the Greenland
Ice Sheet, glaciers and ice caps, implies that 4.1 to 5.8 m of sea level rise during the LIG
was derived from the Antarctic Ice Sheet. These results reemphasize the concern that the
Greenland and especially the Antarctic Ice Sheets may be more sensitive to temperature
than widely thought.
Acknowledgements
We thank Suz Tolwinski-Ward, Joellen Russell, John Fasullo, Toby Ault, and Sarah
Truebe for insightful discussions on the work presented in this paper. Nan Rosenbloom
provided the numbers for the model forcings. NCAR is funded by NSF. We also thank
214
Mark Siddall and Michael Oppenheimer for helpful comments in review of this paper.
References
Anand, P., H. Elderfield, and M. H. Conte (2003), Calibration of Mg/Ca thermometry in
planktonic foraminifera from a sediment trap time series, Paleoceanography, 18,
doi:200310.1029/2002PA000846.
Berger, A., and M. F. Loutre (1991), Insolation values for the climate of the last 10
million years, Quaternary Science Reviews, 10, 297–317.
Bindoff, N. et al. (2007), Observations: oceanic climate change and sea level. In Climate
Change 2007: The Physical Science Basis. Contribution of Working Group I to
the Fourth Assessment Report of the Intergovernmental Panel on Climate
Change, edited by S. Solomon and others, Cambridge University Press,
Cambridge.
CAPE Project Members (2006), Last interglacial Arctic warmth confirms polar
amplification of climate change, Quaternary Science Reviews, 25, 1383–1400.
Clark, P. U., and P. Huybers (2009), Interglacial and future sea level, Nature, 462, 856–
857.
CLIMAP Project Members (1984), The Last Interglacial ocean, Quat. Res., 21, 123–224.
Collins, W. D. et al. (2006), The Community Climate System Model Version 3 (CCSM3),
J. Climate, 19, 2122-2143, doi:10.1175/JCLI3761.1.
EPICA (2004), Eight glacial cycles from an Antarctic ice core, Nature, 429, 623–628.
Hearty, P. J., J. T. Hollin, A. Neumann, M. O’Leary, and M. McCulloch (2007), Global
sea-level fluctuations during the Last Interglaciation (MIS 5e), Quat. Sci. Rev., 26,
2090-2112.
Holden, P. B. et al. (2010), Interhemispheric coupling, the West Antarctic Ice Sheet and
warm Antarctic interglacials, Clim. Past, 6, 431-443.
Huybers, P., and G. Denton (2008), Antarctic temperature at orbital timescales controlled
by local summer duration, Nature Geoscience, 1, 787–792.
Kaspar, F., and U. Cubasch (2007), Simulations of the Eemian interglacial and the
subsequent glacial inception with a coupled ocean-atmosphere general circulation
215
model, in The Climate of Past Interglacials, edited by F. Sirocko, M. Claussen,
M.F. Sánchez Goñi, and T. Litt, pp. 499-515, Elsevier.
Kawamura, K. et al. (2007), Northern Hemisphere forcing of climatic cycles in Antarctica
over the past 360,000 years, Nature, 448, 912–916.
Kopp, R., F. J. Simons, J. X. Mitrovica, A. Maloof, and M. Oppenheimer (2009),
Probabilistic assessment of sea level during the Last Interglacial stage, Nature,
462, 863–867.
LIGA members (1991), Report of 1st discussion group: The Last Interglacial in high
latitudes of the Northern Hemisphere: Terrestrial and marine evidence, Quat. Int.,
10-12, 9-28.
Lorenz, S. J., J. Kim, N. Rimbu, R. R. Schneider, and G. Lohmann (2006), Orbitally
driven insolation forcing on Holocene climate trends: Evidence from alkenone
data and climate modeling, Paleoceanography, 21,
doi:200610.1029/2005PA001152.
Masson-Delmotte, V. et al. (2010), EPICA Dome C record of glacial and interglacial
intensities, Quaternary Science Reviews, 29, 113–128.
Meehl, G. A. et al. (2007), Chapter 10: Global Climate Projections. In Climate Change
2007: The Physical Science Basis. Contribution of Working Group I to the Fourth
Assessment Report of the Intergovernmental Panel on Climate Change, edited by
S. Solomon and others, Cambridge University Press, Cambridge.
Montoya, M., H. von Storch, and T. J. Crowley (2000), Climate simulation for 125 kyr
BP with a coupled ocean-atmosphere general circulation model, Journal of
Climate, 13, 1057–1072.
Morey, A. E., A. C. Mix, and N. G. Pisias (2005), Planktonic foraminiferal assemblages
preserved in surface sediments correspond to multiple environment variables,
Quaternary Science Reviews, 24, 925-950, doi:10.1016/j.quascirev.2003.09.011.
Oerlemans, J., D. Dahl-Jensen, and V. Masson-Delmotte (2006), Ice Sheets and Sea
Level, Science, 313, 1043 -1045, doi:10.1126/science.313.5790.1043c.
Otto-Bliesner, B. L., S. Marshall, J. T. Overpeck, G. Miller, and A. Hu (2006), Simulating
Arctic climate warmth and icefield retreat in the last interglaciation, Science, 311,
1751-1753.
Overpeck, J. T., B. L. Otto-Bliesner, G. H. Miller, D. R. Muhs, R. B. Alley, and J. T.
Kiehl (2006), Paleoclimatic evidence for future ice-sheet instability and rapid sea-
216
level rise, Science, 311, 1747-1750.
Petit, J. R. et al. (1999), Climate and atmospheric history of the past 420,000 years from
the Vostok ice core, Antarctica, Nature, 399, 429–436.
Radić, V., and R. Hock (2010), Regional and global volumes of glaciers derived from
statistical upscaling of glacier inventory data, Journal of Geophysical Research,
115, doi:201010.1029/2009JF001373.
Saher, M. H. et al. (2009), Western Arabian Sea SST during the penultimate interglacial:
A comparison of U37K′ and Mg/Ca paleothermometry, Paleoceanography, 24,
doi:10.1029/2007PA001557.
Sime, L. C., E. W. Wolff, K. I. C. Oliver, and J. C. Tindall (2009), Evidence for warmer
interglacials in East Antarctic ice cores, Nature, 462, 342–345.
Turney, C. S., and R. T. Jones (2010), Does the Agulhas Current amplify global
temperatures during super-interglacials?, Journal of Quaternary Science, 25, 839–
843.
Vermeer, M., and S. Rahmstorf (2009), Global sea level linked to global temperature,
Proceedings of the National Academy of Sciences, 106, 21527-21532.
Weldeab, S., R. R. Schneider, and P. Müller (2007), Comparison of Mg/Ca-and alkenonebased sea surface temperature estimates in the fresh water–influenced Gulf of
Guinea, eastern equatorial Atlantic, Geochemistry Geophysics Geosystems, 8,
Q05P22, 10.1029/2006GC001360.
Yin, J., J. T. Overpeck, S. T. Griffies, A. Hu, J. L. Russell, and R.J. Stouffer (2011),
Different magnitudes of projected subsurface ocean warming around Greenland
and Antarctica, Nature Geoscience, in press.
217
Auxiliary Material
Error Analysis
To quantify the error in our paleoceanographic global mean SST anomalies, we
propagated the SST errors for each SST measurement though each of the averaging steps
(temporal → grid cell → zonal → area-weighted global) in our ocean-area-weighted
average. We used quoted error estimates for each study, except for those that only cited
analytical errors, and not SST-calibration error. For these studies, and those which did not
provide error estimates, we applied proxy-specific error estimates (UK37 = 1.5°C, Mg/Ca =
1.1°C, Faunal = 1.5°C), following Turney and Jones [2010] and Barrows et al. [2006].
To estimate the error associated with the limited spatial range of the
oceanographic proxies, we calculated 1000 random 1-year global SST anomalies from
the ERSST dataset over the period 1910 to 2009 AD (e.g., one iteration could be 1954 –
2003, another 1981 – 1939, and so on). We then calculate the global SST anomalies using
all of the data, and a second estimate using just the grid cells where we have
paleoceanographic data. We consider the difference between these estimates to be the
spatial coverage error, and derive the statistics of the error from the distribution of the
1000 calculated distributions, and then incorporate this into our error analysis (Figure CS1). This is obviously not a perfect analog to our millennial scale analysis, but does
diminish the effect of memory and climatic modes from annual estimates, and can have a
similar magnitude as our global anomaly (0.7°C).
We also evaluated the effect of grid-size on both sources of error, and the results
are presented in Aux. Table 2. Calculated error decreases with larger grid cells,
218
presumably due to the fact the we do not incorporate the error associated with scaling up
site-specific SST anomalies to grid-cell mean anomalies. Consistent with our
conservative approach, the 10° by 10° we focus on in this study are on the high end of the
range of errors.
Bias Analysis
To evaluate some of the potential biases in our database, we subsampled our
database by proxy type and seasonality. The ocean-area-weighted global ΔSST for the
Uk37 and Mg/Ca proxies are higher (1.6°C, n=28, and 1.4°C, n=10, respectively) than the
average for faunal assemblage proxies (0.1°C, n=124). Some of the difference may be
due to the different spatial coverage of the proxies; there are far fewer Uk37 and Mg/Ca
records, and they tend to be located near coasts. Regionally, ΔSST appears generally
consistent between proxies; however, in regions that appear to have been cooler during
the LIG, the few available Mg/Ca- and Uk37-based reconstructions often suggest warmerthan-modern SSTs (Figure C1a).
