Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
From deep lakes to deserts: Plio-Pleistocene paleoenvironmental transitions
(USC Thesis Other)
From deep lakes to deserts: Plio-Pleistocene paleoenvironmental transitions
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
i
FROM DEEP LAKES TO DESERTS: PLIO-PLEISTOCENE PALEOENVIRONMENTAL
TRANSITIONS
By
Mark Donald Peaple
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(GEOLOGICAL SCIENCES)
August 2022
Copyright 2022 Mark Donald Peaple
ii
Acknowledgments
I’m grateful for the support of my advisor Prof. Sarah Feakins, my co-authors and committee
members, fellow Feakins lab members past and present (especially including Christine, Hyejung,
Hannah, Emily, Rachel, Christoph, Patrick, and Efrain), my partner Alan as well as friends and
family, throughout my PhD journey.
iii
Table of Contents
Acknowledgments…...……………………………………………………………………………………….ii
List of Tables ................................................................................................................................ vii
List of Figures .............................................................................................................................. viii
Abstract ........................................................................................................................................ xxi
Chapter 1: Introduction ................................................................................................................... 1
1.1 Motivation for studying Plio-Pleistocene paleoenvironmental change ............................ 1
1.1.1 Approaches to reconstructing Pliocene and Pleistocene terrestrial environments.... 1
1.2 Biomarker proxies ............................................................................................................ 6
1.2.1 Plant wax ................................................................................................................... 6
1.2.2 GDGTs .................................................................................................................... 10
1.3 Dissertation publications ................................................................................................ 13
References ................................................................................................................................. 14
Chapter 2: Identifying plant wax inputs in lake sediments using machine learning..................... 24
Abstract ..................................................................................................................................... 24
2.1 Introduction .................................................................................................................... 25
2.1.1 Study location and sampling ................................................................................... 27
2.2 Methods .......................................................................................................................... 32
2.2.1 Laboratory ............................................................................................................... 32
2.2.2 Machine Learning ................................................................................................... 34
2.3 Results and Discussion ................................................................................................... 37
2.3.1 Modern Plant Leaf molecular abundance ............................................................... 37
2.3.2 SLAPP core plant waxes ......................................................................................... 40
2.3.3 SLAPP core ACE index .......................................................................................... 42
2.3.4 Model training and testing ...................................................................................... 43
2.3.5 Model vegetation from SLAPP-SRLS17 core ........................................................ 45
iv
2.3.6 Model comparison................................................................................................... 47
2.3.7 Model Validation .................................................................................................... 49
2.4 Conclusions .................................................................................................................... 54
Acknowledgments ..................................................................................................................... 55
References ................................................................................................................................. 55
Chapter 3: Identifying the drivers of GDGT distributions in alkaline soil profiles within the
Serengeti ecosystem ...................................................................................................................... 64
Abstract ..................................................................................................................................... 64
3.1 Introduction .................................................................................................................... 65
3.1.1 Study location and climate ...................................................................................... 68
3.2 Materials and methods.................................................................................................... 71
3.2.1 Soil sampling .......................................................................................................... 71
3.2.2 Soil properties ......................................................................................................... 72
3.2.3 GDGT preparation and analyses ............................................................................. 72
3.3 Results ............................................................................................................................ 74
3.3.1 Soil temperature and pH measurements.................................................................. 74
3.3.2 Depth profiles.......................................................................................................... 75
3.3.3 Surface soil GDGT concentrations ......................................................................... 78
3.3.4 Temperature ............................................................................................................ 80
3.3.5 pH ............................................................................................................................ 82
3.4 Discussion ...................................................................................................................... 83
3.4.1 GDGT concentrations ............................................................................................. 83
3.4.2 ACE and soil salinity .............................................................................................. 85
3.4.3 Factors influencing BayBMT
0
in Serengeti soils.................................................... 87
3.4.4 Offsets between surface and deep brGDGT predicted temperatures ...................... 89
3.4.5 Comparison of organic and carbonate paleothermometers ..................................... 92
3.5 Conclusions .................................................................................................................... 96
Acknowledgments ..................................................................................................................... 97
References ................................................................................................................................. 97
Chapter 4: Biomarker and pollen evidence for late Pleistocene pluvials in the Mojave Desert . 110
Abstract ................................................................................................................................... 110
v
4.1 Introduction .................................................................................................................. 111
4.2 Regional setting ............................................................................................................ 113
4.3 Age Model .................................................................................................................... 114
4.4 Material and methods ................................................................................................... 115
4.4.1 Lipid extraction ..................................................................................................... 115
4.4.2 GDGT analyses ..................................................................................................... 116
4.4.3 Compound-specific isotopic analyses ................................................................... 118
4.4.4 Palynological Analyses ......................................................................................... 119
4.4.5 Correlation analysis............................................................................................... 120
4.5 Results and Discussion ................................................................................................. 120
4.5.1 Vegetation reconstructions from Searles Lake spanning 200 kyr......................... 120
4.5.2 Plant wax evidence for glacially paced changes in hydroclimate......................... 126
4.5.3 Comparison with regional precipitation isotope archives ..................................... 128
4.5.4 Searles Lake salinity and regional moisture availability ...................................... 131
4.5.5 Terrestrial temperatures ........................................................................................ 138
4.6 Conclusions .................................................................................................................. 141
4.7 Acknowledgments ........................................................................................................ 142
References ............................................................................................................................... 143
Chapter 5: Warm and wet Pliocene southwestern North America: evidence from Searles Lake156
Abstract ................................................................................................................................... 156
5.1 Introduction .................................................................................................................. 156
5.2 Materials and Methods ................................................................................................. 159
5.2.1 Sediment core and age model ............................................................................... 159
5.2.2 Lipid purification and analyses ............................................................................. 160
5.2.3 Climate Model Simulations .................................................................................. 161
5.3 Results and Discussion ................................................................................................. 162
5.3.1 Searles Lake Biomarker Reconstruction ............................................................... 162
5.3.2 Constraining moisture sources .............................................................................. 166
5.3.3 Pliocene tectonic context ...................................................................................... 169
5.4 Conclusions .................................................................................................................. 170
Acknowledgements ................................................................................................................. 170
References ............................................................................................................................... 171
vi
Chapter 6: Paleoenvironmental reconstruction of Pliocene hominin habitats at Woranso-Mille179
Abstract ................................................................................................................................... 179
6.1 Introduction .................................................................................................................. 180
6.2 Regional Setting ........................................................................................................... 182
6.3 Methods ........................................................................................................................ 186
6.3.1 Field sampling ....................................................................................................... 186
6.3.2 Laboratory methods .............................................................................................. 186
6.3.3 Pedogenic carbonate isotopic analyses ................................................................. 189
6.4 Results .......................................................................................................................... 190
6.4.1 n-Alkanoic acid abundances ................................................................................. 190
6.4.2 Carbon and hydrogen isotopic compositions of n-alkanoic acids......................... 190
6.4.3 GDGT distributions............................................................................................... 193
6.5 Discussion .................................................................................................................... 194
6.5.1 Biomarkers across facies....................................................................................... 194
6.5.2 Comparing plant wax and pedogenic carbonate proxies ...................................... 197
6.5.3 Comparing marine and terrestrial paleoenvironmental archives .......................... 199
6.5.4 Hominin paleoenvironment reconstruction........................................................... 203
6.5.5 Hominin dietary expansion ................................................................................... 209
6.6 Conclusions .................................................................................................................. 211
Acknowledgments ................................................................................................................... 212
References ............................................................................................................................... 212
Chapter 7: Dissertation conclusions............................................................................................ 222
References ............................................................................................................................... 226
Appendices .................................................................................................................................. 273
Appendix. A Supplementary information for Chapter 2 ......................................................... 274
Appendix. B Supplementary information for Chapter 3 ......................................................... 277
Appendix. C Supplementary information for Chapter 4 ......................................................... 279
Appendix. D Supplementary information for Chapter 5 ......................................................... 307
Appendix. E Supplementary information for Chapter 6 ......................................................... 309
vii
List of Tables
Table 2.1. Performance metrics for each machine learning model............................................. 43
Table 2.2. Accuracy of models when uniformly distributed noise with 3 different
ranges (0–0.01, 0–0.05, and 0–0.1) was added to the test set. Accuracy of models
when uniformly distributed noise with 3 different ranges (0–0.01, 0–0.05, and 0–0.1)
was added to the test set. ............................................................................................................. 45
Table 3.1. Site locations and characteristics ............................................................................... 69
viii
List of Figures
Fig 2.1. Topographic map (black shading) with elevation-based predictions of
vegetation cover within the Searles Lake catchment (colors, see legend), showing the
expected configuration during a pluvial period (a) versus the modern catchment and
vegetation (b). Topography is from a 7.5-arcsecond resolution DEM (Danielson, J.J.,
Gesch, 2011), with drainage basins delineated from topographic highs in ArcMap.
Modern vegetation in the Owens/Searles catchment (California Department of
Forestry and Fire Protection, 2015) is grouped into four categories: Desert, Shrub,
Conifer Woodland, and Conifer Forest. We grouped Desert and Shrub as “desert
shrubs” and grouped Conifer woodland and Conifer Forest as “conifer” for the
machine learning calculations. .................................................................................................... 27
Fig 2.2. Chain length distributions of n-alkanes (blue) and n-alkanoic acids (green) of
(A) Modern plants and (B) SLAPP-SRLS17 core samples, showing samples
containing high modeled (SVM) desert plants, conifers, and macrophytes as well as
mean core distributions. .............................................................................................................. 38
Fig 2.3. Relative macrophyte abundance was calculated using the machine learning
models and the linear mixing model (Gao et al., 2011). Support vector machine (SVM)
and the linear mixing model are highlighted to show similarity. Data have been
normalized by the mean and scaled by unit variance. For comparison, we show
measured variables: B. Carbon preference index (CPI) of n-alkanes and n-alkanoic
acids; C. Average chain length (ACL) of n-alkanes and n-alkanoic acids; D. Archaeol
caldarchaeol ecometric (ACE). ................................................................................................... 42
ix
Fig 2.4. Probability estimates of modeled vegetation types downcore for each machine
learning model. Probabilities are generated by inputting core wax distribution
information into models trained on modern plant wax distributions. Salt in core
highlighted by grey bars.............................................................................................................. 46
Fig 2.5. Correlation matrices showing Spearman’s Rank correlations between
probabilities of the three modeled vegetation types downcore (A. Macrophyte, B.
Desert plant, C. Conifer) and archaeol caldarchaeol ecometric (ACE) salinity index in
SLAPP-SRLS17. Black boxes denote p<0.01. .......................................................................... 48
Fig 2.6. Modeled macrophyte plant proportions using the support vector machine
model and ACE index values against SLAPP-SRLS17 core depth. Data has been
centered to the mean and scaled by unit variance. Modeled vegetation was generated
using wax distributions measured from SLAPP-SRLS17. The top image is a composite
high-resolution photograph of the core. ...................................................................................... 51
Fig 3.1. A) Location of the Serengeti ecosystem within Africa. B) Map of study region
showing the Serengeti ecosystem spanning Tanzania and Kenya, the soil sampling
locations (yellow points) along a transect within the Ngorongoro Conservation Area
and Serengeti National Parks in Tanzania. The vegetation map was simplified from
Reed et al. (2009) to show only grassland, shrubland, woodland, and forest vegetation.
Figure from Zhang et al., (2021) and reprinted with permission. ............................................... 70
Fig 3.2. Serengeti soil data compared to prior global surface soil calibrations. A) Mean
annual air temperature (MAAT) plotted against MBT´
5Me
for the global compilation of
surface soil samples (grey symbols) (Dearing Crampton-Flood et al., 2020) and the
surface (<0.1 m; solid blue symbols) and deep (>0.1 m; open blue symbols) soil
x
samples from this study. B) Soil pH plotted against CBT´
from a global compilation of
surface soil samples (De Jonge et al., 2014) and soils from this study. C) Soil pH
plotted against IR
6Me
in both a global compilation (Dearing Crampton-Flood et al.,
2020) and soils from this study. .................................................................................................. 75
Fig 3.3. Depth profiles of the 11 sites showing A) brGDGT concentrations, B)
isoGDGT concentrations, C) Total organic carbon (TOC) % (Zhang et al., 2021), D)
ΛbrGDGT concentrations, E) ΛisoGDGT concentrations, F) R
i/b
, G) BIT, H) ACE, I)
MBT´
5Me
, J) CBT´, K) IR
6Me
, L) Total organic carbon δ
13
C (TOC δ
13
C) (Zhang et al.,
2021), M) Total dissolved solids (TDS), N) Measured pH, O) C
23-33
alkanes (μg/g)
and P) C
22-32
alkanoic acids (μg/g). .......................................................................................... 77
Fig 3.4. Relationship between mean annual precipitation (MAP) and surface samples
A) R
i/b
, B) BIT, C) ΛbrGDGT concentrations (ng/g OC), D) ΛisoGDGT
concentrations (ng/g OC). ........................................................................................................... 79
Fig 3.5. Comparison of measured mean annual soil temperature (MAST) and predicted
soil temperature, modeled air temperature and pH at both the surface (left column,
filled blue symbols) and at depth (right column, open blue symbols). A) and B). Cross
plots of mean annual air temperature (Fick and Hijmans, 2017) against mean
temperature from months above freezing (MAF) predicted by BayMBT
0
. C. D. Cross
plots of the measured soil temperature at nearest depth to brGDGT sample (Beverly et
al., 2021) against BayMBT
0
. E) and F). Cross plots of measured soil pH against CBT'
predicted pH. ............................................................................................................................... 81
Fig 3.6. Exploring possible factors influencing mean annual air temperature (MAAT)
residuals (MAAT-predicted MAAT) in all soil samples (surface <0.1 m, solid circles;
xi
deep >0.1 m, open symbols), including A) mean annual precipitation (MAP), B)
IR
6Me
, C) measured pH and D) total dissolved solids (TDS). The Gray dashed line
represents 0 residual. ................................................................................................................... 82
Fig 3.7. Comparison of total dissolved solids (TDS) versus A) isoGDGTs, B)
brGDGTs, and C) TOC and sampling depth vs E) isoGDGTs, F) brGDGTs and
G) TOC. We show the amount of variance explained by a power-law regression as
well as the Spearman rank correlation coefficient and corresponding p-value. Solid
blue symbols represent surface samples and white filled symbols represent deep soil
samples. ....................................................................................................................................... 86
Fig 3.8. Cross plot of measured Total dissolved solids (TDS) and calculated ACE
index in soil samples, showing surface samples (solid symbols) and deep (open
symbols). ..................................................................................................................................... 87
Fig 3.9. PCA biplots showing biomarker and environmental indices: Total organic
carbon % (TOC), CBT´, ratio of isoGDGTs to brGDGTs (R
i/b
), measured pH, total
dissolved solids (TDS), mean annual air temperature (MAT; Fick and Hijmans, 2017),
MBT´
5Me
, archaeol caldarchaeol ecometric (ACE), mean annual precipitation (MAP;
Fick and Hijmans, 2017) for A) all surface samples (<0.1 m), B) all deep samples
(>0.1 m)....................................................................................................................................... 92
Fig 3.10. Comparison of brGDGT (this study) and clumped (Beverly et al., 2021)
temperature estimates. A) Violin plot showing distributions of all samples: central
estimates are the same within uncertainties however the range differs. MAAT and
minimum and maximum MAST temperatures for the Serengeti are shown for
comparison. B) Cross-plot of site MAAT (equivalent to MAF) from surface soil
xii
BayMBT
0
(error bars, calibration 1 RMSE) and ∆
47
temperatures within the soil
profile (error bars, propagated analytical and calibration uncertainty, 1 SEM); 1:1 line
(dashed line). ............................................................................................................................... 93
Fig 4.1. Maps showing the location of A) Searles Lake (red star) and climate archives
referred to in the text including Owens Lake (blue circle), ODP 1012/1010 (pink
circles), Devils Hole (orange circle), Leviathan Cave, Lehman Cave, and Pinnacle
Cave (black circles) B) The Lakes connected to Searles Lake during pluvial periods
where M = Mono Lake, O = Owens Lake, C = China Lake, S = Searles Lake, P = Lake
Panamint, M = Lake Manly. C) Map of Searles Lake during pluvial conditions
highlighting inflow and outflow. .............................................................................................. 112
Fig 4.2. Vegetation reconstructions using pollen and plant wax proxies from SLAPP-
SRLS17. A) Proportion of pollen taxa. B) Modelled vegetation types based on SVM
machine learning of plant wax distributions in modern taxa applied to the downcore
record (Peaple et al., 2021b). C) Comparison between modelled desert plant types and
pollen “desert shrubs” (the sum of Amaranth and Asteraceae pollen). D) δ
13
C
28acid
and
δ
13
C
31alk
compared to Amaranth pollen. E) δD
28acid
and
δD
31alk
. ............................................... 122
Fig 4.3. PCA to assess biomarker and pollen covariations (Shrub = sum of
Amaranthaceae and Asteraceae pollen abundance). ................................................................. 123
Fig 4.4. Comparison of Searles Lake plant wax δD
31alk
and calculated δD
precip
to global
climate data across two glacial interglacial cycles showing A) Antarctic pCO
2
record
(Lüthi et al., 2008), B) LR04 δ
18
O benthic foraminifera stack (Lisiecki and Raymo,
2005), C) plant wax C
31
n-alkane D (blue curve) and inferred precipitation D after
apparent fractionation and ice volume correction (black curve). D) BayMBT
0
and E)
xiii
shrub pollen%. Upper labels: “Hol” = Holocene, “LGM” = Last glacial maximum,
“LIG” = Last interglacial, “PGM” = Penultimate glacial maximum. Lower labels:
“MIS” = Marine isotope stage. ................................................................................................. 127
Fig 4.5. Comparison of plant wax and speleothem isotopic records. A) Searles Lake
δD
precip
(black, this study), Leviathan composite record δ
18
O
calcite
(orange; Lachniet et
al., 2016) and Devils Hole δ
18
O
calcite
(red; Moseley et al., 2016) with the
18
O axis
scaled to account for the 8x greater mass dependent fractionation for hydrogen. B)
Searles Lake δD
precip
(black) and summer insolation at 65°N (gray). C) Devils Hole
δ
18
O
calcite
and summer insolation at 65°N (gray). D) Leviathan composite record
δ
18
O
calcite
(left) as in A but showing the individual caves, two of which (Lehman and
Pinnacle), were adjusted for spatial gradients in precipitation isotopes (Lachniet et al.,
2016). Black and white bars represent MIS stages. E – G) Weighted wavelet z
transform frequency spectrum for the records in B, C, and D. H) 5 to 95 % quartile
range for measured values (blue), and after corrections for ice volume (grey), cave
temperature (Leviathan record, black bar) and plant wax
wax/w
(green). The δ
18
O axis
is scaled to account for the 8x difference in mass dependent fractionation between H
and O. Ice volume-corrected Devils Hole shows the smallest range. In contrast, larger
and comparable magnitudes are recorded at the temperature-corrected Leviathan
composite record and Searles Lake........................................................................................... 129
Fig 4.6. Biomarker evidence that the late MIS 6 pluvial was a fresher water lake than
the late MIS 2 pluvial. Water balance reconstructions Searles Lake: A) ACE, B)
IR
6+7me
, C) BIT D) %GDGT-0, and E) Devils Hole water table elevation (Wendt et al.,
2018). Age model without tie point is plotted for all GDGT indices as a thin faint line.
xiv
Terminations 1 and 2 are highlighted with yellow shading and Heinrich 1 and 11 are
highlighted with blue shading. Upper labels: “LGM” = Last glacial maximum, “LIG”
= Last interglacial, “PGM” = Penultimate glacial maximum. Lower labels: “MIS” =
Marine isotope stage. ................................................................................................................ 133
Fig 4.7. Local and regional temperature records over the past 200 kyr. A. Searles Lake
(blue line; this study) and Lake Elsinore (orange line; Feakins et al., 2019) recalibrated
to MAF using Martinez Sosa et al., (2021) brGDGT temperature records, using the
lake MBT´
5Me
BayMBT
0
calibration to mean temperature from months above freezing
(MAF). TEX
86
calibrated to lake surface temperature (black dot) (Tierney et al., 2010).
B) Noble gas derived ground water temperature records (Mojve: Kulongoski et al.,
2009; San Diego: Seltzer et al., 2021). Comparison temperature responsive vegetation
change showing C) shrub pollen % (Amaranthaceae and Asteraceae; this study). D)
Alkenone based sea surface temperature (SST) records (ODP 1012, ODP 893: Herbert
et al., 2001, 1995)...................................................................................................................... 140
Fig 5.1 Site location map showing Searles Lake (this study) and other lakes and
marine core sites mentioned in the text. Published Pliocene marine core sites (circles)
showing the 95% (inner) and 5% (outer) confidence intervals and median (middle) for
Pliocene (3.4–2.7 Ma) SSTs anomalies relative (bottom color bar) to core top
(modern) are calculated from published alkenone U37k' indices at ODP Sites 1010
(LaRiviere et al., 2012), 1012 (Brierley et al., 2009), and (Dekens et al., 2007),
recalibrated using BAYSPLINE (Tierney et al., 2019). Map shading shows modern
sea surface temperature (bottom color bar) from GHRSST (OurOcean Project, 2010)
xv
and precipitation data from CHIRPS Version 2.0 (Funk et al., 2015). Hatching covers
the area with > 50% summer (JAS) precipitation. .................................................................... 158
Fig 5.2. Age model generated using BACON (black line), 95% confidence interval
(grey shading), and paleomagnetic datums (Liddicoat et al., 1980) updated to the
GPTS2020 (Channell et al., 2020; Glen et al., 1999) (red symbols). ....................................... 160
Fig 5.3. a) Summary of lake depth from Lake Andrei (LA) (Knott et al., 2021), Lake
Manly (LM) (Knott et al., 2018) and Searles Lake (SL) (Smith et al., 1983; this study).
b-e) Searles Lake proxy reconstructions from Pliocene (this study) and Pleistocene
(Peaple et al submitted) including: b) ACE index of salinity c) BayMBT0 temperature
reconstruction of mean air temperature for months above freezing d) δ13C value of
C30 alkanoic acid e) δD value of precipitation calculated using δD of C30 alkanoic
acid f) 1st principle component of SSTs anomalies relative to core top (modern),
calculated from published alkenone U37k' indices at ODP Sites 1010 (LaRiviere et al.,
2012), 1012 (Brierley et al., 2009) and ODP 1014 (Dekens et al., 2007), recalibrated
using BAYSPLINE (Tierney et al., 2019). G = Glacial (Peaple et al submitted), IG =
Interglacial (Peaple et al submitted) and P = Pliocene (this study). ......................................... 163
Fig 5.4. Isotope-enabled model output from iCESM1.2. Top left: Anomaly of late
Pliocene precipitation δD relative to modern (black diamond shows the location of
Searles Lake). Top right: precipitation–evaporation (P–E). Bottom left: Pliocene wind
vector map with colors representing amount of integrated water vapor transport (IVT).
Bottom right (top panel): Seasonal precipitation, evaporation, and P–E late Pliocene
anomalies calculated from box in top panels. Bottom right (bottom panel): Seasonal
xvi
precipitation δD from both preindustrial and late Pliocene periods, calculated from box
in top panels. ............................................................................................................................. 166
Fig 5.5. Plio-Pleistocene SST anomalies (black line) and uncertainties (blue shading,
1 ) from ODP Sites 1012 (Brierley et al., 2009) and 1014 (Dekens et al., 2007), where
the original alkenone data from those studies were recalibrated using BAYSPLINE
(Tierney et al., 2019). The top bar shows reconstructed Searles Lake depth from this
study and Smith et al. (1983). ................................................................................................... 167
Fig 6.1 Location of hominin fossil localities Woranso-Mille and Hadar, and DSDP
Site 231 A) regional DEM (ETOPO1, 1 minute resolution) and B) 3D digital elevation
map (Amante and Eakins, 2009) highlighting the modern Mille River drainage basin
(color shading highlighting elevation change). ......................................................................... 185
Fig 6.2 Violin plots showing distributions of a) δ
13
C
and b) δD
data measured from
C
26-32
alkanoic acids.
c) δ
13
C
28acid
and d) δD
28acid
of samples delineated by lithology. ............. 192
Fig 6.3 Correlation coefficient between A) δ
13
C and B) δD measured in C
26-32
alkanoic
acids. ......................................................................................................................................... 192
Fig 6.4 Violin plots of A) δ
13
C
28acid
,
B) δD
28acid
, C) R
i/b
and
D) BayMBT
0
mean annual
air temperature, measured in fluvial/lacustrine, paleosol and paleosol with pedogenic
carbonate sediments. ................................................................................................................. 195
Fig 6.5 Violin plot showing distributions of δ
13
C
28acid
(left) and δ
13
C
PC
(right). δ
13
C
28acid
samples are grouped by facies type........................................................................................... 198
Fig 6.6. A. Insolation at 20 degrees latitude North and eccentricity (Laskar et al.,
2004). B. δ
13
C
wax
records from marine core DSDP 231(Liddy et al., 2016) and
Woranso-Mille. C. Soil carbonate δ
13
C data from Afar compiled by Levin, (2013,
xvii
references therein) and including new δ
13
C data in this publication. D. Bovid enamel
δ
13
C from Afar paleoanthropological sites compiled by Paquette and Drapeau, (2021).
E. Carbonate clumped isotope derived temperatures from paleosols in the Turkana
Basin, Kenya. F. Modelled atmospheric CO
2
concentration (Stap et al., 2016). ..................... 200
Fig 6.7 Violin plots of a) δ
13
C
28acid
b) δD
28acid
c) R
i/b
and d) BayMBT
0
mean annual air
temperature,
with samples delineated by the hominin species present at outcrop
locality....................................................................................................................................... 204
Fig 6.8 Cross plots showing δ
13
C
28acid
versus R
i/b.
A) Color represents different sample
facies. B) Color represents samples from sites associated with a hominin species. ................. 206
Fig 6.9 Woranso-Mille δ
13
C wax (small black circles), at individual sites (for colors,
see legend) with the sample most proximal to hominin fossil (symbol). b) Hominin
tooth enamel δ
13
C for a variety of species (see legend) collected from sites across
eastern Africa (Lee-Thorp et al., 2012; Cerling et al., 2013; Wynn et al., 2013; Levin
et al., 2015)................................................................................................................................ 210
Fig. A.1.Chain length distributions of n-alkanes (blue) and n-alkanoic acids (green) in
(A) Modern desert plants. (B) Modern conifers. (C) Modern macrophytes. Columns A,
B, and C show the relative proportion of n-alkyl lipids. Error bars represent 1σ from
the mean when multiple plants were analyzed.......................................................................... 275
Fig. A.2. Receiver operating characteristics (ROC) graphs of the five machine learning
models. ROC graphs plot the true positive rate (sensitivity) against the false positive
rate (specificity) under different classification threshold settings—models with higher
discriminant capacity plot closer to the upper left-hand corner of the plot. Left column
xviii
shows the ROC curves of models tested against the original data. Right column shows
ROC curves of models tested against original data with uniformly distributed noise
added. Dashed black 1:1 line included for comparison. .......................................................... 276
Fig. B.1. Serengeti surface soil longitudinal transect showing: A) temperature, B) pH,
C) TOC (%), D) TOC δ
13
C, E) ∑brGDGTs (ng/g sediment), F) ∑isoGDGTs (ng/g
sediment), G) ΛbrGDGTs (ng/g TOC), H) ΛisoGDGTs (ng/g TOC). ..................................... 277
Fig. B.2. Standard deviations of mean distributions were created from bootstrapping
the BayMBT
0
and ∆
47
temperature data with different resampling sizes. As the
resampling size increases, the variance associated with the mean of resampled data
decreases. .................................................................................................................................. 278
Fig. C.1. Violin plots showing a) δ
13
C and b) δD values n-alkanoic acids (blue) and n-
alkanes (orange) in SLAPP-SL17 and c) δ
13
C and d) δD distributions for plants
collected in the San Bernadino Mountain (SBM, green) and Searles Valley (SV, pink). ........ 281
Fig. C.2. Changes in the isotopic composition of n-alkanes and n-alkanoic acids with
depth in SLAPP-SL17 core a) δ13C measured from the C27, C29, and C31 alkane
chain lengths. b) δ13C measured from the C24, C26, and C28 alkanoic acid chain
lengths. c) δD measured from the C27, C29, and C31 alkane chain lengths. d) δD
measured from the C24, C26, and C28 alkanoic acid chain lengths. e) Composite core
photo showing the presence of muds (dark) and salts (white) downcore. Thick salt
accumulations without biomarker sampling are indicated by pale pink shading. .................... 286
Fig. C.3. Downcore pollen and plant wax proxies from SLAPP-SRLS17. a) Proportion
of pollen taxa. b) Modelled vegetation types based on SVM machine learning of plant
wax distributions in modern taxa applied to the downcore record (Peaple et al.,
xix
2021b). c) Compare modeled desert plant types and sum Amaranth and Asteraceae
pollen. d) δ
13
C
28acid
and
δ
13
C
31alk
compared to Amaranth pollen. e) δD
28acid
and
δD
31alk
.
f) Precipitation δD estimated from the δD
31alk
with
wax/p
determined as i) “constant” -
93‰ or temporally varying as calculated based on the ii) “pollen,” i.e., the proportion
of pollen taxa in the core, iii), “C
3
v. C
4
” proportion based on δ
13
C
31alk
, and iv) “ML”
where “constant” is modified by mixing with the SVM modeled desert plant
proportions from machine learning. g) Composite core photo. Thick salt
accumulations without biomarker sampling (pale pink shading). ............................................ 289
Fig. C.4. Analysis of major pollen taxa in SLAPP-SRLS17 sediment core. a) Non-
metric multi-dimensional scaling (NMDS) analysis with glacial (blue diamond) and
interglacial (red circle) samples highlighted. b, c, d) Time series plots showing the
pollen percentage and influx rate. ............................................................................................. 295
Fig. C.5. Downcore GDGT concentrations and indices in SLAPP-SRLS17, showing
concentrations of a) summed and selected GDGTs concentrations, b) BIT and
%GDGT-0, c) caldarchaeol and isoGDGT-0, d) ACE and IR
6+7me
index for salinity, e)
pH, f) reconstructed MAF temperatures and g) composite core stratigraphic column.
Overflowing lake conditions with mud facies and low ACE salinity (blue shading),
only the T2 deep lake is associated with high productivity (a) and a well-mixed lake
(b). Thick salt accumulations without biomarker sampling (pale pink shading). ..................... 297
Fig. C.6. Time and frequency response of salinity proxies ACE (blue) and IR
6+7me
(orange)
.
a) Time series of ACE and IR
6+7me
b) Wavelet coherence between ACE and
IR
6+7me.
c) Weighted wavelet Z transform frequency spectrum of ACE (blue) and
IR
6+7me
(orange). ........................................................................................................................ 298
xx
Fig. C.7. Violin plots of brGDGT distributions from soil (Dearing Crampton-Flood et
al., 2020) and lake (Martínez-Sosa et al., 2021) compilations, selecting those entries
with MAT of 14–25 °C, for comparison to brGDGT distributions from Searles Lake
SLAPP samples. Searles appears to have lake-like distributions. ............................................ 301
Fig. D.1 Lake proxy reconstructions for the Late Pliocene. A) Summary of lake depth
from Lake Andrei (Knott et al., 2021), Lake Manly (Knott et al., 2018), and Searles
Lake (Smith et al., 1983). Searles Lake proxy reconstructions from core KM3 (this
study) including: B) ACE index of salinity, C) IR
6+7me
index of salinity, D) BayMBT
0
temperature reconstruction of mean air temperature for months above freezing, E)
Branched and Isoprenoid Tetraether (BIT) index, F) ∑C
22
-
32
alkanoic acid
concentration, G) δ
13
C value of C
30
alkanoic acid and H) δD value of precipitation. .............. 308
Fig. E.1. Location and stratigraphic columns of sites within Woranso-Mille. Numbers
represent field codes of biomarker samples and black circles represent biomarker
sample stratigraphic position. Woranso-Mille sampling site codes are as follows: BRT
= Burtele, NFR = Nefuraytu, LDD = Leado Dido’a, KSD = Korsi Dora, MSD =
Mesgid Dora, MRD = Miro Dora, LHG = Lahaysule Gera, KSA = Kuserale Dora,
GUG = Gugubsi, MRV = Mille River Valley........................................................................... 309
xxi
Abstract
Anthropogenic emissions of CO
2
are leading to global warming, with Earth's surface
temperatures rising to levels unseen since the Pliocene, 3 million years ago. This temperature rise
will impact precipitation patterns and induce large-scale environmental changes that will
challenge human societies. However, predicting the impacts of these changes is difficult and
relies primarily on models. In some regions, including southern California, models disagree on
whether the climate will become wetter or drier as a result of this increase in temperature.
Another approach to determining future changes is to look back at past periods of Earth’s
history, with similar atmospheric CO
2
concentrations and Earth surface temperatures as we will
likely experience soon. The previous interglacial (130 – 115 kyr) and the Pliocene (5.3 – 2.6 Ma)
are two periods with such conditions. Paleoclimatic reconstructions of these periods could
provide insight for predicting future climate change impacts. To reconstruct the paleoclimate
during these periods, I used biomarker proxies sensitive to temperature changes, precipitation,
and landscape vegetation composition. Chapter 2 introduces a machine learning approach to gain
information about vegetation change encoded in the molecular distribution of plant waxes
preserved in ancient lake sediments from Searles Lake, California. I compare my vegetation
reconstruction to an independent proxy of lake salinity and find a correlation between landscape
desert plant abundance and high lake salinity, thereby validating my vegetation modeling
approach.
In Chapter 3 I aim to understand the controlling factors on the distribution of glycerol dialkyl
glycerol tetraethers (GDGTs) in a suite of arid and alkaline soils from the Serengeti National
xxii
Park, Tanzania. GDGTs are produced by bacteria (branched-GDGTs) and archaea (isoprenoid-
GDGTs) and act as lipid biomarkers sensitive to temperature, pH, and moisture changes. As we
wish to apply the GDGT proxy to understand past climates, we must fully understand the process
governing the distribution of different GDGT compounds in these modern soils. We find that
brGDGTs in the surface soil are sensitive to changes in temperature. However, brGDGTs
sampled deeper in the soil profiles were found not responsive to temperature and thus most likely
represent compounds produced by a different bacterial community than is present in the surface.
This study thus can guide future paleoenvironment studies focused on temperature
reconstructions from ancient soil (paleosols).
In Chapter 4 I present a paleoclimatic reconstruction of southwestern North America from an
expanded suite of biomarkers and pollen preserved in Searles Lake sediments between 200 – 5
ka. We find that the hydrogen isotopic composition of precipitation (δD of precipitation)
increased during interglacial periods and decreased during glacial periods, which agrees with
other regional records of precipitation isotopes. We also find that Searles Lake showed the
largest changes in lake depth in association with terminal glacial phases, with termination 2
being characterized by a very deep overturning lake that overflowed into the adjacent basin.
Crucially, we find that the previous interglacial was wetter than the present Holocene and a
permanent lake was present in Searles Valley under a warmer climate. Chapter 5 presents
biomarker data from Pliocene sediments deposited in Searles Valley. Previous sedimentological
work indicated that a deep lake environment existed between 2.7 – 3.2 Ma in Searles Valley.
Recent modeling studies suggest that this deep lake existed due to higher southern California
coastal sea surface temperatures driving an increase in the North American Monsoon, which
xxiii
increased summer season precipitation. We tested this hypothesis using the hydrogen isotopic
composition of our plant waxes. Moisture from the North American Monsoon region typically
has a more positive δD, whereas winter precipitation from the North Pacific has a more negative
δD. We find that the plant waxes in the Pliocene Searles Lake sediments are relatively positive,
which is indicative of an increase in North American Monsoon precipitation, supporting the sea
surface temperature “warmer = wetter” hypothesis.
Chapter 6 aims to both reconstruct the local paleoenvironments of early hominin species who
lived in Woranso-Mille, Ethiopia, and to assess whether the paleoenvironment of Woranso-Mille
was following regional trends forced through changes in paleoclimate or whether it was more
sensitive to local tectonic and volcanic influences. By studying the isotopic composition of plant
waxes preserved in sediments, we found that the early hominins were inhabiting mixed
landscapes, including both open savannah and closed woodland environments. Interestingly,
there was a shift to more C
4
grass/shrub vegetation between the 3.8 – 3.2 Ma in Woranso-Mille
which parallels the changes reported from a marine core record of terrestrial plant waxes
offshore in the Arabian Gulf. Additionally, we find that an expansion of C
4
plants in Woranso-
Mille which occurred 3.8 – 3.6 Ma occurs at the same time as Hominin tooth enamel carbon
isotopes shift to more positive values indicating an expansion of dietary breadth. This suggests
the change in hominin diet was forced by a change in their habitat’s vegetation composition.
xxiv
1
Chapter 1: Introduction
1.1 Motivation for studying Plio-Pleistocene paleoenvironmental change
The Pliocene (5.4–2.4 Ma) is a crucial epoch in the history of our human evolutionary journey.
Important morphological and behavioral transitions occurred, including the emergence and
diversification of the bipedal Australopithecus (Haile-Selassie et al., 2010, 2015; Melillo et al.,
2021), to the eventual nexus of our own genus Homo (Spoor et al., 2015). These evolutionary
transitions occurred against the backdrop of Pliocene climate changes. Through anthropogenic
global warming, we are returning earth to a Pliocene-like climate, with warmer conditions forced
by high levels of CO
2
, causing changes to precipitation patterns (e.g. Fowler and Hennessy,
1995; Chou and Neelin, 2004; Giorgi et al., 2019), extreme weather events (e.g. Huang et al.,
2020; Bartusek et al., 2021) and ocean-atmosphere teleconnections (e.g. Bartusek et al., 2021).
Thus, the Pliocene climate is relevant 1) to understand the environment in which our hominin
ancestors developed and to understand if and how it shaped our genus (Homo) and 2) to attempt
to predict future climate changes and their impacts. Likewise, the Pleistocene climate
experienced dramatic swings in temperature forced by changes in pCO
2
and ice sheet feedbacks.
Some periods, such as the previous interglacial period, were likely warmer than present (Peaple
et al in review) and thus have the potential to serve as analogs to future climate changes.
1.1.1 Approaches to reconstructing Pliocene and Pleistocene terrestrial environments
Marine cores provide the majority of our knowledge on Pliocene climate and environments due
to the continuous nature of sedimentation (usually) and the (relative) ease of dating them makes
the proxies contained within easier to interpret (Burls and Fedorov, 2017). Terrestrial
2
environments are comparatively poorly sampled with only a few studies reporting high-
resolution paleoenvironmental data (Adam et al., 1989; Magill et al., 2013; Lupien et al., 2019).
Challenges involve dating, access to outcropping sediments, preservation of proxies, and
complications in interpreting a climate signal from the proxies (Chapter 3,4,5). As such, whilst
we have comparatively good high-resolution continuous records of Pliocene sea surface
temperature along the coast of California (Dekens et al., 2007), we have few continuous climate
records of terrestrial changes further inland.
The development of relatively new proxies, such as GDGTs and plant waxes, offers
opportunities (e.g. Dearing Crampton-Flood et al., 2018) to glean terrestrial climate and
environmental information from sediments in which older generation proxies could not be
applied. Thus, scientists are starting to revisit Pliocene deposits to extract additional information
in addition to more traditional analysis (notably including sediment stratigraphy and lithological
logging) that have been applied in past studies. Aided by advances in geochronological
techniques (e.g. Ar/Ar, Deino et al., 2010), we now have the potential to generate continuous
records of terrestrial climate change.
Notably, lacustrine deposits offer the most continuous records of terrestrial Pliocene
sedimentation and are thus attractive targets for reanalysis. Some proxies found in lake sediments
are integrated from the entire drainage basin of the lake, allowing us the opportunity to study
regional climate change, similarly to coastal marine cores. Lake records also usually offer good
preservation of different proxies on account of their cool (usually) bottom water temperatures,
fine sediment size (usually), and anoxic sediment environments. However, Pliocene lake records
3
are incredibly rare in some terrestrial environments (Pound et al., 2014a) and thus lakes can thus
only offer limited spatial coverage. Also, lakes are by definition wet environments, which
presents a climate bias to proxies contained in their sediments. Access to lake sediments can also
be challenging. Pliocene lake sediments are usually deeply buried under overlying Pleistocene
sediment deposits, and thus require expensive drilling operations to access. Thus, whilst lakes are
attractive targets for terrestrial climate reconstruction they are limited by occurrence, a “wet
bias” and logistics.
Sediment outcrops are also useful archives of paleoenvironmental proxies. On the surface,
outcrops appear to solve many of the problems inherent in lake records. Outcrops do not require
equipment to access, they are (usually) relatively accessible and they offer a higher degree of
spatial coverage across paleo landscapes. If we wish to understand the paleoenvironment of an
extinct hominin, what better strategy is there than to directly sample the outcrops from/proximal
to where a hominin fossil was preserved? Such specificity in paleoenvironmental and
paleoclimatic information can only be obtained from outcrop studies. Notable examples of
studies include Magill et al., (2013) who sampled across a continuous paleosol outcrop in
Eastern Africa to establish the degree of water availability on a paleo landscape to the hominin
whose fossils are preserved in the paleosol. However, outcrops also present substantial
challenges to paleoenvironmental reconstruction. Unlike a lake or marine core, they are not
continuous records of sedimentation. Unconformities can be present, complicating stratigraphic
interpretation. Dating outcrops is a challenge and relies on both paleomagnetism (can be applied
to a range of sediments, usually low resolution) and radiometric dating of volcanic material (can
only be applied to volcanic materials, high precision) which are only present proximal to
4
volcanoes and often show poor preservation in the geological record (Fontijn et al., 2014).
Understanding the temporal relationships between different outcrops can be challenging and
requires substantial amounts of geological mapping work. Furthermore, the preservation of
organic proxies can be relatively poor in outcrops and many proxies have been calibrated in
depositional environments not commonly preserved in the geological record. For instance, in
Chapter 3 I examine how brGDGT-based temperature indices (MBT'
5Me
) change down profile,
as these have traditionally only been calibrated in soil surfaces and we lack understanding of the
controls on MBT'
5Me
deeper in soils, which is relevant to correctly interpret MBT'
5Me
in
outcropping paleosols. Exposure to oxygen, microbes, and high temperatures (volcanic eruptions
and deposits) alter the signal of a proxy, although we can attempt to counter this by employing
multiple proxies in unison to detect the signal in the noise.
The changing nature of the sedimentary deposition can result in fluvial, aeolian, lacustrine, and
soil sediments being present in the same outcrop, with some sedimentary depositional
environments overprinting on older deposits. As such, establishing clear climatic and
environmental information from proxies is challenged by the potential changes in (temporal and
spatial) scales that they were integrating, differences in diagenesis and sedimentation rates
obscuring proxy signal interpretation, and unknowns concerning proxy formation and deposition.
Again, multi-proxy approaches can alleviate some of these challenges, as demonstrated by
Saylor et al., (2019) who employed both pollen and plant wax to constrain paleovegetation
change through time across Pliocene fluviolacustrine sediments in Woranso-Mille, Ethiopia.
5
Many traditional and widely deployed proxies measured in outcrops can only be formed in
certain sediments. For example, pedogenic carbonates can only form in soils between 200-2500
mm/yr of precipitation and which have a wet and dry season (Zamanian et al., 2016). This
imparts biases to climatic reconstructions based solely on this proxy and prevents its application
across multiple sediment types. Biomarker proxies are not constrained to the same extent by
sediment type, although the interpretations of the proxy signal can depend heavily on the
sediment type.