Subdividing seasonally, records interpreted to record annual SSTs had an oceanarea-weighted global average difference of 1.3°C (n=42). Records interpreted to record
boreal summertime (JJA) SSTs had an average difference of 0.2°C (n=57), and 0.0°C
(n=63) for boreal winter (DJF) temperature estimates. The higher ΔSST averages in
annual records primarily reflects differences between proxies, since most Mg/Ca and Uk37
derived records are interpreted to reflect annual SST, whereas most of the JJA- and DJFsensitive records are derived from faunal assemblages. The lower ΔSST estimate for the
219
boreal winter is consistent with negative insolation anomalies during that interval.
Finally, the ΔSST estimates from the CLIMAP project [CLIMAP Project Members,
1984], were slightly lower (0.1°C, n=94) than the other faunal assemblage estimates in
the database (0.3°C, n=30). This small offset is may be due to the selection of the
warmest 5 kyr interval for the LIG reference for the non-CLIMAP data.
Finally, the impact of latitude-longitude grid size on global ΔSST was evaluated
by calculating ΔSST for latitude and longitude grid sizes ranging from 5° to 60°
(Auxiliary Table 2). Increasing longitude grid sizes had a minimal effect on ΔSST, but
increasing latitude grid sizes resulted in lower global ΔSST estimates. To sample the
dataset at higher resolution, and to determine a conservatively high estimate for ΔSST, we
chose 10° by 10° boxes for the analyses in this study.
130 ka climate model simulation
In addition to our model simulation for 125 ka, we have a second simulation with orbital
parameters and greenhouse gas concentrations consistent with 130 ka (CO2: 300 ppm,
CH4: 720 ppb). Comparing this simulation to our simulation for 125 ka allows us to
evaluate whether our conclusions and the data-model mismatch are only specific to 125
ka, or are also characteristic of earlier parts the LIG. Figure C-S2 shows the difference in
surface temperatures between the two simulations. There are regional differences
between the two simulations, but the global mean temperatures are within 0.1°C,
suggesting that our conclusions are not limited to the interval 125 ka.
220
Auxiliary References
de Abreu, L., N. J. Shackleton, J. Schönfeld, M. Hall, and M. Chapman (2003),
Millennial-scale oceanic climate variability off the Western Iberian margin during
the last two glacial periods, Marine Geology, 196(1-2), 1–20.
Bard, E., F. Rostek, and C. Sonzogni (1997), Interhemispheric synchrony of the last
deglaciation inferred from alkenone palaeothermometry, Nature, 385(6618), 707–
710.
Barrows, T. T., S. Juggins, P. De Decker , E. Calvo, and C. Pelejero (2007). Long-term
sea surface temperature and climate change in the Australian–New Zealand
region, Paleoceanography 22 doi:10.1029/2006PA001328.
Becquey, S., and R. Gersonde (2003), A 0.55-Ma paleotemperature record from the
Subantarctic zone: Implications for Antarctic Circumpolar Current development,
Paleoceanography, 18(1), 1014.
Budziak, D. (2001), Late Quaternary monsoonal climate and related variations in
paleoproductivity and alkenone-derived sea-surface temperatures in the western
Arabian Sea, Universität Bremen.
Clemens, S. C., W. L. Prell, Y. Sun, Z. Liu, and G. Chen (2008), Southern Hemisphere
forcing of Pliocene δ18O and the evolution of Indo-Asian monsoons,
Paleoceanography, 23(4).
CLIMAP Project Members (1984), The Last Interglacial ocean, Quaternary Research, 21,
123–224.
Cortese, G., and A. Abelmann (2002), Radiolarian-based paleotemperatures during the
last 160 kyr at ODP Site 1089 (Southern Ocean, Atlantic Sector),
Palaeogeography, Palaeoclimatology, Palaeoecology, 182(3-4), 259–286.
Cortese, G., A. Abelmann, and R. Gersonde (2007), The last five glacial-interglacial
transitions: A high-resolution 450,000-year record from the subantarctic Atlantic,
Paleoceanography, 22(4).
Crosta, X., A. Sturm, L. Armand, and J. J. Pichon (2004), Late Quaternary sea ice history
in the Indian sector of the Southern Ocean as recorded by diatom assemblages,
Marine Micropaleontology, 50(3-4), 209–223.
Crundwell, M., G. Scott, T. Naish, and L. Carter (2008), Glacial-interglacial ocean
climate variability from planktonic foraminifera during the Mid-Pleistocene
transition in the temperate Southwest Pacific, ODP Site 1123, Palaeogeography,
Palaeoclimatology, Palaeoecology, 260(1-2), 202–229.
Herbert, T. D., and J. D. Schuffert (2000), 16. Alkenone unsaturation estimates of seasurface temperatures at site 1002 over a full glacial cycle, in Proceedings of the
Ocean Drilling Program, Scientific Results, vol. 165, pp. 239–247.
Herbert, T. D., J. D. Schuffert, D. Andreasen, L. Heusser, M. Lyle, A. Mix, A. C. Ravelo,
221
L. D. Stott, and J. C. Herguera (2001), Collapse of the California Current during
glacial maxima linked to climate change on land, Science, 293(5527), 71.
Imbrie, J. et al. (1992), On the structure and origin of major glaciation cycles 1. Linear
responses to Milankovitch forcing, Paleoceanography, 7(6), 701–738.
Jasper, J. P., J. M. Hayes, A. C. Mix, and F. G. Prahl (1994), Photosynthetic fractionation
of 13C and concentrations of dissolved CO2 in the central equatorial Pacific
during the last 255,000 years, Paleoceanography, 9(6), 781–798.
Kandiano, E. S., and H. A. Bauch (2003), Surface ocean temperatures in the north-east
Atlantic during the last 500 000 years: evidence from foraminiferal census data,
Terra Nova, 15(4), 265-271, doi:10.1046/j.1365-3121.2003.00488.
Kiefer, T., and M. Sarnthein (1998), Distribution of planktic foraminifers and SST of
sediment core GIK15637-1, , doi:10.1594/PANGAEA.202139.
Kirst, G. J., R. R. Schneider, P. J. Müller, I. von Storch, and G. Wefer (1999), Late
Quaternary temperature variability in the Benguela Current system derived from
alkenones, Quaternary Research, 52(1), 92–103.
Lawrence, K. T., and T. D. Herbert (2005), Late Quaternary sea-surface temperatures in
the western Coral Sea: Implications for the growth of the Australian Great Barrier
Reef, Geology, 33(8), 677.
Lea, D. W., D. K. Pak, C. L. Belanger, H. J. Spero, M. A. Hall, and N. J. Shackleton
(2006), Paleoclimate history of Galapagos surface waters over the last 135,000 yr,
Quaternary Science Reviews, 25(11-12), 1152–1167.
Lea, D. W., D. K. Pak, and H. J. Spero (2000), Climate impact of late Quaternary
equatorial Pacific sea surface temperature variations, Science, 289(5485), 1719.
Martinson, D. G., N. G. Pisias, J. D. Hays, J. Imbrie, T. C. Moore, and N. J. Shackleton
(1987), Age dating and the orbital theory of the ice ages: Development of a highresolution 0 to 300,000-year chronostratigraphy, Quaternary Research, 27(1), 129, doi:10.1016/0033-5894.
Martrat, B., J. O. Grimalt, C. Lopez-Martinez, I. Cacho, F. J. Sierro, J. A. Flores, R. Zahn,
M. Canals, J. H. Curtis, and D. A. Hodell (2004), Abrupt temperature changes in
the western Mediterranean over the past 250,000 years, Science, 306(5702), 1762.
Martrat, B., J. O. Grimalt, N. J. Shackleton, L. de Abreu, M. A. Hutterli, and T. F. Stocker
(2007), Four climate cycles of recurring deep and surface water destabilizations
on the Iberian margin, Science, 317(5837), 502.
Mashiotta, T. A., D. W. Lea, and H. J. Spero (1999), Glacial-interglacial changes in
Subantarctic sea surface temperature and δ18O-water using foraminiferal Mg,
Earth and Planetary Science Letters, 170(4), 417–432.
Müller, P. J., R. Schneider, and G. Ruhland (1994), Late Quaternary pCO2 variations in
the Angola Current: Evidence from organic carbon d13C and alkenone
222
temperatures, Carbon Cycling in the Glacial Ocean: Constraints on the Ocean’s
Role in Global Change, 343–366.
Nürnberg, D., A. Müller, and R. R. Schneider (2000), Paleo-sea surface temperature
calculations in the equatorial east Atlantic from Mg/Ca ratios in planktic
foraminifera: A comparison to sea surface temperature estimates from U37K′,
oxygen isotopes, and foraminiferal transfer function, Paleoceanography, 15(1),
124–134.
Nürnberg, D., M. Ziegler, C. Karas, R. Tiedemann, and M. W. Schmidt (2008),
Interacting Loop Current variability and Mississippi River discharge over the past
400 kyr, Earth and Planetary Science Letters, 272(1-2), 278–289.
Pahnke, K., and J. P. Sachs (2006), Sea surface temperatures of southern midlatitudes 0–
160 kyr BP, Paleoceanography, 21(2).
Pahnke, K., R. Zahn, H. Elderfield, and M. Schulz (2003), 340,000-year centennial-scale
marine record of Southern Hemisphere climatic oscillation, Science, 301(5635),
948.
Pailler, D., and E. Bard (2002), High frequency palaeoceanographic changes during the
past 140 000 yr recorded by the organic matter in sediments of the Iberian
Margin, Palaeogeography, Palaeoclimatology, Palaeoecology, 181(4), 431–452.
Pichon, J. J., L. D. Labeyrie, G. Bareille, M. Labracherie, J. Duprat, and J. Jouzel (1992),
Surface water temperature changes in the high latitudes of the Southern
Hemisphere over the last glacial-interglacial cycle, Paleoceanography, 7(3), 289–
318.
Pisias, N. G., and A. C. Mix (1997), Spatial and temporal oceanographic variability of the
eastern equatorial Pacific during the Late Pleistocene: Evidence from
radiolaria microfossils, Paleoceanography, 12(3), PP. 381-393,
doi:199710.1029/97PA00583.
Rickaby, R. E. M., and H. Elderfield (1999), Planktonic foraminiferal Cd/Ca:
Paleonutrients or paleotemperature?, Paleoceanography, 14(3), 293–303.
Ruddiman, W. F., M. E. Raymo, D. G. Martinson, B. M. Clement, and J. Backman
(1989), Pleistocene evolution: Northern hemisphere ice sheets and North Atlantic
Ocean, Paleoceanography, 4(4), 353–412.
Schaefer, G., J. S. Rodger, B. W. Hayward, J. P. Kennett, A. T. Sabaa, and G. H. Scott
(2005), Planktic foraminiferal and sea surface temperature record during the last 1
Myr across the Subtropical Front, Southwest Pacific, Marine Micropaleontology,
54(3-4), 191–212.
Schmidt, M. W., H. J. Spero, and D. W. Lea (2004), Links between salinity variation in
the Caribbean and North Atlantic thermohaline circulation, Nature, 428(6979),
160–163.
Schneider, R. R., P. J. Müller, and G. Ruhland (1995), Late Quaternary surface circulation
223
in the east equatorial South Atlantic: Evidence from alkenone sea surface
temperatures, Paleoceanography, 10(2), 197–219.
Sowers, T., M. Bender, L. Labeyrie, D. Martinson, J. Jouzel, D. Raynaud, J. J. Pichon,
and Y. S. Korotkevich (1993), A 135,000-year Vostok-Specmap Common
temporal framework, Paleoceanography, 8(6), 737, doi:10.1029/93PA02328.
Stein, R., J. Hefter, J. Grützner, A. Voelker, and B. D. Naafs (2009), Variability of surface
water characteristics and Heinrich-like events in the Pleistocene midlatitude North
Atlantic Ocean: Biomarker and XRD records from IODP Site U1313 (MIS 16–9),
Paleoceanography, 24(2).
Weldeab, S., D. W. Lea, R. R. Schneider, and N. Andersen (2007), 155,000 years of West
African monsoon and ocean thermal evolution, Science, 316(5829), 1303.
Wolff, T., B. Grieger, W. Hale, A. Dürkoop, S. Mulitza, J. Pätzold, and G. Wefer (1999),
On the reconstruction of paleosalinities, Use of Proxies in Paleoceanography:
Examples From the South Atlantic, 315–344.
Yamamoto, M., R. Suemune, and T. Oba (2005), Equatorward shift of the subarctic
boundary in the northwestern Pacific during the last deglaciation, Geophysical
Research Letters, 32(5), L05609.
Yamamoto, M., M. Yamamuro, and Y. Tanaka (2007), The California current system
during the last 136,000 years: response of the North Pacific High to precessional
forcing, Quaternary Science Reviews, 26(3-4), 405–414.
224
Faunal
Uk37
Mg/Ca
Cd/Ca
Ann
DJF
JJA
Latitude
A
-5
-5
0
5
0
ΔSST (°C)
5
Δ (Warmest 5 kyr of LIG - Late Holocene) SST (°C)
All proxies
Latitude
B
-5
-9
0
Δ (LIG - modern) SST (°C)
9
0
5
ΔSST (°C)
Figure C1. Maps of global ΔSST values in A) our database, where symbols indicate proxy
type (see legend) and B) the synthesis of Turney and Jones [2010]. Note that in both maps,
the locations of the symbols were adjusted slightly for visibility. To the right each map, ΔSST
values are plotted by latitude. For our database (A), records interpreted to reflect annual,
austral summer, and boreal summer temperatures are shown with different symbols.
225
Δ (LIG - Preindustrial)
A
B
-3
-5 -4 -3 -2 -1 -.5 -.25.25 .5 1 2 3 4 5
Surface temperature (°C)
0
3
-5 -4 -3 -2 -1 -.5 -.25.25 .5 1 2 3 4 5
200 m potential ocean temperature (°C)
-2
0
2
-10
0
10
C
D
60N
30N
0
30S
60S
J
F
M
A
M
J
J
A
S
O
N
D
J -5
0
5
-20 -15 -10 -6 -4 -2 -1 1 2 4 6 10 15 20
-30-20 -15 -10 -5 -3 -1 0 1 3 5 10 15 20 30
-2
Incoming solar radiation (W m )
-2
Downwelling longwave radiation at the surface (W m )
E
F
-30 -20 -15 -10 -6 -4 -2 2 4 6 10 15 20 30
-1
-2
Specific humidity (g kg x 10 )
-1 0 1
-15 -10 -7 -4 -2 -1 -.25 .25 1 2 4 7 10 15
-2
Outgoing longwave radiation (W m )
-10
0
Figure C2. LIG simulation - preindustrial control anomalies in our global climate model simulation
parameters. The parameters include: A) surface air temperature, B) potential temperature
averaged over the top 200 m of the ocean, C) incoming solar radiation, by latitude and month, D)
downwelling longwave radiation at the surface, E) specific humidity, averaged over all layers of the
atmosphere and F) outgoing longwave radiation at the top of the model. Zonal average anomalies
are plotted to the right of each map.
10
226
Aux. Table 1. Site names, locations and reconstruction details for paleoceanographic synthesis.
Core Site
K-11
K-11
V28-56
V28-56
V27-86
V27-86
V28-14
V28-14
V27-20
V27-20
GIK23414-9
GIK23414-9
GIK23414-9
GIK23414-9
V23-82
V23-82
K708-1
K708-1
GIK15612-2
GIK15612-2
V29-179
V29-179
ODP1020
IODP-U1313B
V30-97
V30-97
MD95-2040
MD95-2040
ODP607
ODP607
MD012443
MD95-2042
V21-146
V21-146
V32-128
V32-128
MD01-2421
V32-126
V32-126
ODP977A
Lat
71.8
71.8
68.0
68.0
66.6
66.6
64.8
64.8
54.0
54.0
53.5
53.5
53.5
53.5
52.6
52.6
50.0
50.0
44.4
44.4
44.0
44.0
41.0
41.0
41.0
41.0
40.6
40.6
40.0
40.0
37.9
37.8
37.7
37.7
36.5
36.5
36.0
35.3
35.3
35.0
Long
1.6
1.6
-6.1
-6.1
1.1
1.1
-29.6
-29.6
-46.2
-46.2
-20.3
-20.3
-20.3
-20.3
-21.9
-21.9
-23.7
-23.7
-26.5
-26.5
-24.5
-24.5
-126.4
-33.0
-32.9
-32.9
-9.9
-9.9
-33.0
-33.0
-10.2
-10.2
163.0
163.0
117.2
117.2
141.8
117.9
117.9
-2.0
Proxy*
Faunal
Faunal
Faunal
Faunal