One goal of terrestrial paleoreconstructions should be linking all three of these different proxy
archive locations together to tell rich paleoenvironmental stories across multiple spatial and
temporal scales. For instance, in this dissertation, I will attempt to establish in Chapter 6 to what
extent outcrop-based terrestrial paleoenvironmental reconstructions in the Afar triangle agree
with terrestrial reconstructions based on offshore marine cores. This allows us to see if Woranso-
Mille, Ethiopia, is showing the same paleoenvironmental trend as the broader regional trend
through time, the amplitude of the trend, and the degree of heterogeneity that could be present in
marine core source regions. Additionally, in Chapter 6 I will compare marine records of
temperature change with lacustrine records of temperature and hydroclimate change and
determine to what extent both are connected.
Pleistocene paleoenvironmental records are comparatively easier to produce since many of the
challenges we face in the Pliocene are either eliminated or greatly reduced. For instance, there
are many more Pleistocene than Pliocene lake deposits, and as such the spatial scale available to
us is finer (e.g. Pound et al., 2014b; Reheis et al., 2014). Additionally, proxy preservation is
6
almost uniformly better in the Pleistocene and new types of proxy archives (e.g., ice cores,
speleothem records, rat middens) become available for us to extract proxy information from.
Dating is also easier, with tools such as carbon dating (for sediments < 45 kyr) and optically
stimulated luminescence dating (<400 kyr) becoming available in addition to all the other dating
methods discussed prior. The contrast between Pliocene and Pleistocene studies is clear when
comparing Chapters 4 and 5 of this thesis. Both are from Searles Lake sediments, but in Chapter
4 I discuss Pleistocene climate and environmental changes whereas Chapter 5 focuses on the
Pliocene. Chapter 4 (and associated studies of SLAPP-17 sediments) makes use of multiple
environmental proxies (biomarkers, pollen, salt thermometry) as well as a relatively well-
constrained age model U/Th dating. However, in the Pliocene sediments of Chapter 5, we only
have biomarkers augmented by qualitative grain size information in addition to a lower
resolution paleomagnetic age model (Liddicoat et al., 1980).
1.2 Biomarker proxies
1.2.1 Plant wax
The epicuticular layer of plants protects them from disease, weather, insect attack, and water loss
(Koch et al., 2009). It is the first barrier between the plant and its outside environment, and it is
crucial to its survival. Composed primarily of long-chain n-alkanes, n-alkanoic acids, and n-
alkanols, these molecules are relatively resistant to degradation and are among the last
compounds from a plant to exist post-plant death (Eglington and Hamilton, 1963). Plant waxes
can be preserved for millions of years in sedimentary rocks and paleoenvironmental signals can
still be extracted following exposure to relatively high temperatures (200-300 °C) (Diefendorf et
al., 2015b). In the following studies, we make use of three aspects of plant waxes to discern
7
climate and environmental information: their molecular distributions (Eglinton and Hamilton,
1967) and their carbon isotopic composition, and hydrogen isotopic composition (Sessions,
2006). Molecular distributions involve studying the relative abundance of different chain lengths
present in either one or multiple, plant wax compound classes. The most common measures of
plant wax distributions are the ratio of even to odd (or odd to even) chain lengths and the average
chain length. Whilst these have some usage in assessing the degree to which plant waxes have
been degraded (CPI close to 1 = more petrogenic waxes, e.g. Jeng, 2006) they offer little
information besides this, even in the face of repeated attempts to find correlations to
environmental parameters. Part of the reason for this is that even closely related plants can have
very different chain length distributions. For instance, Juniper and Pine are both members of the
Conifer family, and yet the alkane chain length distribution of Pine is more reminiscent of a lake
dwelling macrophyte plant than a Juniper (Gao et al., 2011; Diefendorf et al., 2015a). Another
potential challenge to the usage of plant wax distributions is the taphonomic filter through which
plant waxes travel during their diagenesis. Plant waxes are degraded by microbes that break
down long-chain length molecules before consumption and can add shorter chain lengths to the
sediment wax pool through the production of their own short to mid-chain length lipids (Wu et
al., 2019 and references therein). Combined, these processes can obfuscate any useful
information contained in sedimentary plant wax chain length distributions. As such, new
techniques are needed to tease out phylogenetic information from the noise. I have proposed one
potential technique in Chapter 2 (Peaple et al., 2021b), which involves training machine learning
algorithms to detect plant identities from chain length information from multiple compound
classes of plant wax, even in the presence of substantial noise.
8
Carbon and hydrogen isotopes of plant waxes are comparatively more challenging to obtain yet
offer greater potential as paleoenvironmental and paleoclimatic proxies. The proportion of
13
C to
12
C in plant waxes is dependent upon the photosynthetic pathway of the plant (C
3
vs C
4
vs CAM)
(Collister et al., 1994), the moisture availability of the environment (Diefendorf et al., 2010), the
CO
2
concentration of the atmosphere (Körner et al., 1991; Ehleringer et al., 1992), and salinity of
soil (Sandquist and Ehleringer, 1995). Upon wax transportation and deposition, biological
degradation (G. Li et al., 2018; Wu et al., 2019) can further alter plant waxes in addition to
inputs from petrogenic sources (Li et al., 2009). Once lithified, plant wax δ
13
C is thought to be
relatively stable on geological timescales, although once exposed to temperatures >300 °C, δ
13
C
can be affected (Diefendorf et al., 2015b). Interpreting the main cause of δ
13
C measured from
sedimentary deposits is simplified by possessing some rudimentary understanding of the
location, time, and geological history of the deposits, etc. For instance, in a high latitude
Holocene wax deposit (pre-industrial), we can rule out contributions of large changes in CO
2
and
the presence of C
4
plants contributing to the δ
13
C signal. As such, changes in moisture
availability are the most likely influence on the net δ
13
C wax signal. However, the range in δ
13
C
wax in C
3
plants living in identical environmental conditions can be large. For instance, Juniper
has a relatively negative δ
13
C (-32‰) compared to pines (-26‰) (Diefendorf et al., 2015a) so a
large landscape-scale replacement of one conifer with another has the potential to alter the net
wax signal preserved in a sediment archive. In the environments I have worked in during my
PhD, the interpretation is not quite as simple. These environments have radically changing
precipitation-evaporation amounts, vegetation changes (including C
4
vegetation), and changes in
the mean atmospheric composition of CO
2
. As such, information gained from this proxy is by
itself ambiguous, and any interpretation should consider other available proxies to constrain any
9
extraneous influencing factor on the δ
13
C signal. Compound-specific hydrogen isotopes of plants
are predominantly affected by the isotopic composition of the soil water that the plants consume,
the salinity of the environment, transpiration loss of water as it moved within the plant, and
biosynthetic fractionation in the production of the plant wax (Sachse et al., 2012 and references
therein).
Studies in modern environments have demonstrated that the hydrogen isotopic composition of
plant waxes is highly correlated to the isotopic composition of precipitation used by the plant. As
such, measuring the isotopic composition of fossil waxes can provide hydrological information
during the time the wax was produced. The hydrogen isotopic composition of plants suffers from
the same problems facing the δ
13
C composition, namely a large biological fractionation which
occurs during the production of the wax which obfuscates the real signal scientists target, the
isotopic composition of precipitation. However, whilst this biological fractionation can differ
greatly between different plant groups (Sachse et al., 2012), when waxes from different plants
are mixed in sediment, the different biological fractions (often but not always) average out and
reduce the complication of accounting for this effect in our measured δD of plant wax (Sachse et
al., 2012). Another potential complication in interpreting δD (in addition to δ
13
C and chain
length information) is the contribution from non-terrestrial plant sources. We will discuss this
challenge, especially in the context of Chapters 2 and 4, which aim to reconstruct the vegetation
and precipitation change over the past 200 kyr from Searles Lake, USA. Both n-alkanes and n-
alkanoic acids were measured in the same samples extracted from the SLAPP-17 core for δD.
However, both compound classes produced markedly different δD values that did not correlate
with each other which, along with machine learning of plant wax chain length information,
10
suggests that the n-alkanoic acids were representing a mixed source of conifers and aquatic
sourced inputs.
This work serves as a reminder of the utility of measuring multiple wax compound classes and
examining the wax chain length distributions to tease out the confounding influences on our
proxies. We also must be mindful of the impact D/H exchange can have on the isotopic
composition of plant waxes. Whilst high temperatures are unlikely to pose risks for the
interpretation of δ
13
C, δD can be affected by the isotopic exchange between plant waxes and
water-mediated by free radical reactions during organic matter maturation (Pedentchouk et al.,
2006). Given that we are usually studying thermally immature sediments this is not often an
issue and it would be possible to screen samples for a particularly low CPI to check for evidence
of thermal alteration.
1.2.2 GDGTs
Glycerol dibiphytanyl glycerol tetraethers (GDGTs) are membrane lipids produced by both
archaea and bacteria. These compounds are commonly divided into two groups that are
structurally different and sourced from different producers. BrGDGTs are membrane lipids
composed of a branched alkyl chain with 4 to 6 methyl groups (De Jonge et al., 2014b, p.98) that
are produced by bacteria and are found in a variety of environments including soils, peats, lakes,
and coastal marine environments. However, the identity of the bacteria which produce brGDGTs
is still mostly unknown (Weijers et al., 2006, 2009; Sinninghe Damsté et al., 2011, 2014, 2018).
Branched glycerol dialkyl glycerol tetraether (brGDGTs) lipids are reliable recorders of
temperature in modern surface soils and peats (Weijers et al., 2007; De Jonge et al., 2014a; Naafs
et al., 2017; Dearing Crampton-Flood et al., 2020) as well as lakes (Martínez-Sosa et al., 2021).
11
This is likely due to bacteria changing the degree of methylation of membrane lipids under
different temperature conditions to adjust the rigidity of their membranes (Weijers et al., 2007;
Naafs et al., 2021). In the studies that follow I apply the brGDGT proxy in both soils and lakes to
reconstruct modern and paleo temperatures. Sourcing of brGDGTs in lakes can have a significant
impact on the temperatures reconstructed from the MBT'
5Me
proxy. BrGDGTs produced in soils
and lakes have different relationships to temperature, expressed through the MBT'
5Me
index. If a
significant % of brGDGTs is soil-derived this can result in temperature reconstructions that are
too high. Whilst this appears to not be a large problem on a global scale, it can occur when the
production of brGDGT-producing bacteria is suppressed in high alkaline and saline
environments, for instance, modern Mono Lake (Martínez-Sosa et al., 2021). We must be
careful, therefore, when interpreting MBT'
5Me
temperatures from Searles Lake (Chapter 4) which
likely had similar alkalinity and salinity conditions during episodes of the past 200 kyr. One way
to constrain the inputs of soil vs lake brGDGTs is to compare the overall distribution of brGDGT
molecules of samples with compilations of the lake and soils brGDGTs, to examine whether the
sample is closer to a soil or a lake distribution (when screened for comparable temperature
conditions) (Martínez-Sosa et al., 2021). This was performed in Chapter 4, and the brGDGTs in
Searles Lake sediments more closely align with the global lake compilation. Other more
elaborate methods of constraining sourcing are being developed, including machine learning
approaches, following in the footsteps of similar approaches pioneered on plant wax taxonomic
identification (Chapter 2). The sourcing is usually simpler to determine in soils, with the majority
of brGDGTs assumed to be soil-derived. However, care must be taken in soils located in flood
plains which could receive inputs from distal produced soil brGDGTs in addition to brGDGTs
produced in the river (Tierney and Russell, 2009). Additionally, further challenges associated
12
with changing environmental composition and bacterial community composition complicates the
application of temperature and pH proxies in soils. Whilst the surface layer of soils is known to
respond to temperature and pH, little is known about the behavior of these proxies at depth. This
knowledge is crucial however when applying the brGDGT proxies in past sediments to
reconstruct pH and temperature. Often in outcrops the surface layer of the paleosol is missing
(Kraus, 1999), so opportunistic sampling of deeper paleosol strata is common. Additionally,
scientists may be guided in the sampling strategy of paleosols by the appearance of pedogenic
carbonate nodules (which typically form in the Bk layer of a soil) and thus attempt to collect
organic biomarker samples adjacent to a pedogenic carbonate sample given that both carbonate
and brGDGTs can provide temperature information. As such, in Chapter 3 we sampled a suite of
paleosols at a range of depths to assess the response of brGDGTs to changes in depth, in the
hope of guiding future paleoenvironmental studies.
IsoGDGTs are composed of two C
40
isoprenoid chains with several cyclopentane and
cyclohexane rings connected by ether bonds to two terminal glycerol groups (Li et al., 2016,
p.114). IsoGDGTs are membrane lipids produced by archaea and occur in broadly the same
environment as bacterial-derived brGDGTs (Schouten et al., 2013 and references therein). The
relative abundance of different isoGDGTs can be altered by environmental temperature, methane
abundance, oxygen abundance, light availability, and nutrient availability (Schouten et al., 2013).
Classically, the isoGDGTs have been useful to investigate past temperature changes in marine
and large lake settings (e.g. Kim et al., 2008; Tierney et al., 2010). The TEX
86
proxy measures
the relative abundance of crenarchaeol, which is thought to be exclusively produced by
Thaumarcheotal (Schouten et al., 2013). However, one of the challenges in applying TEX
86
in
13
the lakes present in this dissertation is that crenarchaeol can be sourced not only from
Thaumarcheota living in water but also from Thaumarcheota present in catchment soils (Pester et
al., 2011). One traditional way of screening for soil inputs is through the BIT index which
measures the relative abundance of brGDGT to isoGDGT, under the assumption that the
brGDGT is solely sourced from the soils. A high BIT index would suggest that the overall
GDGT budget of the lake is dominated by soil inputs. However, this metric has fallen out of
favor recently given the discovery that significant brGDGT quantities can be produced from
lakes' bacterial communities (e.g. Tierney and Russell, 2009; Bechtel et al., 2010; Martínez-Sosa
and Tierney, 2019; Baxter et al., 2021). Regardless, the low abundance of crenarchaeol in many
modern lakes makes source attribution challenges, and thus the interpretation of Tex
86
is difficult
(Sinninghe Damsté et al., 2012a; Baxter et al., 2021). Only in one sample from Pleistocene
Searles Lake sediments (from the study presented in Chapter 4) was there a sufficiently high
abundance of crenarchaeol to interpret the TEX
86
proxy as capturing a lake temperature signal,
although this sample proved a useful check against the more widely applied brGDGT based
temperature reconstruction used in the study.
1.3 Dissertation publications
This dissertation contains the following published and in-preparation manuscripts:
Peaple, M. D., Tierney, J. E., McGee, D., Lowenstein, T. K., Bhattacharya, T., & Feakins, S. J.
(2021). Identifying plant wax inputs in lake sediments using machine learning. Organic
Geochemistry, 156, 104222. (Chapter 2)
The co-authors contributed to project conceptualization, methodology, funding acquisition and
reviewing of the manuscript.
14
Peaple, M.D., Beverly, E.J., Garza, B., Baker, S., Levin, N.E., Tierney, J.E., Häggi, C., Feakins,
S.J. (2022). Identifying the drivers of GDGT distributions in alkaline soil profiles within the
Serengeti ecosystem. Organic Geochemistry, 169, 104433 (Chapter 3)
The co-authors contributed to project conceptualization, methodology, funding acquisition and
reviewing of the manuscript.
Peaple, M.D., Bhattacharya, T., Lowenstein, T.K., McGee, D., Olson, K.J., Stroup, J.S., Tierney,
J.E., Feakins, S.J (in review). Biomarker and pollen evidence for late Pleistocene pluvials in the
Mojave Desert. (Chapter 4)
The co-authors contributed to project conceptualization, methodology, funding acquisition and
reviewing of the manuscript.
Peaple, M.D., Bhattacharya, T., Feng, R., Lowenstein, T.K., Tierney, J.E., Feakins, S.J (in prep).
Warm and wet Pliocene southwestern North America: evidence from Searles Lake (Chapter 5)
The co-authors contributed to project conceptualization, methodology, funding acquisition and
reviewing of the manuscript.
Peaple, M.D., Langston, J., Levin, N. E., Haile-Selassie, Y., Saylor, B. Z., Tierney, J.E., Feakins,
S.J (in prep). Paleoenvironmental reconstruction of Pliocene hominin habitats at Woranso-Mille
(Chapter 6).
The co-authors contributed to project conceptualization, methodology, funding acquisition and
reviewing of the manuscript.
References
Adam, D.P., Sarna-Wojcicki, A.M., Rieck, H.J., Bradbury, J.P., Dean, W.E., Forester, R.M.,
1989. Tulelake, California: The last 3 million years. Palaeogeography, Palaeoclimatology,
15
Palaeoecology 72, 89–103.
Bartusek, S.T., Seo, H., Ummenhofer, C.C., Steffen, J., 2021. The role of nearshore air‐sea
interactions for landfalling atmospheric rivers on the US West Coast. Geophysical Research
Letters 48, e2020GL091388.
Baxter, A.J., van Bree, L.G.J., Peterse, F., Hopmans, E.C., Villanueva, L., Verschuren, D.,
Sinninghe Damsté, J.S., 2021. Seasonal and multi-annual variation in the abundance of
isoprenoid GDGT membrane lipids and their producers in the water column of a meromictic
equatorial crater lake (Lake Chala, East Africa). Quaternary Science Reviews 273, 107263.
Bechtel, A., Smittenberg, R.H., Bernasconi, S.M., Schubert, C.J., 2010. Distribution of branched
and isoprenoid tetraether lipids in an oligotrophic and a eutrophic Swiss lake: Insights into
sources and GDGT-based proxies. doi:10.1016/j.orggeochem.2010.04.022
Burls, N.J., Fedorov, A. V, 2017. Wetter subtropics in a warmer world: Contrasting past and
future hydrological cycles. Proceedings of the National Academy of Sciences 114, 12888–
12893.
Chou, C., Neelin, J.D., 2004. Mechanisms of global warming impacts on regional tropical
precipitation. Journal of climate 17, 2688–2701.
Collister, J.W., Rieley, G., Stern, B., Eglinton, G., Fry, B., 1994. Compound-specific δ 13C
analyses of leaf lipids from plants with differing carbon dioxide metabolisms. Organic
geochemistry 21, 619–627.
De Jonge, C., Hopmans, E.C., Zell, C.I., Kim, J.H., Schouten, S., Sinninghe Damsté, J.S., 2014a.
Occurrence and abundance of 6-methyl branched glycerol dialkyl glycerol tetraethers in
soils: Implications for palaeoclimate reconstruction. Geochimica et Cosmochimica Acta
141, 97–112.
16
De Jonge, C., Stadnitskaia, A., Hopmans, E.C., Cherkashov, G., Fedotov, A., Sinninghe Damsté,
J.S., 2014b. In situ produced branched glycerol dialkyl glycerol tetraethers in suspended
particulate matter from the Yenisei River, Eastern Siberia. Geochimica et Cosmochimica
Acta 125, 476–491.
Dearing Crampton-Flood, E., Peterse, F., Munsterman, D., Sinninghe Damsté, J.S., 2018. Using
tetraether lipids archived in North Sea Basin sediments to extract North Western European
Pliocene continental air temperatures. Earth and Planetary Science Letters 490, 193–205.
Dearing Crampton-Flood, E., Tierney, J.E., Peterse, F., Kirkels, F.M.S.A., Sinninghe Damsté,
J.S., 2020. BayMBT: A Bayesian calibration model for branched glycerol dialkyl glycerol
tetraethers in soils and peats. Geochimica et Cosmochimica Acta 268, 142–159.
Deino, A.L., Scott, G.R., Saylor, B., Alene, M., Angelini, J.D., Haile-Selassie, Y., 2010.
40Ar/39Ar dating, paleomagnetism, and tephrochemistry of Pliocene strata of the hominid-
bearing Woranso-Mille area, west-central Afar Rift, Ethiopia. Journal of Human Evolution
58, 111–126.
Dekens, P.S., Ravelo, A.C., McCarthy, M.D., 2007. Warm upwelling regions in the Pliocene
warm period. Paleoceanography 22. doi:10.1029/2006PA001394
Diefendorf, A.F., Leslie, A.B., Wing, S.L., 2015a. Leaf wax composition and carbon isotopes
vary among major conifer groups. Geochimica et Cosmochimica Acta 170, 145–156.
Diefendorf, A.F., Mueller, K.E., Wing, S.L., Koch, P.L., Freeman, K.H., 2010. Global patterns in
leaf 13C discrimination and implications for studies of past and future climate. Proceedings
of the National Academy of Sciences of the United States of America 107, 5738–43.
Diefendorf, A.F., Sberna, D.T., Taylor, D.W., 2015b. Effect of thermal maturation on plant-
derived terpenoids and leaf wax n-alkyl components. Organic Geochemistry 89–90, 61–70.
17
Dragoni, M., D’Onza, F., Tallarico, A., 2002. Temperature distribution inside and around a lava
tube. Journal of Volcanology and Geothermal Research 115, 43–51.
Eglington, G., Hamilton, R.J., 1963. Thedistribution of Alkanes. Chemical Plant Taxonomy, T.
Swain, ed. Academic Press, London and New York. p.
Eglinton, G., Hamilton, R.J., 1967. Leaf epicuticular waxes. Science 156, 1322–1335.
Ehleringer, J.R., Phillips, S.L., Comstock, J.P., 1992. Seasonal Variation in the Carbon Isotopic
Composition of Desert Plants. Functional Ecology 6, 396.
Fontijn, K., Lachowycz, S.M., Rawson, H., Pyle, D.M., Mather, T.A., Naranjo, J.A., Moreno-
Roa, H., 2014. Late Quaternary tephrostratigraphy of southern Chile and Argentina.
Quaternary Science Reviews 89, 70–84.
Fowler, A.M., Hennessy, K.J., 1995. Potential impacts of global warming on the frequency and
magnitude of heavy precipitation. Natural Hazards 11, 283–303.
Gao, L., Hou, J., Toney, J., MacDonald, D., Huang, Y., 2011. Mathematical modeling of the
aquatic macrophyte inputs of mid-chain n-alkyl lipids to lake sediments: Implications for
interpreting compound specific hydrogen isotopic records. Geochimica et Cosmochimica
Acta 75, 3781–3791.
Giorgi, F., Raffaele, F., Coppola, E., 2019. The response of precipitation characteristics to global
warming from climate projections. Earth System Dynamics 10, 73–89.
Haile-Selassie, Y., Gibert, L., Melillo, S.M., Ryan, T.M., Alene, M., Deino, A., Levin, N.E.,
Scott, G., Saylor, B.Z., 2015. New species from Ethiopia further expands Middle Pliocene
hominin diversity. Nature 521, 483–488.
Haile-Selassie, Y., Latimer, B.M., Alene, M., Deino, A.L., Gibert, L., Melillo, S.M., Saylor,
B.Z., Scott, G.R., Lovejoy, C.O., 2010. An early Australopithecus afarensis postcranium
18
from Woranso-Mille, Ethiopia. Proceedings of the National Academy of Sciences 107,
12121–12126.
Huang, X., Swain, D.L., Hall, A.D., 2020. Future precipitation increase from very high
resolution ensemble downscaling of extreme atmospheric river storms in California.
Science Advances 6, eaba1323.
Jeng, W.L., 2006. Higher plant n-alkane average chain length as an indicator of petrogenic
hydrocarbon contamination in marine sediments. Marine Chemistry 102, 242–251.
Kim, J.H., Schouten, S., Hopmans, E.C., Donner, B., Sinninghe Damsté, J.S., 2008. Global
sediment core-top calibration of the TEX86 paleothermometer in the ocean. Geochimica et
Cosmochimica Acta 72, 1154–1173.
Koch, K., Bhushan, B., Barthlott, W., 2009. Multifunctional surface structures of plants: An
inspiration for biomimetics. Progress in Materials Science 54, 137–178.
Körner, C., Farquhar, G.D., Wong, S.C., 1991. Carbon isotope discrimination by plants follows
latitudinal and altitudinal trends. Oecologia 88, 30–40.
Kraus, M.J., 1999. Paleosols in clastic sedimentary rocks: their geologic applications. Earth-
Science Reviews 47, 41–70.
Li, C., Sessions, A.L., Kinnaman, F.S., Valentine, D.L., 2009. Hydrogen-isotopic variability in
lipids from Santa Barbara Basin sediments. Geochimica et Cosmochimica Acta 73, 4803–
4823.
Li, G., Li, L., Tarozo, R., Longo, W.M., Wang, K.J., Dong, H., Huang, Y., 2018. Microbial
production of long-chain n-alkanes: Implication for interpreting sedimentary leaf wax
signals. Organic Geochemistry 115, 24–31.
Li, J., Pancost, R.D., Naafs, B.D.A., Yang, H., Zhao, C., Xie, S., 2016. Distribution of glycerol
19
dialkyl glycerol tetraether (GDGT) lipids in a hypersaline lake system. Organic
Geochemistry 99, 113–124.
Liddicoat, J.C., Opdyke, N.D., Smith, G.I., 1980. Palaeomagnetic polarity in a 930-m core from
Searles Valley, California. Nature 286, 22–25.
Lupien, R.L., Russell, J.M., Yost, C.L., Kingston, J.D., Deino, A.L., Logan, J., Schuh, A.,
Cohen, A.S., 2019. Vegetation change in the Baringo Basin, East Africa across the onset of
Northern Hemisphere glaciation 3.3–2.6 Ma. Palaeogeography, Palaeoclimatology,
Palaeoecology 109426.
Magill, C.R., Ashley, G.M., Freeman, K.H., 2013. Ecosystem variability and early human
habitats in eastern Africa. Proceedings of the National Academy of Sciences of the United
States of America 110, 1167–1174.
Martínez-Sosa, P., Tierney, J.E., 2019. Lacustrine brGDGT response to microcosm and
mesocosm incubations. Organic Geochemistry 127, 12–22.
Martínez-Sosa, P., Tierney, J.E., Stefanescu, I.C., Dearing Crampton-Flood, E., Shuman, B.N.,
Routson, C., 2021. A global Bayesian temperature calibration for lacustrine brGDGTs.
Geochimica et Cosmochimica Acta 305, 87–105.
Melillo, S.M., Gibert, L., Saylor, B.Z., Deino, A., Alene, M., Ryan, T.M., Haile-Selassie, Y.,
2021. New Pliocene hominin remains from the Leado Dido’a area of Woranso-Mille,
Ethiopia. Journal of Human Evolution 153, 102956.
Molnar, P., Cane, M.A., 2002. El Niño’s tropical climate and teleconnections as a blueprint for
pre-Ice Age climates. Paleoceanography 17, 11–1.
Naafs, B.D.A., Gallego-Sala, A. V., Inglis, G.N., Pancost, R.D., 2017. Refining the global
branched glycerol dialkyl glycerol tetraether (brGDGT) soil temperature calibration.
20
Organic Geochemistry 106, 48–56.
Naafs, B.D.A., Oliveira, A.S.F., Mulholland, A.J., 2021. Molecular dynamics simulations
support the hypothesis that the brGDGT paleothermometer is based on homeoviscous
adaptation. Geochimica et Cosmochimica Acta 312, 44–56.
Peaple, M.D., Tierney, J.E., McGee, D., Lowenstein, T.K., Bhattacharya, T., Feakins, S.J., 2021.
Identifying plant wax inputs in lake sediments using machine learning. Organic
Geochemistry 156, 104222.
Pedentchouk, N., Freeman, K.H., Harris, N.B., 2006. Different response of δD values of n-
alkanes, isoprenoids, and kerogen during thermal maturation. Geochimica et Cosmochimica
Acta 70, 2063–2072.
Pester, M., Schleper, C., Wagner, M., 2011. The Thaumarchaeota: an emerging view of their
phylogeny and ecophysiology. Current Opinion in Microbiology 14, 300–306.
Pound, M.J., Tindall, J., Pickering, S.J., Haywood, A.M., Dowsett, H.J., Salzmann, U., 2014a.
Late Pliocene lakes and soils: A global data set for the analysis of climate feedbacks in a
warmer world. Climate of the Past 10, 167–180.
Pound, M.J., Tindall, J., Pickering, S.J., Haywood, A.M., Dowsett, H.J., Salzmann, U., 2014b.
Late Pliocene lakes and soils: a global data set for the analysis of climate feedbacks in a
warmer world. Climate of the Past 10, 167–180.
Reheis, M.C., Adams, K.D., Oviatt, C.G., Bacon, S.N., 2014. Pluvial lakes in the Great Basin of
the western United States—a view from the outcrop. Quaternary Science Reviews 97, 33–
57.
Sachse, D., Billault, I., Bowen, G.J., Chikaraishi, Y., Dawson, T.E., Feakins, S.J., Freeman,
K.H., Magill, C.R., Mcinerney, F.A., Van Der Meer, M.T.J.J., Polissar, P., Robins, R.J.,
21
Sachs, J.P., Schmidt, H.-L., Sessions, A.L., White, J.W.C., West, J.B., Kahmen, A., 2012.
Molecular Paleohydrology: Interpreting the Hydrogen-Isotopic Composition of Lipid
Biomarkers from Photosynthesizing Organisms 40. doi:10.1146/annurev-earth-042711-
105535
Sandquist, D.R., Ehleringer, J.R., 1995. Carbon isotope discrimination in the C₄ shrub Atriplex
confertifolia along a salinity gradient. The Great Basin Naturalist 135–141.
Saylor, B.Z., Gibert, L., Deino, A., Alene, M., Levin, N.E., Melillo, S.M., Peaple, M.D., Feakins,
S.J., Bourel, B., Barboni, D., Novello, A., Sylvestre, F., Mertzman, S.A., Haile-Selassie, Y.,
2019. Age and context of mid-Pliocene hominin cranium from Woranso-Mille, Ethiopia.
Nature 573. doi:10.1038/s41586-019-1514-7
Schouten, S., Hopmans, E.C., Sinninghe Damsté, J.S., 2013. The organic geochemistry of
glycerol dialkyl glycerol tetraether lipids: A review. Organic Geochemistry 54, 19–61.
Sessions, A.L., 2006. Isotope‐ratio detection for gas chromatography. Journal of Separation
Science 29, 1946–1961.
Sinninghe Damsté, J.S., Ossebaar, J., Schouten, S., Verschuren, D., 2012. Distribution of
tetraether lipids in the 25-ka sedimentary record of Lake Challa: extracting reliable TEX86
and MBT/CBT palaeotemperatures from an equatorial African lake. Quaternary Science
Reviews 50, 43–54.
Sinninghe Damsté, J.S., Rijpstra, W.I.C., Foesel, B.U., Huber, K.J., Overmann, J., Nakagawa, S.,
Kim, J.J., Dunfield, P.F., Dedysh, S.N., Villanueva, L., 2018. An overview of the
occurrence of ether- and ester-linked iso-diabolic acid membrane lipids in microbial
cultures of the Acidobacteria: Implications for brGDGT paleoproxies for temperature and
pH. Organic Geochemistry 124, 63–76.
22
Sinninghe Damsté, J.S., Rijpstra, W.I.C., Hopmans, E.C., Foesel, B.U., Wüst, P.K., Overmann,
J., Tank, M., Bryant, D.A., Dunfield, P.F., Houghton, K., Stott, M.B., 2014. Ether- and
ester-bound iso-diabolic acid and other lipids in members of Acidobacteria subdivision 4.
Applied and Environmental Microbiology 80, 5207–5218.
Sinninghe Damsté, J.S., Rijpstra, W.I.C., Hopmans, E.C., Weijers, J.W.H., Foesel, B.U.,
Overmann, J., Dedysh, S.N., 2011. 13,16-Dimethyl octacosanedioic acid (iso-Diabolic
Acid), a common membrane-spanning lipid of Acidobacteria subdivisions 1 and 3. Applied
and Environmental Microbiology 77, 4147–4154.
Spoor, F., Gunz, P., Neubauer, S., Stelzer, S., Scott, N., Kwekason, A., Dean, M.C., 2015.
Reconstructed Homo habilis type OH 7 suggests deep-rooted species diversity in early
Homo. Nature 2015 519:7541 519, 83–86.
Tierney, J.E., Mayes, M.T., Meyer, N., Johnson, C., Swarzenski, P.W., Cohen, A.S., Russell,
J.M., 2010. Late-twentieth-century warming in Lake Tanganyika unprecedented since AD
500. Nature Geoscience 2010 3:6 3, 422–425.
Tierney, J.E., Russell, J.M., 2009. Distributions of branched GDGTs in a tropical lake system:
Implications for lacustrine application of the MBT/CBT paleoproxy. Organic Geochemistry
40, 1032–1036.
Weijers, J.W.H., Panoto, E., van bleijswijk, J., Schouten, S., Rijpstra, W.I.C., Balk, M., Stams,
A.J.M., Damsté, J.S.S., 2009. Constraints on the Biological Source(s) of the Orphan
Branched Tetraether Membrane Lipids. http://dx.doi.org/10.1080/01490450902937293 26,
402–414.
Weijers, J.W.H., Schouten, S., van den Donker, J.C., Hopmans, E.C., Sinninghe Damsté, J.S.,
2007. Environmental controls on bacterial tetraether membrane lipid distribution in soils.
23
Geochimica et Cosmochimica Acta 71, 703–713.
Weijers, J.W.H.H., Schouten, S., Hopmans, E.C., Geenevasen, J.A.J.J., David, O.R.P.P.,
Coleman, J.M., Pancost, R.D., Sinninghe Damste, J.S., Sinninghe Damsté, J.S., 2006.
Membrane lipids of mesophilic anaerobic bacteria thriving in peats have typical archaeal
traits. Environmental Microbiology 8, 648–657.
Wu, M.S., West, A.J., Feakins, S.J., 2019. Tropical soil profiles reveal the fate of plant wax
biomarkers during soil storage. Organic Geochemistry 128, 1–15.
Zamanian, K., Pustovoytov, K., Kuzyakov, Y., 2016. Pedogenic carbonates: Forms and
formation processes. Earth-Science Reviews 157, 1–17.
24
Chapter 2: Identifying plant wax inputs in lake sediments using machine
learning
This chapter was published in Organic Geochemistry in 2021 as:
Peaple, M. D., Tierney, J. E., McGee, D., Lowenstein, T. K., Bhattacharya, T., & Feakins, S. J.
(2021). Identifying plant wax inputs in lake sediments using machine learning. Organic
Geochemistry, 156, 104222.
Abstract
This study aims to evaluate whether machine learning techniques can be successfully applied to
process the complex information contained within the molecular abundance distributions of plant
wax n-alkane and n-alkanoic acid homologous series. We trained five vegetation identification
models using plant wax chain length distributions from modern plants in the Mojave Desert
(hyperarid) and the San Bernardino Mountains (conifer forest) and previously published data for
macrophytes from Blood Pond (USA) and Mt Kenya (Kenya). All vegetation identification
models proved accurate (mean classification accuracy = 0.81) at classifying the modern plant
wax chain length distributions into desert plants, conifer, and macrophyte categories. We then
applied the models to fossil waxes extracted from a 76 m lacustrine sediment core drilled in
Searles Valley, CA with an approximate age range of 10 to 150 kyrs (SLAPP-SRLS17) to
reconstruct the proportion of desert plants, conifer woodland, and lake vegetation. We compared
our machine learning models with a previously published linear mixing model and validated our
modeled plant type distributions by comparing the results with the archaeol caldarchaeol
ecometric (ACE), a proxy for lake salinity, measured in the same core. We found a moderate
positive correlation (r = 0.40) between the modeled desert plant proportion and high lake salinity
25
in our models as well as a negative correlation (r = –0.45) between modeled macrophyte plants
and ACE, validating the ability of the machine learning techniques to detect both xeric and
macrophyte plant communities. Our results suggest that machine learning of plant wax molecular
abundance distributions has the potential to reconstruct past plant communities, given
information from two compound classes and highly differentiated vegetation types.
Keywords: machine learning; plant wax; ACE; glacial; xeric
2.1 Introduction
Plant waxes include long-chain n-alkyl molecules that are found on the outer surface of plant
leaves (Eglinton and Hamilton, 1967). The waxes limit water loss (Koch et al., 2009) and protect
leaves from UV light (Shepherd and Griffiths, 2006), insects, and fungal attacks (Koch et al.,
2009). Plant wax chain length distributions have long been of interest (Eglinton et al., 1962),
although their use as a taxonomic diagnostic tool is limited (Rommerskirchen et al., 2006).
Quantitative uses of chain length distributions focus on averages and ratios of different chain
lengths. These parameters have some predictive power in determining the growing season
temperature of plants (Bush and McInerney, 2015), but many plants have similar chain length
averages (Bush and McInerney, 2013) and thus these parameters are not suitable for plant
classification. Past biomarker-based vegetation modeling efforts (e.g., Jansen et al., 2010, 2013;
Gao et al., 2011) have used linear regression approaches to reconstruct vegetation in peat and
lacustrine archives, validating these results with independent records has proved challenging.
Machine learning has the power to extract taxonomic information from chain length
distributions beyond the qualitative analysis of bivariate plots or reducing the chain length
distribution to a single number. Machine learning techniques have been successfully applied to a
variety of geochemical data sets in igneous geology (Ueki et al., 2018), aqueous geochemistry
26
(Engle and Brunner, 2019) as well as organic geochemistry, where they were applied to
alkenones (Zheng et al., 2019). We seek to evaluate the usefulness of machine learning models to
plant taxonomic identification using both modern and ancient leaf wax molecular abundance
distributions.
Here we develop and evaluate five machine learning models trained on n-alkane and n-
alkanoic acid relative abundances sampled from plants in the Mojave Desert, the San Bernardino
forest, CA, as well as previously published data for aquatic macrophytes from Blood Pond, MA
(Gao et al., 2011) and Mt Kenya, Kenya (Ficken et al., 2000). We group plants into categories
“desert plants” (600 – 2000 m asl), “conifer” (2000 – 3500 m asl), and aquatic macrophytes, and
train the models to predict the plant category proportions based on their plant wax molecular
abundance distributions. We compare the different models’ ability to classify both test data as
well as test data with added noise. We then evaluate the efficacy of applying the models to
reconstruct paleovegetation using fossil plant waxes extracted from sediments of Searles Lake,
CA, and compare our methods to previously published models based on linear mixing (Gao et
al., 2011). We calculate the probability of “Desert plants”, “Conifer” and “Macrophyte” from the
archived plant waxes in the late Pleistocene sediment core. We then validate the models by
comparing reconstructions of vegetation from the plant wax molecular abundance distributions
in downcore samples to that of the ACE index, a proxy for lake salinity based on the relative
proportion of archaeol and caldarchaeol molecules in the same samples. As the processes of
vegetation turnover and lake salinity variations differ, this presents an independent test of
hydrologic conditions to assess the utility of the machine learning (ML) approach to infer
paleoenvironmental conditions.
27
2.1.1 Study location and sampling
2.1.1.1 Searles Lake
Searles Valley in the Mojave Desert, south-eastern California, contains an endorheic paleolake
(presently a dry lake, or playa) with a salt pan surface and elevated paleoshorelines indicating
previously deep lake conditions (Fig 2.1). Lacustrine sediments, predominantly carbonate mud
and salts (Smith et al., 1983), have been deposited throughout the last 3.2 Ma (Liddicoat et al.,
1980). Lacustrine sediments are underlain by alluvial sand/gravel and quartz monazite bedrock.
Accommodation space was generated through late Pleistocene slip on normal faults bounding the
eastern side of Searles Valley (Numelin et al., 2007). Runoff from the eastern slopes of the Sierra
Nevada Mountains drains into endorheic basins to the east and at times formed a chain of lakes:
Owens Lake, which drains to China Lake Basin, and onto Searles Lake (Fig 2.1). During pluvial
periods (e.g., 11 – 16 ka, Fig 2.1), the Searles Lake catchment had an area of 18,168 km
2
and
encompassed the drainage basin of upstream Owens Lake (Smith, 2009), in contrast to the
smaller catchment today without stream flow between basins (Fig 2.1).
Fig 2.1. Topographic map (black shading) with elevation-based predictions of vegetation cover
within the Searles Lake catchment (colors, see legend), showing the expected configuration
during a pluvial period (a) versus the modern catchment and vegetation (b). Topography is from
28
a 7.5-arcsecond resolution DEM (Danielson, J.J., Gesch, 2011), with drainage basins delineated
from topographic highs in ArcMap. Modern vegetation in the Owens/Searles catchment
(California Department of Forestry and Fire Protection, 2015) is grouped into four categories:
Desert, Shrub, Conifer Woodland, and Conifer Forest. We grouped Desert and Shrub as “desert
shrubs” and grouped Conifer woodland and Conifer Forest as “conifer” for the machine learning
calculations.
Sediment cores SLAPP-SRLS17-1A and 1B (35.7372ºN, 117.33ºW, 494.75 m asl) were drilled
from the present salt pan surface (Fig 2.1) of Searles Lake in 2017 using a sonic drill rig with
95% recovery. Core 1A was drilled from the surface to 76 m and 1B extends from 21 – 38 m
below the lake floor. A continuous core stratigraphy was generated from the overlapping cores;
the splice was based on well-defined stratigraphic boundaries identified from high-resolution
core images. Almost all samples were obtained from Core 1A, with just 4 samples collected from
the spliced sections of 1B. Cores were split and sampled at the National Lacustrine Core Facility
(LacCore). 120 samples were collected for biomarker analyses. We sampled every 60 cm in the
mud, which is equivalent to a temporal resolution of ~2 ka based on the previous dating of
Searles Valley sediments (Bischoff et al., 1985). Few samples were collected within the salt units
(salts dominate above 33 m in the core), that contain low concentrations of plant wax.
2.1.1.2 Regional vegetation
The vegetation in the Searles basin today includes desert shrubs and herbs in the lowland below
2000 m asl (Fig 2.1). Open conifer woodland containing predominantly pinyon-juniper occurs
between 2000 – 2750 m asl, including some minor representation within the local catchment (Fig
2.1). Further afield, montane conifer forests (e.g, Pinus jeffreyi and Abies concolor) are found in
the Sierra Nevada Mountains between 2750 – 3500 m asl today. Alpine tundra above 3500 m
29
including perennial herbs (e.g. Oxyria digyna, Polemonium eximium, and Hulsea algida; Chabot
and Billings, 1972). During past glacial pluvial periods, the conifer forests likely were able to
thrive with increased moisture at lower altitudes, however, their upper treeline likely descended
to 2500 m reflecting the minimum glacial altitude (Moore and Moring, 2013) leading to an
overall range contraction for these conifer forests as glaciers advanced. In contrast, pinyon-
juniper woodland (also conifers) likely expanded its range during glacial pluvials (Litwin et al.,
1999; Heusser, 2000; Woolfenden, 2003; Koehler et al., 2005; Holmgren et al., 2010; Heusser et
al., 2015), descending from a lower limit of 2000 m today, to fill the lowlands in the Owens Lake
basin (Koehler and Anderson, 1994) and to the Searles Lake pluvial shoreline at 690 m asl, based
on a tufa cast of a juniper tree (Olsen 2020, personal communication)(Koehler and Anderson,
1994)(Koehler and Anderson, 1994)(Koehler and Anderson, 1994)(Koehler and Anderson,
1994)(Koehler and Anderson, 1994)(Koehler and Anderson, 1994). Thus pinyon-juniper
woodland is thought to dominate the lake catchments during pluvials as illustrated in Fig 2.1.
Topographic map (black shading) with elevation-based predictions of vegetation cover within
the Searles Lake catchment (colors, see legend), showing the expected configuration during a
pluvial period (a) versus the modern catchment and vegetation (b). Topography is from a 7.5-
arcsecond resolution DEM (Danielson, J.J., Gesch, 2011), with drainage basins delineated from
topographic highs in ArcMap. Modern vegetation in the Owens/Searles catchment (California
Department of Forestry and Fire Protection, 2015) is grouped into four categories: Desert, Shrub,
Conifer Woodland, and Conifer Forest. We grouped Desert and Shrub as “desert shrubs” and
grouped Conifer woodland and Conifer Forest as “conifer” for the machine learning calculations
Modern desert shrub vegetation around Searles Lake was sampled within the catchment (Fig
2.1). We sampled the C
3
shrubs creosote bush, Larrea tridentata (n = 2) and rabbit brush,
30
Ericameria nauseosa (n = 1) as well as the C
4
saltbush (Atriplex hymenelytra, Atriplex
confertifoli, and Atriplex canescens n = 4). The valley walls have a greater diversity of C
3
shrubs,
and we sampled wild buckwheat, Eriogonum pusillum (n = 1), spiny menodora, Menodora
spinescens (n = 1) as well as crassulacean acid metabolism (CAM) plants including the prickly
pear, Opuntia chlorotica (n = 1). The sampled plants make up the majority of plants by area in
Searles Valley. The present-day climate at the base of Searles Valley is hyperarid, with a mean
annual precipitation of 100 mm between 1920–2016 (Western Regional Climate Center, 2020).
The mean maximum temperature in Searles Valley is 27.4 °C with a mean minimum temperature
of 11.4 °C. Modern taxa are largely similar to vegetation growing during warm and dry periods
of the last glacial cycles as well (Litwin et al., 1999), although aridification during the early to
middle Holocene permitted the expansion of xeric-adapted species including L. tridentata
(Koehler et al., 2005; Holmgren et al., 2010).
We also sampled more widely to collect taxa that would have grown in the Searles Lake
catchment under past pluvial conditions (Fig 2.1). Pollen records indicate that pinyon-juniper
woodland expanded into Searles Valley during the glacial pluvials (Heusser, 2000, Heusser et
al., 2015; Litwin et al., 1999b; Woolfenden, 2003). Today these various conifer taxa were found
at ~2,100 m asl. in the San Bernardino Mountains. We sampled modern conifers including
Jeffrey pine, Pinus jeffreyi (n = 9); western juniper, Juniperus occidentalis (n = 4), and white fir
(Abies concolor) (n = 3) from the San Bernardino Mountains, CA (165 km to the south of Searles
Lake) where analog vegetation could be conveniently sampled to characterize the pinyon-juniper
and conifer forest vegetation zone, collectively known as ‘Conifer forest/woodland’ in this study.
The location had mean annual precipitation of 555 mm between 1960 – 2016 (Western Regional
31
Climate Center., 2020). The monthly mean maximum temperature is 16.6 °C and the mean
monthly minimum temperature is –0.5 °C.
When Searles Lake contained water, there may also have been plants within the waters of the
lake. Aquatic macrophytes (e.g., Potamogeton pectinatus L., Ruppia occidentalis) are common
in saline lakes in North America (Reynolds and Reynolds, 1975; Hammer and Heseltine, 1988;
Wollheim and Lovvorn, 1996) and as such it is likely that they would have been present during
pluvial periods of Searles Lake, especially at times of freshening within the lake or upstream
areas, when there was inter-basin connectivity. As there are no suitable aquatic areas in the
desert today, we use previously published macrophyte n-alkane and n-alkanoic acid chain length
abundance data from freshwater lakes in the USA (Gao et al., 2011) and Kenya (Ficken et al.,
2000) to train the macrophyte endmember in our model. Gao et al. (2011) sampled emergent (n =
7), floating (n = 4), and submerged macrophyte (n = 3) species between 2005–2008 from Blood
Pond, a freshwater lake located in Massachusetts, USA. Ficken et al. (2000) sampled emergent
(n = 7) and submerged (n = 4) species plants from Lake Nkunga, Sacred Lake, Lake Rutundu,
and Lake Ellis surrounding Mt Kenya, Kenya. These plants are the only published paired n-
alkane and n-alkanoic chain length distribution of macrophytes and thus represent the
“macrophyte” category in our model analysis. The modern macrophytes include the genera
Potamogeton, Typha, Utricularia and Lemna which are also found in slightly saline lakes in the
Canadian prairies (Hammer and Heseltine, 1988) suggesting that these macrophytes are widely
distributed and are likely to have been present in Searles Lake during a pluvial period. Although
some genera (such as Potamogeton) include species adapted to a range of salinities, most
hypersaline macrophytes are from different genera than the lower salinity macrophytes included
in our model. This may decrease model plant identification accuracy during periods of lake
32
hypersalinity. Only the relative abundance of the dominant chain lengths of both lipid classes
were available. Whilst we measured all major and minor chain lengths from the modern plants in
the San Bernardino mountains and Searles Valley we were restricted to using only the major
chain lengths (C
23-33
alkanes, C
22-32
alkanoic acids) to ensure compatibility with the macrophyte
data in our models. Data synthesized from the literature also included prior averaging of
individuals of a species (Blood Pond macrophytes; Gao et al., 2011) and averaging of submerged
and emergent types (Mt Kenya; Ficken et al., 2000). Scope to expand training datasets for other
plant types are limited, as few studies report molecular abundance data on both compound
classes, important for diagnostic differentiation between taxa.
2.2 Methods
2.2.1 Laboratory
2.2.1.1 Lipid extraction
Modern plant leaves (~1 g) were dried at 50 °C for 48 hours then chopped and extracted by
immersion in 9:1 dichloromethane (DCM):methanol (MeOH) three times, to yield the Total
Lipid Extract (TLE). Lipids were extracted from the ground, dried sediments (~ 20 g) using a
Dionex Accelerated Solvent Extraction system with 9:1 dichloromethane (DCM):methanol
(MeOH) at 100
º
C and 1500 psi to yield the TLE. The TLEs were then separated into neutral and
acid fractions using columns packed with NH
2 -
coated silica gel. Neutral fractions were eluted
using a 2:1 DCM:isopropanol and acid fractions were eluted with 4% formic acid in diethyl
ether. Carboxylic acids were methylated with 95:5 MeOH:hydrochloric acid, at 70 °C for 12
hours. Fatty acid methyl esters (FAMEs) were then separated from non-polar molecules using a
5% deactivated silica gel column eluted with DCM. Saturated compounds were then removed
from samples by using a silver nitrate silica gel column, with FAMEs being eluted by DCM.
33
The neutral fraction was separated into n-alkanes and the glycerol dialkyl glycerol tetraethers
(GDGTs) fraction using columns packed with 5% deactivated silica gel. n-Alkanes were eluted
with hexane and the GDGTs were eluted with DCM and methanol. n-Alkanes extracted from
sediments were passed through a copper column to remove elemental sulfur.
2.2.1.2 Leaf wax quantification
n-Alkanoic acids and n-alkanes were identified using an Agilent 6890 Gas Chromatograph (GC)
equipped with a Rxi®-5ms (30 m x 0.25 mm, film thickness 0.25 μm) column connected to an
Agilent 5973 MSD mass spectrometer (MS), and quantified using the flame ionization detector
(FID). Samples were dissolved in 500 μL of hexane and 1 μL was injected into a split/splitless
(S/SL) inlet in SL mode using an Agilent 7683 programmable autosampler. Compounds were
identified using MS spectra and retention times in comparison to library and standards.
Quantification of FAMEs was conducted through a comparison of sample peak areas in the FID
chromatograms to peak areas of an in-house standard containing known concentrations of n-
alkanoic acids. Carbon preference index (CPI) and average chain length (ACL) were calculated
to represent chain length distributions using the following equations:
ACL = ∑(n × [C
n
]) / ∑[C
n
] (1)
CPI = 2[C
n
]/([C
n−1
] + [C
n+1
]) (2)
where the chain length (n) refers to the C
23
to C
35
n-alkanes and C
22
to C
32
n-alkanoic acids.
2.2.1.3 Archaeol and Caldarchaeol analyses
The GDGT fractions were dissolved in hexane:isopropanol (99:1) and filtered through a 0.45 μm
polytetrafluoroethylene filter. Samples were then injected into an Agilent 1260 High-
Performance Liquid Chromatograph (HPLC) coupled to an Agilent 6120 mass spectrometer.
Separation of the GDGTs was accomplished using two Ethylene Bridged Hybrid (BEH)
34
Hydrophilic Interaction (HILIC) silica columns (2.1 mm × 150 mm, 1.7 μm; Waters) following
the methods of Hopmans et al. (2016). Archaeol, the isoprenoidal GDGTs, and branched GDGTs
were detected in single ion monitoring mode and quantified by comparison to a C
46
internal
standard. We calculated the ACE index with the equation:
ACE = [archaeol]/[archaeol + caldarchaeol] × 100 (3)
The ACE index was developed as a proxy for salinity (Turich and Freeman, 2011) based on the
calibration of marine and freshwater samples. A correlation between ACE and salinity was also
observed in highland Tibetan plateau lakes (Wang et al., 2013), suggesting that this proxy can
apply to lacustrine environments. This relationship has not been observed in a transect of
Himalayan lakes (Günther et al., 2014), possibly due to the dataset being very skewed toward
lower salinities (skewness = 3). The ACE index has been applied to the Miocene age
Mediterranean Messinian Salinity crisis (Turich and Freeman, 2011) as well as the last glacial
period in Lake Elsinore, Southern California (Feakins et al., 2019), which has a continuous
record of lacustrine deposition from 36 kyr to the present.
2.2.2 Machine Learning
The Python package Scikit-learn (Pedregosa et al., 2011) was used to construct machine learning
models. The machine learning techniques used were support vector machines, random forests,
Gaussian process classifiers, logistic regressions, and neural networks; these were selected
because they are widely used and applicable to multiclass classifications, i.e., classifying our
chain length distributions as belonging to one of three vegetation categories. By using multiple
models, we can determine which has the best performance and assess consistency between
models.
35
Support vector machines are useful for problems where the number of features is similar to the
number of samples, such as in the present study (number of modern plants = 49, number of n-
alkyl chain lengths = 12). Support vector machines work by positioning a hyperplane through the
data in space to divide the data into two categories (Boser et al., 1992). In 2D space, this means
drawing a line between two clusters of data to best separate them. As SVM is a binary classifier,
we use the “one vs rest” system, where one classifier is sequentially fitted against the other
classes to implement a multiclass classification. In our case, the input data are the proportion of
chain lengths, and the hyperplane is constructed to separate chain lengths that belong to desert
plants, conifers, and macrophytes. We used the radial basis function kernel, which adds an extra
dimension to the data to fit the plane. The regularization hyperparameter C and the gamma
parameter (influence of a single training sample) were chosen through a grid search with cross-
validation and are 9.6 and 2.7. All other settings are Scikit-learn defaults (Pedregosa et al.,
2011).
Random forests have been used in many Earth science applications including lithological
mapping (Harris and Grunsky, 2015), hydrology (Zabihi et al., 2016), and geomorphology
(Zhang et al., 2017) and this approach is recommended when there is limited training data
(Cracknell and Reading, 2014). Random forests act as a collection of decision trees created
through bootstrap aggregation or “bagging” whereby individual trees are created from randomly
selected data, and then the consensus aggregate of all of the trees is taken (Breiman, 2001). As
decision trees have inherently low bias and high variance, the voting consensus of many trees (n
= 300 in this study) results in a low bias low variance output. The random forest algorithm
hyperparameters include the number of decision trees constructed, the maximum depth of
individual decision trees (maximum path length of tree), and maximum number of features to
36
consider when forming a split. A grid search with cross validation determined that the Scikit-
learn defaults were appropriate model hyperparameters.
Within the Earth sciences, artificial neural networks have been widely used in seismology (Pulli
and Dysart, 1990; Wang, 1992; Perol et al., 2018) where they are useful for classifying
seismograms. Artificial neural networks are similar to biological neural networks, in that they are
structured to be composed of neurons, which receive an input signal and output a signal to other
neurons. Neural networks are composed of multiple layers of artificial neurons, with each layer
only in communication with the layers above and below. The first layer of the network is the
input layer which receives the input from the dataset, followed by the hidden layers, and then
finally the output layer. In the hidden layer, neurons receive an input which is passed through an
activation function that modulates the strength of the output signal from the neuron. Neurons
have a weight associated with them which is optimized by the model (through gradient descent)
to minimize their associated error during training (Rumelhart et al., 1986). We constructed a
multi-layer perceptron classifier neural network containing three hidden layers with 100 neurons
each and using the rectified linear activation function. Hyperparameters were determined by a
grid search algorithm maximizing classification accuracy on training data.
We used a Gaussian process classifier to generate a class probability and imposed a Gaussian
distribution using a Laplacian approximation (Williams and Barber, 1998). Gaussian process
classifiers output a probability that is easily interpretable, i.e p = (y|x), as opposed to support
vector machines whose output requires calibration to generate a probability estimate or Random
Forests whose probability estimates are based on the votes of the trees (Rasmussen and
Williams, 2006). Like support vector machines, Gaussian process classifiers are inherently
binary and thus we performed “one versus rest” classifications to perform a multiclass
37
classification. We also constructed a logistic regression model, where the dependant variable is
Bernoulli distributed. A sigmoidal function is then fit to achieve classification probabilities. We
used an L2 penalty (C = 35, selected through grid search), for regularization, which acts to
decrease the weighting given to less important features to prevent model overfitting. To train all
the models, we used a shuffle split method (Pedregosa et al., 2011), training on 66% of the
modern plant data and testing on the remaining 34%, repeating the shuffling 50 times to achieve
different splits, and taking the mean accuracy of all iterations. The models were run on a
computer using an intel core i5-7200U CPU and 8 GB of RAM. The training and testing dataset
contained n = 17 conifers (this study), n = 16 desert plants (this study), and n = 16 macrophytes
(Ficken et al., 2000; Gao et al., 2011).
2.3 Results and Discussion
2.3.1 Modern Plant Leaf molecular abundance
We report mid to long-chain n-alkyl compounds, specifically n-alkanes in the range C
23
-
35
and n-
alkanoic acids in the range, C
22
-
32
for modern plants and the lake sediment core (summarized in
Fig 2.2.) Desert plants had an n-alkane abundance which ranged from 10 to 1121 ng/g, with
CPI
alk
ranging from 8.9 to 59.7 with a mean of 13 and ACL
alk
from 27.4 to 33.0 with a mean of
29. Desert plant n-alkanoic abundance ranged from 40 to 3209 ng/g with CPI
acid
ranging from
3.5 to 62.9 with a mean of 9 and ACL
acid
from 26.4 to 28.3 with a mean of 28. Conifers had an n-
alkane abundance that ranged from 2 to 855 ng/g, with CPI
alk
ranging from 3.5 to 21.1 with a
mean of 9 and ACL
alk
from 27.1 to 33.0 with a mean of 29. Conifer n-alkanoic abundance ranged
from 7 to 256 ng/g with CPI
acid
ranging from 5.5 to 37.4 with a mean of 11 and ACL
acid
ranging
from 26.4 to 28.1 with a mean of 27.
38
Fig 2.2. Chain length distributions of n-alkanes (blue) and n-alkanoic acids (green) of (A) Modern plants
and (B) SLAPP-SRLS17 core samples, showing samples containing high modeled (SVM) desert plants,
conifers, and macrophytes as well as mean core distributions.
Considering the molecular abundance distributions for the n-alkane compound class, we find
desert plants are on average (Fig 2.2) dominated by C
27
, C
29,
and C
31
n-alkanes, with some
species having high proportions of very long chains (C
33
and C
35
) with no species having a
significant proportion of mid-chain lengths (C
23
to C
25
). This is in contrast to conifers and
macrophytes, both of which include a large proportion of species having a comparable mid-chain
length (C
23
to C
25
) and long-chain lengths with generally low C
31
concentrations. It is notable
that we find two conifer species P. jeffreyi and A. concolor, with dominant C
25
, C
27,
and C
29
n-
alkanes (Fig. A.2. Receiver operating characteristics (ROC) graphs of the five machine learning
39
models. ROC graphs plot the true positive rate (sensitivity) against the false positive rate
(specificity) under different classification threshold settings—models with higher discriminant
capacity plot closer to the upper left-hand corner of the plot. Left column shows the ROC curves
of models tested against the original data. Right column shows ROC curves of models tested
against original data with uniformly distributed noise added. Dashed black 1:1 line included for
comparison. ) that produce significant proportions of the mid-chain (C
23
and C
25
) n-alkanes that
have been historically used to infer aquatic macrophytes inputs in lakes and/or as indicators of
microbial production in soils. We also wish to emphasize that we report plant wax n-alkanes in
modern samples of P. jeffreyi, J. occidentalis, and A. concolor from the San Bernadino
Mountains. While most North American conifers have low production of n-alkanes (Diefendorf
et al., 2011) this pattern is not true universally (Diefendorf et al., 2015a). We find here that P.
jeffreyi and A. concolor have low alkane concentrations, but that J. occidentalis has high alkane
concentrations (240-844 ng/g) and this species is regionally significant in the southwestern
United States, making this a potentially useful observation for other regional paleoenvironmental
studies. Furthermore, J. occidentalis has a strongly unimodal distribution with a C
33
n-alkane
dominance and also includes C
35.
Since this compound (C
35
n-alkane) is below detection in most
other species, it may be a useful diagnostic for the presence of J. occidentalis.
Now we describe the molecular abundance distribution of the n-alkanoic acid compound class
that is generally less frequently reported than n-alkanes in plants. We find desert plant n-alkanoic
distributions are variable, with desert plants including bimodal (O. chlorotica), uniform (M.
spinescens), and normal (Atriplex) chain length distributions. On average C
28
n-alkanoic acid is
the modal chain length, as it is in Atriplex, A. tridentata, and L. tridentata but in other species,
chain lengths across C
22
to C
32
may be dominant or co-dominant. Conifers and macrophytes
40
produce C
22
to C
32
n-alkanoic acids. Conifers display bimodality peaking at C
24
and C
30
, overall,
whereas macrophytes distributions have a C
24
modal chain length and are generally skewed
toward shorter chain lengths than desert plants and conifers. Yet, also in this compound class,
interpretations of aquatic macrophytes (or microbial production) or the mid-chain lengths must
be tempered by the observation that many plant species make mid-chain (C
22
, C
24
) n-alkanoic
acids, as also noted elsewhere (e.g. Dion-Kirschner et al., 2020).
The mean CPI and ACL of both n-alkanes and n-alkanoic acids in the modern conifers and the
desert plants are similar highlighting the poor suitability of these metrics for tracking vegetation
change in this location. However, there is a large amount of variability between individual
species' chain length distributions (Fig. A.1.Chain length distributions of n-alkanes (blue) and n-
alkanoic acids (green) in (A) Modern desert plants. (B) Modern conifers. (C) Modern
macrophytes. Columns A, B, and C show the relative proportion of n-alkyl lipids. Error bars
represent 1σ from the mean when multiple plants were analyzed.) which is not described by the
CPI and ACL metrics. For instance, the shape of the distribution of peak heights includes
normal, bimodal, uniform, and Poisson across all plant types, with no relationship between the
distribution of n-alkanes and n-alkanoic acids for an individual species. For example, the desert
plant O. chlorotica has Poisson distributed n-alkane chain lengths and bimodally distributed n-
alkanoic acid chain lengths. As such neither ACL and/nor CPI would be very informative as to
the plant wax distributions, however, machine learning approaches on the full molecular
abundance distributions across both compound classes provide strong diagnostic potential.
2.3.2 SLAPP core plant waxes
n-Alkane abundance ranged from 0.2 to 11 ng/g. CPI ranged from 6.6 to 17.3 and ACL ranged
from 27.3 to 31.1 (Fig 2.3ab). n-Alkanoic abundances ranged from 3 to 244 ng/g, an order of
41
magnitude greater than that of n-alkanes. Mean molecular chain length distributions are shown in
(Fig. A.1.Chain length distributions of n-alkanes (blue) and n-alkanoic acids (green) in (A)
Modern desert plants. (B) Modern conifers. (C) Modern macrophytes. Columns A, B, and C
show the relative proportion of n-alkyl lipids. Error bars represent 1σ from the mean when
multiple plants were analyzed.). CPI ranged from 6.1 to 52.7, and ACL ranged from 23.9 to 27.1
(Fig 2.3ab). The CPI and ACL range in both n-alkyl groups is similar between modern plants and
core samples, indicative that significant diagenesis has not substantially altered the plant waxes
within the sedimentary environment.
42
Fig 2.3. Relative macrophyte abundance was calculated using the machine learning models and the linear
mixing model (Gao et al., 2011). Support vector machine (SVM) and the linear mixing model are
highlighted to show similarity. Data have been normalized by the mean and scaled by unit variance. For
comparison, we show measured variables: B. Carbon preference index (CPI) of n-alkanes and n-alkanoic
acids; C. Average chain length (ACL) of n-alkanes and n-alkanoic acids; D. Archaeol caldarchaeol
ecometric (ACE).
2.3.3 SLAPP core ACE index
We measured both archaeol and caldarchaeol in the SLAPP core. The archaeol concentration
ranged from 10 ng/g sediment to 91667 ng/g sediment. Caldarchaeol ranged from 196 ng/g
43
sediment to 1241 ng/g. ACE index values ranged from 1.1 to 98.9 representing fresh to saline
conditions. A quantitative salinity reconstruction is not possible due to a lack of calibration of
ACE in hypersaline lakes as well as a lack of correction for differential ionization of archaeol
and caldarchaeol.
2.3.4 Model training and testing
All the models proved to have high classification accuracy (defined as the fraction of correctly
identified plants from the test set) when classifying the modern plant data, with Gaussian process
classifiers and the neural network performing the best (accuracy = 0.83) and logistic regression
performing the worst (accuracy = 0.78) (Error! Reference source not found.). F1 is a weighted
harmonic mean between the precision e.g., the ability of a model not to label ‘desert’ to a
‘conifer’ sample, and the recall e.g., the ability of a model to identify all “desert” samples. The
neural network had the best F1 score (F1 = 0.82) and logistic regression had the worst (F1 =
0.77). The relatively small differences between the different models’ performance metrics
suggest that they are all suitable for modern vegetation classification using plant wax
distributions.
Model name Accuracy F1
Support vector machines
0.81
0.8
Gaussian process
classifier 0.83 0.81
Random forest 0.81 0.8
Logistic regression 0.78 0.77
Neural network 0.83 0.82
Table 2.1. Performance metrics for each machine learning model.
To evaluate the robustness of our models we calculated their classification accuracy when white
noise had been added to the test sets. We added three fractional levels of noise, which had ranges
of 0–0.01, 0–0.05, and 0–0.1 (Error! Reference source not found.), and then normalized both
44
the n-alkanoic acids and the n-alkanes to unity. We found that at very high noise levels (0.1)
support vector machines and the neural network still had a high classification accuracy (0.77,
0.78), but logistic regression performed poorly (0.71) suggesting it was overfitting the training
data. Support vector machines had the smallest difference (0.04) in accuracy between testing
without noise and testing with noise, suggesting that it is the most robust model when dealing
with noisy data. We also calculated receiver operating characteristics curves (ROC) for the
models which plot the true positive rate (sensitivity) against the false positive rate (specificity)
under different classification threshold settings (Fig. A.2). We generated ROC curves using both
sample data and data with added random noise and calculated the area under the curves. Models
with higher discriminant capacity plot closer to the upper left-hand corner of the plot. Therefore,
the higher area under the curve (AUC) indicates greater model specificity and sensitivity. Four
models perform well without added noise with an AUC greater than 0.8 across all models and
classes. However, logistic regression shows comparatively worse results with the macrophyte
class with an AUC=0.73. There is little degradation in AUC in the models tested on sample data
with added noise, further suggesting that the models are not overfitting the training data. This
observation is surprising given the small amount of training data available to the models and a
large amount of random noise added. Overall, we would recommend constructing multiple
machine learning models and checking that there is consistency between their output. However,
our results show that support vector machines have high-performance metrics and thus we focus
on this model output below.
Model name
Accuracy
0.01 0.05 0.1
Support vector machines 0.79 0.79 0.77
45
Gaussian process
classifier 0.8 0.8 0.76
Random forest 0.8 0.78 0.74
Logistic regression 0.76 0.76 0.71
Neural network 0.79 0.8 0.78
Table 2.2. Accuracy of models when uniformly distributed noise with 3 different ranges (0–0.01, 0–0.05,
and 0–0.1) was added to the test set. Accuracy of models when uniformly distributed noise with 3
different ranges (0–0.01, 0–0.05, and 0–0.1) was added to the test set.
2.3.5 Model vegetation from SLAPP-SRLS17 core
Across all five machine learning models, there is a large range in predicted vegetation between
all three classes (Fig 2.4). However, the models are dominated by desert plant vegetation and
macrophyte vegetation, with conifer vegetation generally a minor constituent. This suggests that
the more distal conifer signal is dominated by the more local desert and macrophyte inputs. This
local dominance has been observed in other studies of plant waxes in rivers spanning montane to
lowland areas (Galy et al., 2011; Ponton et al., 2014; Feakins et al., 2018) as well as more limited
data from the nearby Lake Elsinore catchment (Feakins et al., 2019). Macrophyte inputs have
been reported to be substantial in several lakes (Aichner et al., 2010), whereas in other settings
woody plants were found to dominate (Freimuth et al., 2019; McFarlin et al., 2019; Dion-
Kirschner et al., 2020).
46
Fig 2.4. Probability estimates of modeled vegetation types downcore for each machine learning model.
Probabilities are generated by inputting core wax distribution information into models trained on modern
plant wax distributions. Salt in core highlighted by grey bars.
47
All five models show very similar trends through time for the desert and macrophyte classes. The
greatest differences between the models occur within the modeled conifer vegetation with the
random forest algorithm having a very low correlation to the other methods (Fig 2.5). One
explanation is that the conifer wax distributions are comparatively more degraded than desert
plants and macrophyte waxes due to the greater distance and hence transport time (Agrawal et
al., 2014) between the distal montane areas inhabited by conifers and Searles Lake.
Alternatively, the random forest algorithm could be performing worse due to the low proportion
of conifer waxes in the sediment. A systematic sampling of plant waxes from soils and montane
rivers in the Sierra Nevada could provide more information on the concentration and degradation
of waxes in this environment.
Model uncertainty was assessed by running the models 100 times each using randomly selected
training and testing splits. The standard deviation of each modelled core sample was then
calculated across the three vegetation classes ( Random forest = 0.06, logistic regression = 0.12,
gaussian process classifier = 0.04, SVM = 0.09, neural network = 0.2). All models had low
uncertainty except the neural network which is very sensitive to the composition of the train and
test sets.
2.3.6 Model comparison
We compare our machine learning models to a linear mixing model (Gao et al., 2011) which is
constructed by compiling data from vegetation surrounding Blood Pond and various lakes
surrounding Mt Kenya (Ficken et al., 2000). The n-alkane abundances are then normalized to
unity before the data are divided into five groups based upon cluster analysis of chain length
proportion, with macrophytes identified as a distinct cluster. A lake sediment n-alkane
distribution is then inputted, and the least square mixing model calculates the proportion of plant
48
waxes contributed from the macrophyte cluster. This model has been applied to surface sediment
lake samples but has not been applied to waxes in older sediments. We applied the model to our
n-alkane data from the core samples to calculate the macrophyte abundance and compare this to
our machine learning model output. We note that our machine learning models use both the n-
alkanes and n-alkanoic acids whereas the linear mixing model only uses n-alkanes for input
information.
Fig 2.5. Correlation matrices showing Spearman’s Rank correlations between probabilities of the three
modeled vegetation types downcore (A. Macrophyte, B. Desert plant, C. Conifer) and archaeol
caldarchaeol ecometric (ACE) salinity index in SLAPP-SRLS17. Black boxes denote p<0.01.
The linear mixing model has a much larger range of macrophyte proportion than is suggested by
any of the machine learning algorithms (Linear mixing = 0.82, max Logistic regression=0.59,
min Gaussian process classifier=0.19). However, when scaled (normalized to the mean and
scaled by unit variance) the models, especially support vector machines, show similarity given
their very different constructions (Fig 2.5). To better understand this similarity we compared the
measured CPI and ACL of the plant waxes in the sediment core to the modeled macrophyte
abundance (Fig 2.5). Neither the machine learning models nor the linear mixing model was
related to CPI (Fig 2.5B), however, no model includes the non-dominant chain length
information as their concentrations were not available in published plant datasets. There is a
49
strong negative correlation (r = – 0.9) between the linear mixing model macrophyte abundance
and the ACL
alk
downcore (Fig 2.5C). This is likely due to the cluster analysis performed by Gao
et al. (2011) to calculate a macrophyte endmember for their mixing model, which only included
macrophytes with a low ACL
alk
in the macrophyte cluster. The machine learning models had a
weaker correlation with ACL
alk
(mean r = – 0.25) and a stronger correlation with ACL
acid
(mean r
= – 0.67) suggesting that they find greater predictive power in the n-alkanoic chain lengths.
2.3.7 Model Validation
Finally, we compare our various modeled and measured plant wax metrics to the ACE index
measured downcore (Fig 2.5C). High ACE values suggest more saline lake waters (Turich and
Freeman, 2011). As lake water salinity is mainly controlled by the amount of water in the lake,
which is in turn controlled by evaporation and precipitation, we can assume that a more saline
lake reflects a drier/warmer environment. As ACE (lake salinity) increased, we expected that the
relative amount of macrophytes would decrease, as both diversity and areal cover of
macrophytes decreases with higher lake salinity (Wollheim and Lovvorn, 1996). To illustrate the
similarity between the two depth series, we compare the ACE index to the macrophyte output
from the support vector machines model (Fig 2.6). The Spearman’s rank correlation coefficient r
= – 0.45 and p < 0.01 using non-parametric methods accounting for autocorrelation (Ebisuzaki,
1997). These values suggest a moderate negative relationship. Gaussian process classifier,
logistic regression, random forest and neural networks also showed similar correlation values (
r= – 0.47, p < 0.01 r = – 0.53 p < 0.01, r = – 0.34 p < 0.01, r = – 0.44 p < 0.01). We also
calculated the correlation between ACE and the linear mixing model of Gao et al. (2011). We
find that r = – 0.24 and p > 0.01, which is lower than any of the ML algorithms, indicating that
the ML models have better performance. ML models perform better for various reasons: 1) ML
50
models were trained predominantly on local plants in the Searles Lake catchment. 2) The cluster
analysis of the linear mixing model does not discriminate between plant types effectively
compared to ML algorithms. Out of the 22 macrophyte samples included in the model, only 10
are present in the “macrophyte” cluster, with the others included in the remaining clusters. This
results in the “macrophyte” cluster only including less than half of the available information on
macrophytes. As such, the macrophyte end member of the linear mixing model is not very
representative of the population which likely decreases its performance. 3) The ML models used
both n-alkanes and n-alkanoic acids, but the linear mixing model only used n-alkanes. We also
tested the performance of models trained with only n-alkane or n-alkanoic acid data. Using a
single lipid class, the mean correlation between modeled macrophyte and ACE was 0.37 and
0.13 for n-alkanoic acids and n-alkanes respectively, which is lower than the 0.45 correlation
when using both lipid classes.
51
Fig 2.6. Modeled macrophyte plant proportions using the support vector machine model and ACE index
values against SLAPP-SRLS17 core depth. Data has been centered to the mean and scaled by unit
variance. Modeled vegetation was generated using wax distributions measured from SLAPP-SRLS17.
The top image is a composite high-resolution photograph of the core.
We also observe a moderate positive correlation between the ML modelled desert plant
composition and ACE (SVM r=0.41 p<0.01, Gaussian process classifier r=0.40 p<0.01, logistic
regression r=0.40, random forest r=0.33 p<0.01, neural network r=0.45 p<0.01). This suggests
that as Searles Lake became saltier and shallower, there was an expansion of the desert plant
community surrounding the lake driven by increased aridity. This relationship makes intuitive
sense and thus further increases our confidence in the ML model's vegetation prediction.
There is no correlation between ACE and the ML model predictions of conifer abundance. This
is likely due to the complex relationship between conifer abundance on the landscape and
climate (Litwin et al., 1999), with both cooler and warmer periods seeing variable conifer
responses. Cooler periods produce an increase in the abundance of pinyon-juniper forests
downslope, with evidence of in situ juniper tufa casts close to Searles Lake shorelines during
previous lake high stands (Olson 2020, personal communication). Yet pollen evidence suggests
that warmer periods saw an expansion of montane pine forests as a glacial retreat in the Sierra
Nevada mountains provided an increase in habitable area. Additionally, as noted previously,
plant waxes from distal mountains likely take a greater amount of time to be transported and
deposited into Searles lake than those from proximal desert plants and macrophytes. This
temporal lag could contribute to the lack of correlation between ACE and modeled conifer
abundance. However, aeolian transportation of wax can be significant in dry settings (Feakins et
al., 2005) which would act to decrease any time lag between the deposition of the proxies.
There is a significant correlation between ACE and both modeled desert plants and modeled
macrophytes. However, it is possible that these correlations are not independent as the modeled
52
vegetation probabilities sum to 1 and the abundance of conifers is generally stable through time
(Fig 2.4). As such, there may only be information in one of the correlations i.e., between ACE
and desert plants or ACE and macrophytes. However, a drier climate would likely cause both an
increase in desert plants and a decrease in macrophyte abundance, and as such we interpret both
desert plants and macrophytes to ultimately be responding to changes in aridity.
The significant positive correlations between ACE and the five machine learning models suggest
great potential for paleovegetation reconstructions. However, several factors may decrease model
performance. First, diagenesis between leaves and the sedimentary archive altering lipid chain
length distributions may reduce the correspondence between the model and the test data. Plant
waxes may undergo diagenetic alteration during soil storage (Wu et al., 2019) and fluvial
transport (Ponton et al., 2014), although waxes are found to be relatively well-preserved within
Searles Lake, likely due to anoxic lake bottom waters during the last glacial period (Winters et
al., 2013). Of these processes, plant waxes are likely most susceptible to microbial
transformation during aerobic conditions in leaf litter and soil storage, with the production of
mid-chain n-alkanes, and thus an increase in the C
23
/C
29
(Wu et al., 2019). Similarly, laboratory
incubation experiments of soils found the production of mid-chain n-alkanes (centered on C
23
;
Brittingham et al., 2017). In contrast, incubated peat (G. Li et al., 2018) found the production of
short (C
18
) and long (C
29
) n-alkanes, highlighting that the distinct substrate, storage temperature,
and microbial community may alter the rate and nature of the alterations. Ultimately, machine
learning could be employed to study such taphonomic changes in molecular abundance
distributions between plants and soils. Secondly, autochthonous production in the lake may
modify the archived lipids. Bacteria and algae can also produce long-chain n-alkanes (G. Li et
al., 2018) and n-alkanoic acids that may be substantial in settings with high aquatic production
53
and low plants wax fluxes such as crater lake Lake Pavin (Makou et al., 2018) and Antarctic
Lakes (Chen et al., 2019). Assessing the production of algae and bacteria in paleo-Searles Lake,
we note that the halophilic green algae Dunaliella was identified in fluid inclusions in halite,
which establishes that organism as the dominant producer in ancient hypersaline brines (Winters
et al., 2013). Culture studies have shown that Dunaliella produces trace amounts of mid-long
chain n-alkanoic acids, and no n-alkanes (Stiehl et al., 2005), so we do not expect algal or
bacterial production of lipids to confound the leaf wax record during periods of lake
hypersalinity. However, during the fresher periods of Searles Lake, a more diverse algal
population may have contributed more lipids decreasing ML model performance.
Third, the modern plant calibration performed in this study does not contain all the possible
plants that could have been growing in the catchment of Searles Lake during the past 150 kyrs.
The presence of plants that fall outside of our three categories could decrease model
paleovegetation prediction accuracy, as the model categorized all samples as conifer, desert, or
macrophyte. A larger and more comprehensive sampling effort of modern vegetation may
increase the model performance and would be recommended in any future study applying these
techniques.
Fourth, our model only contained odd chain n-alkanes and even chain n-alkanoic acids, which
reduced the amount of information available to the models. Whilst we measured both odd and
even chain lengths across both lipid compound classes in the modern plants we sampled, these
data could not be included in the models to ensure compatibility with the published macrophyte
data, which only contained odd chain n-alkanes and even chain n-alkanoic acids. As the carbon
preference index (even/odd chains over odd/even chains) is likely characteristic of a plant
54
species (Bush and McInerney, 2013) our ML algorithms likely performed worse during
vegetation classification in the absence of these chain lengths.
2.4 Conclusions
We trained five machine learning algorithms to identify contrasting modern plant communities
based upon individual species' n-alkane and n-alkanoic acid chain length distributions. The three
studied modern plant communities are the shrubs of the hyperarid floor of Searles Valley, the
coniferous forests and woodlands of the San Bernadino Mountains, and macrophytes from Blood
Pond and Mt. Kenya. All machine learning models successfully identified modern plant
communities based on our training dataset, although support vector machines, neural networks,
and Gaussian process classifiers outperformed random forests and logistic regression.
We used all five machine learning models to predict the abundance of desert plants, conifers, and
macrophytes, from plant wax distributions analyzed in the Searles Lake SLAPP core spanning 10
kyr – 150 kyr. We compared the machine learning predictions of vegetation type to a previously
published linear mixing model (Gao et al., 2011), and an independent ACE index reconstruction
of lake salinity. We hypothesized that high lake salinity would correspond to the dominance of
desert plants (similar to modern), whereas low lake salinity would allow more macrophytes.
Across all five machine learning models we found that modeled macrophyte abundance was
negatively correlated with the ACE index (mean r = – 0.45) whereas the linear mixing model did
not find a significant correlation. A moderate positive correlation between ACE and modeled
desert plants was found with all of the machine learning models (mean r=0.40). Factors such as
leaf wax diagenesis, autochthonous lipid inputs, a limited training set, and chain length
information may have limited model performance.
55
Although this study focuses on just three vegetation classes in a high-contrast ecosystem, our
results suggest that machine learning methods have substantial potential for reconstructing past
distributions of vegetation. The machine learning methods have an advantage over-simplified
metrics such as CPI and ACL in that they fit models in multi-dimensional space and can handle
instances of unclear and/or overlapping wax distributions. Based on our initial results we
encourage further tests of plant wax applications of machine learning in other ecosystems with
suitable training sets, validation exercises, and comparisons between multiple machine learning
algorithms to determine the scope in other sedimentary archives.
Acknowledgments
This study was supported by U.S. National Science Foundation Grant NSF-EAR-1903665 to
S.F., the Comer Science and Education Foundation Grant to D.M. and T.L., and the Packard
Fellowship for Science and Engineering to J.E.T. We thank Searles Valley Minerals for access
and Jade Zimmermann in particular. The sample material used in this project was provided by
LacCore. We thank the SLAPP team involved in coring and collaborative discussions of
surroundings and paleoenvironment. We thank Jay Quade for plant identification in the Mojave
Desert, Alan Juarez for field assistance in the San Bernardino Mountains, and Patrick Murphy
for assistance preparing and measuring GDGTs.
References
Agrawal, S., Galy, V., Sanyal, P., Eglinton, T., 2014. C4 plant expansion in the Ganga Plain
during the last glacial cycle: Insights from isotopic composition of vascular plant
biomarkers. Organic Geochemistry 67, 58–71.
Aichner, B., Herzschuh, U., Wilkes, H., 2010. Influence of aquatic macrophytes on the stable
carbon isotopic signatures of sedimentary organic matter in lakes on the Tibetan Plateau.
56
Organic Geochemistry 41, 706–718.
Bischoff, J.L., Rosenbauer, R.J., Smith, G.I., 1985. Uranium-series dating of sediments from
Searles Lake: Differences between continental and marine climate records. Science 227,
1222–1224.
Boser, B.E., Guyon, I.M., Vapnik, V.N., 1992. 5th Annual ACM Workshop on COLT.
Breiman, L., 2001. Random forests. Machine Learning 45, 5–32.
Brittingham, A., Hren, M.T., Hartman, G., 2017. Microbial alteration of the hydrogen and carbon
isotopic composition of n-alkanes in sediments. Organic Geochemistry 107, 1–8.
Bush, R.T., Mcinerney, F.A., 2015. Influence of temperature and C 4 abundance on n-alkane
chain length distributions across the central USA. doi:10.1016/j.orggeochem.2014.12.003
Bush, R.T., McInerney, F.A., 2013. Leaf wax n-alkane distributions in and across modern plants:
Implications for paleoecology and chemotaxonomy. Geochimica et Cosmochimica Acta
117, 161–179.
California Department of Forestry and Fire Protection, 2015. Vegetation (fveg) - CAL FIRE
FRAP [ds1327] .
Chabot, B.F., Billings, W.D., 1972. Origins and Ecology of the Sierran Alpine Flora and
Vegetation. Ecological Monographs 42, 163–199.
Chen, X., Liu, X., Wei, Y., Huang, Y., 2019. Production of long-chain n-alkyl lipids by
heterotrophic microbes: New evidence from Antarctic lakes. Organic Geochemistry 138,
103909.
Cracknell, M.J., Reading, A.M., 2014. Geological mapping using remote sensing data: A
comparison of five machine learning algorithms, their response to variations in the spatial
distribution of training data and the use of explicit spatial information. Computers and
57
Geosciences 63, 22–33.
Danielson, J.J., Gesch, D.B., 2011. Global Multi-resolution Terrain Elevation Data 2010
(GMTED2010), U.S. Geological Survey Open-File Report 2011-1073.
Diefendorf, A.F., Freeman, K.H., Wing, S.L., Graham, H. V., 2011. Production of n-alkyl lipids
in living plants and implications for the geologic past. Geochimica et Cosmochimica Acta
75, 7472–7485.
Diefendorf, A.F., Leslie, A.B., Wing, S.L., 2015. Leaf wax composition and carbon isotopes
vary among major conifer groups. Geochimica et Cosmochimica Acta 170, 145–156.
Dion-Kirschner, H., McFarlin, J.M., Masterson, A.L., Axford, Y., Osburn, M.R., 2020. Modern
constraints on the sources and climate signals recorded by sedimentary plant waxes in west
Greenland. Geochimica et Cosmochimica Acta 286, 336–354.
Ebisuzaki, W., 1997. A method to estimate the statistical significance of a correlation when the
data are serially correlated. Journal of Climate 10, 2147–2153.
Eglinton, G., Gonzalez, A.G., Hamilton, R.J., Raphael, R.A., 1962. Hydrocarbon constituents of
the wax coatings of plant leaves: A taxonomic survey. Phytochemistry 1, 89–102.
Eglinton, G., Hamilton, R.J., 1967. Leaf epicuticular waxes. Science 156, 1322–1335.
Engle, M.A., Brunner, B., 2019. Considerations in the application of machine learning to
aqueous geochemistry: Origin of produced waters in the northern U.S. Gulf Coast Basin.
Applied Computing and Geosciences 3–4, 100012.
Feakins, S.J., deMenocal, P.B., Eglinton, T.I., 2005. Biomarker records of late Neogene changes
in northeast African vegetation. Geology 33, 977–980.