Faunal
Faunal
Faunal
Faunal
Faunal
Faunal
Faunal (M)
Faunal (M)
Faunal (T)
Faunal (T)
Faunal
Faunal
Faunal
Faunal
Faunal (F)
Faunal (F)
Faunal
Faunal
Uk37
Uk37
Faunal (F)
Faunal (F)
Faunal (F)
Uk37
Faunal (F)
Faunal (F)
Uk37
Uk37
Faunal
Faunal
Faunal
Faunal
Uk37
Faunal
Faunal
Uk37
Season ΔSST** Reference
JJA
1.1
CLIMAP Project Members [1984]
DJF
1.5
CLIMAP Project Members [1984]
JJA
-0.7 CLIMAP Project Members [1984]
DJF
-0.9 CLIMAP Project Members [1984]
JJA
-1.2 CLIMAP Project Members [1984]
DJF
-1.3 CLIMAP Project Members [1984]
JJA
0.4
CLIMAP Project Members [1984]
DJF
-0.1 CLIMAP Project Members [1984]
JJA
-0.2 CLIMAP Project Members [1984]
DJF
1.7
CLIMAP Project Members [1984]
DJF
0.4
Kandiano and Bauch [2003]
JJA
0.9
Kandiano and Bauch [2003]
DJF
0.8
Kandiano and Bauch [2003]
JJA
1.0
Kandiano and Bauch [2003]
JJA
1.6
CLIMAP Project Members [1984]
DJF
0.3
CLIMAP Project Members [1984]
JJA
1.6
Imbrie et al. [1992]
DJF
0.5
Imbrie et al. [1992]
JJA
1.8
Kiefer and Sarnthein [1998]
DJF
2.0
Kiefer and Sarnthein [1998]
JJA
2.6
CLIMAP Project Members [1984]
DJF
2.8
CLIMAP Project Members [1984]
Annual
2.9
Herbert et al. [2001]
Annual
0.4
Stein et al. [2009]
JJA
3.4
Ruddiman et al. [1989]
DJF
1.0
Ruddiman et al. [1989]
JJA
-2.5 de Abreu et al. [2003]
Annual
2.4
Pailler and Bard [2002]
DJF
0.2
Ruddiman et al. [1989]
JJA
2.5
Ruddiman et al. [1989]
Annual
2.8
Martrat et al. [2004]
Annual
1.5
Pailler and Bard [2002]
JJA
2.8
CLIMAP Project Members [1984]
DJF
3.8
CLIMAP Project Members [1984]
JJA
1.0
CLIMAP Project Members [1984]
DJF
-0.5 CLIMAP Project Members [1984]
Annual
3.0
Yamamoto et al. [2005]
JJA
1.8
CLIMAP Project Members [1984]
DJF
1.2
CLIMAP Project Members [1984]
Annual
3.3
Martrat et al. [2007]
227
Aux Table 1. (Continued)
Core Site
ODP1016C
RC8-145
RC8-145
ODP1014A
ODP1012
MD02-2575
LPAZ21P
TR126-29
TR126-29
TR126-23
TR126-23
ODP1146
v12-122
v12-122
v34-88
v34-88
GeoB3007-1
TY93-929/P
ODP999A
V28-127
V28-127
ODP1002C
RC12-339
RC12-339
V29-29
V29-29
GeoB1523-1
GeoB1523-1
MD85674
MD03-2707
TR163-19
V25-59
V25-59
V28-238
V28-238
W8402A-14
TR163-22
ODP806B
A180-73
A180-73
Lat
34.5
33.6
33.6
32.8
32.3
29.0
23.0
21.4
21.4
20.5
20.5
19.5
17.0
17.0
16.5
16.5
16.2
13.7
12.8
11.7
11.7
10.7
9.1
9.1
5.1
5.1
3.8
3.8
3.2
2.5
2.3
1.4
1.4
1.0
1.0
1.0
0.5
0.3
0.2
0.2
Long
-122.3
-62.4
-62.4
-118.9
-118.4
-87.1
-109.5
-94.0
-94.0
-95.6
-95.6
116.3
-74.4
-74.4
59.5
59.5
59.8
53.3
-78.7
-80.1
-80.1
-65.2
90.0
90.0
77.6
77.6
-41.6
-41.6
50.4
9.4
-91.0
-33.5
-33.5
160.5
160.5
-139.0
-92.4
159.4
-23.0
-23.0
Proxy*
Uk37
Faunal
Faunal
Uk37
Uk37
Mg/Ca
Uk37
Faunal
Faunal
Faunal
Faunal
Uk37
Faunal
Faunal
Faunal
Faunal
Uk37
Uk37
Mg/Ca
Faunal
Faunal
Uk37
Faunal
Faunal
Faunal
Faunal
Faunal (F)
Faunal (F)
Uk37
Mg/Ca
Mg/Ca
Faunal
Faunal
Faunal
Faunal
Uk37
Mg/Ca
Mg/Ca
Faunal
Faunal
Season ΔSST** Reference
Annual
1.5
Yamamoto et al. [2007]
JJA
1.0
CLIMAP Project Members [1984]
DJF
1.2
CLIMAP Project Members [1984]
Annual
1.1
Yamamoto et al. [2007]
Annual
2.7
Herbert et al. [2001]
Annual
2.6
Nürnberg et al. [2008]
Annual
1.7
Herbert et al. [2001]
JJA
-1.3 CLIMAP Project Members [1984]
DJF
-1.2 CLIMAP Project Members [1984]
JJA
-0.8 CLIMAP Project Members [1984]
DJF
-1.8 CLIMAP Project Members [1984]
Annual
0.5
Clemens et al. [2008]
JJA
-0.8 CLIMAP Project Members [1984]
DJF
-2.3 CLIMAP Project Members [1984]
JJA
0.6
CLIMAP Project Members [1984]
DJF
-0.1 CLIMAP Project Members [1984]
Annual
1.2
Budziak [2001]
Annual
1.5
Bard et al. [1997]
Annual
0.9
Schmidt et al. [2004]
JJA
0.4
CLIMAP Project Members [1984]
DJF
-1.3 CLIMAP Project Members [1984]
Annual
1.4
Herbert and Schuffert [2000]
JJA
-1.5 CLIMAP Project Members [1984]
DJF
-1.0 CLIMAP Project Members [1984]
JJA
1.1
CLIMAP Project Members [1984]
DJF
0.2
CLIMAP Project Members [1984]
JJA
0.3
Wolff et al. [1999]
DJF
0.3
Wolff et al. [1999]
Annual
0.9
Bard et al. [1997]
Annual
1.1
Weldeab et al. [2007]
Annual
2.1
Lea et al. [2000]
JJA
0.0
CLIMAP Project Members [1984]
DJF
-2.4 CLIMAP Project Members [1984]
JJA
-0.1 CLIMAP Project Members [1984]
DJF
-3.6 CLIMAP Project Members [1984]
Annual
0.3
Jasper et al. [1994]
Annual
2.7
Lea et al. [2006]
Annual
0.6
Lea et al. [2000]
JJA
-0.1 CLIMAP Project Members [1984]
DJF
-0.6 CLIMAP Project Members [1984]
228
Aux Table 1. (Continued)
Core Site
VNTR1-8
MD85668
V22-182
V22-182
V22-182
V22-182
RC13-115
GeoB1105
RC13-205
RC13-205
RC13-205
RC13-205
V19-29
V19-29
GeoB1112
GeoB10083
Y71_9_101
RC11-230
RC11-230
V22-38
V22-38
V22-38
V22-38
V22-174
V22-174
V22-174
V22-174
GeoB1016-3
GeoB1016-3
Y71-6-12
Y71-6-12
ODP820
v19-53
v19-53
V28-345
V28-345
GeoB10285
RC13-228
RC13-228
RC13-228
Lat
0.0
0.0
-0.6
-0.6
-0.6
-0.6
-1.4
-1.7
-2.3
-2.3
-2.3
-2.3
-3.6
-3.6
-5.8
-6.6
-6.6
-8.8
-8.8
-9.6
-9.6
-9.6
-9.6
-10.1
-10.1
-10.1
-10.1
-11.8
-11.8
-16.4
-16.4
-16.6
-17.0
-17.0
-17.7
-17.7
-20.1
-22.3
-22.3
-22.3
Long
-110.5
46.0
-17.3
-17.3
-17.3
-17.3
-104.5
-12.4
5.2
5.2
5.2
5.2
-83.2
-83.2
-10.7
10.3
-107.0
-110.8
-110.8
-34.3
-34.3
-34.3
-34.3
-12.8
-12.8
-12.8
-12.8
11.7
11.7
-77.9
-77.9
146.3
-113.5
-113.5
118.0
118.0
9.2
11.2
11.2
11.2
Proxy*
Faunal (R)
Uk37
Faunal (C)
Faunal (F)
Faunal (C)
Faunal (F)
Faunal (R)
Mg/Ca
Faunal (F)
Faunal (R)
Faunal (F)
Faunal (R)
Faunal (R)
Faunal (R)
Mg/Ca
Uk37
Faunal (R)
Faunal
Faunal
Faunal (C)
Faunal (F)
Faunal (C)
Faunal (F)
Faunal (C)
Faunal (F)
Faunal (C)
Faunal (F)
Uk37
Uk37
Faunal
Faunal
Uk37
Faunal
Faunal
Faunal
Faunal
Uk37
Faunal (C)
Faunal (F)
Faunal (R)
Season ΔSST** Reference
Annual
-0.8 Pisias and Mix [1997]
Annual
1.3
Bard et al. [1997]
JJA
1.8
CLIMAP Project Members [1984]
JJA
-2.4 CLIMAP Project Members [1984]
DJF
0.6
CLIMAP Project Members [1984]
DJF
-3.0 CLIMAP Project Members [1984]
Annual
0.0
Pisias and Mix [1997]
JJA
2.1
Nürnberg et al. [2000]
JJA
2.0
CLIMAP Project Members [1984]
JJA
1.0
CLIMAP Project Members [1984]
DJF
3.4
CLIMAP Project Members [1984]
DJF
0.8
CLIMAP Project Members [1984]
Annual
-0.5 Pisias and Mix [1997]
Annual
-0.5 Pisias and Mix [1997]
JJA
1.1
Nürnberg et al. [2000]
Annual
1.2
Schneider et al. [1995]
Annual
-1.6 Pisias and Mix [1997]
JJA
-0.9 CLIMAP Project Members [1984]
DJF
-1.5 CLIMAP Project Members [1984]
JJA
2.0
CLIMAP Project Members [1984]
JJA
1.1
CLIMAP Project Members [1984]
DJF
0.6
CLIMAP Project Members [1984]
DJF
0.5
CLIMAP Project Members [1984]
JJA
-1.7 CLIMAP Project Members [1984]
JJA
-0.8 CLIMAP Project Members [1984]
DJF
-0.3 CLIMAP Project Members [1984]
DJF
-0.7 CLIMAP Project Members [1984]
Annual
0.9
Müller et al. [1994]
Annual
0.9
Schneider et al. [1995]
JJA
-4.9 CLIMAP Project Members [1984]
DJF
-5.3 CLIMAP Project Members [1984]
Annual
0.7
Lawrence and Herbert [2005]
JJA
3.1
CLIMAP Project Members [1984]
DJF
1.0
CLIMAP Project Members [1984]
JJA
-0.9 CLIMAP Project Members [1984]
DJF
-0.8 CLIMAP Project Members [1984]
Annual
2.7
Schneider et al. [1995]
JJA
1.9
CLIMAP Project Members [1984]
JJA
-4.1 CLIMAP Project Members [1984]
JJA
2.8
CLIMAP Project Members [1984]
229
Aux Table 1. (Continued)
Core Site
RC13-228
RC13-228
RC13-228
GeoB1712-4
GeoB1711-4
GeoB1710-3
RC13-229
RC13-229
RC13-229
RC13-229
RC11-86
RC11-86
RC11-86
RC11-86
RC12-294
RC12-294
RC12-294
RC12-294
MD97-2121
MD97-2121
ODP1089
ODP1089
ODP1123
PS2489-2
RC8-39
RC8-39
V22- 108
V22- 108
MD84-527
RC11-120
RC11-120
RC11-120
RC11-120
MD73025
MD73025
DSDP 90-594
MD97-2120
MD88-770
MD88-770
SO136-111
Lat
-22.3
-22.3
-22.3
-23.3
-23.3
-23.4
-25.5
-25.5
-25.5
-25.5
-35.8
-35.8
-35.8
-35.8
-37.3
-37.3
-37.3
-37.3
-40.4
-40.4
-40.9
-40.9
-41.8
-42.9
-42.9
-42.9
-43.2
-43.2
-43.5
-43.5
-43.5
-43.5
-43.5
-43.8
-43.8
-45.5
-45.5
-46.0
-46.0
-50.7
Long
Proxy*
11.2
Faunal (C)
11.2
Faunal (F)
11.2
Faunal (R)
12.8
Uk37
12.4
Uk37
11.7
Uk37
11.3
Faunal (F)
11.3
Faunal (R)
11.3
Faunal (F)
11.3
Faunal (R)
18.5
Faunal (C)
18.5
Faunal (F)
18.5
Faunal (C)
18.5
Faunal (F)
-10.1 Faunal (C)
-10.1 Faunal (F)
-10.1 Faunal (C)
-10.1 Faunal (F)
178.0
Uk37
178.0
Uk37
9.9
Faunal (R)
9.9
Faunal (R)
-171.5
Faunal
9.0
Faunal (F)
42.4
Faunal (R)
42.4
Faunal (R)
-3.3
Faunal
-3.3
Faunal
51.2 Faunal (F)
79.9
Faunal (R)
79.9
Faunal (R)
79.9
Mg/Ca
79.9 Faunal (F)
51.3
Faunal
51.3
Faunal
174.9 Faunal (F)
174.9
Mg/Ca
96.5
Cd/Ca
96.5 Faunal (D)
160.2 Faunal (D)
Season ΔSST** Reference
DJF
2.1
CLIMAP Project Members [1984]
DJF
-3.3 CLIMAP Project Members [1984]
DJF
2.1
CLIMAP Project Members [1984]
Annual
1.6
Kirst et al. [1999]
Annual
1.9
Kirst et al. [1999]
Annual
1.1
Kirst et al. [1999]
JJA
-0.9 CLIMAP Project Members [1984]
JJA
3.9
CLIMAP Project Members [1984]
DJF
-0.3 CLIMAP Project Members [1984]
DJF
4.7
CLIMAP Project Members [1984]
JJA
2.7
CLIMAP Project Members [1984]
JJA
-1.8 CLIMAP Project Members [1984]
DJF
2.5
CLIMAP Project Members [1984]
DJF
-3.2 CLIMAP Project Members [1984]
JJA
2.7
CLIMAP Project Members [1984]
JJA
1.1
CLIMAP Project Members [1984]
DJF
3.4
CLIMAP Project Members [1984]
DJF
2.1
CLIMAP Project Members [1984]
Annual
2.8
Pahnke et al. [2006]
DJF
2.5
Pahnke et al. [2006]
DJF
2.0
Cortese and Abelmann [2002]
DJF
2.6
Cortese et al. [2007]
Annual
2.0
Crundwell et al. [2008]
DJF
1.3
Becquet and Gersonde [2003]
JJA
-0.8 CLIMAP Project Members [1984]
DJF
-0.7 CLIMAP Project Members [1984]
JJA
-0.3 CLIMAP Project Members [1984]
DJF
-0.7 CLIMAP Project Members [1984]
DJF
0.6
Pichon et al. [1992]
JJA
-2.7 CLIMAP Project Members [1984]
DJF
-3.7 CLIMAP Project Members [1984]
Annual
-0.8 Mashiotta et al. [1999]
DJF
0.8
Martinson et al. [1987]
JJA
1.2
CLIMAP Project Members [1984]
DJF
1.3
CLIMAP Project Members [1984]
DJF
1.5
Schaefer et al. [2005]
SON
2.1
Pahnke et al. [2003]
Annual
3.0
Rickaby and Elderfield [1999]
Annual
0.1
Sowers et al. [1993]
DJF
0.2
Crosta et al. [2004]
230
Aux Table 1. (Continued)
Core Site
Lat
Long
Proxy*
Season ΔSST** Reference
V18-68
-54.6
-77.9
Faunal
JJA
-0.1 CLIMAP Project Members [1984]
V18-68
-54.6
-77.9
Faunal
DJF
-0.3 CLIMAP Project Members [1984]
MD84-551
-55.0
73.2 Faunal (F)
DJF
0.4
Pichon et al. [1992]
* In cases where more than one type of faunal proxy was used, further details are provided;
C = Coccoliths, F = Foraminifera, R = Radiolaria, D = Diatoms.
In Kandiano and Bauch [2003], M= Modern Analog Technique, T = Transfer Function Technique
** All ΔSST values were determined by subtracting late Holocene (5-0 ka) average SST from the
average SST for the 5 kyr interval centered on the warmest value between 135 and 118 ka for each
record, except for those from CLIMAP Project Members [1984]. CLIMAP ΔSST values were
calculated as the difference between MIS 5e SST and core top SST estimates for each record.
231
Auxiliary Table 2. The effect of grid size on ocean-area weighted
global mean ΔSST. 10° x 10° spacing was used in the study
Longitude grid
spacing (°)
5
Latitude grid spacing (°)
10
15
20
30
10
0.7 ± 0.7
0.7 ± 0.6
0.6 ± 0.5
0.5 ± 0.5
0.4 ± 0.4
20
0.7 ± 0.6
0.6 ± 0.5
0.5 ± 0.4
0.4 ± 0.4
0.4 ± 0.3
30
0.7 ± 0.6
0.7 ± 0.5
0.5 ± 0.4
0.4 ± 0.4
0.5 ± 0.3
45
0.7 ± 0.6
0.7 ± 0.5
0.5 ± 0.4
0.4 ± 0.4
0.5 ± 0.3
60
0.7 ± 0.6
0.6 ± 0.5
0.5 ± 0.4
0.5 ± 0.4
0.5 ± 0.3
Global SST data Coverage
error, 1−yr means, 10 x 10 grids
140
1
120
St. Dev.= 0.08
100
# iterations
Global weighted−mean SST
anomaly calculated using data
from grids in paleo analysis only (°C)
232
0
80
60
40
−1
−1
0
1
Global weighted−mean
SST anomaly calculated
using all data (°C)
20
0
−0.4 −0.2
0
0.2
Global SST Error
10 x 10 grid data coverage
Figure C-S1. Data coverage error analysis. Top left: relationship between global SST
calculated using full data coverage vs coverage limited to the sites of the
paleooceanographic data. Red line shows 1:1 relationship. Top right: Histogram of the SST
errors (residuals from top left). Bottom: Spatial coverage of the paleoceanographic data
(green cells) using 10° by 10° grid cells, as used in the study.
Figure C-S2. Surface temperature simulated for 130 ka (top left), 125 ka (top right), the difference between the two (bottom left) and the
significances of the differences (bottom right). The area weighted global surface temperature difference between the two simulations is 0.09°C.