Feakins, S.J., Wu, M.S., Ponton, C., Galy, V., West, A.J., 2018. Dual isotope evidence for
sedimentary integration of plant wax biomarkers across an Andes-Amazon elevation
58
transect. Geochimica et Cosmochimica Acta 242, 64–81.
Feakins, S.J., Wu, M.S., Ponton, C., Tierney, J.E., 2019. Biomarkers reveal abrupt switches in
hydroclimate during the last glacial in southern California. Earth and Planetary Science
Letters 515, 164–172.
Ficken, K.J., Li, B., Swain, D.L., Eglinton, G., 2000. An n-alkane proxy for the sedimentary
input of submerged/floating freshwater aquatic macrophytes, in: Organic Geochemistry.
Pergamon, pp. 745–749.
Freimuth, E.J., Diefendorf, A.F., Lowell, T. V., Wiles, G.C., 2019. Sedimentary n-alkanes and n-
alkanoic acids in a temperate bog are biased toward woody plants. Organic Geochemistry
128, 94–107.
Galy, V., Eglinton, T., France-Lanord, C., Sylva, S., 2011. The provenance of vegetation and
environmental signatures encoded in vascular plant biomarkers carried by the Ganges-
Brahmaputra rivers. Earth and Planetary Science Letters 304, 1–12.
Gao, L., Hou, J., Toney, J., MacDonald, D., Huang, Y., 2011. Mathematical modeling of the
aquatic macrophyte inputs of mid-chain n-alkyl lipids to lake sediments: Implications for
interpreting compound specific hydrogen isotopic records. Geochimica et Cosmochimica
Acta 75, 3781–3791.
Günther, F., Thiele, A., Gleixner, G., Xu, B., Yao, T., Schouten, S., 2014. Distribution of
bacterial and archaeal ether lipids in soils and surface sediments of Tibetan lakes:
Implications for GDGT-based proxies in saline high mountain lakes. Organic Geochemistry
67, 19–30.
Hammer, U.T., Heseltine, J.M., 1988. Aquatic macrophytes in saline lakes of the Canadian
prairies. Hydrobiologia 158, 101–116.
59
Harris, J.R., Grunsky, E.C., 2015. Predictive lithological mapping of Canada’s North using
Random Forest classification applied to geophysical and geochemical data. Computers and
Geosciences 80, 9–25.
Heusser, L.E., 2000. Rapid oscillations in western North America vegetation and climate during
oxygen isotope stage 5 inferred from pollen data from Santa Barbara Basin (Hole 893A),
Palaeogeography, Palaeoclimatology, Palaeoecology 161, 407-421.
Heusser, L.E., Kirby, M.E., Nichols, J.E., 2015. Pollen-based evidence of extreme drought
during the last Glacial (32.6-9.0 ka) in coastal southern California. Quaternary Science
Reviews 126, 242–253.
Holmgren, C.A., Betancourt, J.L., Rylander, K.A., 2010. A long-term vegetation history of the
Mojave-Colorado desert ecotone at Joshua Tree National Park. Journal of Quaternary
Science 25, 222–236.
Hopmans, E.C., Schouten, S., Sinninghe Damsté, J.S., 2016. The effect of improved
chromatography on GDGT-based palaeoproxies. Organic Geochemistry 93, 1–6.
Jansen, B., de Boer, E.J., Cleef, A.M., Hooghiemstra, H., Moscol-Olivera, M., Tonneijck, F.H.,
Verstraten, J.M., 2013. Reconstruction of late Holocene forest dynamics in northern
Ecuador from biomarkers and pollen in soil cores. Palaeogeography, Palaeoclimatology,
Palaeoecology 386, 607–619.
Jansen, B., van Loon, E.E., Hooghiemstra, H., Verstraten, J.M., 2010. Improved reconstruction
of palaeo-environments through unravelling of preserved vegetation biomarker patterns.
Palaeogeography, Palaeoclimatology, Palaeoecology 285, 119–130.
Koch, K., Bhushan, B., Barthlott, W., 2009. Multifunctional surface structures of plants: An
inspiration for biomimetics. Progress in Materials Science 54, 137–178.
60
Koehler, P.A., Anderson, R.S., 1994. Full-glacial shoreline vegetation during the maximum
highstand at Owens Lake, California. Great Basin Naturalist 54, 142–149.
Koehler, P.A., Anderson, R.S., Spaulding, W.G., 2005. Development of vegetation in the Central
Mojave Desert of California during the late Quaternary. Palaeogeography,
Palaeoclimatology, Palaeoecology 215, 297–311.
Li, G., Li, L., Tarozo, R., Longo, W.M., Wang, K.J., Dong, H., Huang, Y., 2018. Microbial
production of long-chain n-alkanes: Implication for interpreting sedimentary leaf wax
signals. Organic Geochemistry 115, 24–31.
Liddicoat, J.C., Opdyke, N.D., Smith, G.I., 1980. Palaeomagnetic polarity in a 930-m core from
Searles Valley, California. Nature 286, 22–25.
Litwin, R.J., Smoot, J.P., Durika, N.J., Smith, G.I., 1999. Calibrating Late Quaternary terrestrial
climate signals: Radiometrically dated pollen evidence from the southern Sierra Nevada,
USA. Quaternary Science Reviews. doi:10.1016/S0277-3791(98)00111-5
Makou, M., Eglinton, T., McIntyre, C., Montluçon, D., Antheaume, I., Grossi, V., 2018. Plant
Wax n ‐Alkane and n ‐Alkanoic Acid Signatures Overprinted by Microbial Contributions
and Old Carbon in Meromictic Lake Sediments. Geophysical Research Letters 45, 1049–
1057.
McFarlin, J.M., Axford, Y., Masterson, A.L., Osburn, M.R., 2019. Calibration of modern
sedimentary δ2H plant wax-water relationships in Greenland lakes. Quaternary Science
Reviews 225, 105978.
Moore, J.G., Moring, B.C., 2013. Rangewide glaciation in the Sierra Nevada, California.
Geosphere 9, 1804–1818.
Numelin, T., Kirby, E., Walker, J.D., Didericksen, B., 2007. Late Pleistocene slip on a low-angle
61
normal fault, Searles Valley, California. Geosphere 3, 163.
Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M.,
Prettenhofer, P., Weiss, R., Dubourg, V., Vanderplas, J., Passos, A., Cournapeau, D.,
Brucher, M., Perrot, M., Duchesnay, E., 2011. Scikit-learn: Machine Learning in Python,
Journal of Machine Learning Research.
Perol, T., Gharbi, M., Denolle, M., 2018. Convolutional neural network for earthquake detection
and location. Science Advances 4, e1700578.
Ponton, C., West, A.J., Feakins, S.J., Galy, V., 2014. Leaf wax biomarkers in transit record river
catchment composition. Geophysical Research Letters 41, 6420–6427.
Pulli, J.J., Dysart, P.S., 1990. An experiment in the use of trained neural networks for regional
seismic event classification. Geophysical Research Letters 17, 977–980.
Rasmussen, C.E., Williams, C.K.I., 2006. Gaussian processes for machine learning. MIT Press.
Reynolds, J., Reynolds, S., 1975. Aquatic angiosperms of some British Columbia saline lakes.
Syesis 8, 291–295.
Rommerskirchen, F., Plader, A., Eglinton, G., Chikaraishi, Y., Rullkötter, J., 2006.
Chemotaxonomic significance of distribution and stable carbon isotopic composition of
long-chain alkanes and alkan-1-ols in C4 grass waxes. Organic Geochemistry 37, 1303–
1332.
Rumelhart, D.E., Hinton, G.E., Williams, R.J., 1986. Learning representations by back-
propagating errors. Nature 323, 533–536.
Shepherd, T., Wynne Griffiths, D., 2006. The effects of stress on plant cuticular waxes. New
Phytologist 171, 469–499.
Smith, G.I., 2009. Late Cenozoic geology and lacustrine history of Searles Valley, inyo and San
62
Bernardino counties, California. US Geological Survey Professional Paper.
Smith, G.I., Barczak, V.J., Moulton, G.F., Liddicoat, J.C., 1983. Core KM-3, a surface-to-
bedrock record of late Cenozoic sedimentation in Searles Valley, California, Professional
Paper. doi:10.3133/PP1256
Stiehl, T., Rullkötter, J., Nissenbaum, A., 2005. Molecular and isotopic characterization of lipids
in cultured halophilic microorganisms from the Dead Sea and comparison with the sediment
record of this hypersaline lake. Organic Geochemistry 36, 1242–1251.
Turich, C., Freeman, K.H., 2011. Archaeal lipids record paleosalinity in hypersaline systems.
Organic Geochemistry 42, 1147–1157.
Ueki, K., Hino, H., Kuwatani, T., 2018. Geochemical Discrimination and Characteristics of
Magmatic Tectonic Settings: A Machine-Learning-Based Approach. Geochemistry,
Geophysics, Geosystems 19, 1327–1347.
Wang, H., Liu, W., Zhang, C.L., Jiang, H., Dong, H., Lu, H., Wang, J., 2013. Assessing the ratio
of archaeol to caldarchaeol as a salinity proxy in highland lakes on the northeastern
Qinghai–Tibetan Plateau. Organic Geochemistry 54, 69–77.
Wang, L.X., 1992. A Neural Detector for Seismic Reflectivity Sequences. IEEE Transactions on
Neural Networks 3, 338–340.
Western Regional Climate Center [WWW Document], 2020. URL https://wrcc.dri.edu/cgi-
bin/cliMAIN.pl?ca9035
Williams, C.K.I., Barber, D., 1998. Bayesian classification with gaussian processes. IEEE
Transactions on Pattern Analysis and Machine Intelligence 20, 1342–1351.
Winters, Y.D., Lowenstein, T.K., Timofeeff, M.N., 2013. Identification of Carotenoids in
Ancient Salt from Death Valley, Saline Valley, and Searles Lake, California, Using Laser
63
Raman Spectroscopy. Astrobiology 13, 1065–1080.
Wollheim, W.M., Lovvorn, J.R., 1996. Effects of macrophyte growth forms on invertebrate
communities in saline lakes of the Wyoming High Plains. Hydrobiologia 323, 83–96.
Woolfenden, W.B., 2003. A 180,000-year pollen record from Owens Lake, CA: Terrestrial
vegetation change on orbital sales. Quaternary Research 59, 430–444.
Wu, M.S., West, A.J., Feakins, S.J., 2019. Tropical soil profiles reveal the fate of plant wax
biomarkers during soil storage. Organic Geochemistry 128, 1–15.
Zabihi, M., Pourghasemi, H.R., Pourtaghi, Z.S., Behzadfar, M., 2016. GIS-based multivariate
adaptive regression spline and random forest models for groundwater potential mapping in
Iran. Environmental Earth Sciences 75, 1–19.
Zhang, K., Wu, X., Niu, R., Yang, K., Zhao, L., 2017. The assessment of landslide susceptibility
mapping using random forest and decision tree methods in the Three Gorges Reservoir area,
China. Environmental Earth Sciences 76, 1–20.
Zheng, Y., Heng, P., Conte, M.H., Vachula, R.S., Huang, Y., 2019. Systematic chemotaxonomic
profiling and novel paleotemperature indices based on alkenones and alkenoates: Potential
for disentangling mixed species input. Organic Geochemistry 128, 26–41.
64
Chapter 3: Identifying the drivers of GDGT distributions in alkaline soil
profiles within the Serengeti ecosystem
This chapter has been published in Organic Geochemistry as:
Mark D. Peaple, Emily J. Beverly, Brittany Garza, Samantha Baker, Naomi E. Levin, Jessica E.
Tierney, Christoph Häggi, Sarah J. Feakins (2022). Identifying the drivers of GDGT distributions
in alkaline soil profiles within the Serengeti ecosystem. Organic Geochemistry, 169, 104433
Abstract
Surface soil glycerol dialkyl glycerol tetraether (GDGT) distributions are influenced by mean
annual air temperature as well as soil pH. However, the controls on GDGT distributions with
depth in soil profiles are less well-known. We report a study of soil profiles in warm, carbonate-
precipitating, alkali soils in the Serengeti ecosystem, Tanzania. Measurements of temperature,
pH, salinity, and complementary data available from carbonates and organics from the same soil
pits provide an interpretive framework for the observed patterns in branched (br-) and
isoprenoidal (iso-) GDGTs in soil profiles. While brGDGT distributions at the soil surface
primarily reflect mean annual temperature, a warm bias at depth indicates additional sub-surface
controls on brGDGT distributions. We consider whether degradation or in situ production in
response to alkaline pH and salinity also modulate brGDGTs. We find that the Archaeol
Caldarchaeol Ecometric (ACE) index correlates with soil salinity, which both increase with
depth. These results support in situ microbial production in deeper soil settings, with pH and
salinity controlling the microbial community composition. We also compared brGDGT-predicted
65
mean annual air temperatures (MAAT) to published clumped isotope thermometry on carbonates
in the same soils and found that the median temperatures of both proxies were the same at 23 °C.
We suggest a further comparison of proxy performance in carbonate-bearing soils and geological
archives. Differences in the nature of the proxy recorders may broaden sample availability for
paleothermometry and help to identify confounding factors in each proxy system.
Keywords: soil profile; Serengeti; GDGT; carbon isotope; temperature.
3.1 Introduction
Branched glycerol dialkyl glycerol tetraether (brGDGTs) are membrane lipids composed of a
branched alkyl chain with 4 to 6 methyl groups (De Jonge et al., 2014a, p.98) which are
produced by bacteria and are found in a variety of environments including soils, peats, lakes, and
coastal marine environments. However, the identity of the bacteria which produce brGDGTs is
still mostly unknown (Weijers et al., 2006, 2009; Sinninghe Damsté et al., 2011, 2014, 2018).
Branched glycerol dialkyl glycerol tetraether (brGDGTs) lipids have been shown to be
reliable recorders of temperature in modern surface soils and peats (Weijers et al., 2007; De
Jonge et al., 2014a; Naafs et al., 2017; Dearing Crampton-Flood et al., 2020). This is likely due
to bacteria changing the degree of methylation of membrane lipids under different temperature
conditions to adjust the rigidity of their membranes (Weijers et al., 2007; Naafs et al., 2021).
Whilst much progress has been made on how different environmental variables (e.g.
temperatures, pH, and soil moisture content) affect the distributions of brGDGTs in surface soils
(e.g. Peterse et al., 2009; Huguet et al., 2010; Dirghangi et al., 2013; Menges et al., 2014; Dang
et al., 2016), our knowledge of brGDGTs in deeper soils (> 0.1 m) is limited (Davtian et al.,
2016). Interpretations of controlling factors can be divided into two classes: temporal (i.e., to
what extent does modern production contribute to deep soil brGDGTs) and environmental (how
66
would chemical conditions deeper in soil affect brGDGTs). In surface soils,
13
C-labelling studies
have shown turnover of brGDGTs on the timescales of decades (Weijers et al., 2010; Huguet et
al., 2017). In deeper soil horizons, compound-specific radiocarbon analyses of brGDGTs have
yielded mean turnover times of 1400–5900 years in mid-latitude soils which is a slower turnover
rate compared to most other particulate organic carbon components (Gies et al., 2021).
Although surface soils have been the focus of most brGDGT calibration efforts (De Jonge et
al., 2014b; Naafs et al., 2017; Dearing Crampton-Flood et al., 2020), understanding the processes
that control brGDGT distributions at depth, including the transfer of extractable GDGTs to a
non-extractable pool, organo-metallic complexations in Podzols, growth depth and post-
depositional effects on GDGT distributions (Huguet et al., 2010; Zech et al., 2012; Yamamoto et
al., 2016) is important when applying brGDGT proxies in paleosols and loess-paleosol sequences
(Peterse et al., 2014) for both paleoclimate and paleoaltimetry studies (Coffinet et al., 2017; Bai
et al., 2018; Y. Li et al., 2018; Feng et al., 2019) where the deeper soils are often all that is
preserved (Cleveland et al., 2007; Tabor and Myers, 2015; Beverly et al., 2018).
In addition to the brGDGTs that have been relatively well-studied in global surface soils,
isoprenoidal GDGTs (isoGDGTs) produced by archaea are abundant in dryland alkali surface
soils (Yang et al., 2014; H. Wang et al., 2017), and thus might be useful for diagnosing paleo soil
moisture content (Xie et al., 2012; Yang et al., 2014). IsoGDGTs are composed of two C
40
isoprenoid chains with several cyclopentane and cyclohexane rings connected by ether bonds to
two terminal glycerol groups (Li et al., 2016, p.114). Here, we measure the distribution of both
brGDGTs and isoGDGTs at multiple soil sites and depths across a precipitation gradient in the
Serengeti grassland. The 5-methyl brGDGT index (MBT′
5Me
) correlates with the MAAT of soils
on both global (De Jonge et al., 2014b) and local scales (Yang et al., 2015). MBT′
5Me
is similar to
67
the methylation (MBT′) index
,
except that MBT′
5Me
excludes 6-methyl brGDGTs and
consequently shows a stronger relationship with MAAT (De Jonge et al., 2014b). We derived
brGDGT temperature estimates using a global soil calibration of the MBT′
5Me
index to MAAT
(Dearing Crampton-Flood et al., 2020). We then compared the known MAAT (Fick and
Hijmans, 2017) between surface samples (<0.1 m) and deeper samples (0.1 to 1.6 m), to
determine if deeper soil samples are dominated by 1) modern brGDGT production reflective of
surface environmental conditions, 2) modern production reflecting deep soil environmental
conditions or 3) by fossil brGDGTs reflecting past surface or deep soil conditions. To
discriminate between these options, we evaluate changes in GDGT concentrations and compare
them to a suite of complementary environmental and biomarker information including
temperature, pH, salinity, and organic (biomarker) concentrations. We also calculate predicted
soil pH based on CBT′ (De Jonge et al., 2014b) and compare this to measured soil pH in both
surface and subsurface samples.
Additionally, these soil profiles have already been studied for clumped isotopes (Beverly et
al., 2021) allowing us to compare the carbonate and organic paleothermometers. Clumped
isotope thermometry records the formation temperature of carbonate and is based on the
abundance of the doubly substituted isotopologue
13
C
18
O
16
O (reported as
47
), which is
proportional to the temperature-dependent ordering of
13
C and
18
O in carbonate minerals (Ghosh
et al., 2006; Schauble et al., 2006). Both brGDGT and clumped isotope thermometry have been
calibrated in modern environments (Kelson et al., 2017, 2020; Dearing Crampton-Flood et al.,
2020) and are relatively stable over geological time (Finnegan et al., 2011; Kemp et al., 2014).
However, to our knowledge, these proxies have not been directly compared at the same site.
Comparison in the same soil profiles allows us to constrain how each proxy responds to the same
68
environmental conditions. We use these modern soil profile observations to offer suggestions to
guide the future use of these proxies in paleoenvironmental studies.
3.1.1 Study location and climate
The Serengeti ecosystem spans 30,000 km
2
in Tanzania and Kenya, in eastern Africa. A transect
of 11 sites was used to study the soils in the region (34–36°E, 1–2°S, 1,153–1,667 m above sea
level; Fig 3.1). Mean annual air temperatures (MAAT) at these sites vary from 19.2–22.8 °C and
mean annual precipitation (MAP) ranges from 499–846 mm (Table 1; Fick and Hijmans, 2017).
69
Table 3.1. Site locations and characteristics
Site Latitude Longitude
a
Elevation
(masl)
b
Vegetation
c
Soil
Order
d
MAAT
(°C)
e
MTWQ
(°C)
e
MAP
(mm/yr)
e
PET
(mm/yr)
f
AI
f
WD
(mm/yr)
f
Malambo Road 2.96 35.43 1354 Dense
grassland
Inceptisol 20.9 22.0 499 1583 0.33 1084
Shifting Sands 2.93 35.24 1549 Dense
shrubbed
grassland
Inceptisol 19.8 20.9 558 1521 0.38 963
Naabi Hill 2.83 35.02 1677 Closed
shrubbed
grassland
Mollisol 19.2 20.1 734 1486 0.52 752
Simba Kopjes 2.61 34.89 1637 Dense to
closed
grassland
Mollisol 19.6 20.5 805 1521 0.54 716
Makoma 2.49 34.75 1549 Closed treed
shrubland
Inceptisol 20.2 21.0 829 1560 0.55 731
Nyaruswiga 2.34 34.82 1451 Open treed
grassland to
closed
grassland
Mollisol 20.8 21.6 832 1613 0.51 781
Banagi 2.32 34.84 1425 Mixed open
grassland to
woodland
Inceptisol 20.9 21.8 819 1622 0.51 803
Kemarishe 2.24 34.64 1315 Open grassed
woodland
Alfisol 21.6 22.3 834 1659 0.51 825
Musabi 2.27 34.53 1278 Closed
grassland
Mollisol 21.9 22.5 830 1659 0.51 829
Kirawira 2.18 34.23 1215 Dense to
closed
grassland
Mollisol 22.4 22.9 838 1655 0.53 817
Ndabaka 2.16 33.97 1153 Dense to
closed
grassland
Vertic
Mollisol
22.8 23.2 846 1642 0.55 796
70
There are two rainy seasons: March to May, and October to November (Norton‐Griffiths et al.,
1975). The vegetation is mostly grassland plains, dissected by riverine forest and woodlands
based on Landsat remote sensing and ground-truthing in 1998–2002 (Reed et al., 2009) and
corresponds to observed conditions in 2018 when samples were collected. Most sites are
grasslands (<10% woody cover), except for three locations (Banagi, 80–100% shrubs;
Kemarishe and Makoma; 20–50% tree canopy cover).
Fig 3.1. A) Location of the Serengeti ecosystem within Africa. B) Map of study region showing the
Serengeti ecosystem spanning Tanzania and Kenya, the soil sampling locations (yellow points) along a
transect within the Ngorongoro Conservation Area and Serengeti National Parks in Tanzania. The
71
vegetation map was simplified from Reed et al. (2009) to show only grassland, shrubland, woodland, and
forest vegetation. Figure from Zhang et al., (2021) and reprinted with permission.
With some exceptions (e.g. Anderson et al., 2014; Pérez-Angel et al., 2020), most global soil
calibrations have been performed in the absence of detailed soil temperature logging, and thus
soils are calibrated to MAAT obtained from global temperature products. Similarly, we obtain
MAAT from WorldClim2 (Fick and Hijmans, 2017), which has a regional (eastern Africa)
average temperature RMSE of <0.75 °C and a spatial resolution of 1 km. In these Serengeti soils,
we additionally have in situ measurements of temperatures in soil profiles at four sites (Ndabaka,
Musabi, Kemarishe, and Naabi Hill) using nine HOBO 64 K Pendant
®
temperature loggers
(accuracy of ±0.53 °C) buried at depths ranging from 20–150 cm (Beverly et al., 2021) between
22/02/2018 to 26/01/2019.
3.2 Materials and methods
3.2.1 Soil sampling
Soil samples were collected in February 2018, as part of a prior study of soil carbonates (Beverly
et al., 2021). Soil profiles (n = 11) were studied by digging soil pits till refusal (up to 160 cm),
often to indurated petrocalcic horizons. Soils were classified in the field according to USDA soil
order descriptors and comprise Incepticols (n = 4), Alfisol (n = 1), and Mollisols (n = 6).
Carbonates were present below 40 cm at all but 2 sites and these nodules were sampled and
studied for clumped isotopes as reported in Beverly et al., (2021).
Samples for lab pH analyses were collected at all sites at 10 cm resolution. Samples for
biomarker analyses were collected from the top 10 cm of soil (below the litter layer), in the
middle of the profiles at ~60 cm, and in the deepest sector of the profiles at around 100–160 cm.
Each sample averages three sub-samples from the 1.5 m width of the soil pit. The soil samples
designated for organic analyses were previously studied for total organic carbon concentration
72
(TOC), bulk carbon isotopic composition, as well as plant wax n-alkane and n-alkanoic acid
concentrations and isotopes (Zhang et al., 2021). Here we use the neutral polar fraction, purified
from the total lipid extracts in the same soil samples, to study the GDGTs.
3.2.2 Soil properties
Soil pH, Electrical Conductivity (EC), and total dissolved solids (TDS) were measured at the
University of Houston using a Hanna Instruments HI 9813-6 meter. Before measurements the
meter was equilibrated to ambient temperature, pH was calibrated using Hanna Instruments
buffer solution HI7007 (pH 7.01) and EC/TDS were calibrated using EC calibration solution HI
70442 (1.41 mS/cm). TDS is determined by conversion from EC within the instrument and thus
TDS range is dependent upon the calibration of EC. A slurry of each soil with distilled water (1:2
v/v) was stirred vigorously to break up any soil particle aggregates. The pH, EC, and TDS were
measured after 15 minutes to allow the solution to equilibrate and particulate matter to settle. The
methodology is similar to that of Weijers et al., (2007).
3.2.3 GDGT preparation and analyses
Lipid extractions from these soils were previously described (Zhang et al., 2021). Briefly, in that
study, soil samples were freeze-dried and lipids were extracted from 26 to 85 g of soil by
Accelerated Solvent Extraction at the University of Southern California. The neutral polar
fraction, containing GDGTs, was isolated by column chemistry and the purified fractions were
stored at –20 °C. For this study, the neutral polar fractions were dissolved in hexane:isopropanol
(99:1) and filtered through a 0.45 μm polytetrafluoroethylene filter. Samples were then injected
into an Agilent 1260 High-Performance Liquid Chromatograph (HPLC) coupled to an Agilent
6120 mass spectrometer at the University of Arizona. Separation of the GDGTs was
accomplished using two Ethylene Bridged Hybrid (BEH) Hydrophilic Interaction (HILIC)
73
silica columns (2.1 mm × 150 mm, 1.7 μm; Waters) following the method of Hopmans et al.
(2016). Single Ion Monitoring (SIM) of the protonated molecules (M + H
+
ions) was used to
detect and quantify GDGTs with abundances determined by comparison to a C
46
internal
standard at m/z 744 (Huguet et al., 2006). As the relative response factor between the internal
standard and brGDGTs was not monitored, these concentrations should be considered semi-
quantitative.
We report concentrations of summed branched (brGDGT) and isoprenoidal (isoGDGTs)
GDGTs per unit mass of soil (ng/g units) and normalized to OC in ng/g OC as summed branched
(ΛbrGDGT) and isoprenoidal (ΛisoGDGTs), as well as the concentrations of individual
compounds that are further used in a range of index calculations as follows. To assess the
temperature information in the brGDGT abundances, we calculate the MBT´
5Me
(De Jonge et al.,
2014b, Hopmans et al., 2016) where
MBT
5Me
′
=
Ia+Ib+Ic
Ia+Ib+Ic+IIa+IIb+IIc+IIIa
(1)
We also calculate the pH-sensitive CBT´ index (De Jonge et al., 2014) where
CBT
′
= log
10
[
Ic+IIa
′
+IIb
′
+IIc
′
+IIIa
′
+IIIb
′
+IIIc
′
Ia+IIa+IIIa
] (2)
CBT´ has been globally calibrated to pH:
pH = 7.15 + 1.59 × CBT
′
(3)
The IR
6Me
index expresses the relative amount of C6 vs C5 methylated brGDGTs (De Jonge et
al., 2014b) and is thought to be higher in soils with lower moisture content (Dang et al., 2016)
and has shown a significant correlation to soil pH (Naafs et al. 2017).
IR
6Me
= ∑ (C6 − methylated brGDGTs)/∑ (C6 − methylated brGDGTs + C5 −
methylated brGDGTs) (4)
74
We also quantify the isoprenoid GDGTs. In addition to the isoGDGTs, we calculate the
Branched and Isoprenoidal Tetraether (BIT) index:
BIT =
Ia+IIa+IIa
′
+IIIa+IIIa′
Ia+IIa+IIa
′
+IIIa+IIIa′+Cren
(5)
where Ia, IIa, and IIIa represent the abundances of both the 5´ and 6´ methyl isomers of the non-
cyclic terrestrial brGDGTs (from soil acidobacteria) and Cren represents the abundance of
crenarchaeol (Hopmans et al., 2004). BIT is commonly used in aquatic depositional settings, to
detect terrestrial soil inputs assuming soils have a BIT 1, however in arid soils values <1 have
been reported (Dirghangi et al., 2013; Yang et al., 2014). The relative amount of isoGDGTs in
soils has been proposed to be an index for aridity (Xie et al., 2012) where:
R
i/b
= ∑ isoGDGTs/∑ brGDGTs (6)
The archaeol caldarchaeol ecometric (ACE) was developed as a proxy for salinity in aquatic
settings, and is based on the relative abundance of archaeol (considered to represent halophilic
archaea) versus caldarchaeol (also referred to as isoGDGT-0) (Turich and Freeman, 2011):
ACE =
archaeol
archaeol+caldarchaeol
× 100 (7)
Here, we assess the ACE response to soil salinity.
3.3 Results
3.3.1 Soil temperature and pH measurements
In situ measurements of soil, temperatures range from 16.7 to 30.2 °C at depths ranging from 20
to 30 cm, as previously reported by Beverly et al. (2021). Soil temperatures are warmer than the
interpolated mean annual temperatures which range from 19.2 to 22.8 °C (Fick and Hijmans,
2017; Beverly et al., 2021) and measured MAAT of 21.0 °C (Jager, 1982). These soil samples,
therefore, contribute to our understanding of proxy performance and environmental
75
interpretations at the warm end of global soil calibrations (Fig 3.2A) and grassland or bare soil
settings where soil temperatures can be hotter than air temperatures. In addition, we measured
soil pH and find surface soil pH values ranging from 6.6 to 9 with a median value of 8.7 (Fig
3.2B; Pangaea dataset; Peaple et al., 2021a). Deeper samples typically have a higher pH than
surface samples with a range of 7.4–10.3 and a median value of 9.1. Our data fills a gap in
current GDGT compilations which contain few data from high pH soils.
Fig 3.2. Serengeti soil data compared to prior global surface soil calibrations. A) Mean annual air
temperature (MAAT) plotted against MBT´5Me for the global compilation of surface soil samples (grey
symbols) (Dearing Crampton-Flood et al., 2020) and the surface (<0.1 m; solid blue symbols) and deep
(>0.1 m; open blue symbols) soil samples from this study. B) Soil pH plotted against CBT´ from a global
compilation of surface soil samples (De Jonge et al., 2014) and soils from this study. C) Soil pH plotted
against IR6Me in both a global compilation (Dearing Crampton-Flood et al., 2020) and soils from this
study.
3.3.2 Depth profiles
We report the GDGT compositions at 11 sites, and 26 samples across the Serengeti, including
relative and absolute abundances (Pangaea dataset; Peaple et al., 2021a). Within soil profiles at
individual sites, brGDGTs (Fig 3.3A) and isoGDGTs (Fig 3.3B) decrease with depth,
mirroring the declines from the surface downward in the TOC from <3% at the surface to 0.14%
at depth (Fig 3.3C) (Zhang et al., 2021). brGDGTs decline from 4–168 ng/g at the surface to 2–
76
29 ng/g below the surface (Fig 3.3A). Normalizing brGDGTs
for TOC, we find ΛbrGDGTs (Fig
3.3D) range from 427 to 9009 ng/g OC at the surface to 348–11,561 ng/g OC at depth (deepest
samples at 1–1.6 m). isoGDGTs decline from surface concentrations of 1–10 ng/g to <1 – 5.
ng/g at depth (Fig 3.3B). ΛisoGDGTs range from 79 to 803 ng/g OC at the surface to 87–883
ng/g OC at depth (Fig 3.3E, deepest samples at 1–1.6 m). The decline in isoGDGT
concentration is proportional to that of TOC such that ΛisoGDGTs are relatively invariant. The
ratio between isoGDGTs and brGDGTs (R
i/b
, Fig 3.3F) is relatively invariant with depth, with all
sites except Malambo Road having an R
i/b
of < 0.51 demonstrating the greater abundance of
brGDGT molecules relative to isoGDGT. Malambo Road has an R
i/b
of 0.89 at the surface and
1.1 at 0.8 m depth, indicating an almost equal abundance of brGDGT and isoGDGT compounds.
BIT is also relatively low at the surface of Malambo Road (0.57), although it increases to 0.91 at
0.8 m depth. Apart from Shifting Sands which has a surface BIT of 0.67, all other sites have BIT
values greater than 0.8 (Fig 3.3G). Plant wax concentrations measured from paired samples
(Zhang et al., 2021) show decreases with depth (Fig 3.3O, P) that are proportional to the
decreases in TOC%. There is no consistent trend in changes in IR
6Me
with depth (Fig 3.3K), with
some sites showing large >0.2 increases (e.g., Banagi and Nyaruswiga) and other sites showing
little change (<0.1) with depth (e.g., Makoma and Kirawira and other sites showing small
decreases (e.g., Musabi).
77
Fig 3.3. Depth profiles of the 11 sites showing A) brGDGT concentrations, B) isoGDGT
concentrations, C) Total organic carbon (TOC) % (Zhang et al., 2021), D) ΛbrGDGT concentrations, E)
ΛisoGDGT concentrations, F) Ri/b, G) BIT, H) ACE, I) MBT´5Me, J) CBT´, K) IR6Me, L) Total organic
carbon δ
13
C (TOC δ
13
C) (Zhang et al., 2021), M) Total dissolved solids (TDS), N) Measured pH, O) C23-
33 alkanes (μg/g) and P) C22-32 alkanoic acids (μg/g).
78
The temperature-sensitive MBT′
5Me
proxy is observed to increase with depth in most profiles
(Fig 3.3I),
with surface sample (mean = 0.82, standard deviation = 0.07) vs deeper samples
(mean = 0.89, standard deviation = 0.07). These values predict BayMBT
0
temperatures of 20.6
°C (surface) and 23.2 °C (deep), implying a warming trend with depth that is not expected from
theory, nor represented by ambient measurements of soil temperature by soil-temperature
dataloggers (Beverly et al., 2021). The community index (De Jonge et al., 2019) (not shown)
correlates strongly with MBT′
5Me
(r = 0.99, p < 0.0001) and consequently shows similar trends
with soil depth (Fig 3.3I). Community index values are > 0.6 throughout the data set and are
thereby consistently in the warm cluster.
The pH-sensitive CBT' shows an increase of >0.25 down profile in Ndabaka, Nyrauswiga, and
Kirawira with relative invariance with depth in the other sites (Fig 3.3J). Measured pH increases
with depth at all sites except Musabi where it is invariant (Fig 3.3N). The ACE index (salinity
proxy) (Fig 3.3H) increases with depth at all sites (except Kemarishe), with surface samples
having a mean value of 12 compared to deeper samples which have a mean value of 53.
Similarly, TDS (a measure of salinity) (Fig 3.3M) values also generally show large increases
with depth (deepest samples 1.7–29.5 times greater TDS than surface samples), except for
Kirawira which sees a large increase followed by a decrease at the bottom of the profile. For
comparison, we also show that TOC δ
13
C increases with depth in most profiles (Fig 3.3L), with
Makoma, Banagi, and Kemarishe showing especially large > 3‰ increases in δ
13
C with depth
(Zhang et al., 2021).
3.3.3 Surface soil GDGT concentrations
Surface soil ∑brGDGT concentrations range from 4–168 ng/g, with a mean of 32 ng/g (Fig
3.3A). Concentrations of ∑isoGDGTs range from 1–10 ng/g with a mean of 3 ng/g (Fig 3.3B).
79
The R
i/b
and BIT range from 0.01 to 0.89 and from 0.57 to 0.99, respectively (Fig 3.3F and G).
The BIT index shows a positive correlation with MAP (Fig 3.4A) (r = 0.87, p <0.01) and the R
i/b
shows a negative correlation with MAP (Fig 3.4B) (r = –0.83, p <0.01). TOC % normalized
isoGDGTs correlate negatively with precipitation (r = –0.85, p <0.01) (Fig 3.4D) with the two
lowest precipitation sites (Shifting Sands and Malambo) having on average 4 times the amount
of isoGDGTs compared to the wetter sites, with the correlation driven by lower TOC values at
the drier sites. brGDGTs show a weaker non-significant positive correlation with MAP (r =
0.29, p = 0.38) (Fig 3.4C).
Fig 3.4. Relationship between mean annual precipitation (MAP) and surface samples A) Ri/b, B) BIT, C)
ΛbrGDGT concentrations (ng/g OC), D) ΛisoGDGT concentrations (ng/g OC).
80
3.3.4 Temperature
Based on the relative abundances of the brGDGTs in each soil sample, we calculated MBT´
5Me
index and then predicted temperatures using BayMBT
0
, a Bayesian model calibrated to months
above freezing (MAF), which is the average temperature of all months >0 °C (Dearing
Crampton-Flood et al., 2020), which is the same as MAAT in tropical Africa. BayMBT
0
calculated surface air temperatures fall within the MAAT climatology for the region (Fig 3.5A)
but are cooler (2–4 °C) than the corresponding measured soil temperature (Fig 3.5C).
Considering the Serengeti soils in isolation, the temperature span of the transect is limited,
nevertheless, BayMBT
0
temperatures from surface soils (0–0.1 m) show a strong correlation with
site MAAT (r = 0.73, p = 0.01) and a low RMSE (1.8) (Fig 3.5A). If added to the global
compilation these new surface soil data would bolster observations for the under-sampled warm
end of the global calibration (20–25 °C). While the global compilation only considers surface
soils, brGDGTs at depth are not yet well described. Here we find in deep soil (>0.1 m) there is
no correlation between MAAT and the BayMBT
0
predicted temperatures (Fig 3.5B).
81
Fig 3.5. Comparison of measured mean annual soil temperature (MAST) and predicted soil temperature,
modeled air temperature and pH at both the surface (left column, filled blue symbols) and at depth (right
column, open blue symbols). A) and B). Cross plots of mean annual air temperature (Fick and Hijmans,
2017) against mean temperature from months above freezing (MAF) predicted by BayMBT0. C. D. Cross
plots of the measured soil temperature at nearest depth to brGDGT sample (Beverly et al., 2021) against
BayMBT0. E) and F). Cross plots of measured soil pH against CBT' predicted pH.
There is a weak correlation between the mean annual precipitation and the temperature
regression residuals although the small number of low precipitation sites (n = 2) precludes
82
further comment on a causal relationship (Fig 3.6A). The ratio of 6 versus 5 methyl groups in
Serengeti soil samples (IR
6Me
) is relatively high (median = 0.72) and is correlated with the
MAAT absolute residuals (measured-predicted temperature) r = 0.69, p <0.01 (Fig 3.6B).
Across the longitudinal transect, the BayMBT
0
predicted temperatures from surface soils show a
similar pattern to the MAAT, with lower temperatures in the higher elevation eastern part of the
transect and higher temperatures in the lower elevation west section (Fig. B.1). There are no
clear trends in the residuals associated with longitude.
Fig 3.6. Exploring possible factors influencing mean annual air temperature (MAAT) residuals (MAAT-
predicted MAAT) in all soil samples (surface <0.1 m, solid circles; deep >0.1 m, open symbols),
including A) mean annual precipitation (MAP), B) IR6Me, C) measured pH and D) total dissolved solids
(TDS). The Gray dashed line represents 0 residual.
3.3.5 pH
The CBT´ index yields inferred pH estimates that are lower than measured values, and CBT´
values for these alkali soils (up to pH 9) are lower than those measured at pH >6 elsewhere in the
world, demonstrating an issue with the application of the global calibration at our sites, and
emphasizing the need for studies in alkali soils (Fig 3.2B). Surface samples showed a weak non-
statistically significant positive relationship between measured pH and predicted pH (r = 0.45, p
= 0.17, Fig 3.5E), with an RMSE of 1.4. Measured samples are uniformly more alkaline than
83
CBT´
based pH predictions. Samples collected at depth >0.1 m had a weaker, non-significant
relationship between measured pH and predicted pH (r = 0.39, p = 0.17) (Fig 3.5F) and plotted
outside of the global calibration range (Fig 3.2B).
The pH residual is greatest at both the far eastern and western ends of the transect (Fig. B.1).
This residual correlates with TOC in the surface soil (r = –0.83, p < 0.0001), likely a
consequence of the strong correlation between TOC % and measured pH (r = –0.78, p < 0.0001).
The two driest sites (Shifting Sands and Malambo) both have the highest measured pH in the soil
surface (Fig 3.3N), although high pH is not exclusive to low precipitation sites (i.e., the mean pH
for MAP >550 mm/year = 7.5 whilst the mean pH for sites with MAP < 560 mm/year = 9.7). A
positive correlation (r = 0.67, p < 0.05) exists between IR
6Me
and pH in surface samples (Fig
3.2C) although this is not present in deeper more alkaline samples (r = -0.44, p = 0.12).
3.4 Discussion
3.4.1 GDGT concentrations
There is a relationship between the relative amounts of brGDGT and isoGDGT (R
i/b
) and the
mean annual precipitation at the soil sites (Fig 3.4B). The R
i/b
shows a negative correlation with
MAP (r = –0.73, p < 0.0001 correlation driven by the two driest sites) and the BIT index shows a
positive correlation with MAP (r = 0.83, p < 0.0001). As MAP is loosely related to soil moisture
content, we interpret that the soil moisture content is controlling R
i/b
in the surface soils. The
same relationship between MAP and both R
i/b
and BIT has been observed in some arid and semi-
arid Chinese and North American soil sites (Yang et al., 2014; H. Wang et al., 2017) where it
was primarily driven by changes in the brGDGT concentration, with the absolute isoGDGT
concentrations less responsive to changing moisture conditions. We observe the same
relationship here (Fig 4. C, D), with the two lowest MAP soil sites (Shifting Sands and Malambo
84
Road) having substantially lower mean brGDGT than the wetter sites (5 vs 39 ng/g
respectively) (Figs. 3A, 4C). In down-profile samples, the relationship of decreased BIT values
in arid regions with a MAP < 600 mm/yr does not persist as values converge to 1 in the lower
soil horizons (Fig 3.3G). Likewise, the elevated R
i/b
values in arid regions also decrease towards
0 in most profiles (Fig 3.3F). This pattern can likely be explained by the disproportionate
decrease in isoGDGTs resulting from a decrease in the activity of isoGDGT-producing aerobic
archaea of the phylum Thaumarchaeota due to lower oxygen availability. This interpretation is
in line with previous studies that have suggested that the low BIT and high R
i/b
values in arid
areas result from the favorable conditions for NH
3
oxidation by Thaumarchaeota in these
environments (Xie et al., 2012; Dirghangi et al., 2013).
As in most soil profiles, TOC declines with depth and so do the biomarker concentrations (Fig
3.3A-E, O, P) linked to the processes of surface inputs and downward mixing and decay. We
find the concentration of isoGDGTs in the soil profiles is strongly correlated with the TOC% of
the sample (r = 0.75, p < 0.0001) although brGDGTs show a much weaker and non-significant
correlation (r = 0.35, p = 0.09), suggesting that brGDGT production or preservation is only partly
related to that of isoGDGTs and TOC. The magnitude of the decline with depth differs with
GDGTs declining proportionally less than TOC (median 62% decline of isoGDGTs, 58% of
brGDGTs vs 70% for TOC), suggesting that these molecules are relatively resistant to
degradation versus other components of organic matter, as also reported in previous studies on
soils (Huguet et al., 2010a) as well as in loess-paleosol sequences (H. Wang et al., 2017). An
increase in the amount of brGDGTs with depth was recorded in 7 out of 15 samples, which we
interpret as representing in situ production of brGDGTs in addition to a fossil brGDGT
component. Whilst downward mixing could in theory result in the transfer of more organic-rich
85
surface soil deeper into a soil profile, we see declines in the TOC% with depth at every soil site
indicating that this explanation is unlikely.