233
234
APPENDIX D
ON THE POTENTIAL OF RAMAN-SPECTROSCOPY-BASED
CARBONATE MASS SPECTROMETRY
Submitted for publication in the professional journal: Journal of Raman Spectroscopy
235
ON THE POTENTIAL OF RAMAN-SPECTROSCOPY-BASED
CARBONATE MASS SPECTROMETRY
Nicholas P. McKay1, David L. Dettman1, Robert T. Downs1 and Jonathan T. Overpeck1,2,3
1
Department of Geosciences, University of Arizona
Institute of the Environment, University of Arizona
3
Department of Atmospheric Science, University of Arizona
2
236
Abstract
The potential for using Raman spectroscopy to measure 18O-16O oxygen isotopic
ratios in carbonates is evaluated by measuring the Raman spectra and isotope ratios of a
suite of sixty synthesized, isotopically-enriched calcite crystals ranging in composition
from natural (0.2% 18O) to 1.2% 18O. We determined the Raman-inferred isotopic ratios
(RRaman) by fitting curves to the ν1 symmetric stretching peak at 1086 cm-1 and the smaller
satellite peak, associated with the ν1 stretching mode of singly-substituted carbonate
groups (C16O218O) at 1065 cm-1. The ratio of the two peak areas shows a 1:1
correspondence with the 18O/16O ratios measured by the water and carbonate mass
spectrometry, confirming that the relative intensities of the ν1 symmetric stretching peaks
is a direct measure of the isotopic ratio in the carbonates. The 1-sigma uncertainties of the
RRaman values of the individual crystals were 0.00079 (384‰ PDB), and 0.00043 (210‰
PDB) for the 4-crystal sample means. Although this level of uncertainty is too high to
provide significant estimates of natural variability, there are multiple prospects for
improving the accuracy and precision of the technique. Carbon isotope ratios in
carbonates cannot be measured by our approach, however our results do highlight the
potential of Raman-based mass spectrometry for C and other elements in minerals and
organic compounds.
Introduction
Raman spectroscopy can potentially be used as a non-destructive, high-resolution
mass spectrometer for a variety of substances, because differences in mass within a
237
molecule can be observed as shifts in the intensity or position of peaks in Raman spectra.
Such an approach has been utilized for carbon isotopic composition in CO2 fluid
inclusions in minerals (Arakawa et al., 2007), and pronounced shifts in the Raman spectra
of organic molecules allow 13C to be used as a tracer of carbon uptake in individual
microbes (Huang et al., 2007). Here, we investigate the potential of Raman spectroscopy
as a method to directly measure O isotopic ratios in carbonates. The Raman spectra of
carbonates are typically characterized by a well-defined ν1 symmetric stretching peak near
1100 cm-1 (1086 cm-1 in calcite) (Rutt and Nicola, 1974; White, 1974). Typically, a
second, much smaller peak is also observed around 20 wavenumbers lower (1065 cm-1 in
calcite). This satellite peak is the ν1 symmetric stretching peak associated with singlysubstituted carbonate groups (C16O218O) (Cloots et al., 1991; Gillet et al., 1996).
Furthermore, the intensity of the C16O218O peak relative to the C16O3 peak is higher in
synthetic carbonates grown in waters enriched in 18O (Gillet et al., 1996). For brevity, in
this paper, we will refer to the ν1 symmetric stretching peak of the unsubstituted carbonate
groups (C16O3) as the ν116O peak, and the ν1 symmetric stretching peak of the singlysubstituted carbonate groups (C16O218O) as the ν118O peak.
The relative amount of 18O in carbonate is a widely used indicator of geologic and
environmental variability, both in modern and past systems. The relative amount is
typically expressed as the per mil deviation of the 18O/16O ratio from a standard (typically
the Pee Dee Belemnite), or δ18OPDB. The δ18O of carbonates is a primary means of
inferring past climatic and hydrologic variability from a wide variety of materials and
environments, including cave deposits (speleothems), corals, carbonate shells from
238
marine and freshwater organisms, and inorganic and organic carbonates deposited in lake
sediments and soils. A Raman-spectroscopy-based carbonate mass spectrometer could
allow for rapid, non-destructive analysis at extremely high resolution (micron-scale), that
would have abundant applications in environmental and geologic research. This study
takes the first step towards this goal by determining the relationship between the
intensities of the two Raman peaks and calcite 18O/16O, the precision with which this ratio
can be determined from Raman spectra, and the potential application of this Raman mass
spectrometer to natural systems.
Experimental Procedure
Calcite synthesis
To create a set of a calcites with a wide range of oxygen isotopic ratios we
synthesized 15 calcite samples in waters with oxygen isotope ratios (18O/16O) ranging
from 0.002 (natural) to 0.012. To generate the waters, we initially diluted 3 ml of 97% 18O
water from Sigma-Aldrich with 172 ml of distilled water. After precipitation of the mostenriched calcite, the water was further diluted to produce the remaining samples.
To precipitate calcite from the enriched waters, 1-2 grams of laboratory-grade
calcite were dissolved in the waters while they were bubbled with CO2 for several hours.
The waters were filtered, and then capped and allowed to sit for two days to allow
complete equilibration of the oxygen isotopes with HCO3- in the water. Next, the water
was bubbled with air for two days to drive off dissolved CO2, driving the precipitation of
calcite. Finally, the waters were sonicated and filtered, and the filter was dried, while the
239
remaining water was preserved for isotopic analysis, and further dilution and calcite
synthesis. The synthesized calcites were generally euhedral crystals ranging from 25 to
50 μm in size.
Raman Spectroscopy
The Raman spectra of the synthesized calcites were collected on a Thermo
Almega microRaman system, using a solid-state laser with an excitation wavelength of
780 nm at 100% power, and a thermoelectrically cooled CCD detector. The spectra were
collected under 10x magnification, with 0.5 cm-1 resolution and a 1-μm spot size. For
each sample, four individual calcite crystals were analyzed; Raman spectra were
collected for fifteen minutes on each crystal. To increase the signal to noise ratio, only
crystals with one face oriented orthogonal to the laser were analyzed, beyond this
constraint, crystal orientation was random. Raman peak positions and intensities were
determined by fitting pseudoVoigt curves to the ν118O and ν116O peaks centered at near
1065 and 1086 cm-1 respectively, corresponding to the ν1 symmetric stretching modes of
the unsubstituted and singly-substituted carbonate groups (Cloots et al., 1991; Gillet et
al., 1996). The ν116O peak, as measured by the Thermo Almega microRaman system is
asymmetric, and required three pseudoVoigt curves centered on ~1086, ~1084 and ~1076
cm-1, to properly fit (Figure D1). This asymmetry is not observed in carbonates measured
on other Raman systems (e.g., Cloots et al., 1991; Gillet et al., 1996), and is likely due to
incomplete polarization of the laser interacting with the orientation of the crystal. The
much smaller ν118O peak, centered near 1065 cm-1, can be fit with a single pseudoVoigt
240
curve. Before fitting peaks to the spectra, the background was removed from each
spectrum using the software package CrystalSleuth (Laetsch and Downs, 2006). All four
peaks in the ν1 stretching region were then fit simultaneously using a weighted nonlinear
least squares fitting algorithm in Matlab 7.12 (The Mathworks, Inc.; Figure D1). The
Raman-inferred 18O/16O ratio (RRaman) is then calculated as
1150
∫
R Raman = 1150 1000
∫∑
1000
I ( ν 18
1 O)
∗
I ( ν 16
1
O 1,2,3 )
1
3
(1)
Where I(ν118O) are the intensities of the ν118O peak, and the intensities of the three ν116O
peaks (ν116O1, ν116O2, ν116O3) are summed I(ν116O1,2,3) before numerically integrating using
Matlab (Mathworks, Inc.). RRaman is one third of the integrated intensity ratio because only
one of the three atoms in C16O218O is 18O. To investigate the potential of using Raman
spectra to infer changes in calcite 13C/12C ratios, Raman spectra were also collected from
a 13C-enriched (97% 13C) calcite and a natural-abundance reagent calcite (1.1% 13C)
calcite from J.T. Baker Chemical.
Isotope notation
We use several notations common to isotopic studies in this paper, which we
define here. R is the relative abundance of two isotopes, in this case 18O/16O. Isotopic
fractionation, α, is the fractional difference between the R of two substances (A and B),
defined as:
241
α A−B =
RA
RB
(2)
Finally, delta notation is used to express per mil (‰) differences from a standard, such
that for oxygen:
δ 18 O=
(
R sample −R std
∗1000
R std
)
(3)
where Rstd is the R of Vienna Standard Mean Ocean Water (VSMOW) for waters and
Vienna Pee Dee Belemnite (VPDB) for carbonates.
Water Mass Spectrometry
To measure the oxygen isotope ratios in the waters used for calcite synthesis, we
diluted the enriched waters with distilled water (-8.9‰ VSMOW) in proportions resulting
in a mixture with a predicted δ18O value less than +50‰. The diluted waters were then
analyzed for δ18O using a dual inlet mass spectrometer (Delta-S, Thermo-Finnegan,
Bremen, Germany) using an automated CO2-H2O equilibration unit. Standardization is
based on internal standards referenced to VSMOW and VSLAP. Precision is better than
±0.08‰ for δ18O. The resulting δ18O values along with the measured masses of the two
waters were used to determine the oxygen isotope ratios of the waters used to synthesize
calcite. Because the mass spectrometer responds linearly between the calibration points
of -55.5 (VSLAP) and 0‰ (VSMOW), we assume that the system maintains a linear
response up to +50‰ VSMOW.