3.4.2 ACE and soil salinity
In these dryland soils, with salts and carbonate precipitating, we find an increase in total
dissolved solids with depth (Fig 3.3M) as well as variations between sites, with the more alkaline
sites (e.g., Shifting Sands) showing higher salt concentrations. TDS variations between sites and
with depths covary with organic concentrations (Fig 3.7) with deeper, more saline site samples
having less TOC. We find an inverse power law relationship between total dissolved solids, a
proxy for salinity, and bulk organics (Fig 3.7C) and the concentration of isoGDGTs (Fig 3.7A)
but not with brGDGTs (Fig 3.7B). Soil salinity can affect soil organic carbon content in two
ways 1) by decreasing the input of plant and microbial-derived carbon into the soil or 2) by
reducing the rate of decomposition and thus increasing organic carbon (Setia et al., 2013).
However, TDS also increases with depth (Fig 3.3H) so this relationship could be explained by
deeper samples having undergone more organic degradation (Fig 3.7D-F). However, we find a
stronger relationship between TDS and isoGDGTs (Fig 3.7A) than depth and isoGDGTs (Fig
3.7D) (r
2
= 0.64 vs 0.34) suggesting that TDS content is the controlling variable. Additionally,
we calculated a partial correlation coefficient of r = -0.67, p <0.01 between TDS and
isoGDGTs whilst controlling for depth, showing that TDS and isoGDGTs are correlated even
when accounting for depth. Whilst TOC has a stronger relationship with depth than TDS (r
2
=
0.66 vs 0.52) when considering all the samples if the surface samples are excluded there is no
relationship between depth and TOC, yet there is between TDS and TOC (r
2
= 0.01 vs 0.35)
suggesting that again TDS is the controlling variable. We also calculated a partial correlation
coefficient of r = -0.40, p = 0.045 between TOC and TDS whilst controlling for depth. As such,
86
it appears that increased salinity leads to increased organic carbon degradation and a power-law
relationship explains both the inverse relationship with depth as well as the variations between
sites for TOC and isoGDGTs but to a lesser extent that of the brGDGTs.
Fig 3.7. Comparison of total dissolved solids (TDS) versus A) isoGDGTs, B) brGDGTs, and C) TOC
and sampling depth vs E) isoGDGTs, F) brGDGTs and G) TOC. We show the amount of variance
explained by a power-law regression as well as the Spearman rank correlation coefficient and
corresponding p-value. Solid blue symbols represent surface samples and white filled symbols represent
deep soil samples.
We also see a change in the ACE index with depth (proportion of halophilic-biomarker archaeol
vs halophobic-biomarker caldarchaeol), with surface samples having notably lower ACE values
than deeper samples (mean surface = 12, 1 = 18, mean deep = 53, 1 = 20, Fig 3.3H). The
ACE index has traditionally been applied to marine (Turich and Freeman, 2011) and lacustrine
water samples (Wang et al., 2013; He et al., 2020; Peaple et al., 2021b) to track the relative
proportion of halophilic archaea and Euryarchaeota which produce both archaeol and GDGT-0,
87
“caldarchaeol”. As halophilic archaea are known to inhabit saline soils (e.g., Lizama et al.,
2001), we assessed ACE in what we believe is the first such application to soils. Over all
samples and sites, we see a moderate positive correlation of r = 0.73, p < 0.002 between ACE
and total dissolved solids in our samples (Fig 3.8), which suggests that the proportion of
halophilic archaea in soils and thus the ACE index responds linearly to salinity, similarly to the
behavior of archaea and ACE in hypersaline, aquatic settings.
Fig 3.8. Cross plot of measured Total dissolved solids (TDS) and calculated ACE index in soil samples,
showing surface samples (solid symbols) and deep (open symbols).
3.4.3 Factors influencing BayBMT
0
in Serengeti soils
Low mean annual precipitation (MAP) has been reported to lead to higher MAAT residuals
(Peterse et al., 2012; De Jonge et al., 2014b; Menges et al., 2014; Dang et al., 2016). The recent
global soil compilation (Dearing Crampton-Flood et al., 2020) reported no correlation between
MAP and MAAT residuals; however, there was a greater variance in samples from MAP <500
mm/year. In this study, we see that both surface and deep soil samples from lower precipitation
sites have positive temperature residuals (1.8 °C vs 2.8 °C, respectively, Fig 3.6A) and the mean
88
absolute residual is slightly higher than from the higher precipitation samples although the low
sample size for the <560 mm/yr (n = 4) precludes firm conclusions.
Additionally, the proportion of 5 to 6 methylated compounds (IR
6Me
) may affect the MBT´
5Me
temperature proxy skill, with samples with a greater proportion of 6 methylated compounds (IR
>0.5) having brGDGT distributions that were less responsive to temperature (Dang et al., 2016;
Naafs et al., 2017). However, recent global compilations show that the effect is limited to IR>0.8
(Dearing Crampton-Flood et al., 2020). Here we also find evidence for a threshold at IR
6Me
= 0.8
among the deep samples, with samples with an IR
6Me
>0.8 having a large negative MAAT
residual (Fig 3.6B) implying the BayMBT
0
predicted temperatures
have a warm bias (mean =
5.2, n = 4) versus small residuals in those with IR<0.8 (mean = 1.7, n = 21).
pH can have a significant although not unidirectional effect on MBT´
5Me
values (De Jonge et al.,
2019, 2021; Halffman et al., 2022) due to the influence of changing bacterial community
composition on brGDGT lipid relative abundance. Among the surface samples in our study, pH
does not have a relationship with MAT residuals, however, there are trends in the MAT residuals
with regards to the deeper samples with less alkaline samples having more negative residuals,
implying a bias toward warmer predicted temperatures (Fig 3.6C). We tested the hypothesis that
the mean temperature residuals were different between deep high pH samples (pH >9) and deep
lower pH (pH<9) by performing a one-way Anova test. We found that the mean residual for the
deep low pH samples was greater than the high pH samples (4.0 vs 1.9 respectively) and that this
difference was statistically meaningful (F value (5.2) > F critical value (4.7) and p = 0.04).
Salinity has been demonstrated to impact MBT´
5Me
values in lakes (Wang et al., 2021) by
controlling the isomerization of brGDGTs. Although some previous soil-based brGDGT studies
found a negative correlation between MBT´
5Me
and soil salinity (Zang et al., 2018), other studies
89
did not (Dang et al., 2016). In this study, we do not find salinity affects the MAAT residuals (Fig
3.6D), suggesting it is not playing a significant role in determining brGDGT distributions in
these Serengeti soils.
3.4.4 Offsets between surface and deep brGDGT predicted temperatures
Whilst brGDGTs samples from surface soils are reflective of modern MAAT, and to a lesser
extent pH (Fig 3.5A, C, E), there is no relationship between deeper samples and modern surface
environmental conditions (Fig 3.5B, D, F). Previous studies have also reported significant
variability in soil MBT′ (Davtian et al., 2016) and MBT′
5Me
with depth (Pei et al., 2021) with
potential explanations focused on modern production under lower oxygen conditions in deeper
soils (Pei et al., 2021), land-use changes, degradation and varying chemical compositions down
profile (Davtian et al., 2016). In a collection of Vietnamese soils (Davtian et al., 2016), podzols
were identified as showing large decreases in MBT´ with depth relative to the surface soils,
although other soil types did not show this change. Changes in CBT (pH-sensitive cyclization of
branched tetraethers index) with depth were variable and did not depend on soil type (Davtian et
al., 2016). Lower oxygen conditions were linked to higher MBT' and lower CBT in soils sampled
from Zhoukoudian and Mt. Changbai, China through modulation of the bacterial community (Pei
et al., 2021).
The distributions of core brGDGTs (measured in this study) are thought to be resistant to
degradation (Schouten et al., 2004) and stable over geological time scales (e.g., M. Wang et al.,
2017; Lu et al., 2019). As such, whilst the overall concentration of GDGTs in our study
decreased with depth (Fig 3.3A, B) there is no reason to assume that this will impart a bias on the
MBT´
5Me
index, given the structural similarity of the brGDGT molecules involved in its
calculation. As such degradation is unlikely to be a large factor in down profile changes in
90
MBT´
5Me.
In addition, given that the soil temperature loggers show that there is no change in the
mean soil temperature with depth (Beverly et al., 2021) it is unlikely that temperature differences
between surface and deep soils would cause the offset in MBT´
5Me.
Whilst low oxygen levels
have previously been implicated in high MBT´
5Me
bias in deep soils (Pei et al., 2021) we can also
rule out the presence of anoxic conditions in our Serengeti soils due to the lack of gleyed colors
and there is no evidence of Fe reduction.
We examined a suite of environmental and biomarker information to attempt to determine if in
situ production at depth could explain the predicted temperature offset from the surface. We find
that in principal component space, MAP and TOC are positively correlated with MBT´
5Me
among
deep soil samples (Fig 3.9). Whilst MAP is not correlated with MAT residuals in the global
BayMBT model, previous studies have shown that MAP <600 mm/year can lead to a cold bias
(De Jonge et al., 2014) in temperature reconstructions. Two dry sites with low TOC (Malambo
Road and Shifting Sands) have deep samples with relatively low MBT´
5Me
(0.73), which could
be explained by the low MAP. However, most samples show an increase in
MBT´
5Me
with depth
(Fig 3.9B). We note the deep samples MBT´
5Me
also have a negative correlation with total
dissolved solids and pH in principal component space. This corresponds to more saline and alkali
soils having lower MBT′
5Me
values and more acidic, less saline soils having higher MBT´
5Me
.
Changes in pH have been demonstrated to influence the MBT´
5Me
of surface soils in the absence
of a change in temperature, by altering the brGDGT-producing bacterial community (De Jonge et
al., 2019, 2021; Halffman et al., 2022). Additionally, out of the three soil profiles studied by Pei
et al., (2021), the largest increase in MBT´
5Me
occurred in the highest pH soil (pH > 8). As the
soils in this study are all alkaline this possibly suggests that deep soil with high pH conditions
biases MBT´
5Me
higher relative to the surface soil. Given that we find MBT´
5Me
is responsive to
91
changes in pH only in deep samples, compared to surface samples which respond primarily to
MAAT (Fig 3.9A), this could suggest that we see modern production of brGDGTs throughout
our soil profiles, with different bacterial communities at different depths responsive to different
environmental stimuli. We do not see a large change in the CBT' with depth, with both shallow
and deep samples having statistically insignificant correlations between CBT' and soil pH and
CBT' based pH models (De Jonge et al., 2014b) underestimating soil pH. A systematic
underestimation of pH using a CBT' soil calibration (Dearing Crampton-Flood et al., 2020)
compared to a global compilation (Fig 3.2B) suggests that the relative abundance of 6-methyl
and/or cyclic brGDGT compounds are controlled in part by something other than pH in our
samples. We note that a similar pH underestimation occurs in a suite of dry, alkaline soils
sampled from Inner Mongolia (Guo et al., 2022), although CBT' does correlate with soil pH in
these samples. Both of these factors are also present in this study, possibly explaining the
underestimation of CBT' derived pH we see here (Fig 3.5E, F). There is a moderate correlation (r
= 0.67, p = 0.02) between IR
6Me
and pH in our surface soil samples and these values are broadly
consistent with a global compilation (Dearing Crampton-Flood et al., 2020) (Fig 3.2C) which
suggests that the isomerization is responsive to pH. However, the lack of a significant correlation
between IR
6Me
and pH in the deeper samples could be caused by the relative proportion of 6-
methyl brGDGTs being controlled by other factors possibly including changing bacterial
community composition.
92
Fig 3.9. PCA biplots showing biomarker and environmental indices: Total organic carbon % (TOC),
CBT´, ratio of isoGDGTs to brGDGTs (Ri/b), measured pH, total dissolved solids (TDS), mean annual
air temperature (MAT; Fick and Hijmans, 2017), MBT´5Me, archaeol caldarchaeol ecometric (ACE), mean
annual precipitation (MAP; Fick and Hijmans, 2017) for A) all surface samples (<0.1 m), B) all deep
samples (>0.1 m).
3.4.5 Comparison of organic and carbonate paleothermometers
3.4.5.1 Proxy comparison in the entire Serengeti dataset
The paired sampling for organics and carbonates within the same soil profiles affords a
comparison of brGDGT thermometry with carbonate clumped isotope thermometry (Beverly et
al., 2021). Across both proxies we find all temperature estimates fall within the range of modern
Serengeti soil temperatures of 17–30 °C and the median temperature (across all samples)
generated from each proxy is identical (23 °C, Fig 3.10A). Despite proxy analytical uncertainty,
possible differences in time integration of each proxy and inter-sample variability (aspects that
will be discussed further in Section 5.5.4), both proxies on average generate temperatures that
are comparable and correlate with MAAT (Fig 3.10B). We note that the Serengeti experiences
small seasonal air temperature changes (Fick and Hijmans, 2017) on account of its equatorial
position, which likely precludes any complications from proxy formation seasonal biases.
Further, the general stability of regional air temperatures over the past few thousand years
93
(Tierney et al., 2008; Berke et al., 2012) would not suggest a bias if the time span of each
recorder differs.
Fig 3.10. Comparison of brGDGT (this study) and clumped (Beverly et al., 2021) temperature estimates.
A) Violin plot showing distributions of all samples: central estimates are the same within uncertainties
however the range differs. MAAT and minimum and maximum MAST temperatures for the Serengeti are
shown for comparison. B) Cross-plot of site MAAT (equivalent to MAF) from surface soil BayMBT0
(error bars, calibration 1 RMSE) and ∆47 temperatures within the soil profile (error bars, propagated
analytical and calibration uncertainty, 1 SEM); 1:1 line (dashed line).
3.4.5.2 Inter- and intra-site proxy comparison
Within the Serengeti transect, there is approximately 4°C spatial variation in MAAT and both
proxies capture the temperature gradient across the region (Fig 3.10B). The brGDGT-
temperatures from deeper soils present a complication in these saline soils, and we find evidence
for in situ production. Within depth profiles, we find no significant correlation (r = 0.16, p =
0.60) between predicted brGDGTs and MAAT or carbonate-formation temperatures –
intentionally sampled from the same position in the soil for this comparison (Beverly et al.,
2021). We also found no correlation (r = 0.29, p = 0.45) between surface brGDGT samples (in
the more organic-rich horizons) with the clumped isotope temperatures derived from soil
carbonates (that are found at depth in the profile) – which would reflect how each proxy would
94
be typically sampled in a soil-based on organic and carbonate substrate distribution. Even though
both proxies show a positive correlation with MAAT (clumped: r = 0.60, p = 0.09: brGDGT: r =
0.73, p = 0.01), they do not correlate with each other. This lack of correlation is perhaps
unsurprising given the size of the uncertainties (instrumental and calibration), the small
temperature range within the Serengeti, and the different processes of formation for each proxy
recorder, which we discuss next.
3.4.5.3 Process-based differences between proxy recorders
Both proxies are known to capture different temperature signals, with the brGDGT MBT′
5Me
showing a strong positive relationship with months above freezing mean annual air temperature
(Dearing Crampton-Flood et al., 2020), with salinity at depth creating a confounding signal in
these soils.
47
records the temperature of carbonate precipitation which often reflects warm-
season temperatures, and which can be modulated by soil grain size, vegetation cover,
precipitation seasonality, and depth in soil (Kelson et al., 2020). Of these factors, equatorial
climates allow us to eliminate concerns over seasonal differences between proxy recorders in this
setting. Whilst BayMBT
0
temperatures are similar to MAAT, ∆
47
temperatures are typically
warmer (Fig 3.10B) which is in agreement with the previous ∆
47
(Kelson et al., 2020) and
MBT′
5Me
temperature compilations (Dearing Crampton-Flood et al., 2020).
A clear difference between the two analytes is the difference in their distribution in the profiles:
the carbonates are located from 0.4–1.6 m and below, whereas the brGDGTs decline in
concentrations with depth from 0–1.6 m (Fig 3.3A). We find that the surface soil brGDGT-
based, BayMBT
0
-calibrated MAF temperatures range from 17 to 25 °C with a mean of 21 °C
(SEM = 0.8 °C, n = 11), whereas samples at depth >0.1 m range from 17.0 to 26.8 °C with a
mean of 22.8 °C, in this case, warm-biased by high salinities. Clumped isotopes yield a mean of
95
23.3 °C (SEM = 5.5 °C, n = 27) that is indistinguishable from the organic proxy within error, but
the samples have a larger temperature range (14 to 31 °C, Beverly et al., 2021).
The analytical and calibration uncertainties of each proxy differ. BrGDGTs have a trivial
analytical uncertainty (approximately 0.2 °C, Fleming and Tierney, 2016), and are thus measured
singly and the uncertainty is dominated by the calibration uncertainty for BayMBT
0
to MAF,
reported as an RMSE of 3.8 °C (Dearing Crampton-Flood et al., 2020).
47
records the assumed-
thermodynamic formation temperature of carbonate and has been calibrated in laboratory
settings (e.g. Ghosh et al., 2006; Bonifacie et al., 2017; Petersen et al., 2019) and a range of
environments. For clumped isotope measurements, the combined instrumental and analytical
uncertainties on samples and standard replicates indicate a 1 SEM typically on the order of ~4 °C
(5.5 °C in these particular samples; Beverly et al., 2021).
Soils are notoriously spatially heterogeneous, and we find that both pedogenic carbonate nodules
and biomarkers in this study show a large variance within soil profiles and across the Serengeti
transect which is notable given the relatively small variance in regional spatial and seasonal
temperature (Fig 3.10A). Meso and macro fauna homogenize soil microbial communities over
multiple meters (Vos et al., 2013) thus this may also be the scale over which molecular
biomarkers may be mobile. Similarly, soil carbonates form with waters and pore space
atmospheres that likely mix over the scale of meters (Zamanian et al., 2016), but localized
nodule conditions may differ on a finer scale. While soil carbonate nodules are actively growing
in deeper horizons (Beverly et al., 2021), in stable soils such as these they have likely formed
over 10
2
to 10
3
years (Gallagher and Sheldon, 2016). Soil brGDGT ages are only recently
receiving attention, but they are likely modern in surface horizons, and initial compound-specific
radiocarbon studies in Switzerland indicate they may reach mean ages of 10
3
years at depth (Gies
96
et al., 2021). Microbial production is most active in the high TOC upper soils, with both older
fossil biomarkers and living microbial communities present deeper in the profile.
Thus, while the spatial and temporal scales of mixing may be similar, differences in the
processes of formation of each proxy, soil heterogeneity, and proxy uncertainty likely limit the
correspondence that can be achieved at fine scales. We encourage large sample sizes across
multiple depths and locations as in this study n>5 was found to significantly reduce the standard
deviation of the distribution of means from bootstrapped data (Fig. C.2.). Such an approach is
necessary to capture mean conditions with both proxies and helps to average out the observed
heterogeneity caused by the inherent spatial heterogeneity of soils as well as the process of
formation differences and analytical uncertainty in each. In modern soils, the full profile can be
studied, but in paleosols, the surface soil horizon is sometimes missing in the stratigraphic record
(Kraus, 1999), and thus the ability to fully characterize a paleosol profile and thus account for
depth-dependent signals identified here (e.g. warm, saline bias at depth) may be curtailed. This
emphasizes the importance of soil profile studies to fully characterize and study the processes
affecting the spatial signatures of GDGT proxies' in-depth distributions as relevant to paleosol
applications.
3.5 Conclusions
We analyzed the GDGT composition of a suite of soil samples collected across the Serengeti
National Park, Tanzania. Surface soil BayMBT
0
-predicted temperatures correlated with MAAT
(equivalent to MAF in the tropics). We found deep soil brGDGTs were more responsive to high
soil pH and salinity than to mean annual temperature and represent in situ production. We also
compared the organic temperature proxy to previously published clumped isotope thermometry
on soil carbonates in the same soils. We find that both proxies predict temperatures in the range
97
of modern recorded temperatures, and with large sample sizes, their mean values agree. Both
proxies detect the temperature gradient across the transect in surface soils, but clumped
temperatures are warm-biased as expected in grasslands where soils are often warmer than air
temperatures, and deep soil samples have salinity-induced warm-biases for the brGDGT proxy.
Further comparison between the two proxies would be aided by studies in cool climates, and
carbonate-bearing soils to compare the recorders across a broader range of temperatures and
pedogenic settings. These Serengeti soil comparisons show broad proxy agreement, however,
care should be taken when interpreting brGDGT-derived temperatures below the soil surface, as
well as low numbers of ∆
47
samples. Based on our findings here, we encourage more dual proxy
paleothermometry and emphasize the need for a large number of analyses (n>5) across multiple
soil locations to achieve robust comparison and to explore confounding factors. Notwithstanding
the need for further calibration, the potential of dual proxy applications is in their joint
application to a range of modern and ancient soils.
Acknowledgments
This work was supported in part by the National Science Foundation EAR Postdoctoral
Fellowship (EAR-PF-1725621) to EB. Samples were collected under permits issued to EB by the
Tanzanian government under COSTECH Permit #2018-39-NA-2018-17, TANAPA Research
Permit #: TNP/HQ/C.10/13, and TAWIRI Permit #: TWRI/RS-342/2016/116. Thanks to Joseph
Masoy and Honest Ndoro for field assistance; Yannick Matia, Efrain Vidal, Daolai Zhang, Nick
Rollins, and Patrick Murphy for laboratory assistance.
References
Anderson, V.J., Shanahan, T.M., Saylor, J.E., Horton, B.K., Mora, A.R., 2014. Sources of local
and regional variability in the MBT′/CBT paleotemperature proxy: Insights from a modern
98
elevation transect across the Eastern Cordillera of Colombia. Organic Geochemistry 69, 42–
51.
Bai, Y., Chen, C., Xu, Q., Fang, X., 2018. Paleoaltimetry Potentiality of Branched GDGTs From
Southern Tibet. Geochemistry, Geophysics, Geosystems 19, 551–564.
Berke, M.A., Johnson, T.C., Werne, J.P., Grice, K., Schouten, S., Sinninghe Damsté, J.S., 2012.
Molecular records of climate variability and vegetation response since the Late Pleistocene
in the Lake Victoria basin, East Africa. Quaternary Science Reviews 55, 59–74.
Beverly, E.J., Lukens, W.E., Stinchcomb, G.E., 2018. Paleopedology as a Tool for
Reconstructing Paleoenvironments and Paleoecology. Vertebrate Paleobiology and
Paleoanthropology 151–183.
Beverly, E.J., Levin, N.E., Passey, B.H., Aron, P.G., Yarian, D.A., Page, M., Pelletier, E.M.,
2021. Triple oxygen and clumped isotopes in modern soil carbonate along an aridity
gradient in the Serengeti, Tanzania. Earth and Planetary Science Letters 567, 116952.
Bonifacie, M., Calmels, D., Eiler, J.M., Horita, J., Chaduteau, C., Vasconcelos, C., Agrinier, P.,
Katz, A., Passey, B.H., Ferry, J.M., Bourrand, J.J., 2017. Calibration of the dolomite
clumped isotope thermometer from 25 to 350 °C, and implications for a universal
calibration for all (Ca, Mg, Fe)CO3 carbonates. Geochimica et Cosmochimica Acta 200,
255–279.
Cleveland, D.M., Atchley, S.C., Nordt, L.C., 2007. Continental Sequence Stratigraphy of the
Upper Triassic (Norian–Rhaetian) Chinle Strata, Northern New Mexico, U.S.A.: Allocyclic
and Autocyclic Origins of Paleosol-Bearing Alluvial Successions. Journal of Sedimentary
Research 77, 909–924.
Coffinet, S., Huguet, A., Anquetil, C., Derenne, S., Pedentchouk, N., Bergonzini, L., Omuombo,
99
C., Williamson, D., Jones, M., Majule, A., Wagner, T., 2017. Evaluation of branched
GDGTs and leaf wax n-alkane δ2H as (paleo) environmental proxies in East Africa.
Geochimica et Cosmochimica Acta 198, 182–193.
Dang, X., Yang, H., Naafs, B.D.A., Pancost, R.D., Xie, S., 2016. Evidence of moisture control
on the methylation of branched glycerol dialkyl glycerol tetraethers in semi-arid and arid
soils. Geochimica et Cosmochimica Acta 189, 24–36.
Davtian, N., Ménot, G., Bard, E., Poulenard, J., Podwojewski, P., 2016. Consideration of soil
types for the calibration of molecular proxies for soil pH and temperature using global soil
datasets and Vietnamese soil profiles. Organic Geochemistry 101, 140–153.
De Jonge, C., Stadnitskaia, A., Hopmans, E.C., Cherkashov, G., Fedotov, A., Sinninghe Damsté,
J.S., 2014a. In situ produced branched glycerol dialkyl glycerol tetraethers in suspended
particulate matter from the Yenisei River, Eastern Siberia. Geochimica et Cosmochimica
Acta 125, 476–491.
De Jonge, C., Hopmans, E.C., Zell, C.I., Kim, J.H., Schouten, S., Sinninghe Damsté, J.S., 2014.
Occurrence and abundance of 6-methyl branched glycerol dialkyl glycerol tetraethers in
soils: Implications for palaeoclimate reconstruction. Geochimica et Cosmochimica Acta
141, 97–112.
De Jonge, C., Radujković, D., Sigurdsson, B.D., Weedon, J.T., Janssens, I., Peterse, F., 2019.
Lipid biomarker temperature proxy responds to abrupt shift in the bacterial community
composition in geothermally heated soils. Organic Geochemistry 137, 103897.
De Jonge, C., Kuramae, E.E., Radujković, D., Weedon, J.T., Janssens, I.A., Peterse, F., 2021.
The influence of soil chemistry on branched tetraether lipids in mid- and high latitude soils:
Implications for brGDGT- based paleothermometry. Geochimica et Cosmochimica Acta
100
310, 95–112.
Dearing Crampton-Flood, E., Tierney, J.E., Peterse, F., Kirkels, F.M.S.A., Sinninghe Damsté,
J.S., 2020. BayMBT: A Bayesian calibration model for branched glycerol dialkyl glycerol
tetraethers in soils and peats. Geochimica et Cosmochimica Acta 268, 142–159.
Dirghangi, S.S., Pagani, M., Hren, M.T., Tipple, B.J., 2013. Distribution of glycerol dialkyl
glycerol tetraethers in soils from two environmental transects in the USA. Organic
Geochemistry 59, 49–60.
Feng, X., D’Andrea, W.J., Zhao, C., Xin, S., Zhang, C., Liu, W., 2019. Evaluation of leaf wax
δD and soil brGDGTs as tools for paleoaltimetry on the southeastern Tibetan Plateau.
Chemical Geology 523, 95–106.
Fick, S.E., Hijmans, R.J., 2017. WorldClim 2: new 1-km spatial resolution climate surfaces for
global land areas. International Journal of Climatology 37, 4302–4315.
Finnegan, S., Bergmann, K., Eiler, J.M., Jones, D.S., Fike, D.A., Eisenman, I., Hughes, N.C.,
Tripati, A.K., Fischer, W.W., 2011. The Magnitude and Duration of Late Ordovician–Early
Silurian Glaciation. Science 331, 903–906.
Fleming, L.E., Tierney, J.E., 2016. An automated method for the determination of the TEX86
and U37K′ paleotemperature indices. Organic Geochemistry 92, 84–91.
Gallagher, T.M., Sheldon, N.D., 2016. Combining soil water balance and clumped isotopes to
understand the nature and timing of pedogenic carbonate formation. Chemical Geology 435,
79–91.
Ghosh, P., Adkins, J., Affek, H., Balta, B., Guo, W., Schauble, E.A., Schrag, D., Eiler, J.M.,
2006. 13C-18O bonds in carbonate minerals: A new kind of paleothermometer. Geochimica
et Cosmochimica Acta 70, 1439–1456.
101
Gies, H., Hagedorn, F., Lupker, M., Montluçon, D., Haghipour, N., Sophia Van Der Voort, T.,
Ian Eglinton, T., 2021. Millennial-age glycerol dialkyl glycerol tetraethers (GDGTs) in
forested mineral soils: 14C-based evidence for stabilization of microbial necromass.
Biogeosciences 18, 189–205.
Guo, J., Ma, T., Liu, N., Zhang, X., Hu, H., Ma, W., Wang, Z., Feng, X., Peterse, F., 2022. Soil
pH and aridity influence distributions of branched tetraether lipids in grassland soils along
an aridity transect. Organic Geochemistry 164, 104347.
Halffman, R., Lembrechts, J., Radujković, D., De Gruyter, J., Nijs, I., De Jonge, C., 2022. Soil
chemistry, temperature and bacterial community composition drive brGDGT distributions
along a subarctic elevation gradient. Organic Geochemistry 163, 104346.
Hargreaves, G.H., Samani, Z.A., 1985. Reference Crop Evapotranspiration from Temperature.
Applied Engineering in Agriculture 1, 96–99.
He, Y., Wang, H., Meng, B., Liu, H., Zhou, A., Song, M., Kolpakova, M., Krivonogov, S., Liu,
W., Liu, Z., 2020. Appraisal of alkenone- and archaeal ether-based salinity indicators in
mid-latitude Asian lakes. Earth and Planetary Science Letters 538, 116236.
Hopmans, E.C., Weijers, J.W.H., Schefuß, E., Herfort, L., Sinninghe Damsté, J.S., Schouten, S.,
2004. A novel proxy for terrestrial organic matter in sediments based on branched and
isoprenoid tetraether lipids. Earth and Planetary Science Letters 224, 107–116.
Hopmans, E.C., Schouten, S., Sinninghe Damsté, J.S., 2016. The effect of improved
chromatography on GDGT-based palaeoproxies. Organic Geochemistry 93, 1–6.
Huguet, C., Hopmans, E.C., Febo-Ayala, W., Thompson, D.H., Sinninghe Damsté, J.S.,
Schouten, S., 2006. An improved method to determine the absolute abundance of glycerol
dibiphytanyl glycerol tetraether lipids. Organic Geochemistry 37, 1036–1041.
102
Huguet, A., Fosse, C., Metzger, P., Fritsch, E., Derenne, S., 2010. Occurrence and distribution of
extractable glycerol dialkyl glycerol tetraethers in podzols. Organic Geochemistry 41, 291–
301.
Huguet, A., Meador, T.B., Laggoun-Défarge, F., Könneke, M., Wu, W., Derenne, S., Hinrichs,
K.U., 2017. Production rates of bacterial tetraether lipids and fatty acids in peatland under
varying oxygen concentrations. Geochimica et Cosmochimica Acta 203, 103–116.
Jager, T., 1982. Soils of the Serengeti woodlands, Tanzania. Pudoc Wageningen.
Kelson, J.R., Huntington, K.W., Schauer, A.J., Saenger, C., Lechler, A.R., 2017. Toward a
universal carbonate clumped isotope calibration: Diverse synthesis and preparatory methods
suggest a single temperature relationship. Geochimica et Cosmochimica Acta 197, 104–
131.
Kelson, J.R., Huntington, K.W., Breecker, D.O., Burgener, L.K., Gallagher, T.M., Hoke, G.D.,
Petersen, S. V., 2020. A proxy for all seasons? A synthesis of clumped isotope data from
Holocene soil carbonates. Quaternary Science Reviews 234, 106259.
Kemp, D.B., Robinson, S.A., Crame, J.A., Francis, J.E., Ineson, J., Whittle, R.J., Bowman, V.,
O’Brien, C., 2014. A cool temperate climate on the Antarctic Peninsula through the latest
Cretaceous to early Paleogene. Geology 42, 583–586.
Kraus, M.J., 1999. Paleosols in clastic sedimentary rocks: their geologic applications. Earth-
Science Reviews 47, 41–70.
Li, J., Pancost, R.D., Naafs, B.D.A., Yang, H., Zhao, C., Xie, S., 2016. Distribution of glycerol
dialkyl glycerol tetraether (GDGT) lipids in a hypersaline lake system. Organic
Geochemistry 99, 113–124.
Li, Y., Zhao, S., Pei, H., Qian, S., Zang, J., Dang, X., Yang, H., 2018. Distribution of glycerol
103
dialkyl glycerol tetraethers in surface soils along an altitudinal transect at cold and humid
Mountain Changbai: Implications for the reconstruction of paleoaltimetry and paleoclimate.
Science China Earth Sciences 61, 925–939.
Lizama, C., Monteoliva-Sánchez, M., Prado, B., Ramos-Cormenzana, A., Weckesser, J.,
Campos, V., 2001. Taxonomic study of extreme halophilic archaea isolated from the “Salar
de Atacama”, Chile. Systematic and Applied Microbiology 24, 464–474.
Lu, H., Liu, W., Yang, H., Wang, H., Liu, Z., Leng, Q., Sun, Y., Zhou, W., An, Z., 2019. 800-
kyr land temperature variations modulated by vegetation changes on Chinese Loess Plateau.
Nature Communications 2019 10:1 10, 1–10.
Menges, J., Huguet, C., Alcañiz, J.M., Fietz, S., Sachse, D., Rosell-Melé, A., 2014. Influence of
water availability in the distributions of branched glycerol dialkyl glycerol tetraether in soils
of the Iberian Peninsula. Biogeosciences 11, 2571–2581.
Molnar, P., 2022. Differences between soil and air temperatures: Implications for geological
reconstructions of past climate. Geosphere 18, 800–824.
Naafs, B.D.A., Gallego-Sala, A. V., Inglis, G.N., Pancost, R.D., 2017. Refining the global
branched glycerol dialkyl glycerol tetraether (brGDGT) soil temperature calibration.
Organic Geochemistry 106, 48–56.
Naafs, B.D.A., Oliveira, A.S.F., Mulholland, A.J., 2021. Molecular dynamics simulations
support the hypothesis that the brGDGT paleothermometer is based on homeoviscous
adaptation. Geochimica et Cosmochimica Acta 312, 44–56.
Norton‐Griffiths, M., Herlocker, D., Pennycuick, L., 1975. The patterns of rainfall in the
Serengeti Ecosystem, Tanzania. African Journal of Ecology 13, 347–374.
Peaple, M.D., Beverly, E.J., Garza, B., Baker, S., Levin, N.E., Tierney, J.E., Häggi, C., Feakins,
104
S.J., 2021a. Abundances and indices for soil microbial biomarkers (brGDGTs and
isoGDGTs) in eleven soil profiles across a Serengeti transect. Pangaea.
Peaple, M.D., Tierney, J.E., McGee, D., Lowenstein, T.K., Bhattacharya, T., Feakins, S.J.,
2021b. Identifying plant wax inputs in lake sediments using machine learning. Organic
Geochemistry 156, 104222.
Pei, H., Zhao, S., Yang, H., Xie, S., 2021. Variation of branched tetraethers with soil depth in
relation to non-temperature factors: Implications for paleoclimate reconstruction. Chemical
Geology 572, 120211.
Pérez-Angel, L.C., Sepúlveda, J., Molnar, P., Montes, C., Rajagopalan, B., Snell, K., Gonzalez-
Arango, C., Dildar, N., 2020. Soil and Air Temperature Calibrations Using Branched
GDGTs for the Tropical Andes of Colombia: Toward a Pan-Tropical Calibration.
Geochemistry, Geophysics, Geosystems 21, e2020GC008941.
Peterse, F., Kim, J.-H., Schouten, S., Kristensen, D.K., Koç, N., Sinninghe Damsté, J.S., 2009.
Constraints on the application of the MBT/CBT palaeothermometer at high latitude
environments (Svalbard, Norway). Organic Geochemistry 40, 692–699.
Peterse, F., van der Meer, J., Schouten, S., Weijers, J.W.H., Fierer, N., Jackson, R.B., Kim, J.H.,
Sinninghe Damsté, J.S., 2012. Revised calibration of the MBT–CBT paleotemperature
proxy based on branched tetraether membrane lipids in surface soils. Geochimica et
Cosmochimica Acta 96, 215–229.
Peterse, F., Martínez-García, A., Zhou, B., Beets, C.J., Prins, M.A., Zheng, H., Eglinton, T.I.,
2014. Molecular records of continental air temperature and monsoon precipitation
variability in East Asia spanning the past 130,000 years. Quaternary Science Reviews 83,
76–82.
105
Petersen, S. V., Defliese, W.F., Saenger, C., Daëron, M., Huntington, K.W., John, C.M., Kelson,
J.R., Bernasconi, S.M., Colman, A.S., Kluge, T., Olack, G.A., Schauer, A.J., Bajnai, D.,
Bonifacie, M., Breitenbach, S.F.M., Fiebig, J., Fernandez, A.B., Henkes, G.A., Hodell, D.,
Katz, A., Kele, S., Lohmann, K.C., Passey, B.H., Peral, M.Y., Petrizzo, D.A., Rosenheim,
B.E., Tripati, A., Venturelli, R., Young, E.D., Winkelstern, I.Z., 2019. Effects of Improved
17O Correction on Interlaboratory Agreement in Clumped Isotope Calibrations, Estimates
of Mineral-Specific Offsets, and Temperature Dependence of Acid Digestion Fractionation.
Geochemistry, Geophysics, Geosystems 20, 3495–3519.
Reed, D.N., Anderson, T.M., Dempewolf, J., Metzger, K., Serneels, S., 2009. The spatial
distribution of vegetation types in the Serengeti ecosystem: the influence of rainfall and
topographic relief on vegetation patch characteristics. Journal of Biogeography 36, 770–
782.
Schauble, E.A., Ghosh, P., Eiler, J.M., 2006. Preferential formation of 13C-18O bonds in
carbonate minerals, estimated using first-principles lattice dynamics. Geochimica et
Cosmochimica Acta 70, 2510–2529.
Schouten, S., Hopmans, E.C., Sinninghe Damsté, J.S., 2004. The effect of maturity and
depositional redox conditions on archaeal tetraether lipid palaeothermometry. Organic
Geochemistry 35, 567–571.
Setia, R., Gottschalk, P., Smith, P., Marschner, P., Baldock, J., Setia, D., Smith, J., 2013. Soil
salinity decreases global soil organic carbon stocks. Science of The Total Environment 465,
267–272.
Sinninghe Damsté, J.S., Rijpstra, W.I.C., Hopmans, E.C., Weijers, J.W.H., Foesel, B.U.,
Overmann, J., Dedysh, S.N., 2011. 13,16-Dimethyl octacosanedioic acid (iso-Diabolic
106
Acid), a common membrane-spanning lipid of Acidobacteria subdivisions 1 and 3. Applied
and Environmental Microbiology 77, 4147–4154.
Sinninghe Damsté, J.S., Rijpstra, W.I.C., Hopmans, E.C., Foesel, B.U., Wüst, P.K., Overmann,
J., Tank, M., Bryant, D.A., Dunfield, P.F., Houghton, K., Stott, M.B., 2014. Ether- and
ester-bound iso-diabolic acid and other lipids in members of Acidobacteria subdivision 4.
Applied and Environmental Microbiology 80, 5207–5218.
Sinninghe Damsté, J.S., Rijpstra, W.I.C., Foesel, B.U., Huber, K.J., Overmann, J., Nakagawa, S.,
Kim, J.J., Dunfield, P.F., Dedysh, S.N., Villanueva, L., 2018. An overview of the
occurrence of ether- and ester-linked iso-diabolic acid membrane lipids in microbial
cultures of the Acidobacteria: Implications for brGDGT paleoproxies for temperature and
pH. Organic Geochemistry 124, 63–76.
Tabor, N.J., Myers, T.S., 2015. Paleosols as indicators of paleoenvironment and paleoclimate.
Annual Review of Earth and Planetary Sciences 43, 333–361.
Tierney, J.E., Russell, J.M., Huang, Y., Damsté , J.S.S., Hopmans, E.C., Cohen, A.S., 2008.
Northern Hemisphere Controls on Tropical Southeast African Climate During the Past
60,000 Years. Science 322, 252–255.
Turich, C., Freeman, K.H., 2011. Archaeal lipids record paleosalinity in hypersaline systems.
Organic Geochemistry 42, 1147–1157.
Vos, M., Wolf, A.B., Jennings, S.J., Kowalchuk, G.A., 2013. Micro-scale determinants of
bacterial diversity in soil. FEMS Microbiology Reviews 37, 936–954.
Wang, H., Liu, W., Zhang, C.L., Jiang, H., Dong, H., Lu, H., Wang, J., 2013. Assessing the ratio
of archaeol to caldarchaeol as a salinity proxy in highland lakes on the northeastern
Qinghai–Tibetan Plateau. Organic Geochemistry 54, 69–77.
107
Wang, H., Liu, W., Lu, H., Zhang, C., 2017. Potential degradation effect on paleo-moisture
proxies based on the relative abundance of archaeal vs. bacterial tetraethers in loess-
paleosol sequences on the Chinese Loess Plateau. Quaternary International 436, 173–180.
Wang, H., Liu, W., He, Y., Zhou, A., Zhao, H., Liu, H., Cao, Y., Hu, J., Meng, B., Jiang, J.,
Kolpakova, M., Krivonogov, S., Liu, Z., 2021. Salinity-controlled isomerization of
lacustrine brGDGTs impacts the associated MBT5ME’ terrestrial temperature index.
Geochimica et Cosmochimica Acta 305, 33–48.
Wang, M., Zheng, Z., Man, M., Hu, J., Gao, Q., 2017. Branched GDGT-based paleotemperature
reconstruction of the last 30,000 years in humid monsoon region of Southeast China.
Chemical Geology 463, 94–102.
Weijers, J.W.H.H., Schouten, S., Hopmans, E.C., Geenevasen, J.A.J.J., David, O.R.P.P.,
Coleman, J.M., Pancost, R.D., Sinninghe Damste, J.S., Sinninghe Damsté, J.S., 2006.
Membrane lipids of mesophilic anaerobic bacteria thriving in peats have typical archaeal
traits. Environmental Microbiology 8, 648–657.
Weijers, J.W.H., Schouten, S., van den Donker, J.C., Hopmans, E.C., Sinninghe Damsté, J.S.,
2007. Environmental controls on bacterial tetraether membrane lipid distribution in soils.
Geochimica et Cosmochimica Acta 71, 703–713.
Weijers, J.W.H., Panoto, E., van bleijswijk, J., Schouten, S., Rijpstra, W.I.C., Balk, M., Stams,
A.J.M., Damsté, J.S.S., 2009. Constraints on the Biological Source(s) of the Orphan
Branched Tetraether Membrane Lipids. http://dx.doi.org/10.1080/01490450902937293 26,
402–414.
Weijers, J.W.H., Wiesenberg, G.L.B., Bol, R., Hopmans, E.C., Pancost, R.D., 2010. Carbon
isotopic composition of branched tetraether membrane lipids in soils suggest a rapid
108
turnover and a heterotrophic life style of their source organism(s). Biogeosciences 7, 2959–
2973.
Xie, S., Pancost, R.D., Chen, L., Evershed, R.P., Yang, H., Zhang, K., Huang, J., Xu, Y., 2012.
Microbial lipid records of highly alkaline deposits and enhanced aridity associated with
significant uplift of the Tibetan Plateau in the Late Miocene. Geology 40, 291–294.
Yamamoto, Y., Ajioka, T., Yamamoto, M., 2016. Climate reconstruction based on GDGT-based
proxies in a paleosol sequence in Japan: Postdepositional effect on the estimation of air
temperature. Quaternary International 397, 380–391.
Yang, H., Pancost, R.D., Dang, X., Zhou, X., Evershed, R.P., Xiao, G., Tang, C., Gao, L., Guo,
Z., Xie, S., 2014. Correlations between microbial tetraether lipids and environmental
variables in Chinese soils: Optimizing the paleo-reconstructions in semi-arid and arid
regions. Geochimica et Cosmochimica Acta 126, 49–69.