242
Carbonate Mass Spectrometry
The fractionation of oxygen isotopes during the precipitation of calcite is
relatively well known (α=1.0288 at 25°C; Friedman and O'Neill, 1977; or α=1.0285 at
25°C. Kim and O'Neill, 1997), and uncertainties are small compared to the uncertainty
associated with calculating the oxygen isotope ratio from Raman spectra. Given the low
precision of the Raman-spectra-derived ratios, calculating the oxygen isotope
composition of the precipitated calcites from the δ18O value of the waters should be
sufficiently precise. To verify this assumption, we measured the δ18O of five of the
synthesized carbonate samples. Samples HC14 and HC15 had oxygen isotope ratios
approaching natural abundances and could be measured without dilution. The δ18O and
δ13C of these carbonates were measured using an automated carbonate preparation device
(KIEL-III) coupled to a gas-ratio mass spectrometer (Finnigan MAT 252). Powdered
samples were reacted with dehydrated phosphoric acid under vacuum at 70°C. The
isotope ratio measurement is calibrated based on repeated measurements of NBS-19 and
NBS-18 and precision is ±0.10‰ for δ18O and ±0.08‰ for δ13C (1 sigma). Samples HC2,
HC5 and HC8 were too enriched in 18O to be measured directly, and were diluted with
J.T. Baker calcite (δ18Opdb=-15.65‰) such that the resulting mixture would have a mean
δ18Opdb of ~20‰. To have precise dilutions, 10 mg of total calcite was reacted overnight
in sealed glass tubes with dehydrated phosphoric acid at 25°C. The evolved CO2 gas was
cleaned cryogenically and measured on a Finnigan Delta-S gas-ratio mass spectrometer.
The oxygen isotope fractionation between calcite and acid-liberated CO2 was taken from
Swart et al. (1983).
243
Results and Discussion
The Raman-inferred, and mass-spectrometer-measured water and calcite R and
δ18O values for the 15 samples are presented Table D1. The five calcite samples measured
on the mass spectrometer generally agree with the values predicted from the waters,
within the uncertainty associated with the mass dilution analysis procedure. In samples
HC14 and HC15, the δ18O values predicted from the waters underestimate those directly
measured from the calcites by 2.4 and 1.9‰, respectively. These offsets may be
associated with the extrapolation of the water and carbonate calibrations towards more
enriched values, but may also be due to enrichment in the calcites associated with high
dissolved [Ca+] during calcite synthesis (cf., Kim and O'Neill, 1997). Overall, the δ18O
and R values inferred from the water and carbonate mass spectrometry appear to be
sufficiently consistent for comparison with the less-certain Raman-inferred values. For
consistency, in the rest of this paper, we compare the Raman-inferred isotope ratios
(RRaman) with the calcite isotope ratios inferred from the water measurements (Rcalc-inf).
Raman-inferred isotope ratios (RRaman)
The RRaman values as calculated by equation 1 are reasonable for both the
individual crystals and sample means (Table 1). The variability RRaman between crystals of
the same batch displays a range in standard deviations from 0.00012 to 0.00166.
Remarkably, the RRaman values demonstrate a clear 1:1 relationship with the Rcalc-inf values,
especially when comparing the 4-crystal mean RRaman values (Figure D2). The residuals
between the RRaman and Rcalc-inf values are normally distributed, with a standard deviation
244
of 0.00079 (corresponding to 384‰) for the individual crystals, and 0.00043 (210‰) for
the 4-crystal means values. This range of uncertainty is about an order of magnitude
outside the range of natural variability; therefore, this approach is not yet sufficiently
resolved for most geologic and environmental applications. However, it should be noted
that this is an exploratory study, and a number of existing-technology refinements could
be applied that would significantly improve the accuracy and precision of our approach.
These will be discussed below.
Potential sources of error and bias
Because the ν118O peak is small, particularly at lower, closer-to-natural R values, it
would be expected that the signal to noise ratio (SNR) in the ν118O peak would be a
primary factor driving the accuracy and precision of the RRaman values. However, this does
not appear to be the case in these samples. There are no significant relationships between
the residuals and the ν118O peak to background ratio, the ν118O:ν116O ratios, or even the
root-mean squared error (RMSE) of the fit. Additionally, the size and breadth of the third
ν116O peak centered at 1076 cm-1 (ν116O3) controls the separation between the ν116O and
ν118O peaks, and in some cases, such as for spectra from HC6, which have the highest
residuals of any samples, clearly affects the quality of the ν118O peak fit. Despite this,
various metrics of the relative size and influence of the ν116O3 peak show no consistent
relationship with the accuracy of RRaman, either for sample means or individual crystals.
The only metric of SNR that shows a clear relationship with the residuals, for the samples
with small relative ν118O peaks (R < 0.005), is the ratio between ν116O height and
245
background (the height of the nearby baseline, between 1020 and 1040 cm-1, before
adjustment by CrystalSleuth; Laetsch and Downs, 2006) (Figure D3). This result suggests
that the overall signal strength of the ν1 symmetric stretching mode is a primary control
on the precision and accuracy of the RRaman estimates.
Potential for improving the accuracy and precision of RRaman estimates
There is considerable potential for reducing the uncertainty of the RRaman estimates,
and increasing the utility of Raman-spectroscopic carbonate mass spectroscopy, even
with existing techniques and technology. First, it should be emphasized that the Raman
spectra collected here were preliminary, and that virtually no effort was made to screen
crystals based on orientation or signal strength. Still, the fact that the Raman intensity
ratios show a 1:1 relationship with mass spectrometer values, with no apparent biases, is
extremely encouraging for future applications of the technique. There are several simple
ways to potentially reduce uncertainty in the estimates. The incomplete polarization of
the microAlmega Raman system causes the ν116O peak to be split into three smaller
peaks, including a small peak near 1076 cm-1 that can influence the ν118O peak. Using a
fully polarized spectrometer may reduce the overlap between the peaks resulting in a
better fit. Second, the result that the ν116O peak height to background ratio affects the
accuracy indicates that efforts to increase the overall signal to noise ratio of the Raman
spectra, even screening crystals by signal strength, would likely reduce uncertainty in
RRaman. Increasing the spectral resolution by focusing the CCD on a smaller area of the
spectrum should increase the quality of the peak fit, and eliminate aliasing driven by the
246
precise position of the peak (cf. Arakawa et al., 2007). Another approach is resonance
Raman (RR) scattering, which occurs when the wavelength of the exciting laser is chosen
such that its energy corresponds to the electronic transition of specific atomic bonds
within a molecule (Ferraro et al., 2003). RR spectroscopy can greatly increase (by a
factor of 103 to 105) the signal strength of a Raman spectrum (Ferraro et al., 2003), and
could potentially allow for much more precise determination of area of the ν118O peak. A
simpler approach is immersing the sample in liquid nitrogen, which increases the
separation of nearby peaks, and increases the sharpness of the peaks (Figure D5). The
accuracy of RRaman would likely improve with frequent comparison of Raman-inferred
values to an internal standard during sample collection. The use of relative measurements
in isotope ratio mass spectrometry has allowed greatly improved precisions to be attained
(Criss, 1999). Relative deviations from a standard can be measured far more accurately
than the determination of the absolute 18O/16O ratio, as we did here.
Carbon isotopes
The displacement of the carbon atom in the ν1 symmetric stretching mode is not
very large, so there is no significant shift of frequency in the 1050-1100 cm-1 region
between the normal and 13C-enriched calcites. However, the substituted C should be
apparent in stretching modes where the C atom does undergo large motions, such as the
ν13 active translational mode near 281 cm-1. Here, the 13C-enriched calcite appears shifted
to lower wavenumbers (centered near 279 cm-1) (Figure D4). This shift is consistent with
theoretical expectation that the fractional change in position is a function of the square
247
root of the changes in mass (Gillet et al., 1996). In this case for the ν13 mode of Ca13CO3:
ν 13 / ν 13=
√( m
13 C 16
O
/ m 12
C
16
O
) =0 . 9918
(4)
corresponding to a theoretical ν13 peak at 279 cm-1. Because the change in mass is small,
and the peak is at low wavenumbers, small changes in the size of the Ca13CO3 ν13 peak
would be completely obscured by the Ca12CO3 peak. This does not mean that the 13C/12C
ratios could not be measured in other minerals or molecules. For example, the C-C
stretching mode in the calcium oxalate weddellite (CaC2O4·2H2O) has a satellite peak at
~870 cm-1 (Figure D6). The position (following equation 2) and intensity of the peak
suggest that the satellite peak is associated with a 13C-12C bond.
Potential for quantifying isotope ratios in other molecules
Following the discussion of the previous section on the inability to quantify
changes in 13C/12C ratios in calcite Raman spectra, it makes sense to generalize the
conditions necessary to directly quantify isotope ratios from Raman spectra. There are
three primary criteria necessary to effectively fit and quantify peaks in Raman spectra,
and to relate those peaks to isotopic ratios in the compound of interest. First, to avoid the
complication of the effect of sample orientation on relative intensities, the two peaks
must be associated with the same vibrational mode. This constraint drives the second
criterion: the shift in the wavenumber associated with the change in mass must be large
enough so that the secondary peak is distinct and identifiable from the primary peak.
248
Following equation 4, this is a function of both the relative change in mass in the
molecule, and the position of the stretching peak, where larger changes in mass and
vibrational modes at higher wavenumbers result in larger shifts. Finally, the rare isotope
of interest must be sufficiently abundant that the associated peak is above the background
intensity. Despite these constraints, it is likely that isotopic ratios in a wide variety of
organic and inorganic compounds could be quantified from Raman spectra, with variable
precision.