Yang, H., Lü, X., Ding, W., Lei, Y., Dang, X., Xie, S., 2015. The 6-methyl branched tetraethers
significantly affect the performance of the methylation index (MBT′) in soils from an
altitudinal transect at Mount Shennongjia. Organic Geochemistry 82, 42–53.
Zamanian, K., Pustovoytov, K., Kuzyakov, Y., 2016. Pedogenic carbonates: Forms and
formation processes. Earth-Science Reviews 157, 1–17.
Zang, J., Lei, Y., Yang, H., 2018. Distribution of glycerol ethers in Turpan soils: implications for
use of GDGT-based proxies in hot and dry regions. Frontiers of Earth Science 2018 12:4
12, 862–876.
Zech, R., Gao, L., Tarozo, R., Huang, Y., 2012. Branched glycerol dialkyl glycerol tetraethers in
Pleistocene loess-paleosol sequences: Three case studies. Organic geochemistry 53, 38–44.
Zhang, D., Beverly, E.J., Levin, N.E., Vidal, E., Matia, Y., Feakins, S.J., 2021. Carbon isotopic
109
composition of plant waxes, bulk organics and carbonates from soils of the Serengeti
grasslands. Geochimica et Cosmochimica Acta. doi:10.1016/J.GCA.2021.07.005
Zomer, R.J., Bossio, D.A., Trabucco, A., Yuanjie, L., Gupta, D.C., Singh, V.P., 2007. Trees and
water: smallholder agroforestry on irrigated lands in Northern India. IWMI.
110
Chapter 4: Biomarker and pollen evidence for late Pleistocene pluvials
in the Mojave Desert
This chapter is in review at AGU Paleoceanography and Paleoclimatology:
Mark D. Peaple, Tripti Bhattacharya, Tim K. Lowenstein, David McGee, Kristian J. Olson,
Justin S. Stroup, Jessica E. Tierney, Sarah J. Feakins. Biomarker and pollen evidence for late
Pleistocene pluvials in the Mojave Desert. Paleoceanography and Paleoclimatology
Abstract
The climate of southwestern North America has experienced profound changes between wet and
dry phases over the past 200 kyr. To better constrain the timing, magnitude, and
paleoenvironmental impacts of these changes in hydroclimate, we conducted a multiproxy
biomarker study from samples collected from a new 76 m sediment core (SLAPP-SRLS17)
drilled in Searles Lake, California. Here, we use biomarkers and pollen to reconstruct vegetation,
lake conditions and climate. We find that δD values of long-chain n-alkanes are dominated by
glacial to interglacial changes that match nearby Devils Hole calcite δ
18
O variability, suggesting
both archives predominantly reflect precipitation isotopes. However, precipitation isotopes do
not simply covary with evidence for wet-dry changes in vegetation and lake conditions,
indicating a partial disconnect between large scale atmospheric circulation tracked by
precipitation isotopes and landscape moisture availability. Increased crenarchaeol production and
decreased evidence for methane cycling reveal a 10 kyr interval of a fresh, productive and well-
mixed lake during Termination II, corroborating evidence for a paleolake highstand from
shorelines and spillover deposits in downstream Panamint Basin during the end of the
penultimate (Tahoe) glacial (140 –130 ka). At the same time brGDGTs yield the lowest
111
temperature estimates (mean months above freezing = 9 ± 3°C) of the 200 kyr record. These
limnological conditions are not replicated elsewhere in the 200 kyr record, suggesting that the
Heinrich stadial 11 highstand was wetter than that during the last glacial maximum and Heinrich
1 (18–15 ka).
4.1 Introduction
There is considerable concern over water availability in southwestern North America and
uncertainties around precipitation in climate model projections (Pierce et al., 2013). Proxy
reconstructions of past moisture availability under different temperature regimes can help to
understand the changing water balance (P–E) during periods of climate change (McGee, 2020),
including evidence for water table rise and fall in southwestern North America detected during
recent glacial cycles (Wendt et al., 2018).
However, available proxy evidence from southwestern North America suggests different
magnitudes of variability and climate change during the Late Pleistocene. For instance, Devils
Hole and the Leviathan composite record (Fig 4.1a) are high resolution speleothem δ
18
O records
that record glacial-interglacial changes in δ
18
O of precipitation over two glacial cycles (Lachniet,
2016; Moseley et al., 2016). However, the magnitudes of variability are larger and precessional
pacing is more strongly represented in the Leviathan composite record than in the Devils Hole
calcite. It is likely that differences in aquifer mixing, karstic dissolution and calcite precipitation
processes (including temperature) lead to differences between δ
18
O
calcite
in the different cave
systems and speleothem types. Independent evidence for precipitation isotopic composition for
the last glacial is available from groundwater, studied further south in San Diego, but only for the
last glacial (Kulongoski et al., 2009; Seltzer et al., 2021). Lake sediments provide a longer
archive of precipitation isotopes, for example, the plant wax δD record back to 33 ka from Lake
112
Elsinore, California (Fig 4.1a) (Feakins et al., 2019). Biomarker studies of Lake Elsinore
sediments, specifically bacterial membrane lipids, also yielded evidence of previously
unrecognized, highly-variable lake temperatures during the last glacial period (Feakins et al.,
2019). Fossil pollen in sediment cores from Lake Elsinore and Searles Lake provide evidence for
past vegetation and yield insights into past hydroclimate (Litwin et al., 1999; Heusser et al.,
2015). However, vegetation composition can be influenced by multiple variables (e.g.,
temperature, pCO
2
, and rainfall).
Fig 4.1. Maps showing the location of A) Searles Lake (red star) and climate archives referred to in the
text including Owens Lake (blue circle), ODP 1012/1010 (pink circles), Devils Hole (orange circle),
Leviathan Cave, Lehman Cave, and Pinnacle Cave (black circles) B) The Lakes connected to Searles
Lake during pluvial periods where M = Mono Lake, O = Owens Lake, C = China Lake, S = Searles Lake,
P = Lake Panamint, M = Lake Manly. C) Map of Searles Lake during pluvial conditions highlighting
inflow and outflow.
Here we revisit lacustrine sediments from Searles Lake (Fig 4.1c), to generate a 200 kyr
biomarker and pollen reconstruction of limnology as well as the regional climate and
environmental changes. The combination of plant wax and pollen allows us to independently
infer changes in regional precipitation δD and vegetation, which act as tracers for changes in
rainfall seasonality in this region. In addition, we analyze a suite of microbial biomarkers to
reconstruct aridity and inform on lake salinity, depth, and temperature. The multi-proxy dataset
uniquely yields new insights into the timing and magnitude of past changes between aridity and
113
pluvials that filled the chain of lakes to the east of the Sierra Nevada Mountains (Fig 4.1b) in
what is today part of the hyperarid Mojave Desert.
4.2 Regional setting
Searles Valley is an endorheic basin located in the Mojave Desert in southeastern California (Fig
4.1). Below the evaporites on the valley floor, there are lacustrine muds from past deep lake
conditions. Shoreline tufa deposits indicate the lake was formerly up to ~300 m deep (Smith et
al., 1983). During past wet climate states, the Owens River carried spill-over from the upstream
Owens Lake to China Lake and Searles Lake (Fig 4.1c). Owens Lake receives snowmelt runoff
from the eastern flanks of the Sierra Nevada Mountains (Bischoff and Cummins, 2001). Over the
past 200 ka, the Owens River has been almost continuously inflowing into Searles Lake, with
only the late Holocene and 6 brief (<1 ka) periods during the late glacial receiving no inflow
(Bacon et al., 2020). Once between 190 and 130 ka, the catchment may have briefly expanded to
include that of Mono Lake (Reheis et al., 2002). When Searles Lake reached 696 m, it would
also reach the spillway into Panamint Basin and ultimately Death Valley (Forester et al., 2005).
The present-day climate in Searles Valley is hyperarid, with a mean annual precipitation of 100
mm between 1920 – 2016 (Western Regional Climate Center, 2022). Modern monthly mean
temperature averages 27.4°C in summer (JJA) and 11.4°C in winter (DJF) with recorded
temperature extremes of 41.0°C and -0.8°C (Western Regional Climate Center, 2022). Hot, dry,
and often windy, conditions promote high potential evaporation ~2000 mm/yr, far over
precipitation. Sporadic precipitation is winter-dominated with DJF and JAS monthly means of 18
mm and 4 mm, respectively (Western Regional Climate Center, 2022). During past pluvial
conditions, Searles would receive precipitation falling on the Eastern Sierra Nevada Mountains
through Owens River inflow. Modern Eastern Sierra precipitation also has a winter dominance
114
with DJF and JAS monthly means of 67 mm and 16 mm, respectively (Lake Sabrina ) (Western
Regional Climate Center, 2022). Local winter precipitation is sourced from storms that derive
from the North Pacific in addition to sub-tropical sourced atmospheric rivers (Friedman et al.,
1992). Summer rain is sourced from the Gulf of California and Gulf of Mexico at the northern
limits of the incursion of the North American Monsoon, with a small contribution from North
Pacific moisture (Friedman et al., 1992). Precipitation from northerly winter and summer storms
is typically more D-depleted than southerly sourced moisture in either winter or summer
(Friedman et al., 2002), with Searles Valley precipitation having mean summer (March to
September) and winter (October to April) δD values of -56‰ and -76‰ respectively (collection
dates 1982–1989, Friedman et al., 1992). Measured winter precipitation in Owens Valley is more
D-depleted (mean October to April = –96‰) than in Searles Valley (mean October to April = –
74‰) due to Rayleigh distillation in rainout over the Sierra Nevada topographic barrier (~4 km).
Additionally, moisture can leak to the south of the mountain range across the Mojave Desert
(Friedman et al., 1992) and can become enriched by evaporation during raindrop descent
(Friedman et al., 2002). Summer precipitation isotopic compositions reported from Searles
Valley (mean, –57‰) and Owens Valley (mean, –62‰) are similar (Friedman et al., 1992). The
relative enrichment of summer rainfall could reflect a greater proportion of convective rainfall in
summer in addition to the re-evaporation of raindrops as they fall in a hot, low humidity
environment (Friedman et al., 1992; Berkelhammer et al., 2012).
4.3 Age Model
Sediment cores SLAPP-SRLS17-1A and 1B (35.7372ºN, 117.33ºW, 495 m asl) were drilled
from Searles Lake in 2017 with 95% recovery extending to 78 m below the lake floor (Fig 4.1b).
U/Th dating of evaporite minerals (Stroup et al., in prep) indicate the recovered sediments span
115
200 ka BP. Stroup et al., (in prep) use 37 U-Th ages to construct a Bayesian age model using
BACON. The model takes into consideration the mineralogy, stratigraphic superposition, and
boundaries between lithological units. To constrain a portion of the mud horizons without salt
minerals suitable for dating, a tie point near Termination 2 (T2) was identified linking the δD
31alk
record (generated in this study) to the Leviathan composite record δ
18
O
calcite
record, following the
approach of Wang et al., (2022). The data were scaled and interpolated before applying a low
pass filter to both records to remove high-frequency variability. We then calculated the
second derivatives to identify a match point at a gradient of 0. An age constraint of
126.5 ± 0.5 ka from the Leviathan composite record was applied to the feature found at 54.5 m
depth in SLAPP-SRLS17-1A. This tie point assumes that changes in the speleothem δ
18
O in
Nevada and leaf wax δ D in Searles Basin should closely correspond with each other; this
assumption is supported by the good agreement between regional speleothem records and Searles
basin δD
wax
over the last 100 ka when the records are anchored by independent U-Th-based age
models (section 3.2). Between 200–50 ka the accumulation rate of lacustrine carbonate muds
was 0.2 m/ka (95% CI, ± 3.5 ka). After 50 ka sediments and salts accumulated more rapidly (1.3
m/ka). The late glacial and deglacial age model is well constrained (± 0.9 ka), but the late
Holocene is less well resolved due to slowed deposition since the lake desiccated completely and
mining disturbed the record in the upper salts.
4.4 Material and methods
4.4.1 Lipid extraction
Lacustrine muds were sampled in 2018 for biomarkers and pollen roughly every 60 cm (~2 ka),
avoiding salts that dominate the upper 33 m of the core. As previously described in Peaple et al.
(2021), 120 sediment samples (~20 g) were dried, ground, and extracted using a Dionex
116
Accelerated Solvent Extraction system at the University of Southern California with 9:1
dichloromethane (DCM):methanol (MeOH) at 100ºC and 1500 psi to yield the Total Lipid
Extract (TLE). Briefly, the TLEs were separated into neutral and acid fractions by column
chromatography. The neutral fraction was further separated using columns packed with 5%
deactivated silica gel, eluting n-alkanes with hexanes, and the polar fraction with DCM followed
by methanol. n-Alkanes were treated with copper to remove elemental sulfur before GC
analyses. Fatty acids were methylated (to FAMEs) using 95:5 MeOH:hydrochloric acid at 70ºC
for 12 h, using MeOH of known isotopic composition (methyl group
13
C of –24.7‰ and D of
–187‰).
4.4.2 GDGT analyses
The neutral polar fraction was analyzed by an Agilent 1260 High-Performance Liquid
Chromatography (HPLC) coupled with an Agilent 6120 mass spectrometer at the University of
Arizona, following the methods of Hopmans et al. (2016). Compounds were detected in single
ion monitoring mode and quantified relative to a C
46
internal standard. Concentrations of
archaeol, caldarchaeol and the ACE index for salinity were previously reported (Peaple et al.,
2021). Here we report concentrations of individual and summed () isoGDGTs and brGDGTs
and calculate temperature, pH, and methane sensitive indicators.
We calculate the branched isoprenoid tetraether (BIT) index:
𝐵𝐼𝑇 =
𝐼𝑎 +𝐼𝐼𝑎 +𝐼𝐼 𝑎 ′
+𝐼𝐼𝐼𝑎 +𝐼𝐼𝐼𝑎 ′
𝐼𝑎 +𝐼𝐼𝑎 +𝐼𝐼 𝑎 ′
+𝐼𝐼𝐼𝑎 +𝐼𝐼𝐼𝑎 ′+𝑐𝑟𝑒𝑛 (1)
where brGDGTs Ia IIa and IIIa, including both the 5´ and 6´ methyl isomers, are compared with
the abundance of crenarchaeol (Hopmans et al., 2004). In lakes, BIT has traditionally been
interpreted to represent the balance between soil inputs of brGDGTs and lake production of
crenarchaeol (e.g., Verschuren et al., 2009). However, interpretations may differ as bacterial
117
production may dominate in many lakes, and changes in oxycline depth may control the
abundance of creanarchaeol-producing Thaumarchaeota (Baxter et al., 2021). As an additional
measure of stratification, we calculate %GDGT-0 (Sinninghe Damsté et al., 2012a), which
measures the proportion of isoGDGT-0, which is produced by Thaumarcheota (e.g., Sinninghe
Damsté et al., 2012b; Schouten et al., 2013), anaerobic methane-oxidizing archaea (Pancost et
al., 2001; Schouten et al., 2001) and methanogenic Euryarchaeota (Schouten et al., 2013, and
references therein) relative to crenarchaeol which is produced uniquely by Thaumarchaeota (e.g.,
Sinninghe Damsté et al., 2002; Schouten et al., 2013) :
%GDGT − 0 =
[𝐺𝐷𝐺𝑇 −0]
[𝐺𝐷𝐺𝑇 −0]+[Crenarchaeol ]
× 100 (2)
We calculate the CBT´ index (De Jonge et al., 2014) where:
𝐶𝐵𝑇 ′
= 𝑙𝑜𝑔 10
[
𝐼𝑐 +𝐼𝐼 𝑎 ′
+𝐼𝐼 𝑏 ′
+𝐼𝐼 𝑐 ′
+𝐼𝐼𝐼 𝑎 ′
+𝐼𝐼𝐼 𝑏 ′
+𝐼𝐼𝐼 𝑐 ′
𝐼𝑎 +𝐼𝐼𝑎 +𝐼𝐼𝐼𝑎
]
(3)
CBT´ has been calibrated to pH in east African lakes (Russell et al., 2018):
𝑝𝐻 = 7.15 − 1.59 ∗ 𝐶𝐵 𝑇 ′
(4)
The temperature-sensitive MBT′
5Me
index is the relative methylation of the 5′ isomers of the
brGDGTs (De Jonge et al., 2014, Hopmans et al., 2016) and is expressed as:
𝑀𝐵𝑇 ´
5𝑀𝐸
=
(𝐼𝑎 +𝐼𝑏 +𝐼𝑐 )
(𝐼𝑎 + 𝐼𝑏 + 𝐼𝑐 +𝐼𝐼𝑎 +𝐼𝐼𝑏 +𝐼𝐼𝑐 +𝐼𝐼𝐼𝑎 )
(5)
The Type I, II, and III brGDGTs have four, five, and six methyl groups, respectively, and the
Type a, b, and c brGDGTs have zero, one, and two rings, respectively. Duplicate analyses and
analyses of an internal laboratory standard throughout the runs yielded an error of 0.009
MBT´
5Me
units (1). To convert MBT´
5Me
to temperature, we use the Bayesian BayMBT
0
model,
which was generated by calibrating MBT´
5Me
against the mean temperature of the months above
118
freezing from a global lake dataset (Martínez-Sosa et al., 2021), including lakes over a range of
pH (4.3 to 10), salinity (0–275 psu) and temperature (1.6 to 28.1°C).
We calculate IR
6+7Me,
an index sensitive to changes in lake salinity (Wang et al., 2021):
𝐼𝑅
6+7𝑀𝑒
= [
𝐼𝐼 𝑎 ′
+𝐼𝐼 𝑏 ′
+𝐼𝐼 𝑐 ′
+𝐼𝐼𝐼 𝑎 ′
+𝐼𝐼𝐼 𝑏 ′
+𝐼𝐼𝐼 𝑐 ′
𝐼𝐼𝑎 +𝐼𝐼𝑏 +𝐼𝐼𝑐 +𝐼𝐼𝐼𝑎 +𝐼𝐼𝐼𝑏 +𝐼𝐼𝐼𝑐 +𝐼𝐼 𝑎 ′
+𝐼𝐼 𝑏 ′
+𝐼𝐼 𝑐 ′
+𝐼𝐼𝐼 𝑎 ′
+𝐼𝐼𝐼 𝑏 ′
+𝐼𝐼𝐼 𝑐 ′
+
𝐼𝐼𝐼 𝑎 ′′′
+𝐼𝐼 𝑎 ′′′
𝐼𝐼𝐼𝑎 +𝐼𝐼𝐼 𝑎 ′
+𝐼𝐼𝐼 𝑎 ′′′
+𝐼𝐼𝑎 +𝐼𝐼 𝑎 ′
+𝐼𝐼 𝑎 ′′′
] × 0.5 (6)
We also calculate TEX
86
for all samples (Schouten et al., 2002):
𝑇𝐸𝑋 86
=
([𝐺𝐷𝐺𝑇 −2]+[𝐺𝐷𝐺𝑇 −3]+[𝑐𝑟𝑒𝑛 ′])
([𝐺𝐷𝐺𝑇 −1]+[𝐺𝐷𝐺𝑇 −2]+[𝐺𝐷𝐺𝑇 −3]+[𝑐𝑟𝑒𝑛 ′])
(7)
and convert to lake surface temperature (LST) using the calibration (Tierney et al., 2010):
𝐿𝑆𝑇 = 𝑇𝐸𝑋 86
× 38.874 − 3.4992 (8)
in a single sample where BIT <0.3 and %GDGT-0 <50, indicating high thaumarcheotal relative
abundance.
4.4.3 Compound-specific isotopic analyses
n-Alkanoic acids and n-alkanes were identified using an Agilent 6890 Gas Chromatograph (GC)
connected to an Agilent 5973 MSD mass spectrometer (MS) and quantified by flame ionization
detector (FID). Abundances, average chain length (ACL), and carbon preference index (CPI)
were previously reported in (Peaple et al., 2021). The carbon and hydrogen isotopic composition
of n-alkanoic acids and n-alkanes were measured for this study using a Thermo Scientific Trace
GC equipped with a Rxi®–5 ms column (30 m × 0.25 mm, film thickness 0.25 μm) with a PTV
injector in solvent-split mode, coupled via an Isolink combustion/pyrolysis furnace
(1000/1400°C) to a Thermo Scientific Delta V Plus isotope ratio mass spectrometer (IRMS) at
the University of Southern California. Reference gas linearity was assessed daily across 1–8 V,
for
13
C (1 = 0.04‰), and for D (H
3
+
factor = 10.6 ppm/mV). A standard containing C
16
-C
30
119
n-alkanes of known isotopic compositions (A6 mix supplied by A. Schimmelmann, University of
Indiana;
13
C values from –25.9 to –33.7‰ and D values from –17 to –256‰) was measured
daily, allowing for normalization to Vienna Standard Mean Ocean Water (VSMOW) and Vienna
Pee Dee Belemnite (VPDB), respectively. Reported
13
C and D values for n-alkanoic acids
were corrected to account for the contribution of the methyl group.
4.4.4 Palynological Analyses
Pollen assemblages were studied for 113 samples at Syracuse University; for detailed sample
processing methodology, see the Supplementary Information. Pollen samples were counted on
400x and 1000x magnification and compared to known pollen keys (Kapp et al., 2000). Our
counts found 22 unique taxa, though samples were dominated by Pinus pollen (e.g., greater than
40% of each sample). Pollen assemblages are expressed in percentages and pollen influx rates
(grains/cm
2
/yr). The similarity of the broad trends across these two ways of expressing the pollen
data increases our confidence that the patterns in our data are robust. For our analysis, we
exclude one sample at 27.49 m associated with a tephra layer. To identify the patterns of
variability in the pollen data, we calculated the Bray-Curtis dissimilarity index between samples,
using pollen taxa that were present in 2 or more samples at a percentage greater than 2%. This
index calculates the compositional dissimilarity between two ecological samples in space or time
and minimizes the contribution of rare taxa to the dissimilarity between samples (Faith et al.,
1987). We used a matrix of pairwise Bray-Curtis indices between samples to perform a non-
metric multidimensional scaling (NMDS). NMDS iteratively moves all samples in 2-dimensional
ordination space so that their final distance from each pairwise sample is proportional to the
Bray-Curtis dissimilarity between those two samples. It is analogous to principle components
analysis in that the distance between samples on the plot provides a guide to their dissimilarity
120
but is more robust for assemblages containing rare taxa (Fasham, 1977; Faith et al., 1987). The
results from this NMDS analysis are used to guide our interpretation of specific plant taxa in the
pollen record.
4.4.5 Correlation analysis
All correlations between time series use non-parametric methods that account for serial
correlation (Ebisuzaki, 1997).
4.5 Results and Discussion
4.5.1 Vegetation reconstructions from Searles Lake spanning 200 kyr
We present a multi-proxy biomarker and pollen study of vegetation change as recovered from the
sediments of Searles Lake in the SLAPP-SRLS17 sediment core. All vegetation-related data
obtained from the core are shown in stratigraphic context (Fig. C.4) and on the age scale (Fig
4.2). Pollen reconstructions are dominated by Pinus (Fig 4.2) because of their long pollen
dispersal distances and high pollen production (Campbell et al., 1999; Wood, 2000). The Owens
Lake pollen record suggests that during glacials, pines may have expanded into lowlands while
being restricted in the uplands (Litwin et al., 1999; Woolfenden, 2003). Pollen from other taxa
can offer more diagnostic climatic information: during cooler/wetter glacial periods,
Taxodiaceae-Cupressaceae-Taxaceae (TCT), mostly Juniper, increase. Elsewhere, glacial
increases in Juniperus-type pollen have been noted from the Gulf of California to the Great Basin
(Byrne, 1982; Davis, 1998). Packrat middens confirm this expansion across southwestern North
America at the LGM (Thompson and Anderson, 2000; Koehler et al., 2005). Middens in the
Central Mojave specifically identify the glacial expansion of J. osteosperma, which is more
sensitive to water stress than other taxa in this genus, e.g., J. californica (Koehler et al., 2005;
Willson et al., 2008; Holmgren et al., 2010) indicating more moisture availability in the
121
lowlands. During interglacials, Juniper declined, and herbaceous taxa like Asteraceae and
Amaranthaceae increased in the Searles basin (Fig 4.2a). We sum Asteraceae and
Amaranthaceae together to represent the total number of desert shrubs in the record. NMDS
analysis reveals that glacial and interglacial samples from Searles Lake show distinct pollen
assemblages and that these changes are primarily driven by changing proportions of desert shrub
pollen, as well as Juniperus-type pollen (Fig. C.4). Desert shrub proportions were previously
modeled by machine learning on n-alkane, and n-alkanoic acid homologs in the same sediments
(Peaple et al., 2021; Figure. 2b), and the comparison with desert shrubs reconstructed by pollen
(Figure. 2c) indicates similar long-term trends when high-frequency changes are removed (r =
0.42, p>0.01). Pollen aids paleovegetation interpretations as it reveals the specific plant taxa
present, but it does not inform on plant wax sourcing since pollen and waxes may have distinct
transport mechanisms to the lake basin. For example, Juniper produces a distinct molecular
abundance distribution with modal C
33
n-alkanes (Diefendorf et al., 2015a). However, we find
this compound is not abundant in Searles Lake sediments even during glacials when pollen and
packrat midden macrobotanicals show that they thrived.
122
Fig 4.2. Vegetation reconstructions using pollen and plant wax proxies from SLAPP-SRLS17. A)
Proportion of pollen taxa. B) Modelled vegetation types based on SVM machine learning of plant wax
distributions in modern taxa applied to the downcore record (Peaple et al., 2021b). C) Comparison
between modelled desert plant types and pollen “desert shrubs” (the sum of Amaranth and Asteraceae
pollen). D) δ
13
C28acid and δ
13
C31alk compared to Amaranth pollen. E) δD28acid and δD31alk.
Pollen microfossil assemblages were analyzed in different core samples than the biomarkers, so
each was linearly interpolated onto 2 kyr sampling resolution to assess shared variance by
principal component analysis (PCA; Fig 4.3). The PCA analysis identifies a negative relationship
between dominant pine pollen and TCT (Taxodiaceae-Cupressaceae-Taxaceae, mostly
Juniperus spp.). Juniper is associated with Artemisia, denoting their glacial co-occurrence
(Figures 2 and 3). The ACE salinity index and desert taxa Amaranthaceae show a
123
correspondence between lake salinity and desert plants, similar to their association with salty
lowland areas of the valley today. Shrub pollen increases with warming, consistent with prior
reports that shrub vegetation is temperature responsive (Lyle et al., 2010).
Fig 4.3. PCA to assess biomarker and pollen covariations (Shrub = sum of Amaranthaceae and Asteraceae
pollen abundance).
Carbon isotope evidence from plant wax biomarkers reveals additional information about
vegetation. The δ
13
C
31alk
and δ
13
C
28acid
covary (Fig 4.3), although the range of values differs (Fig
4.2), reflecting similar pacing but potential differences in sourcing, with C
31alk
broadly reflecting
terrestrial plants and C
28acid
reflecting conifers and/or aquatic inputs. δ
13
C increases under
warmer conditions, as shown through similar correlations with PC 1 and 2 (Fig 4.2), likely
driven by increasing water stress on C
3
plants. δ
13
C also increases with higher Amaranthaceae
(C
4
and C
3
members) pollen percentages as well as influx rates (Figures 2d, 3 and Fig. C.4),
which suggests that temperature-controlled evaporative demand, rather than pCO
2
, is the
dominant selection pressure on the prevalence of C
4
woody taxa in this region. In contrast,
grasses are nearly absent from the pollen record during interglacial intervals, making it unlikely
that the δ
13
C signal reflects C
4
grasses (Fig 4.2a). The C
4
pathway is used in some woody,
124
halophilic desert plants sampled in the catchment today, including plants in the Atriplex and
Suaeda genera. These plants are phreatophytes and thrive in locations with shallow groundwater
(Patten et al., 2008). Warmer temperatures may drive the expansion of these taxa by increasing
the area of seasonal environments C
4
Amaranthaceae taxa occupy. Warmer conditions would
likely also restrict drought-sensitive Juniperus species' habitat and promote conifer-dominated
environments' expansion (Fig 4.2).
The lack of correlation between the δD
28acid
and δD
31alk
indicates a difference in the signals
captured by these two compound classes in Searles Lake. The D
28acid
is puzzling as it records D-
enrichment during glacials, opposite to hydroclimate expectations, suggesting producer
complications. D
28acid
anticorrelates with pines (Fig 4.3), likely reflecting their abundant
production of fatty acids. During the LGM (locally termed the Tioga glaciation), the upper limit
of tree production descended from 3.5 to 2.5 km as glaciers and snowpacks accumulated (Moore
and Moring, 2013). Eliminating the highest elevation conifer forests during glacials could
increase the D value of exported plant wax n-alkanoic acids by at most 10‰ based on the
expected altitude effect (Feakins et al., 2018). A glacial expansion of lowland conifers could
further add D-enriched lowland production. However, it seems unlikely that the altitude source
effect could explain the 40‰ variability observed downcore (Fig 4.2e). Machine learning has
suggested the possibility of aquatic macrophyte inputs (Peaple et al., 2021b), although unverified
locally, given the lack of modern surface water. There is some upstream macrophyte evidence
from palynology of Owens Lake (Woolfenden, 2003), although none in Searles Lake (this
study). The δD signal of aquatic production (whether by macrophytes or microbial production)
would be affected by changing lake water D and lake salinity effects on fractionation (Sachse et
al., 2012). Rather than attempt to further theorize about multiple unknowns, we suggest that
125
upland conifer and aquatic production may produce complications, confounding the D
28acid
signal here.
In contrast, the n-alkanes yield a clear D-depleted glacial and D-enriched interglacial pattern (Fig
4.2e). δD
31alk
has a close phasing with desert shrub pollen (sum of Amaranthaceae and
Asteraceae) and temperature (Figures 2c and 3), and this covariation of proxies suggests a
common driver which will be explored when compared to regional and global climate (in Section
4.2). During arid climates, like today, we assume that desert shrubs dominate the n-alkane
record. Although the details are necessarily unconstrained for past pluvial climates in southern
California, trees in modern temperate North American forests and woodlands are prolific
producers of n-alkanes. They have been shown to contribute strongly to lakes rather than the
marginal plants (Freimuth et al., 2019). We thus infer that plant wax n-alkanes may have been
supplied by wind transport to Searles Lake from the woody shrubs and trees of the surrounding
lowlands. We reconstruct D
precip
using the constant local fractionation by plants (
31alk/p,
-93‰),
determined from regional calibration across the modern aridity gradient (Feakins and Sessions,
2010). Sensitivity tests that assess the effect of changing vegetation based on pollen and plant
wax
13
C (Fig. C.2) lead to confidence in the constant fractionation and hydroclimate
interpretations here. Climate model experiments support theoretical expectations of D-depletion
associated with condensation at colder temperatures and as ice versus liquid cloud droplets and
as would be expected in a glacial climate (Jasechko et al., 2015), together with changing storm
tracks introducing more D-depleted North Pacific sourced moisture (Oster et al., 2015) and a
decrease in enriched North American Monsoon sourced precipitation (Bhattacharya et al., 2018).
126
4.5.2 Plant wax evidence for glacially paced changes in hydroclimate
The Searles Lake δD
31alk
record (Fig 4.4) is dominated by glacial to interglacial variability, with
interglacials characterized by more positive values and glacials by more negative values. After
accounting for the ice volume corrections for seawater δD, and the apparent fractionation by
plants, we can interpret plant wax δD
31alk
as precipitation isotopic variations (Fig 4.4c, see
Supplemental Information for method details). δD
precip
during interglacials averages –87‰ and
during glacials averages –127‰. The Searles δD
precip
closely matches global climate records of
glacial to interglacial changes in pCO
2
(Fig 4.4b), ice volume, and deep ocean temperature
changes interpreted from benthic foraminiferal oxygen isotopes (Fig 4.4c) across two glacial-
interglacial cycles.
127
Fig 4.4. Comparison of Searles Lake plant wax δD31alk and calculated δDprecip to global climate data across
two glacial interglacial cycles showing A) Antarctic pCO2 record (Lüthi et al., 2008), B) LR04 δ
18
O
benthic foraminifera stack (Lisiecki and Raymo, 2005), C) plant wax C31 n-alkane D (blue curve) and
inferred precipitation D after apparent fractionation and ice volume correction (black curve). D)
BayMBT0 and E) shrub pollen%. Upper labels: “Hol” = Holocene, “LGM” = Last glacial maximum,
“LIG” = Last interglacial, “PGM” = Penultimate glacial maximum. Lower labels: “MIS” = Marine
isotope stage.
128
A comparison of the two glacial cycles in Searles Lake records suggests that the penultimate
glacial maximum (PGM) was cooler and wetter compared to the last glacial maximum (LGM),
which is in contrast to records of global climate change that show similar magnitudes of changes
during both glacial maxima (Fig 4.4). The δD
precip
is lower (Fig 4.4c), the BayMBT
0
temperature
is 5º C lower (Fig 4.4d), and shrub pollen reaches a 200 ky minimum (Fig 4.4e) during the later
stages of the penultimate glaciation compared to the LGM. The glacial-interglacial variations at
Searles Lake are captured by changes in three independent, climate-sensitive proxies: plant wax,
bacterial membrane lipids, and pollen microfossils. The climate changes that produce variations
in these proxies are explored and evaluated further, in discussions of regional precipitation
archives (Section 4.5.3), past water availability (Section 4.5.4), and past temperatures (Section
4.5.5).
4.5.3 Comparison with regional precipitation isotope archives
We compare the new 200 kyr Searles Lake plant wax reconstruction of δD
precip
to regional
speleothem δ
18
O
calcite
records (Fig 4.5). Devils Hole (located 120 km NE of Searles Valley) is a
flooded cave with calcite deposition on the cave walls (Moseley et al., 2016). The Leviathan
composite record (Lachniet, 2016), is comprised of stalagmite samples collected in Leviathan
(270 km NE), Pinnacle (200 km E), and Lehman Caves (450 km NE), in Nevada (Fig 4.1a).
These caves are located in a similar precipitation isotope region, the Sierra Nevada rain shadow
(Friedman et al., 2002), which justifies comparison with the plant wax record from Searles Lake.
The cave records have been used extensively to determine the timing and amplitude of glacial-
interglacial periods and their relationship to orbital cycles (Winograd et al., 1992; Lachniet et al.,
2014; Moseley et al., 2016). Here we add an independent 200 kyr record from the plant wax
129
precipitation isotope proxy (Fig 4.5a).
Fig 4.5. Comparison of plant wax and speleothem isotopic records. A) Searles Lake δDprecip (black, this
study), Leviathan composite record δ
18
Ocalcite (orange; Lachniet et al., 2016) and Devils Hole δ
18
Ocalcite
(red; Moseley et al., 2016) with the
18
O axis scaled to account for the 8x greater mass dependent
fractionation for hydrogen. B) Searles Lake δDprecip (black) and summer insolation at 65°N (gray). C)
Devils Hole δ
18
Ocalcite and summer insolation at 65°N (gray). D) Leviathan composite record δ
18
Ocalcite
(left) as in A but showing the individual caves, two of which (Lehman and Pinnacle), were adjusted for
spatial gradients in precipitation isotopes (Lachniet et al., 2016). Black and white bars represent MIS
stages. E – G) Weighted wavelet z transform frequency spectrum for the records in B, C, and D. H) 5 to
95 % quartile range for measured values (blue), and after corrections for ice volume (grey), cave
temperature (Leviathan record, black bar) and plant wax wax/w (green). The δ
18
O axis is scaled to account
for the 8x difference in mass dependent fractionation between H and O. Ice volume-corrected Devils Hole
shows the smallest range. In contrast, larger and comparable magnitudes are recorded at the temperature-
corrected Leviathan composite record and Searles Lake.
The cave records and the plant wax precipitation isotope proxy all show similar glacial to
interglacial pacing (Fig 4.5), with higher δ
18
O and δD values during interglacial periods and
lower values during glacials. Climate modelling studies have linked this isotopic signature to a
southward displacement of the North Pacific low-level jet leading to increased cool season
precipitation (Oster et al., 2015; Tabor et al., 2021). The lower δD
precip
values from plant waxes
130
during the penultimate glaciation compared to the last glacial were not observed in the Devils
Hole or Leviathan cave calcite records, perhaps because Searles Lake captured high altitude
Sierra Nevada winter precipitation. In contrast, the speleothems were recharged by lower altitude
Great Basin precipitation more prone to changes in precipitation seasonality (Friedman et al.,
1992). Both MIS 2 and 6 have similar insolation seasonality (Laskar et al., 2004), potentially
explaining the similarity between the speleothem records. It should also be noted that the two
glacial minima in the Leviathan composite record come from different caves hundreds of
kilometers apart (Fig 4.5d). Spectral analyses of each record (Fig 4.5e–g) show that Searles Lake
plant waxes and Devils Hole are paced predominantly by obliquity, whereas the Leviathan
composite record shows a precession signal.
For the obliquity pacing recorded in each archive, summer insolation maxima correspond to
higher δ
18
O and δD, and the magnitude of change can be compared after accounting for the mass
dependent fractionation scaling of 8 (Fig 4.5h). The amplitude of variability at Devils Hole, less
than half that of the Leviathan composite record, was attributed to aquifer averaging (Lachniet et
al., 2017) and the slow rate of carbonate deposition (Moseley et al., 2016) relative to the
stalagmite records from the Leviathan composite record. In addition, cave temperature can
modulate the amplitude of δ
18
O
calcite
through its control on equilibrium fractionation between
water and calcite (Hendy, 1971; Kim and O’Neil, 1997). There is some evidence for this here, as
recent studies of triple oxygen isotopes have shown sensitivity to mineralization temperature at
Leviathan and evaporation at Lehman (Huth et al., 2022), although Devils Hole was not included
in that study. A different study using clumped isotope methods suggested Devils Hole cave water
remained within ±1°C over the past 600 kyrs because of the large aquifer size (Bajnai et al.,
131
2021). We calculate a 6–10°C cooling (consistent with our temperature reconstruction) during
glacial maxima on the range in δ
18
O
calcite
in the Leviathan composite record (black bar Fig 4.5h).
Now we can compare the plant wax δD record at Searles Lake, which shows a similar amplitude
to the Leviathan composite record (Fig 4.5a, h). The correspondence of the glacial-interglacial
changes and obliquity pacing with an independent proxy, such as plant wax in lake sediments,
provides independent corroboration of the importance of obliquity pacing on large-scale
hydroclimate and atmospheric circulation. We note that obliquity and eccentricity are the
dominant components of North American ice volume (Bintanja and Van De Wal, 2008). As
such, changes in ice sheet extent may have been a forcing of hydroclimate with glacial-
interglacial and obliquity signals recorded in both the Searles Lake and Devils Hole precipitation
isotope archives. This means that the dominant precessional swings in the Leviathan composite
δ
18
O record may reflect cave air temperature changes that affect calcite fractionation. Plant wax
δD
precip
is not considered temperature sensitive, but it carries uncertainty associated with
fractionation, aridity, and plant type. The similarity between the plant wax and cave records
supports the obliquity pacing of precipitation isotopes, but their climate significance is less clear.
Although precipitation isotopes are valued hydrological tracers that capture the obliquity pacing
and glacial-interglacial climate, they remain an indirect proxy for moisture availability on the
landscape, leaving a need for additional proxy constraints on hydroclimate.
4.5.4 Searles Lake salinity and regional moisture availability
4.5.4.1 Salinity proxies
Searles Lake biomarkers contain a record of salinity, which inversely covaries with lake depth in
terminal lakes (Olson and Lowenstein, 2021). We compare results from two indices responsive
to lake salinity, the ACE index (Fig 4.6a) previously reported in Peaple et al. (2021), and the
132
IR
6+7me
(Fig 4.6b) newly measured here for comparison and to differentiate the times of
freshwater conditions. While the ACE index (Turich and Freeman, 2011), is sensitive to lake
hypersalinity (Wang et al., 2013), it loses sensitivity below 60 psu (He et al., 2020). Below 100
psu the IR
6+7me
index has promise, although it performs worse than ACE above that threshold
(He et al., 2020). We test both approaches and find a moderate positive correlation between ACE
and IR
6+7me
in the bottom section of the core (76-50 m, r = 0.43, p > 0.01) within carbonate muds
of a perennial, predominantly saline lake. No correlation exists in the interbedded muds and salts
deposited in perennial saline to hypersaline lakes (50–6 m, r = 0, p = 0.6), above the salinity limit
of IR
6+7me
(Sup Fig. C.6). ACE values range from 0 to 100, and IR
6+7me
range from 0.4 to 0.8.
Low ACE values were found in thick muds (25–22 m and 29–28 m) deposited during deeper
lake phases. High ACE was measured in interbedded muds and salts (36–34 m depth) deposited
in perennial hypersaline lakes that precipitated evaporite salts including trona and burkeite
(Olson and Lowenstein, 2021; Olson et al., in review). IR
6+7me
(0.60) is lower (i.e. fresher) in
hypersaline lake stages than in deeper lake mud units (0.66) suggesting that IR
6+7me
fails to
register the salty conditions.
However, the lowest IR
6+7me
indicates a fresh water lake at 140–130 ka where we also observe
unique BIT and %GDGT-0 values (0.3 and 0.6 respectively) that reflect the highest lake level
and freshest water of the 200 kyr record. ACE is low between 140-130 ka, but several other
periods have similarly low ACE in the Searles Lake record (Fig 4.6a). Both salinity proxies show
precessional pacing, in contrast to the obliquity-dominated δD
precip
(Fig. C.5c). Cross spectral
analysis shows both salinity proxies have phase coherence in the precessional and obliquity
bands (1/19–1/45 kyrs) between 175–90 kyrs (Fig. C.5b). Substantial precessional variability is
also present in the % total organic matter measured in Baldwin Lake (Southern California)
133
sediments, between 125–75 kyrs (Glover et al., 2017), suggesting changes in summer insolation
were significant in controlling regional lake paleoenvironmental conditions during MIS6–MIS5.
Higher frequency coherence exists between 1/10–1/17 kyrs bands than between 90–50 kyrs and
then weaker antiphase coherence between 50–12 kyrs is associated with proxy discrepancy
across changing salinities described. BIT and %GDGT-0 (Fig 4.6c) remain close to 1 and >1,
respectively, throughout much of the 200 kyr record, indicating a stratified lake with a shallow
oxycline (Baxter et al., 2021).
Fig 4.6. Biomarker evidence that the late MIS 6 pluvial was a fresher water lake than the late MIS 2
pluvial. Water balance reconstructions Searles Lake: A) ACE, B) IR6+7me, C) BIT D) %GDGT-0, and E)
Devils Hole water table elevation (Wendt et al., 2018). Age model without tie point is plotted for all
GDGT indices as a thin faint line. Terminations 1 and 2 are highlighted with yellow shading and Heinrich
1 and 11 are highlighted with blue shading. Upper labels: “LGM” = Last glacial maximum, “LIG” = Last
interglacial, “PGM” = Penultimate glacial maximum. Lower labels: “MIS” = Marine isotope stage.
134
4.5.4.2 Comparison to regional moisture availability records
We compare biomarker records of salinity based on ACE, IR
6+7me
, BIT, and %GDGT-0 evidence
for limnological conditions in Searles Lake, to the nearby Devils Hole record of water table
elevation from calcite deposits (Wendt et al., 2018). The Searles Lake ACE (Fig 4.6a) and
IR
6+7me
(Fig 4.6b) records share critical similarities with the Devils Hole water table (Fig 4.6d)
including a more saline lake corresponds with low water table elevations during the previous
interglacial (Eemian), and fresher conditions accompany relatively high water tables during
Heinrich 1 and Heinrich 11.
In both archives we find MIS 6 was wetter than MIS 2. The mean ACE and IR
6+7me
were 10
lower and 0.1 higher during MIS 6 relative to MIS 2 Searles Lake. Evidence for wetter
conditions during MIS 6 at Devils Hole comes from subaqueous calcite (Wendt et al., 2018),
which formed when the water table was at a higher elevation than 9.5 m (maximum height of
calcite collected by Wendt et al., 2018) and are much thicker in MIS 6 compared to MIS 2,
suggesting longer and more frequent water table highstands.
4.5.4.3 Lake outflow
Previous studies have identified 7 periods of outflow from Searles Lake into Panamint Valley
during glacial pluvials during the last 2 Ma (Smith et al., 1983; Jannik et al., 1991). Searles Lake
shoreline deposits indicate brief episodes of outflow occurred between 15–12 kyr (Lin et al.,
1998; Smith, 2009), resulting in a 180–200 m deep lake in Panamint Valley during periods of
MIS 2 (Jayko et al., 2008). During MIS 6, Searles Lake shoreline deposits (Smith, 2009) and
chlorine transfer budget (Jannik et al., 1991) suggest a period of intensive overflow into
Panamint Valley, which resulted in the formation of a >300m deep Lake Panamint which
overspilled into Death Valley (Jayko et al., 2008). This overspill resulted in Lake Manly being
135
deeper during MIS 6 than MIS2, likely due to inflow from Lake Panamint (Roberts and Spencer,
1998; Forester et al., 2005). Further upstream, dates of lake highstands and outflows suggest that
Mono Lake was possibly overspilling into the Owens River catchment during MIS 6 but not
during MIS 2 (Reheis et al., 2002; Reheis pers. comm., 1/20/2022).
We find evidence suggesting unique limnological conditions existed at Searles Lake between
140–130 ka during late MIS 6. In comparison to high BIT and %GDGT-0 indexes (~1 and
>99%, respectively) in most of the 200 kyr record, indicating stratified low productivity lakes,
both indices decrease to medians of 0.72 and 46%, respectively, between 140–130 ka (Fig 4.6c),
and IR
6+7Me
values reach a freshwater minimum. Modern studies of Lake Challa suggest that
crenarchaeol-producing Thaumarchaeota live above a shallow oxycline, and methanogenic
archaea, which produce GDGT-0, occur below the oxycline of an anoxic lake (Baxter et al.,
2021). Searles Lake sediments typically have low crenarchaeol relative to brGDGTs (high BIT)
and GDGT-0 (high %GDGT-0), suggesting salinity stratified and/or low oxygen conditions.
However, biomarkers from 140–130 ka denote freshwater, high lake productivity and a
vigorously mixed water column with deep oxygenation. Lake overturning is enabled in
freshwater systems where winter cooling causes surface waters to sink, also assisted by the
turbulence of water inflow and outflow (Rimmer et al., 2011). While much of SLAPP-SLRS17
consists of laminated aragonite thought to reflect salinity-stratified conditions, this portion of the
core is characterized by massive mud deposits (Fig. C.5). These massive mud deposits likely
reflect well-mixed lake conditions that allowed for bioturbation.
While Searles was likely overflowing during MIS 2, we do not see a decrease in BIT or
%GDGT-0, suggesting that the lake was not well oxygenated and/or fresh. This could suggest
that Searles was not vigorously outflowing for any extended period during MIS 2. Additionally,
136
constant sediment deposition on the lake floor from MIS 6 to the present has made the lake floor
shallower with time, impacting the lake depth necessary to reach the sill elevation. During MIS 6
the lake depth required for spill over was 274m, but during MIS 2 this was reduced to 225m
(Smith, 2009). Given that Searles Lake was vigorously outflowing during late MIS 6 under 45m
deeper water levels than MIS 2, we can infer that inflow into Searles must have been more
significant during MIS 6 than MIS 2.
4.5.4.4 Heinrich Stadials 1 and 11
Benthic δ
18
O values (Lisiecki and Raymo, 2005) and atmospheric pCO
2
(Lüthi et al., 2008) are
broadly similar in amplitude during the last two glacial cycles (Fig 4.4a, b). However, the PGM
has a longer duration than the LGM (Jouzel et al., 1993) which is manifested regionally by the
extended high water table at Devils Hole (Wendt et al., 2018). At Searles Lake, the freshest and
highest lake levels are reconstructed not during the PGM but H11. Terminations 1 and 2 differ
regarding their sea level rise, T2 being a continuous, rapid rise, whereas T1 has a two-step rise
(Clark et al., 2020). T2 has stronger insolation forcing in the Northern Hemisphere summer
solstice maximum (Bova et al., 2021). We note the difference in terms of the wet climate state
captured in the Searles Lake setting: we find peak wetness at T2, not T1. While Sierra Nevada
glacial melt could be a transient contributor, extended wet conditions indicate increased
precipitation. Tracers of cave infiltration including trace element ratios, Sr/Ca,
87
Sr/
86
Sr and
carbon isotopic evidence from Lehman cave, Nevada, also suggest that H11 was wetter than the
preceding MIS6 glacial maximum and terminated rapidly within 2 kyrs. A disconnect between
P–E and δ
18
O
calcite
was identified from Lehman cave (Cross et al., 2015), which matches the lack
of covariance we observe between the Searles Lake δD
precip
and our ACE and IR
6+7me
proxies,
suggesting that this is a regional climatic feature and not constrained by one cave. This
137
disconnect could be related to either increasing temperatures and/or changing seasonality/source
of precipitation (Cross et al., 2015). Model simulations link North Atlantic cooling during
Heinrich stadials to pluvials in southwestern North America (McGee et al., 2018). Freshwater
inputs to the North Atlantic slow the Atlantic Meridional Overturning Circulation, leading to
winter cooling in the Northern Hemisphere, causing the Inter Tropical Convergence Zone to shift
southward as seen in the proxy record (Jacobel et al., 2016). These changes intensify the northern
Hadley Cell, accelerating the subtropical jet and increasing the winter season delivery of
atmospheric river precipitation to southwestern North America (McGee et al., 2018).
Precipitation from tropical/sub-tropical atmospheric rivers is relatively enriched in the heavier
isotopes of D and
18
O compared to North Pacific derived moisture (Berkelhammer et al., 2012).
Thus, the increase in δD
precip
we observe at Searles Lake during H11 and the increase in δ
18
O
calcite
seen in Lehman Cave (Cross et al., 2015) are consistent with this hypothesis. Central Pacific
ITCZ southward migration appears to be substantially greater in T2 than in T1 (Jacobel et al.,
2017), consistent with the deeper lake we detect at Searles. Temperature changes likely play a
secondary role in amplifying the δD
precip
signal
(Dansgaard, 1964).
The biomarker evidence shows that the pluvial associated with Heinrich 11 produced deeper,
fresher lakes than H1. Coastal pollen records from central California marine core ODP Site 1018
corroborate this pluvial comparison, finding a 20% greater decrease in shrub pollen associated
with the T2 extreme wet event than the T1 pluvial. The T2 pluvial is wetter than all other glacial
terminations of the past 600 kyrs as recorded by ODP Site 1018 pollen and by the Searles to
Panamint chlorine transfer budget (Jannik et al., 1991).
138
4.5.4.5 Timing of the T2 pluvial
Regarding the timing of the pluvial close to H11, we wish to note the implications of the age
model selection represented with the comparison in Fig 4.6. The SLAPP-SRLS17 preferred age
model based on U/Th incorporates an age tie point between the Leviathan composite δ
18
O
calcite
and Searles Lake δD of C
31
alkane, with an age of 126.5 kyr, at a gap in the U/Th constraints
(Section 1.2). This age model places the peak of the vigorous overflow event in Searles Valley
(Fig 4.6a, b, c, d, and e) at 131.4 kyr, coinciding with H11 (Cross et al., 2015). Without the tie
point, the U/Th-only age model places the overflow event later at 126.6 kyr. Regional climate
records from southwestern North America uniformly suggest that MIS5e was relatively dry (e.g.,
Litwin et al., 1999; Woolfenden, 2003; Cross et al., 2015; Wendt et al., 2018). Based on the
assumption that the tie point to regional cave records is appropriate, the microbial lipid record
from the Searles Basin supports wet conditions during H11 followed by a shift to drier
conditions at the beginning of MIS5.
4.5.5 Terrestrial temperatures
We can contribute to sparse evidence for terrestrial temperature change on land with the new
biomarker records from Searles Lake (Fig 4.7). We reconstruct the mean annual temperature of
months above freezing (Fig 4.7a) of Searles Lake using the BayMBT
0
calibration of the bacterial
lipid MBT´
5Me
index in global lakes (Martínez-Sosa et al., 2021). This record overlaps with the
33–9 kyr record from Lake Elsinore with the same proxy (Feakins et al., 2019), recalibrated with
the same MAF calibration here (Fig 4.7a). Both lakes show 10˚C glacial-to-Holocene warming
and similar magnitude variability within glacials, with notably warm intervals from 50–30 kyrs
at Searles (22°C), corroborating reports of warm times during the last glacial in the region
(Feakins et al., 2019). While brGDGT reconstructions can suffer from biases induced by shallow
139
lake depth, hypersalinity (He et al., 2020) and high alkalinity (Martínez-Sosa et al., 2021) in part
related to more influence from allochthonous inputs from soil derived brGDGTs in less
productive, saline lakes (Martínez-Sosa et al., 2021), our tests corroborate the use of the
BayMBT
0
lake calibration (see Supplemental Information). We note that reconstructed
temperatures from Searles Lake and Lake Elsinore during the Holocene are similar to modern
measured MAF. Independent corroboration of the magnitude of the terrestrial deglacial warming
comes from noble gas groundwater paleotemperature reconstructions from the Mojave Desert
(Kulongoski et al., 2009) and San Diego (Seltzer et al., 2021) that capture evidence for 7–10°C
deglacial warming (Fig 4.7b).
In the 200 kyr BayMBT
0
record from Searles Lake we identify the penultimate glacial as colder
than the last glacial. That cooling occurred between 215–150 kyr, followed by sharp warming
during T2 (140–130 kyr) and relative temperature stability between 130–50 kyr, pronounced
cooling from 50–18 kyr and then deglacial warming, as previously described. Within the low
BIT interlude (BIT = 0.3) of the penultimate glaciation at 131.4 kyr, we were able to obtain a
single archaeal, isoGDGT-based TEX
86
estimate of lake surface temperature applying the lake
calibration (Tierney et al., 2010) to one sample yielding an estimate of 12 ± 2°C (Fig 4.7a). This
sample also yielded a BayMBT
0
temperature estimate of 14 ± 3°C, equivalent within calibration
uncertainties. We note that the coldest temperatures are also associated with the freshest
conditions in the lake (low ACE, lowest IR
6+7me
) and the indication of overflow into Panamint
based on the %GDGT-0 and BIT. Overturning in lakes increases brGDGT production and export
to sediments (Loomis et al., 2014), which could result in a larger proportion of lake derived
bGDGT compared to allochthonous inputs. Given that soil calibration of MBT ’
5Me
underestimates temperatures when applied to lakes (Martínez-Sosa et al., 2021), a decreased
140
input of soil-derived brGDGTs could lead to a decrease in reconstructed temperatures
independent of a change in air temperature.
Fig 4.7. Local and regional temperature records over the past 200 kyr. A. Searles Lake (blue line; this
study) and Lake Elsinore (orange line; Feakins et al., 2019) recalibrated to MAF using Martinez Sosa et
al., (2021) brGDGT temperature records, using the lake MBT´5Me BayMBT0 calibration to mean
temperature from months above freezing (MAF). TEX86 calibrated to lake surface temperature (black dot)
(Tierney et al., 2010). B) Noble gas derived ground water temperature records (Mojve: Kulongoski et al.,
2009; San Diego: Seltzer et al., 2021). Comparison temperature responsive vegetation change showing C)
shrub pollen % (Amaranthaceae and Asteraceae; this study). D) Alkenone based sea surface temperature
(SST) records (ODP 1012, ODP 893: Herbert et al., 2001, 1995).
141
Glacial-interglacial Fig 4.7 pacing dominates the SSTs (Fig 4.7c), which have a smaller
amplitude (5°C) compared to the terrestrial records, which vary by 10°C between 50–30 ka (Fig
4.7a). Terrestrial changes in vegetation covary with air temperature; for example, warm
interludes around 50 ka (Fig 4.7a) correspond to increased pollen from desert shrub taxa (Fig
4.7c), confirming that hot conditions matter to regional moisture availability. This indicates the
importance of terrestrial temperature reconstructions to understand the relationships between
hydroclimate and vegetation on land.
4.6 Conclusions
We present a new biomarker and pollen record from the SLAPP core drilled in Searles Lake
spanning the past 200 kyr. We show evidence from pollen and plant wax for vegetation change
and find that shrub pollen responds to glacial-interglacial temperature change. We show that the
plant wax n-alkane-based proxy for δD
precip
is characterized by large glacial to interglacial and
obliquity changes, likely driven by variations in ice volume. There is a strong correlation (r =
0.75, p > 0.01) determined by non-parametric methods that account for serial correlation
(Ebisuzaki, 1997) between changes in δD
precip
and changes in δ
18
O
calcite
from the nearby Devils
Hole speleothem. The similar pacing suggests that both archives are recording precipitation
isotopic composition; however the Searles Lake δD
precip
record shows larger amplitude changes.
We also present more direct indicators of moisture availability. The ACE index of lake salinity
as well as IR
6+7me
are consistent with lake core lithology and shoreline markers. We find
similarities between Devils Hole water table and regional lake depths, with pluvials during
glacials and drier interglacial conditions. However, we find that Searles Lake was likely deeper
during the penultimate glacial, MIS 6, compared to MIS 2, with the wettest conditions occurring
during Termination 2, especially Heinrich stadial 11. During H11, Searles Lake was well-mixed
142
and overflowed into Panamint Basin, interpreted from the large decrease in BIT and %GDGT-0.
In comparison, Searles Lake remained a stratified, saline, terminal lake during the last lake
highstand in Heinrich 1.
Both brGDGT-derived temperatures and the proportion of shrub pollen increase during
interglacial periods, although glacial temperature minima differ, with terminal MIS 6 being 4°C
cooler than MIS 2. We find less shrub pollen, a fresher lake and more D-depleted precipitation in
the T2 pluvial, providing confidence that the T2 pluvial was wetter than the T1 pluvial from
these independent lines of evidence from the sediments in the Searles Lake core. This 200 kyr
record reveals differences between the two glacial pluvials and between two interglacials,
highlighting the sensitivity of southwestern North America’s hydroclimate.
4.7 Acknowledgments
The plant wax study and GRA (Peaple) were supported by U.S. National Science
Foundation Grant NSF-EAR-1903665 to S.F., GDGT analyses were supported by the Packard
Fellowship for Science and Engineering to J.T., and the pollen analyses were supported by a sub-
award to T.B. from NSF-EAR-1903659 to T.L. Drilling was supported by the Comer Science
and Education Foundation Grant to D.M. and T.L. We thank Searles Valley Minerals for access
and Jade Zimmermann in particular. The sample material used in this project was provided by
LacCore. We thank the SLAPP team involved in coring and collaborative discussions of
surroundings and paleoenvironment, as well as Marith Reheis for discussion of Lake Russell
overflow, Jay Quade for plant identification in the Mojave Desert, Alan Juarez for field
assistance in the San Bernardino Mountains, and Patrick Murphy for assistance measuring
GDGTs.
143
References
Bacon, S. N., Jayko, A. S., Owen, L. A., Lindvall, S. C., Rhodes, E. J., Schumer, R. A., &
Decker, D. L. (2020). A 50,000-year record of lake-level variations and overflow from
Owens Lake, eastern California, USA. Quaternary Science Reviews, 238, 106312.
https://doi.org/10.1016/j.quascirev.2020.106312
Bajnai, D., Coplen, T. B., Methner, K., Löffler, N., Krsnik, E., & Fiebig, J. (2021). Devils Hole
Calcite Was Precipitated at ±1°C Stable Aquifer Temperatures During the Last Half Million
Years. Geophysical Research Letters, 48(11). https://doi.org/10.1029/2021GL093257
Baxter, A. J., van Bree, L. G. J., Peterse, F., Hopmans, E. C., Villanueva, L., Verschuren, D., &
Sinninghe Damsté, J. S. (2021). Seasonal and multi-annual variation in the abundance of
isoprenoid GDGT membrane lipids and their producers in the water column of a meromictic
equatorial crater lake (Lake Chala, East Africa). Quaternary Science Reviews, 273, 107263.
https://doi.org/10.1016/J.QUASCIREV.2021.107263
Berkelhammer, M., Stott, L., Yoshimura, K., Johnson, K., & Sinha, A. (2012). Synoptic and
mesoscale controls on the isotopic composition of precipitation in the western United
States. Climate Dynamics, 38(3–4), 433–454. https://doi.org/10.1007/s00382-011-1262-3
Bhattacharya, T., Tierney, J. E., Addison, J. A., & Murray, J. W. (2018). Ice-sheet modulation of
deglacial North American monsoon intensification. Nature Geoscience, 1.
https://doi.org/10.1038/s41561-018-0220-7
Bintanja, R., & Van De Wal, R. S. W. (2008). North American ice-sheet dynamics and the onset
of 100,000-year glacial cycles. Nature 2008 454:7206, 454(7206), 869–872.
https://doi.org/10.1038/nature07158
Bischoff, J. L., & Cummins, K. (2001). Wisconsin Glaciation of the Sierra Nevada (79,000-
144
15,000 yr B.P.) as recorded by rock flour in sediments of Owens Lake, California.
Quaternary Research, 55(1), 14–24. https://doi.org/10.1006/qres.2000.2183
Bova, S., Rosenthal, Y., Liu, Z., Godad, S. P., & Yan, M. (2021). Seasonal origin of the thermal
maxima at the Holocene and the last interglacial. Nature 2021 589:7843, 589(7843), 548–
553. https://doi.org/10.1038/s41586-020-03155-x
Byrne, R. (1982). Preliminary pollen analysis of Deep Sea Drilling Project Leg 64 Hole 480
(Cores 1-11).
Campbell, I. D., McDonald, K., Flannigan, M. D., & Kringayark, J. (1999). Long-distance
transport of pollen into the Arctic. Nature, 398(6731), 29–30. https://doi.org/10.1038/19891
Clark, P. U., He, F., Golledge, N. R., Mitrovica, J. X., Dutton, A., Hoffman, J. S., & Dendy, S.
(2020). Oceanic forcing of penultimate deglacial and last interglacial sea-level rise. Nature
2020 577:7792, 577(7792), 660–664. https://doi.org/10.1038/s41586-020-1931-7
Cross, M., McGee, D., Broecker, W. S., Quade, J., Shakun, J. D., Cheng, H., et al. (2015). Great
Basin hydrology, paleoclimate, and connections with the North Atlantic: A speleothem
stable isotope and trace element record from Lehman Caves, NV. Quaternary Science
Reviews, 127, 186–198. https://doi.org/10.1016/J.QUASCIREV.2015.06.016
Dansgaard, W. (1964). Stable isotopes in precipitation. Tellus, 16(4), 436–468.
https://doi.org/10.1111/j.2153-3490.1964.tb00181.x
Davis, O. K. (1998). Palynological evidence for vegetation cycles in a 1.5 million year pollen
record from the Great Salt Lake, Utah, USA. Palaeogeography, Palaeoclimatology,
Palaeoecology, 138(1–4), 175–185. https://doi.org/10.1016/S0031-0182(97)00105-3
Diefendorf, A. F., Leslie, A. B., & Wing, S. L. (2015). Leaf wax composition and carbon
isotopes vary among major conifer groups. Geochimica et Cosmochimica Acta, 170, 145–
145
156. https://doi.org/10.1016/j.gca.2015.08.018
Ebisuzaki, W. (1997). A method to estimate the statistical significance of a correlation when the
data are serially correlated. Journal of Climate, 10(9), 2147–2153.
https://doi.org/10.1175/1520-0442(1997)010<2147:AMTETS>2.0.CO;2
Faith, D. P., Minchin, P. R., & Belbin, L. (1987). Compositional dissimilarity as a robust
measure of ecological distance. Vegetatio 1987 69:1, 69(1), 57–68.
https://doi.org/10.1007/BF00038687
Fasham, M. J. R. (1977). A Comparison of Nonmetric Multidimensional Scaling, Principal
Components and Reciprocal Averaging for the Ordination of Simulated Coenoclines, and
Coenoplanes. Ecology, 58(3), 551–561. https://doi.org/10.2307/1939004
Feakins, S. J., Wu, M. S., Ponton, C., Galy, V., & West, A. J. (2018). Dual isotope evidence for
sedimentary integration of plant wax biomarkers across an Andes-Amazon elevation
transect. Geochimica et Cosmochimica Acta, 242, 64–81.
https://doi.org/10.1016/j.gca.2018.09.007
Feakins, S. J., Wu, M. S., Ponton, C., & Tierney, J. E. (2019). Biomarkers reveal abrupt switches
in hydroclimate during the last glacial in southern California. Earth and Planetary Science
Letters, 515, 164–172. https://doi.org/10.1016/j.epsl.2019.03.024
Forester, R. M., Lowenstein, T. K., & Spencer, R. J. (2005). An ostracode based paleolimnologic
and paleohydrologic history of Death Valley: 200 to 0 ka. GSA Bulletin, 117(11–12), 1379–
1386. https://doi.org/10.1130/B25637.1
Freimuth, E. J., Diefendorf, A. F., Lowell, T. V., & Wiles, G. C. (2019). Sedimentary n-alkanes
and n-alkanoic acids in a temperate bog are biased toward woody plants. Organic
Geochemistry, 128, 94–107. https://doi.org/10.1016/j.orggeochem.2019.01.006
146
Friedman, I., Smith, G. I., Gleason, J. D., Warden, A., & Harris, J. M. (1992). Stable isotope
composition of waters in southeastern California 1. Modern precipitation. Journal of
Geophysical Research, 97(D5), 5795. https://doi.org/10.1029/92JD00184
Friedman, I., Harris, J. M., Smith, G. I., & Johnson, C. A. (2002). Stable isotope composition of
waters in the Great Basin, United States 1. Air-mass trajectories. Journal of Geophysical
Research: Atmospheres, 107(D19), ACL 14-1. https://doi.org/10.1029/2001JD000565
Glover, K. C., MacDonald, G. M., Kirby, M. E., Rhodes, E. J., Stevens, L., Silveira, E., et al.
(2017). Evidence for orbital and North Atlantic climate forcing in alpine Southern
California between 125 and 10 ka from multi-proxy analyses of Baldwin Lake. Quaternary
Science Reviews, 167, 47–62. https://doi.org/10.1016/J.QUASCIREV.2017.04.028
He, Y., Wang, H., Meng, B., Liu, H., Zhou, A., Song, M., et al. (2020). Appraisal of alkenone-
and archaeal ether-based salinity indicators in mid-latitude Asian lakes. Earth and
Planetary Science Letters, 538, 116236.
Hendy, C. H. (1971). The isotopic geochemistry of speleothems—I. The calculation of the
effects of different modes of formation on the isotopic composition of speleothems and their
applicability as palaeoclimatic indicators. Geochimica et Cosmochimica Acta, 35(8), 801–
824. https://doi.org/10.1016/0016-7037(71)90127-X
Herbert, T. D., Yasuda, M., & Burnett, C. (1995). Glacial-Interglacial Sea-Surface Temperature
Record Inferred from Alkenone Unsaturation indices, Site 893, Santa Barbara Basin. In
Proceedings of the Ocean Drilling Program, 146 Part 2 Scientific Results (Vol. 146).
https://doi.org/10.2973/odp.proc.sr.146-2.301.1995
Herbert, T. D., Schuffert, J. D., Andreasen, D., Heusser, L., Lyle, M., Mix, A., et al. (2001).
Collapse of the California current during glacial maxima linked to climate change on land.
147
Science, 293(5527), 71–76. https://doi.org/10.1126/SCIENCE.1059209
Heusser, L. E., Kirby, M. E., & Nichols, J. E. (2015). Pollen-based evidence of extreme drought
during the last Glacial (32.6-9.0 ka) in coastal southern California. Quaternary Science
Reviews, 126, 242–253. https://doi.org/10.1016/j.quascirev.2015.08.029
Holmgren, C. A., Betancourt, J. L., & Rylander, K. A. (2010). A long-term vegetation history of
the Mojave-Colorado desert ecotone at Joshua Tree National Park. Journal of Quaternary
Science, 25(2), 222–236. https://doi.org/10.1002/jqs.1313
Hopmans, E. C., Schouten, S., & Sinninghe Damsté, J. S. (2016). The effect of improved
chromatography on GDGT-based palaeoproxies. Organic Geochemistry, 93, 1–6.
https://doi.org/10.1016/j.orggeochem.2015.12.006
Huth, T. E., Passey, B. H., Cole, J. E., Lachniet, M. S., McGee, D., Denniston, R. F., et al.
(2022). A framework for triple oxygen isotopes in speleothem paleoclimatology.
Geochimica et Cosmochimica Acta, 319, 191–219.
https://doi.org/10.1016/J.GCA.2021.11.002
Jacobel, A. W., McManus, J. F., Anderson, R. F., & Winckler, G. (2017). Climate-related
response of dust flux to the central equatorial Pacific over the past 150 kyr. Earth and
Planetary Science Letters, 457, 160–172. https://doi.org/10.1016/J.EPSL.2016.09.042
Jacobel, Allison W, McManus, J. F., Anderson, R. F., & Winckler, G. (2016). Large deglacial
shifts of the Pacific intertropical convergence zone. Nature Communications, 7(1), 1–7.
Jannik, N. O., Phillips, F. M., Smith, G. I., & Elmore, D. (1991). A 36Cl chronology of
lacustrine sedimentation in the Pleistocene Owens River system. Geological Society of
America Bulletin, 103(9), 1146–1159.
Jasechko, S., Lechler, A., Pausata, F. S. R., Fawcett, P. J., Gleeson, T., Cendon, D. I., et al.
148
(2015). Late-glacial to late-Holocene shifts in global precipitation δ18O. Climate of the
Past, 11(10), 1375–1393. https://doi.org/10.5194/CP-11-1375-2015
Jayko, A. S., Forester, R. M., Kaufman, D. S., Phillips, F. M., Yount, J. C., McGeehin, J., &
Mahan, S. A. (2008). Late Pleistocene lakes and wetlands, Panamint Valley, Inyo County,
California. In Special Paper of the Geological Society of America (Vol. 439, pp. 151–184).
https://doi.org/10.1130/2008.2439(07)
De Jonge, C., Hopmans, E. C., Zell, C. I., Kim, J. H., Schouten, S., & Sinninghe Damsté, J. S.
(2014). Occurrence and abundance of 6-methyl branched glycerol dialkyl glycerol
tetraethers in soils: Implications for palaeoclimate reconstruction. Geochimica et
Cosmochimica Acta, 141, 97–112. https://doi.org/10.1016/j.gca.2014.06.013
Jouzel, J., Barkov, N. I., Barnola, J. M., Bender, M., Chappellaz, J., Genthon, C., et al. (1993).
Extending the Vostok ice-core record of palaeoclimate to the penultimate glacial period.
Nature 1993 364:6436, 364(6436), 407–412. https://doi.org/10.1038/364407a0
Kim, S. T., & O’Neil, J. R. (1997). Equilibrium and nonequilibrium oxygen isotope effects in
synthetic carbonates. Geochimica et Cosmochimica Acta, 61(16), 3461–3475.
https://doi.org/10.1016/S0016-7037(97)00169-5
Koehler, P. A., Anderson, R. S., & Spaulding, W. G. (2005). Development of vegetation in the
Central Mojave Desert of California during the late Quaternary. Palaeogeography,
Palaeoclimatology, Palaeoecology, 215(3–4), 297–311. Retrieved from
https://www.researchgate.net/publication/222663179_Development_of_vegetation_in_the_
Central_Mojave_Desert_of_California_during_the_late_Quaternary
Kulongoski, J. T., Hilton, D. R., Izbicki, J. A., & Belitz, K. (2009). Evidence for prolonged El
Nino-like conditions in the Pacific during the Late Pleistocene: a 43 ka noble gas record
149
from California groundwaters. Quaternary Science Reviews, 28(23–24), 2465–2473.
https://doi.org/10.1016/J.QUASCIREV.2009.05.008
Lachniet, M., Asmerom, Y., Polyak, V., & Denniston, R. (2017). Arctic cryosphere and
Milankovitch forcing of Great Basin paleoclimate. Scientific Reports 2017 7:1, 7(1), 1–10.
https://doi.org/10.1038/s41598-017-13279-2
Lachniet, M. S. (2016). A Speleothem Record of Great Basin Paleoclimate: The Leviathan
Chronology, Nevada. In Developments in Earth Surface Processes (Vol. 20, pp. 551–569).
Elsevier B.V. https://doi.org/10.1016/B978-0-444-63590-7.00020-2
Lachniet, Matthew S., Denniston, R. F., Asmerom, Y., & Polyak, V. J. (2014). Orbital control of
western North America atmospheric circulation and climate over two glacial cycles. Nature
Communications, 5(1), 3805. https://doi.org/10.1038/ncomms4805
Laskar, J., Robutel, P., Joutel, F., Gastineau, M., Correia, A. C. M., & Levrard, B. (2004). A
long-term numerical solution for the insolation quantities of the Earth. Astronomy &
Astrophysics, 428(1), 261–285. https://doi.org/10.1051/0004-6361:20041335
Lin, J. C., Broecker, W. S., Hemming, S. R., Hajdas, I., Anderson, R. F., Smith, G. I., et al.
(1998). A Reassessment of U-Th and 14C Ages for Late-Glacial High-Frequency
Hydrological Events at Searles Lake, California. Quaternary Research, 49, 11–23.
https://doi.org/10.1006/qres.1997.1949
Lisiecki, L. E., & Raymo, M. E. (2005). A Pliocene-Pleistocene stack of 57 globally distributed
benthic δ18 O records. Paleoceanography, 20(1), n/a-n/a.
https://doi.org/10.1029/2004PA001071
Litwin, R. J., Smoot, J. P., Durika, N. J., & Smith, G. I. (1999). Calibrating Late Quaternary
terrestrial climate signals: Radiometrically dated pollen evidence from the southern Sierra
150
Nevada, USA. Quaternary Science Reviews. https://doi.org/10.1016/S0277-3791(98)00111-
5
Loomis, S. E., Russell, J. M., Heureux, A. M., D’Andrea, W. J., & Sinninghe Damsté, J. S.
(2014). Seasonal variability of branched glycerol dialkyl glycerol tetraethers (brGDGTs) in
a temperate lake system. Geochimica et Cosmochimica Acta, 144, 173–187.
https://doi.org/10.1016/j.gca.2014.08.027
Lüthi, D., Le Floch, M., Bereiter, B., Blunier, T., Barnola, J.-M., Siegenthaler, U., et al. (2008).
High-resolution carbon dioxide concentration record 650,000–800,000 years before present.
Nature, 453(7193), 379–382. https://doi.org/10.1038/nature06949
Lyle, M., Heusser, L., Ravelo, C., Andreasen, D., Olivarez Lyle, A., & Diffenbaugh, N. (2010).
Pleistocene water cycle and eastern boundary current processes along the California
continental margin. Paleoceanography, 25(4). https://doi.org/10.1029/2009PA001836
Martínez-Sosa, P., Tierney, J. E., Stefanescu, I. C., Dearing Crampton-Flood, E., Shuman, B. N.,
& Routson, C. (2021). A global Bayesian temperature calibration for lacustrine brGDGTs.
Geochimica et Cosmochimica Acta, 305, 87–105.
https://doi.org/10.1016/J.GCA.2021.04.038
McGee, D., Moreno-Chamarro, E., Marshall, J., & Galbraith, E. D. (2018). Western U.S. lake
expansions during Heinrich stadials linked to Pacific Hadley circulation. Science Advances,
4(11), eaav0118. https://doi.org/10.1126/sciadv.aav0118
McGee, David. (2020, January 3). Glacial-Interglacial Precipitation Changes. Annual Review of
Marine Science. Annual Reviews. https://doi.org/10.1146/annurev-marine-010419-010859
Moore, J. G., & Moring, B. C. (2013). Rangewide glaciation in the Sierra Nevada, California.
Geosphere, 9(6), 1804–1818. https://doi.org/10.1130/GES00891.1
151
Moseley, G. E., Edwards, R. L., Wendt, K. A., Cheng, H., Dublyansky, Y., Lu, Y., et al. (2016).
Reconciliation of the Devils Hole climate record with orbital forcing. Science (New York,
N.Y.), 351(6269), 165–8. https://doi.org/10.1126/science.aad4132
Olson, K. J., & Lowenstein, T. K. (2021). Searles Lake evaporite sequences: Indicators of late
Pleistocene/Holocene lake temperatures, brine evolution, and pCO2. GSA Bulletin.
https://doi.org/10.1130/B35857.1
Oster, J. L., Ibarra, D. E., Winnick, M. J., & Maher, K. (2015). Steering of westerly storms over
western North America at the Last Glacial Maximum. Nature Geoscience, 8(3), 201–205.
https://doi.org/10.1038/ngeo2365
Pancost, R. D., Hopmans, E. C., & Sinninghe Damsté, J. S. (2001). Archaeal lipids in
Mediterranean cold seeps: molecular proxies for anaerobic methane oxidation. Geochimica
et Cosmochimica Acta, 65(10), 1611–1627. https://doi.org/10.1016/S0016-7037(00)00562-7
Patten, D. T., Rouse, L., & Stromberg, J. C. (2008). Isolated spring wetlands in the Great Basin
and Mojave deserts, USA: Potential response of vegetation to groundwater withdrawal.
Environmental Management, 41(3), 398–413. https://doi.org/10.1007/S00267-007-9035-9
Peaple, M. D., Tierney, J. E., McGee, D., Lowenstein, T. K., Bhattacharya, T., & Feakins, S. J.
(2021). Identifying plant wax inputs in lake sediments using machine learning. Organic
Geochemistry, 156, 104222. https://doi.org/10.1016/J.ORGGEOCHEM.2021.104222
Pierce, D. W., Cayan, D. R., Das, T., Maurer, E. P., Miller, N. L., Bao, Y., et al. (2013). The Key
Role of Heavy Precipitation Events in Climate Model Disagreements of Future Annual
Precipitation Changes in California. Journal of Climate, 26(16), 5879–5896.
https://doi.org/10.1175/JCLI-D-12-00766.1
Reheis, M. C., Stine, S., & Sarna-Wojcicki, A. M. (2002). Drainage reversals in Mono Basin
152
during the late Pliocene and Pleistocene. Geological Society of America Bulletin, 114(8),
991–1006.
Rimmer, A., Gal, G., Opher, T., Lechinsky, Y., & Yacobi, Y. Z. (2011). Mechanisms of long-
term variations in the thermal structure of a warm lake. Limnology and Oceanography,
56(3), 974–988. https://doi.org/10.4319/LO.2011.56.3.0974
Roberts, S. M., & Spencer, R. J. (1998). A desert responds to Pleistocene climate change: Saline
lacustrine sediments, Death Valley, California, USA. In Quaternary Deserts and Climatic
Change (pp. 357–370). CRC Press. https://doi.org/10.1201/9781003077862-37
Russell, J. M., Hopmans, E. C., Loomis, S. E., Liang, J., & Sinninghe Damsté, J. S. (2018).
Distributions of 5- and 6-methyl branched glycerol dialkyl glycerol tetraethers (brGDGTs)
in East African lake sediment: Effects of temperature, pH, and new lacustrine
paleotemperature calibrations. Organic Geochemistry, 117, 56–69.
https://doi.org/10.1016/J.ORGGEOCHEM.2017.12.003
Sachse, D., Billault, I., Bowen, G. J., Chikaraishi, Y., Dawson, T. E., Feakins, S. J., et al. (2012).
Molecular Paleohydrology: Interpreting the Hydrogen-Isotopic Composition of Lipid
Biomarkers from Photosynthesizing Organisms. Annual Review of Earth and Planetary
Sciences, 40(1). https://doi.org/10.1146/annurev-earth-042711-105535
Schouten, S., Wakeham, S. G., & Damsté, J. S. S. (2001). Evidence for anaerobic methane
oxidation by archaea in euxinic waters of the Black Sea. Organic Geochemistry, 32(10),
1277–1281. https://doi.org/10.1016/S0146-6380(01)00110-3
Schouten, S., Hopmans, E. C., Schefuß, E., & Sinninghe Damsté, J. S. (2002). Distributional
veriations in marine crenarchaeol membrane lipids: a new tool for reconstructing ancient
sea water temperatures? Earth and Planetary Science Letters, 204, 265–274.
153
https://doi.org/10.1016/S0012-821X(03)00193-6
Schouten, S., Hopmans, E. C., & Sinninghe Damsté, J. S. (2013). The organic geochemistry of
glycerol dialkyl glycerol tetraether lipids: A review. Organic Geochemistry, 54, 19–61.
https://doi.org/10.1016/J.ORGGEOCHEM.2012.09.006
Seltzer, A. M., Ng, J., Aeschbach, W., Kipfer, R., Kulongoski, J. T., Severinghaus, J. P., & Stute,
M. (2021). Widespread six degrees Celsius cooling on land during the Last Glacial
Maximum. Nature, 593(7858), 228–232. https://doi.org/10.1038/S41586-021-03467-6
Sinninghe Damsté, J. S., Schouten, S., Hopmans, E. C., Van Duin, A. C. T., & Geenevasen, J. A.
J. (2002). Crenarchaeol. Journal of Lipid Research, 43(10), 1641–1651.
https://doi.org/10.1194/JLR.M200148-JLR200
Sinninghe Damsté, J. S., Ossebaar, J., Schouten, S., & Verschuren, D. (2012). Distribution of
tetraether lipids in the 25-ka sedimentary record of Lake Challa: extracting reliable TEX86
and MBT/CBT palaeotemperatures from an equatorial African lake. Quaternary Science
Reviews, 50, 43–54. https://doi.org/10.1016/J.QUASCIREV.2012.07.001
Sinninghe Damsté, J. S., Rijpstra, W. I. C., Hopmans, E. C., Jung, M. Y., Kim, J. G., Rhee, S. K.,
et al. (2012). Intact polar and core glycerol dibiphytanyl glycerol tetraether lipids of group
I.1a and I.1b Thaumarchaeota in soil. Applied and Environmental Microbiology, 78(19),
6866–6874. https://doi.org/10.1128/AEM.01681-12/ASSET/7E48339A-F904-4F5F-9251-
4652603B0C1C/ASSETS/GRAPHIC/ZAM9991036760003.JPEG
Smith, G. I. (2009). Late Cenozoic geology and lacustrine history of Searles Valley, inyo and
San Bernardino counties, California. US Geological Survey Professional Paper.
https://doi.org/10.3133/pp1727
Smith, G. I., Barczak, V. J., Moulton, G. F., & Liddicoat, J. C. (1983). Core KM-3, a surface-to-
154
bedrock record of late Cenozoic sedimentation in Searles Valley, California. Professional
Paper. https://doi.org/10.3133/PP1256
Tabor, C., Lofverstrom, M., Oster, J., Wortham, B., de Wet, C., Montañez, I., et al. (2021). A
mechanistic understanding of oxygen isotopic changes in the Western United States at the
Last Glacial Maximum. Quaternary Science Reviews, 274, 107255.
https://doi.org/10.1016/J.QUASCIREV.2021.107255
Thompson, R. S., & Anderson, K. H. (2000). Biomes of western North America at 18,000, 6000
and 0 14C yr BP reconstructed from pollen and packrat midden data. Journal of
Biogeography, 27(3), 555–584. https://doi.org/10.1046/J.1365-2699.2000.00427.X
Tierney, J. E., Mayes, M. T., Meyer, N., Johnson, C., Swarzenski, P. W., Cohen, A. S., &
Russell, J. M. (2010). Late-twentieth-century warming in Lake Tanganyika unprecedented
since AD 500. Nature Geoscience 2010 3:6, 3(6), 422–425.
https://doi.org/10.1038/ngeo865
Verschuren, D., Sinninghe Damsté, J. S., Moernaut, J., Kristen, I., Blaauw, M., Fagot, M., &
Haug, G. H. (2009). Half-precessional dynamics of monsoon rainfall near the East African
Equator. Nature 2009 462:7273, 462(7273), 637–641. https://doi.org/10.1038/nature08520
Wang, H., Liu, W., Zhang, C. L., Jiang, H., Dong, H., Lu, H., & Wang, J. (2013). Assessing the
ratio of archaeol to caldarchaeol as a salinity proxy in highland lakes on the northeastern
Qinghai–Tibetan Plateau. Organic Geochemistry, 54, 69–77.
https://doi.org/10.1016/J.ORGGEOCHEM.2012.09.011
Wang, H., Liu, W., He, Y., Zhou, A., Zhao, H., Liu, H., et al. (2021). Salinity-controlled
isomerization of lacustrine brGDGTs impacts the associated MBT5ME’ terrestrial
temperature index. Geochimica et Cosmochimica Acta, 305, 33–48.
155
https://doi.org/10.1016/J.GCA.2021.05.004
Wang, Z., Zhang, F., Cao, Y., Hu, J., Wang, H., Lu, H., et al. (2022). Linking sedimentary and
speleothem precipitation isotope proxy records to improve lacustrine and marine 14C
chronologies. Quaternary Science Reviews, 282, 107444.
https://doi.org/10.1016/J.QUASCIREV.2022.107444
Wendt, K. A., Dublyansky, Y. V., Moseley, G. E., Edwards, R. L., Cheng, H., & Spötl, C.
(2018). Moisture availability in the southwest United States over the last three glacial-
interglacial cycles. Science Advances, 4(10), eaau1375.
Western Regional Climate Center. (2022). US COOP Station Map. Retrieved October 31, 2018,
from https://wrcc.dri.edu/coopmap/
Willson, C. J., Manos, P. S., & Jackson, R. B. (2008). Hydraulic traits are influenced by
phylogenetic history in the drought-resistant, invasive genus Juniperus (Cupressaceae).
American Journal of Botany, 95(3), 299–314. https://doi.org/10.3732/AJB.95.3.299
Winograd, I. J., Coplen, T. B., Landwehr, J. M., Riggs, A. C., Ludwig, K. R., Szabo, B. J., et al.
(1992). Continuous 500,000-year climate record from vein calcite in Devils Hole, Nevada.
Science, 258(5080), 255–260. https://doi.org/10.1126/science.258.5080.255
Wood, G. D. (2000). Pollen analysis of death valley sediments deposited between 166 and 114
ka. Palynology, 24(1), 49–61. https://doi.org/10.1080/01916122.2000.9989537
Woolfenden, W. B. (2003). A 180,000-year pollen record from Owens Lake, CA: Terrestrial
vegetation change on orbital sales. Quaternary Research, 59(3), 430–444.
https://doi.org/10.1016/S0033-5894(03)00033-4
156
Chapter 5: Warm and wet Pliocene southwestern North America:
evidence from Searles Lake
This chapter is in preparation for submission to AGU Geophysical Research Letters:
Mark D. Peaple, Tripti Bhattacharya, Ran Feng, Tim K. Lowenstein, Jessica E. Tierney, Sarah J.