Conclusions
We have synthesized a suite of 18O-enriched calcites, and measured their 18O/16O
ratios with both traditional mass spectrometry and through their Raman spectra. The
Raman-inferred ratios show a clear 1:1 relationship with the ratio calculated from
traditional mass spectrometry measurements, suggesting that oxygen isotope ratios can be
measured directly in Raman spectra. The measurements of individual crystals were
precise to within a ratio of 0.00079 (384‰) and the 4-crystal means were precise to
within 0.00043 (210‰). These uncertainties are too large to be useful in natural systems,
however this is a preliminary study, and the accuracy and precision could be greatly
improved by increasing the signal to noise ratio of the Raman spectra, using a
spectrometer that maintains the polarization of the laser, and refining our curve-fitting
techniques. Carbon isotope ratios in carbonates cannot be measured by our approach,
however our results do highlight the potential of Raman-based mass spectrometry for C
and other elements in minerals and organic compounds.
249
References
Arakawa, M., J. Yamamoto, and H. Kagi (2007), Developing micro-Raman mass
spectrometry for measuring carbon isotopic composition of carbon dioxide,
Applied spectroscopy, 61(7), 701–705.
Cloots, R. (1991), Raman spectrum of carbonates MIICO3 in the 1100-1000 cm−1 region:
observation of the ν1, mode of the isotopic (C16O182O)2 ion, Spectrochimica Acta
Part A: Molecular Spectroscopy, 47(12), 1745–1750, doi:10.1016/05848539(91)80012-8.
Criss, R. E. (1999), Principles of stable isotope distribution, Oxford University Press,
USA.
Ferraro, J. R. (2002), Introductory Raman Spectroscopy, Second Edition, 2nd ed.,
Academic Press.
Friedman, I., and J. R. O’Neil (1977), Compilation of stable isotope fractionation factors
of geochemical interest, USGPO.
Gillet, P., P. McMillan, J. Schott, J. Badro, and A. Grzechnik (1996), Thermodynamic
properties and isotopic fractionation of calcite from vibrational spectroscopy of
18
O-substituted calcite, Geochimica et cosmochimica acta, 60(18), 3471–3485.
Huang, W. E., K. Stoecker, R. Griffiths, L. Newbold, H. Daims, A. S. Whiteley, and M.
Wagner (2007), Raman-FISH: combining stable-isotope Raman spectroscopy and
fluorescence in situ hybridization for the single cell analysis of identity and
function, Environmental microbiology, 9(8), 1878–1889.
Kim, S., and J. Oneil (1997), Equilibrium and nonequilibrium oxygen isotope effects in
synthetic carbonates, Geochimica et Cosmochimica Acta, 61, 3461–3475,
doi:10.1016/S0016-7037(97)00169-5.
Laetsch, T., and R. Downs (2006), Software for identification and refinement of cell
parameters from powder diffraction data of minerals using the RRUFF project
and American Mineralogist Crystal Structure Databases, 19th General Meeting of
the International Mineralogical Association, Kobe, Japan, pp. 23–28.
Rutt, H. N., and J. H. Nicola (1974), Raman spectra of carbonates of calcite structure,
Journal of Physics C: Solid State Physics, 7, 4522.
Swart, P. K., S. J. Burns, and J. J. Leder (1991), Fractionation of the stable isotopes of
oxygen and carbon in carbon dioxide during the reaction of calcite with
phosphoric acid as a function of temperature and technique, Chemical Geology:
250
Isotope Geoscience section, 86(2), 89–96.
White, W. B. (1974), The carbonate minerals, The infrared spectra of minerals, 4, 227–
284.
251
Table D1. Water and Calcite oxygen isotope data, measured by traditional mass
spectrometry and Raman spectroscopy.
Water and calcite mass spectrometry
Raman-based mass spectrometry
18
Sample Rwaters Rcalc-inf δ18OPDB Rcalc-meas* δ OPDB Rraman X1 Rraman X2 Rraman X3 Rraman X4 Rraman Mean
HC01 0.0121 0.0125 5042.4
0.0143 0.0128 0.0135 0.0127
0.0133
HC02 0.0086 0.0088 3258.2 0.0083 3014.9 0.0082 0.0092 0.0093 0.0093
0.0090
HC03 0.0054 0.0055 1672.8
0.0055 0.0053 0.0060 0.0056
0.0056
HC04 0.0048 0.0050 1406.1
0.0045 0.0053 0.0052 0.0049
0.0050
HC05 0.0038 0.0039 868.2
0.0036 721.8 0.0036 0.0022 0.0044 0.0040
0.0036
HC06 0.0033 0.0034 623.7
0.0054 0.0065 0.0036 0.0029
0.0046
HC07 0.0028 0.0029 386.9
0.0029 0.0012 0.0028 0.0029
0.0025
HC08 0.0027 0.0028 329.2
0.0027 310.5 0.0031 0.0015 0.0036 0.0036
0.0029
HC09 0.0024 0.0025 206.3
0.0030 0.0021 0.0029 0.0028
0.0027
HC10 0.0024 0.0024 169.6
0.0038 0.0023 0.0033 0.0035
0.0032
HC11 0.0022 0.0023 110.2
0.0027 0.0021 0.0021 0.0027
0.0024
HC12 0.0021 0.0022
66.4
0.0023 0.0024 0.0024 0.0022
0.0023
HC13 0.0021 0.0022
53.2
0.0033 0.0026 0.0011 0.0030
0.0025
HC14 0.0021 0.0021
29.6
0.0021
32.0 0.0015 0.0029 0.0018 0.0016
0.0020
HC15 0.0020 0.0021
4.7
0.0021
6.6
0.0021 0.0026 0.0030 0.0030
0.0027
note: Rcalc-inf is the 18O/16O ratio of the calcite inferred from water composition (Kim and O'Neill, 1997),
Rcalc-meas is the 18O/16O ratio of the calcite measured by mass spectrometry. RRaman values are shown for
each of the four individual crystals measured, and their mean ratio.
*Italicized calcite values were measured offline using mass dilution, and are less certain
252
−1
Wavenumber (cm )
1090
1080
1070
1060
Rraman=0.0143
1070
1060
Intensity (normalized)
Rraman=0.0045
1070
1060
Rraman=0.0027
1070
1090
1080
1070
−1
Wavenumber (cm )
1060
1060
Figure D1. Examples of three measured Raman spectra in the ν1 symmetric stretching mode
region, and the associated pseudoVoigt curve fits for each (top: HC1, middle: HC4, bottom:
HC11). For all panels: Raman measurements are shown as black circles, the green, black and
cyan curves show the fits to the three ν116O peaks, ν116O1, ν116O2, ν116O3, respectively; the
dark blue curve shows the ν118O fit, and the red curve shows the sum of all four pseudoVoigt
peaks. The Raman intensities for each sample are normalized to the maximum height of the
ν116O peaks. The inset in each panel shows the ν118O region and associated curves with the
Raman intensity scale amplified by a factor of 10 to better illustrate changes in ν118O peak area.
The RRaman values calculated for each spectrum are shown in each panel.
253
Four−crystal mean 18O/16O
Individual crystal mean
0.005
0
0
O/16O
0.01
RRaman
RRaman
0.01
18
0.005
0.01
Rwaters
0.005
0
0
0.005
0.01
Rwaters
Figure D2. Relation between Raman-inferred 18O/16O ratios (RRaman) and the R values
predicted from the 18O/16O measured in the waters (Rcalc-inf), for both the 4-crystal mean of
each sample (left) and measurements of the individual crystals (right). For each plot, a 1:1
relationship is shown in red.
254
16
Ov1 peak signal to noise ratio
70
60
50
40
30
20
10
0
0
0.001
0.002
0.003
Residual (RRaman − Rwaters)
Figure D3. Relation between 18O/16O residuals (RRaman – Rcalc-inf) and the ν116O peak height
to background ratio in the Raman spectra, for samples with R values < 0.005.
255
C−enriched (97% 13C) calcite
13
Standard (1.1% C) calcite
Raman Intensity
13
300
290
280
270
−1
Wavenumber (cm )
260
Figure D4. Raman spectra of 13C-enriched and standard calcite in the ν13 active translational
mode region.
256
Intensity (normalized)
Room temperature
LN−cooled
1100
1095
1090
1085
1080
Wavenumber (cm−1)
1075
1070
Figure D5. The υ1 symmetric stretching mode region of Raman spectra of
optical calcite before and after immersion in liquid nitrogen (LN).
Intensity (normalized)
257
940
920
900
880
−1
Wavenumber (cm )
860
Figure D6. Raman spectrum of the C-C stretching mode region in the calcium
oxalate weddelite (CaC2O4·2H2O). Based on its position (following eq. 4) and
relative intensity, the satellite peak at ~870 cm-1 likely corresponds to the
13
C-12C bond.
258
APPENDIX E
PERMISSIONS
259
APPENDIX C: THE ROLE OF OCEAN THERMAL EXPANSION IN LAST
INTERGLACIAL SEA LEVEL RISE is reprinted with permission of the American
Geophysical Union.
McKay, N. P., J. T. Overpeck, and B. L. Otto-Bliesner, The role of ocean thermal
expansion in Last Interglacial sea level rise, Geophysical Research Letters, 38(14),
L14605, 2011. Copyright 2011 American Geophysical Union.