Feakins. Warm and wet Pliocene southwestern North America: evidence from Searles Lake.
Abstract
During the greenhouse climate state of the Pliocene, deep lakes, including Searles Lake (~300 m
deep), filled basins in the Mojave Desert, southwestern North America. Here, we present
biomarker evidence from a sediment core collected from paleolake Searles spanning 3.4–2.7 Ma.
Biomarkers from microbial membrane lipids indicate a freshwater lake from 3.4 to 3.2 Ma with
temperatures averaging 25ºC throughout the record. Vegetation was dominated by woody C
3
plants, and plant wax d
2
H records isotopically heavier precipitation isotopes during the late
Pliocene compared to the Pleistocene. Isotope-enabled Earth system model simulations suggest
that north Pacific SST patterns play a crucial role in driving mesic conditions, with increased
southerly moisture transport in winter and summer. This results in a 20% increase in
precipitation from winter storms and summer monsoonal moisture and a heavier isotopic
signature of precipitation. These results suggest that atmospheric circulation changes favoring
southerly moisture sources played a crucial role in sustaining large Pliocene lakes in
southwestern North America
5.1 Introduction
Multiple lines of evidence suggest that southwestern North America has become drier in recent
decades and that this trend may be exacerbated by the projected further rise in CO
2
157
concentrations this century (Seager et al., 2007; Williams et al., 2020). However, large lakes
were found in southwestern North America during the late Pliocene 3.4–2.9 Ma (Ibarra et al.,
2014; Pound et al., 2014b), the last time greenhouse gas concentrations were above 400 ppm
(Martínez-Botí et al., 2015). Water balance estimates suggest that 1.6–2x modern precipitation is
needed to sustain deep lakes and offset higher evaporative demand in hotter climates (Ibarra et
al., 2018).
Warmer eastern Pacific SSTs and weaker Pacific SST gradients (Ford et al., 2015; Burls and
Fedorov, 2017) are linked to a weaker Walker circulation (Tierney et al., 2019) and an increase
in precipitation in southwestern North America (Molnar and Cane, 2007; Haywood et al., 2013;
Winnick et al., 2013; Seager et al., 2014). Climate model simulations (Fu et al., 2022) forced by
high sea surface temperatures indicate an expansion in the North American Monsoon influenced
zone (Fig 5.1), which is corroborated by ODP Site 475 plant wax-derived precipitation isotope
reconstructions (D
precip
) from Baja California (Bhattacharya et al., preprint). In another study,
which forced an atmospheric model with PRISM SST estimates, localized warming of California
margin SSTs required land surface moisture feedbacks to affect wind stress and reduce
upwelling along the California margin (Fu et al., 2021) – a scenario that may be consistent with
evidence for large lakes in the region (Smith et al., 1983).
We revisit evidence for wet conditions during the Pliocene through paleolake Searles.
Interpretations of sedimentology found consistent lacustrine muds that suggested a continuously
deep lake, without interruption by evaporites as in the Pleistocene and modern setting (Smith et
al., 1983). However, the sedimentology is not diagnostic beyond the presence of the lake, and
new multi-proxy biomarker approaches can provide a suite of observations of limnological and
catchment conditions. Here, we present a new multi-proxy record of hydroclimatic change from
158
paleolake Searles in southern California. The vegetation, salinity, and precipitation isotope
evidence track changes in hydroclimate through the late Pliocene (3.4–2.7 Ma; i.e., much of the
Piacenzian 3.6-2.6 Ma). Plant wax-based D
precip
can be used to study the seasonal distribution of
rainfall, and lacustrine GDGT proxies reveal lake salinity and temperature. Using a multi-proxy
biomarker approach (similar to Peaple et al. submitted, for the Pleistocene), we develop a new
terrestrial reconstruction of moisture sources and availability and seek to identify the climate
mechanisms for the wetter climate state in the Pliocene through comparison with the climate
model simulations.
Fig 5.1 Site location map showing Searles Lake (this study) and other lakes and marine core sites
mentioned in the text. Published Pliocene marine core sites (circles) showing the 95% (inner) and 5%
(outer) confidence intervals and median (middle) for Pliocene (3.4–2.7 Ma) SSTs anomalies relative
(bottom color bar) to core top (modern) are calculated from published alkenone U
37
k
′
indices at ODP Sites
1010 (LaRiviere et al., 2012), 1012 (Brierley et al., 2009), and (Dekens et al., 2007), recalibrated using
159
BAYSPLINE (Tierney et al., 2019). Map shading shows modern sea surface temperature (bottom color
bar) from GHRSST (OurOcean Project, 2010) and precipitation data from CHIRPS Version 2.0 (Funk et
al., 2015). Hatching covers the area with > 50% summer (JAS) precipitation.
5.2 Materials and Methods
5.2.1 Sediment core and age model
We studied a Searles Lake sediment core (USGS U234, well KM3, 35.73371ºN, 117.32566ºW,
493 m asl) collected in 1968 (Smith et al., 1983) and archived for 50 years in ambient, dry
storage (USGS Core Research Center, Denver). We generated a Bayesian age model (BACON, r
package) for sediments from 200–693 m (Fig 5.2) using previously identified paleomagnetic
reversals (Liddicoat et al., 1980) with updated age estimates (Channell et al., 2020; Glen et al.,
1999; dataset available at NOAA: Peaple et al., 2022b). We selected 29 samples from Pliocene
age (3.34–2.71 Ma, 541–693 m depth) grey/brown and thinly bedded mudstone (Smith et al.,
1983; Hay et al., 1991) and previously identified as deep lake facies (Smith et al., 1983). The
sediments had a relatively uniform sedimentation rate (0.22 cm/year), consistent with a persistent
lake.
160
Fig 5.2. Age model generated using BACON (black line), 95% confidence interval (grey
shading), and paleomagnetic datums (Liddicoat et al., 1980) updated to the GPTS2020 (Channell
et al., 2020; Glen et al., 1999) (red symbols).
5.2.2 Lipid purification and analyses
Lipids were extracted from the ground, dried sediments from KM3 (~20 g) by Accelerated
Solvent Extraction (Dionex) using 9:1 dichloromethane (DCM):methanol (MeOH) at 100ºC and
1500 psi. Lipids were purified following standard methods recently used for late Pleistocene
sediments in the same basin (Peaple et al., 2021). The n-alkane fraction contains an
uncharacterized complex mixture (hydrocarbon contamination), precluding analysis of the plant
wax n-alkanes. The plant wax-derived n-alkanoic acids were quantified and analyzed for C and
H isotopic composition to reconstruct vegetation and precipitation isotopic composition similar
to regional applications for the Pleistocene (Feakins et al., 2019). The δD value of precipitation
was reconstructed from the δD value of C
30
alkanoic acids by applying a regional
wax/precip
of -
93‰ (Feakins and Sessions, 2010; Feakins et al., 2014, 2019).
We quantified glycerol dialkyl glycerol tetraether (GDGTs), including the branched GDGTs,
derived from bacterial membrane lipids, used for temperature reconstructions using the MBT'
5Me
161
index (Fig. D.1) (De Jonge et al., 2014a). MBT'
5Me
is calibrated to the mean temperature of
months above freezing (MAF) using the BayMBT
0
calibration (Martínez-Sosa et al., 2021).
Archaeal-derived isoprenoid lipids were used for the determination of the Archaeol Caldarchaeol
Ecometric (ACE) (Turich and Freeman, 2011), a salinity indicator (Fig. D.1) used in overlying
sediments in the last 200 kyr (Peaple et al., 2021). IR
6+7me
(Wang et al., 2021) is another proxy
for the salinity of brackish water based on the observed changes in the proportion of 6 and 7
methyl brGDGT molecules (Fig. D.1).
5.2.3 Climate Model Simulations
To understand the drivers of Pliocene hydroclimate change, we analyzed a simulation of the
isotope-enabled version of CESM1.2 (iCESM1.2). This simulation was run in atmosphere-only
mode at 0.9x1.25° horizontal resolution, with 30 vertical layers (Bhattacharya et al., preprint).
While iCESM1.2 has isotopically light precipitation compared to observations, it captures the
relative changes between summer and winter rainfall in Southwestern North America
(Bhattacharya et al., preprint). We use Pliocene SST anomalies inferred from alkenone
paleotemperature records, projected onto the dominant EOFs of observational SSTs to create a
“reduced space” SST reconstruction following the method Tierney et al., 2019 (Bhattacharya et
al., preprint). Reconstructed Pliocene SST anomalies from this reduced space reconstruction
were used to adjust the model’s preindustrial SST field between 165°E–60°W and 40°S–60°N.
Outside this region, SSTs were obtained from a coupled mid-Pliocene simulation of CESM1.2.
The Pliocene simulation used pCO
2
of 400 ppm, and other boundary conditions (e.g.,
topography, land surface conditions), were kept as for the preindustrial to isolate the influence of
SSTs on hydroclimate in the western US. We also note that the hydroclimate changes in this
simulation strongly resemble the dynamical changes seen in another, more idealized modeling
162
study of Pliocene hydroclimate (Fu et al., 2022).
5.3 Results and Discussion
5.3.1 Searles Lake Biomarker Reconstruction
To constrain change in Searles Lake depth (Fig 5.3a) and precipitation-evaporation (P-E) we
analyzed the salinity proxies ACE (Fig 5.3b) and IR
6+7me
(Fig. D.1). Both proxies record low
salinities, likely fresh to brackish conditions, from 3.4–3.25 Ma and variable salinities, including
extreme hypersalinity, between 3.25–2.7 Ma. Low air
temperature (MAF) (Fig 5.3c) and
BIT
(Fig. D.1) between 3.4–3.25 Ma indicates relatively cool conditions (mean = 20.7 °C, n=2) and
that the lake was likely overturning, with warmer conditions (mean= 24.5 °C, n =17) and a
stratified lake between 3.25–2.7 Ma. Grain mounts inspected under a microscope reveal a
transition from fine/medium sand (3.4–3.25 Ma) to more silt and clay (3.25–2.7 Ma). This is
consistent with an increasing abundance of n-alkanoic acids with finer grains (Fig. D.1). From
163
these various indicators we infer Searles was a deep lake between 3.4–3.25 Ma, then shallower
and deeper after 3.25 Ma (Fig 5.3a).
Fig 5.3. a) Summary of lake depth from Lake Andrei (LA) (Knott et al., 2021), Lake Manly (LM) (Knott
et al., 2018) and Searles Lake (SL) (Smith et al., 1983; this study). b-e) Searles Lake proxy
reconstructions from Pliocene (this study) and Pleistocene (Peaple et al submitted) including: b) ACE
index of salinity c) BayMBT0 temperature reconstruction of mean air temperature for months above
freezing d) δ13C value of C30 alkanoic acid e) δD value of precipitation calculated using δD of C30
alkanoic acid f) 1st principle component of SSTs anomalies relative to core top (modern), calculated from
published alkenone U
37
k
′
indices at ODP Sites 1010 (LaRiviere et al., 2012), 1012 (Brierley et al., 2009)
and ODP 1014 (Dekens et al., 2007), recalibrated using BAYSPLINE (Tierney et al., 2019). G = Glacial
(Peaple et al submitted), IG = Interglacial (Peaple et al submitted) and P = Pliocene (this study).
The reduction in Searles Lake depth at 3.25 Ma coincides with the desiccation of regional lakes
(Fig 5.3a) including nearby Lake Manly in Death Valley and follows that of Lake Andrei, in the
164
Eureka Valley 200 km north of Searles (Knott et al., 2021). The strong temporal relationship
between the onset of drier conditions at 3.25 Ma in the three lake basins indicates a regional
decrease in P–E after the termination of the MIS MG5 glaciation at 3.33 Ma (Knott et al., 2018).
Regional pollen (Ballog and Malloy, 1981) and petrified wood (Remeika et al., 1988) studies
indicate that the late Pliocene was relatively wet, with woody C
3
plants expanding during this
period. Likewise, the very negative δ
13
C
30acid
(mean = -30.7‰, 1 = 1.4‰, n = 28)
measured
from KM3 sediments (Fig 5.3d) suggests that the Searles Lake catchment was also dominated by
C
3
vegetation (Fig 5.3d).
Plant wax in the lake records precipitation isotopes, a tracer of hydroclimate. δD
precip
values are
relatively invariant between 3.4–3.2 Ma (mean = -89‰, 1 = 2‰, n = 6), before showing higher
amplitude changes (mean = -89‰, 1 = 6‰, n = 20) between 3.2–2.7 Ma (Fig 5.3e). The
Pliocene KM-3 δD
precip
(mean = -89‰) is more enriched than modern Owens Lake groundwater
δD (mean = –122‰) (Bassett et al., 2008) and glacial Searles Lake δD
precip
(–130‰) (Peaple et
al., submitted) both of which represent dominantly winter precipitation signals (Friedman et al.,
1992, Peaple et al in prep). In contrast, the Pliocene δD
precip
is closer to both interglacial Searles
Lake δD
precip
(–94‰) and modern Searles Valley Summer δD
precip
(–57‰) (Friedman et al.,
1992); both of which likely represent greater proportions of summer precipitation (Friedman et
al., 2002, Peaple et al in prep). Here we see that δD
precip
is more enriched in the Pliocene than in
the Pleistocene (Fig 5.3e) and we will compare to model experiments next to constrain the
moisture sourcing and the drivers of the wetter lake conditions.
Searles Lake δD
precip
is more depleted than reconstructions of δD
precip
reconstructed further south
at DSDP Site 475 from Baja California (Bhattacharya et al., preprint), where an expansion of
summer monsoon rainfall was identified in iCESM experiments in response to warmer coastal
165
temperatures (Bhattacharya et al., preprint). Other modeling work has suggested that warmer
coastal temperatures (Fig 5.3f) in the Eastern Pacific would increase coastal precipitation in
southwestern North America, mainly through increasing summer precipitation and a small
increase in wintertime precipitation (Fu et al., 2022).
We compare climate model experiments to the new Searles Lake reconstructions. These new
Searles Lake biomarker reconstructions indicate conditions between 3.4–2.7 Ma were
significantly wetter than modern. Results from iCESM1.2 corroborate the proxy evidence,
simulating a 31% increase in annual precipitation minus evaporation in the region around Searles
Lake (Fig 5.4). Annual changes in P–E are primarily driven by winter rainfall changes (Fig 5.4).
While precipitation in this simulation increases in both summer and winter, increases in
evaporation in summer mean that winter precipitation contributes more to annual P-E increases
in the region around Searles Lake (Fig 5.4). This contrasts with the region around DSDP Site
475, which primarily receives summer rainfall (Bhattacharya et al., preprint).
166
Fig 5.4. Isotope-enabled model output from iCESM1.2. Top left: Anomaly of late Pliocene precipitation
δD relative to modern (black diamond shows the location of Searles Lake). Top right: precipitation–
evaporation (P–E). Bottom left: Pliocene wind vector map with colors representing amount of integrated
water vapor transport (IVT). Bottom right (top panel): Seasonal precipitation, evaporation, and P–E late
Pliocene anomalies calculated from box in top panels. Bottom right (bottom panel): Seasonal precipitation
δD from both preindustrial and late Pliocene periods, calculated from box in top panels.
5.3.2 Constraining moisture sources
The strength of the summer NAM was greater during the late Pliocene (Bhattacharya et al.,
preprint), partly due to elevated coastal temperatures due to reduced cloud cover and decreased
167
coastal winds along the California margin (Schmidt et al., 2020). Coastal temperatures in this
region are elevated above mean Pliocene temperatures (Tierney et al., 2019), suggesting a
significant decrease in coastal upwelling (Miller and Tziperman, 2017; Fu et al., 2021), as well
as an increase in the temperature structure and heat content of the deep ocean (Burls and
Fedorov, 2017; Shankle et al., 2021). To track changes in coastal temperatures through time, we
calculated the mean between ODP 1014 and 1012 SSTs after having subtracted the mean
tropical/mid latitude SSTs (Tierney et al., 2019 and references therein).
Fig 5.5. Plio-Pleistocene SST anomalies (black line) and uncertainties (blue shading, 1 ) from ODP Sites
1012 (Brierley et al., 2009) and 1014 (Dekens et al., 2007), where the original alkenone data from those
studies were recalibrated using BAYSPLINE (Tierney et al., 2019). The top bar shows reconstructed
Searles Lake depth from this study and Smith et al. (1983).
168
We find that between 3.5–3.2 Ma there was a positive and stable SST anomaly of 5°C (Fig 5.5),
followed by a smaller but more variable anomaly between 3.1–2.6 Ma and a sharp decrease in
the anomaly between 2.6–2.5 Ma (Fig 5.5), which represents a shift to glacial/interglacial
variability. This transition at 2.6–2.5 Ma to cooler coastal SSTs likely resulted in a decrease in
NAM moisture west of Baja California and contributed to the desiccation of Searles Lake. These
changes in relative coastal SST anomalies approximately match trends in lake depth and δD
precip
from Searles Lake, with δD
precip
showing markedly increased variability post 3.2 Ma (Fig 5.3)
and lake depth decreasing at 3.25 Ma, followed by total desiccation at 2.6 Ma. Thus, a strong
NAM was important in maintaining relatively deep lakes in Searles Valley during the late
Pliocene. Additionally, we find that warmer coastal temperatures result in an anomalous
anticyclonic circulation along the coast of western North America (Fig 5.4). This is likely a
direct result of coastal warming weakening the pressure gradients that drive along-shore
northwesterly winds along the California Margin (Schmidt et al., 2020). Higher coastal
temperatures are associated with an increase in latent heat flux to atmospheric rivers, which is
associated with increased precipitation intensity (Bartusek et al., 2021). These changes would
result in more frequent and more intense southerly sourced winter precipitation events, which is
analogous to forecasted future precipitation regimes in California (Huang et al., 2020). Our
modeling suggests that δD
precip
would be enriched in all seasons (Fig 5.4), although δD
precip
anomalies are larger in spring and summer than in winter (Fig 5.4). We suggest proxy
reconstructions of more positive δD
precip
result from an increase in southerly moisture sources,
169
while changes in lake level, reflected in the ACE index and other proxy indicators, most likely
reflect a weighted signal towards winter rainfall.
5.3.3 Pliocene tectonic context
Geomorphology indicates that Pliocene Searles Lake was fed through the inflow of the Owens
River, draining the eastern flank of the Sierra Nevada Mountains (Knott et al., 2008). There has
been vigorous debate surrounding the timing of the uplift of the Sierra Nevada Mountain range
to its present height. Significant Pliocene 5–3 Ma uplift was suggested by some studies (e.g.,
Hammond et al., 2012; McPhillips & Brandon, 2010; Stock et al., 2004), although precipitation
isotopes indicate an isotopic rain shadow in place by the middle Miocene (Hren et al., 2010; Mix
et al., 2019; Mulch et al., 2008), implying that the Sierra Nevada’s were already uplifted by the
Pliocene. Today D-depleted precipitation is most associated with northerly moisture sources
distilled by orographic processes crossing the northern and central Sierra Mountains (Friedman
et al., 2002). The absence of this D-depleted rainfall in the Pliocene proxies requires either the
latest possible Pliocene uplift scenarios or the most parsimonious explanation for southerly
moisture to dominate as it does in the Pliocene models.
Tectonics within the Searles basin may also affect the Pliocene lake reconstructions. At the
basin's eastern margin, radiometric evidence indicates rapid exhumation of the Slate Range from
6-4 Ma (Walker et al., 2014). The deepening of the lake floor may have accompanied the uplift
of the basin sides, although the resulting increased accommodation space would have been
counteracted by the infilling of 300 m of lake sediment dating from 3.5 to 2 Ma. If deepening
and infilling were not smoothly aligned, this may have affected the lake storage capacity, and
potential for spillover, during the Pliocene. Searles Lake has rarely overflowed, for example,
during the Pleistocene glacial Termination 2 (Smith, 2009), but geomorphological evidence is
170
inconclusive on whether outflow occurred in the Pliocene (Knott et al., 2008). While local
tectonics may have had a transient influence on the volume of Pliocene Searles Lake, the
presence of the lake robustly indicates a wet climate state.
5.4 Conclusions
Using biomarker proxies, we have demonstrated that variable lacustrine conditions existed
during the late Pliocene, a period previously interpreted as a continuously deep lake (Smith,
1984). We show that the climate of southwestern North America during the late Pliocene (warm
and wet) is very different from the late Pleistocene (cold and wet), with the Pliocene
characterized by more D-enriched precipitation, C
3
plant dominance and high lake levels,
contrasting with Pleistocene climate reconstructions from the same basin (Peaple et al.,
submitted). Lakes were present in both Searles Valley and Death Valley (Knott et al., 2018) from
3.4 to 3.3 Ma. Death Valley desiccated by 3.2 Ma and Searles Lake became shallower at 3.25
Ma, indicating a regional response to changes in P–E. Comparing plant wax δD
precip
reconstructions to isotope-enabled climate models, we show that Searles Lake likely received
increased southerly precipitation from the North American Monsoon and atmospheric rivers
during the late Pliocene. We provide evidence for the potential of these now-desiccated lake
basins to yield continuous biomarker-based reconstructions of Pliocene terrestrial climate in
southwestern North America. This approach can illustrate and motivate the potential of new
continental drilling efforts to increase hydroclimate spatial and temporal resolution on land.
Acknowledgements
The biomarker study and GRA (Peaple) were supported by U.S. National Science
Foundation Grant NSF-EAR-1903665 to S.F., and the Packard Fellowship for Science and
Engineering to J.T. The climate modelling was supported by NSF-OCE-1903148 to T.B. Sample
171
material used in this project was provided by USGS Core Research Center. We thank Patrick
Murphy for assistance preparing and measuring GDGTs.
References
Ballog, R. A., & Malloy, R. E. (1981). Neogene palynology from the southern California
continental borderland, Site 467, Deep Sea Drilling Project, Leg 63. Initial Reports of the
Deep Sea Drilling Project, Leg 63, Long Beach, California to Mazatlan, Mexico, (U.S.Govt.
Printing Office; U.K. Distributors, IPOD Committee, NERC, Swindon), 565–576.
https://doi.org/10.2973/dsdp.proc.63.116.1981
Bartusek, S. T., Seo, H., Ummenhofer, C. C., & Steffen, J. (2021). The role of nearshore air‐sea
interactions for landfalling atmospheric rivers on the US West Coast. Geophysical Research
Letters, 48(6), e2020GL091388.
Bassett, R. L., Steinwand, A., Jorat, S., Petersen, C., & Jackson, R. (2008). Forensic isotope
analysis to refine a hydrologic conceptual model. Ground Water, 46(3), 372–383.
https://doi.org/10.1111/J.1745-6584.2007.00421.X
Bhattacharya, T., Feng, R., Tierney, J. ., Knapp, S., Burls, N. ., & Fu, M., preprint. Expansion
and intensification of the North American Monsoon during the Pliocene. Earth Arxiv doi:
https://doi.org/https://doi.org/10.31223/X54W8N
Brierley, C. M., Fedorov, A. V., Liu, Z., Herbert, T. D., Lawrence, K. T., & LaRiviere, J. P.
(2009). Greatly expanded tropical warm pool and weakened Hadley circulation in the early
Pliocene. Science (New York, N.Y.), 323(5922), 1714–1718.
https://doi.org/10.1126/SCIENCE.1167625
Burls, N. J., & Fedorov, A. V. (2017). Wetter subtropics in a warmer world: Contrasting past and
future hydrological cycles. Proceedings of the National Academy of Sciences, 114(49),
172
12888–12893. https://doi.org/10.1073/pnas.1703421114
Channell, J. E. T., Singer, B. S., & Jicha, B. R. (2020). Timing of Quaternary geomagnetic
reversals and excursions in volcanic and sedimentary archives. Quaternary Science
Reviews, 228, 106114. https://doi.org/10.1016/J.QUASCIREV.2019.106114
Dekens, P. S., Ravelo, A. C., & McCarthy, M. D. (2007). Warm upwelling regions in the
Pliocene warm period. Paleoceanography, 22(3). https://doi.org/10.1029/2006PA001394
Feakins, S. J., & Sessions, A. L. (2010). Controls on the D/H ratios of plant leaf waxes in an arid
ecosystem. Geochimica et Cosmochimica Acta, 74(7), 2128–2141.
https://doi.org/10.1016/J.GCA.2010.01.016
Feakins, S. J., Kirby, M. E., Cheetham, M. I., Ibarra, Y., & Zimmerman, S. R. H. (2014).
Fluctuation in leaf wax D/H ratio from a southern California lake records significant
variability in isotopes in precipitation during the late Holocene. Organic Geochemistry, 66,
48–59. https://doi.org/10.1016/J.ORGGEOCHEM.2013.10.015
Feakins, S. J., Wu, M. S., Ponton, C., & Tierney, J. E. (2019). Biomarkers reveal abrupt switches
in hydroclimate during the last glacial in southern California. Earth and Planetary Science
Letters, 515, 164–172. https://doi.org/10.1016/j.epsl.2019.03.024
Ford, H. L., Ravelo, A. C., Dekens, P. S., LaRiviere, J. P., & Wara, M. W. (2015). The evolution
of the equatorial thermocline and the early Pliocene El Padre mean state. Geophysical
Research Letters, 42(12), 4878–4887. https://doi.org/10.1002/2015GL064215
Friedman, I., Smith, G. I., Gleason, J. D., Warden, A., & Harris, J. M. (1992). Stable isotope
composition of waters in southeastern California 1. Modern precipitation. Journal of
Geophysical Research, 97(D5), 5795. https://doi.org/10.1029/92JD00184
Friedman, I., Harris, J. M., Smith, G. I., & Johnson, C. A. (2002). Stable isotope composition of
173
waters in the Great Basin, United States 1. Air-mass trajectories. Journal of Geophysical
Research: Atmospheres, 107(D19), ACL 14-1. https://doi.org/10.1029/2001JD000565
Fu, M., Cane, M. A., Molnar, P., & Tziperman, E. (2021). Wetter Subtropics Lead to Reduced
Pliocene Coastal Upwelling. Paleoceanography and Paleoclimatology, 36(10),
e2021PA004243. https://doi.org/10.1029/2021PA004243
Fu, M., Cane, M. A., Molnar, P., & Tziperman, E. (2022). Warmer Pliocene Upwelling Site SST
Leads to Wetter Subtropical Coastal Areas: A Positive Feedback on SST.
Paleoceanography and Paleoclimatology, 37(2). https://doi.org/10.1029/2021PA004357
Glen, J. M. G., Liddicoat, J. C., & Coe, R. S. (1999). A detailed record of paleomagnetic field
change from Searles Lake, California 1. Long-term secular variation bounding the
Gauss/Matuyama polarity reversal. Journal of Geophysical Research: Solid Earth, 104(B6),
12865–12882. https://doi.org/10.1029/1999jb900047
Hammond, W. C., Blewitt, G., Li, Z., Plag, H. P., & Kreemer, C. (2012). Contemporary uplift of
the Sierra Nevada, western United States, from GPS and inSAR measurements. Geology,
40(7), 667–670. https://doi.org/10.1130/G32968.1
Hay, R. L., Guldman, S. G., Matthews, J. C., Lander, R. H., Duffin, M. E., & Kyser, T. K.
(1991). Clay mineral diagenesis in core KM-3 of Searles Lake, California. Clays and Clay
Minerals, 39(1), 84–96. https://doi.org/10.1346/CCMN.1991.0390111
Haywood, A. M., Hill, D. J., Dolan, A. M., Otto-Bliesner, B. L., Bragg, F., Chan, W.-L., et al.
(2013). Large-scale features of Pliocene climate: results from the Pliocene Model
Intercomparison Project. Climate of the Past, 9(1), 191–209. https://doi.org/10.5194/cp-9-
191-2013
Hren, M. T., Pagani, M., Erwin, D. M., & Brandon, M. (2010). Biomarker reconstruction of the
174
early Eocene paleotopography and paleoclimate of the northern Sierra Nevada. Geology,
38(1), 7–10. https://doi.org/10.1130/G30215.1
Huang, X., Swain, D. L., & Hall, A. D. (2020). Future precipitation increase from very high
resolution ensemble downscaling of extreme atmospheric river storms in California. Science
Advances, 6(29), eaba1323.
Ibarra, D. E., Egger, A. E., Weaver, K. L., Harris, C. R., & Maher, K. (2014). Rise and fall of
late Pleistocene pluvial lakes in response to reduced evaporation and precipitation:
Evidence from Lake Surprise, California. Geological Society of America Bulletin, 126(11–
12), 1387–1415. https://doi.org/10.1130/B31014.1
Ibarra, D. E., Oster, J. L., Winnick, M. J., Caves Rugenstein, J. K., Byrne, M. P., & Chamberlain,
C. P. (2018). Warm and cold wet states in the western United States during the Pliocene–
Pleistocene. Geology, 46(4), 355–358. https://doi.org/10.1130/G39962.1
De Jonge, C., Hopmans, E. C., Zell, C. I., Kim, J. H., Schouten, S., & Sinninghe Damsté, J. S.
(2014). Occurrence and abundance of 6-methyl branched glycerol dialkyl glycerol
tetraethers in soils: Implications for palaeoclimate reconstruction. Geochimica et
Cosmochimica Acta, 141, 97–112. https://doi.org/10.1016/j.gca.2014.06.013
Knott, J. R., Wan, E., Deino, A. L., Casteel, M., Reheis, M. C., Phillips, F. M., et al. (2021).
Lake Andrei: A Pliocene pluvial lake in Eureka Valley, eastern California. From Saline to
Freshwater: The Diversity of Western Lakes in Space and Time, 125–142.
https://doi.org/10.1130/2018.2536(08)
Knott, J. R., Machette, M. N., Klinger, R. E., Sarna-Wojcicki, A. M., Liddicoat, J. C., Tinsley, J.
C., et al. (2008). Reconstructing late Pliocene to middle Pleistocene Death Valley lakes and
river systems as a test of pupfish (Cyprinodontidae) dispersal hypotheses. In Special Paper
175
of the Geological Society of America (Vol. 439, pp. 1–26). Geological Society of America.
https://doi.org/10.1130/2008.2439(01)
Knott, J. R., Machette, M. N., Wan, E., Klinger, R. E., Liddicoat, J. C., Sarna-Wojcicki, A. M., et
al. (2018). Late Neogene–Quaternary tephrochronology, stratigraphy, and paleoclimate of
Death Valley, California, USA. GSA Bulletin, 130(7–8), 1231–1255.
https://doi.org/10.1130/B31690.1
LaRiviere, J. P., Ravelo, A. C., Crimmins, A., Dekens, P. S., Ford, H. L., Lyle, M., & Wara, M.
W. (2012). Late Miocene decoupling of oceanic warmth and atmospheric carbon dioxide
forcing. Nature, 486(7401), 97–100. https://doi.org/10.1038/nature11200
Liddicoat, J. C., Opdyke, N. D., & Smith, G. I. (1980). Palaeomagnetic polarity in a 930-m core
from Searles Valley, California. Nature, 286(5768), 22–25.
https://doi.org/10.1038/286022a0
Martínez-Botí, M. A., Foster, G. L., Chalk, T. B., Rohling, E. J., Sexton, P. F., Lunt, D. J., et al.
(2015). Plio-Pleistocene climate sensitivity evaluated using high-resolution CO2 records.
Nature, 518(7537), 49–54. https://doi.org/10.1038/nature14145
Martínez-Sosa, P., Tierney, J. E., Stefanescu, I. C., Dearing Crampton-Flood, E., Shuman, B. N.,
& Routson, C. (2021). A global Bayesian temperature calibration for lacustrine brGDGTs.
Geochimica et Cosmochimica Acta, 305, 87–105.
https://doi.org/10.1016/J.GCA.2021.04.038
McPhillips, D., & Brandon, M. T. (2010). Using tracer thermochronology to measure modern
relief change in the Sierra Nevada, California. Earth and Planetary Science Letters, 296(3–
4), 373–383. https://doi.org/10.1016/j.epsl.2010.05.022
Miller, M. D., & Tziperman, E. (2017). The effect of changes in surface winds and ocean
176
stratification on coastal upwelling and sea surface temperatures in the Pliocene.
Paleoceanography, 32(4), 371–383. https://doi.org/10.1002/2016PA002996
Mix, H. T., Ibarra, D. E., Mulch, A., Graham, S. A., & Page Chamberlain, C. (2016). A hot and
high Eocene Sierra Nevada. Bulletin of the Geological Society of America, 128(3–4), 531–
542. https://doi.org/10.1130/B31294.1
Mix, H. T., Caves Rugenstein, J. K., Reilly, S. P., Ritch, A. J., Winnick, M. J., Kukla, T., &
Chamberlain, C. P. (2019). Atmospheric flow deflection in the late Cenozoic Sierra Nevada.
Earth and Planetary Science Letters, 518, 76–85. https://doi.org/10.1016/j.epsl.2019.04.050
Molnar, P., & Cane, M. A. (2007). Early Pliocene (pre-Ice age) El Niño-like global climate:
Which El Niño? Geosphere, 3(5), 337–365. https://doi.org/10.1130/GES00103.1
Mulch, A., Sarna-Wojcicki, A. M., Perkins, M. E., & Chamberlain, C. P. (2008). A Miocene to
Pleistocene climate and elevation record of the Sierra Nevada (California). Proceedings of
the National Academy of Sciences of the United States of America, 105(19), 6819–6824.
https://doi.org/10.1073/pnas.0708811105
Peaple, M. D., Bhattacharya, T., Lowenstein, T. K., McGee, D., Olson, K. J., Stroup, J. S., et al.,
submitted. Biomarker and pollen evidence for late Pleistocene pluvials in the Mojave
Desert. Paleoceanography and Paleoclimatology
Pound, M. J., Tindall, J., Pickering, S. J., Haywood, A. M., Dowsett, H. J., & Salzmann, U.
(2014). Late Pliocene lakes and soils: a global data set for the analysis of climate feedbacks
in a warmer world. Climate of the Past, 10(1), 167–180. https://doi.org/10.5194/cp-10-167-
2014
Remeika, P., Fischbein, I. W., & Fischbein, S. A. (1988). Lower Pliocene petrified wood from
the Palm Spring Formation, Anza Borrego Desert State Park, California. Review of
177
Palaeobotany and Palynology, 56(3–4), 183–198. https://doi.org/10.1016/0034-
6667(88)90057-7
Schmidt, D. F., Amaya, D. J., Grise, K. M., & Miller, A. J. (2020). Impacts of Shifting
Subtropical Highs on the California Current and Canary Current Systems. Geophysical
Research Letters, 47(15), e2020GL088996. https://doi.org/10.1029/2020GL088996
Seager, R., Ting, M., Held, I., Kushnir, Y., Lu, J., Vecchi, G., et al. (2007). Model projections of
an imminent transition to a more arid climate in southwestern North America. Science,
316(5828), 1181–1184.
Seager, R., Neelin, D., Simpson, I., Liu, H., Henderson, N., Shaw, T., et al. (2014). Dynamical
and thermodynamical causes of large-scale changes in the hydrological cycle over North
America in response to global warming. Journal of Climate, 27(20), 7921–7948.
https://doi.org/10.1175/JCLI-D-14-00153.1
Shankle, M. G., Burls, N. J., Fedorov, A. V., Thomas, M. D., Liu, W., Penman, D. E., et al.
(2021). Pliocene decoupling of equatorial Pacific temperature and pH gradients. Nature,
598(7881), 457–461. https://doi.org/10.1038/S41586-021-03884-7
Smith, G. I. (2009). Late Cenozoic geology and lacustrine history of Searles Valley, inyo and
San Bernardino counties, California. US Geological Survey Professional Paper.
https://doi.org/10.3133/pp1727
Smith, G. I., Barczak, V. J., Moulton, G. F., & Liddicoat, J. C. (1983). Core KM-3, a surface-to-
bedrock record of late Cenozoic sedimentation in Searles Valley, California. Professional
Paper. https://doi.org/10.3133/PP1256
Stock, G. M., Anderson, R. S., & Finkel, R. C. (2004). Pace of landscape evolution in the Sierra
Nevada, California, revealed by cosmogenic dating of cave sediments. Geology, 32(3), 193–
178
196. https://doi.org/10.1130/G20197.1
Tierney, J. E., Haywood, A. M., Feng, R., Bhattacharya, T., & Otto-Bliesner, B. L. (2019).
Pliocene Warmth Consistent With Greenhouse Gas Forcing. Geophysical Research Letters,
46(15), 9136–9144. https://doi.org/10.1029/2019GL083802
Turich, C., & Freeman, K. H. (2011). Archaeal lipids record paleosalinity in hypersaline
systems. Organic Geochemistry, 42(9), 1147–1157.
https://doi.org/10.1016/J.ORGGEOCHEM.2011.06.002
Walker, J. D., Bidgoli, T. S., Didericksen, B. D., Stockli, D. F., & Andrew, J. E. (2014). Middle
Miocene to recent exhumation of the Slate Range, eastern California, and implications for
the timing of extension and the transition to transtension. Geosphere, 10(2), 276–291.
https://doi.org/10.1130/GES00947.1
Wang, H., Liu, W., He, Y., Zhou, A., Zhao, H., Liu, H., et al. (2021). Salinity-controlled
isomerization of lacustrine brGDGTs impacts the associated MBT5ME’ terrestrial
temperature index. Geochimica et Cosmochimica Acta, 305, 33–48.
https://doi.org/10.1016/J.GCA.2021.05.004
Williams, A. P., Cook, E. R., Smerdon, J. E., Cook, B. I., Abatzoglou, J. T., Bolles, K., et al.
(2020). Large contribution from anthropogenic warming to an emerging North American
megadrought. Science, 368(6488), 314–318. https://doi.org/10.1126/SCIENCE.AAZ9600
Winnick, M. J., Welker, J. M., & Chamberlain, C. P. (2013). Climate of the Past Geoscientific
Instrumentation Methods and Data Systems Stable isotopic evidence of El NiñoNi˜Niño-
like atmospheric circulation in the Pliocene western United States. Clim. Past, 9, 903–912.
https://doi.org/10.5194/cp-9-903-2013
179
Chapter 6: Paleoenvironmental reconstruction of Pliocene hominin
habitats at Woranso-Mille
This chapter is in preparation for submission to Palaeo
3
:
Mark D. Peaple, Jada Langston, Naomi E. Levin, Yohannes Haile-Selassie, Beverly Z. Saylor,
Jessica E. Tierney, Sarah J. Feakins
Abstract
Hominin fossil discoveries from, Woranso-Mille, Ethiopia, have demonstrated that multiple
Australopithecus species coexisted during the late Pliocene (3.8 – 3.3 Ma). To provide a
palaeoecological context to this hominin diversity, we reconstruct paleoenvironments from
Woranso-Mille between 3.8 – 3.2 Ma. Specifically, we present a biomarker record of
paleoenvironmental change using plant waxes and GDGTs extracted from multiple sedimentary
facies. We compare paleovegetation reconstructions using both δ
13
C of n-alkanoic acids
(δ
13
C
wax
) and pedogenic carbonate nodules (δ
13
C
pc
) and find that δ
13
C
wax
reconstructs greater
abundances of C
4
vegetation compared to δ
13
C
pc
. We compare the evidence from the outcrops of
Woranso-Mille with the nearby Gulf of Aden marine core DSDP Site 231 to determine if the
same proxy (δ
13
C
wax
) in both locations records a similar paleoenvironmental trend. We find that
Woranso-Mille shows higher amplitude changes, likely due to the smaller source area and spatial
and facies variability between samples; both records covary on >100 kyr timescales. We
compare vegetation reconstructions from strata associated with different hominin species at
Woranso-Mille. We find that Australopithecus afarensis and Australopithecus deyiremeda
hominins existed in C
4
grassland environments, whereas the earlier Australopithecus anamensis
preferred environments with woody C
3
vegetation at Woranso-Mille. We also show that hominin
180
dietary expansion to include C
4
foods coincided with a prolonged period of stable C
4
vegetation
at Woranso-Mille, suggesting that vegetation composition influences dietary preferences among
Au. afarensis.
6.1 Introduction
Understanding early hominin paleoenvironments is crucial to providing context for evolutionary
and behavioral changes documented by paleontological discoveries (Cerling et al., 2011, 2013;
Bedaso et al., 2013; Wynn et al., 2013; Barboni, 2014; Uno et al., 2016a; Saylor et al., 2019).
The Pliocene-age sediments outcropping at Woranso-Mille, Ethiopia, have shown that diverse
hominin species were coexisting between 3.8–3.3 Ma (Haile-Selassie et al., 2015, 2019).
Previous environmental reconstruction at Woranso-Mille has used stable carbon isotopes on
pedogenic carbonate nodules within paleosol horizons (Levin et al., 2015) and analyses of the
faunal ungulate community composition (Haile-Selassie et al., 2016b) which indicate the nature
of the C
3
versus C
4
vegetation cover and grazing mammal ecology, respectively. Recent
applications of plant wax carbon and hydrogen isotopes associated with Australopithecus
anamensis have shown that these techniques can work to resolve vegetation and hydroclimate in
a range of facies (Saylor et al., 2019).
Plant waxes have previously been applied to reconstruct eastern African paleoenvironments
including in outcrop exposures of claystone and diatomite lacustrine sediments in the Turkana
Basin, showing the importance of understanding facies (Feakins et al., 2007). More work was
done using these proxies on long 200 m continental drill cores (Lupien et al., 2018, 2020) finding
that precessional changes in insolation played an important role in modulating terrestrial
hydroclimate and from outcrops of terrestrial sediments (Uno et al., 2016b) suggesting that that
hominin diet is influenced by the vegetation composition of their habit. At Olduvai during the
181
Pliocene, plant wax from lake sediments has shown that environments inhabited by hominins had
a variable paleoenvironment that switched between forested to open grassland on precessional
time scales (Magill et al., 2013; Colcord et al., 2018; Shilling et al., 2019). In our initial work at
Woranso-Mille, we studied biomarkers in fine to sand-sized sediments, including the fluvial sand
in which a hominin cranium was found (Saylor et al., 2019). Broadly, changes in both terrestrial
(Uno et al., 2016b) and marine (Uno et al., 2016a) plant wax δ
13
C have been found to track
changes in pedogenic carbonate δ
13
C through the Pliocene (Levin, 2015; Uno et al., 2016b,
2016a), although regionally, pedogenic carbonate has relatively more negative δ
13
C than plant
waxes (Uno et al., 2016b).
Other organic biomarkers can be readily isolated alongside the plant waxes; these include the
glycerol dialkyl glycerol tetraethers (GDGTs) with the branched compounds made by bacteria
and the isoprenoidal compounds associated with archaea (Schouten et al., 2013) that are
commonly reported as temperature-sensitive indicators in a range of settings globally (De Jonge
et al., 2014a). In eastern Africa, brGDGTs have been studied in modern soils of Mt Kilimanjaro
(F. Peterse et al., 2009) and the Serengeti C
4
grasslands in Tanzania (Peaple et al, accepted) to
establish their use as temperature proxies locally. IsoGDGTs have been reported in Holocene
lake sediments (Berke et al., 2012) and Pliocene marine sediments (Liddy et al., 2016) as records
of temperature. IsoGDGTs have also been studied in soil profiles in carbonate-precipitating
Serengeti soils as modern analogs for the Pliocene paleosols studied here (Peaple et al., in
review). To our knowledge, this study constitutes the first study of GDGT compounds in
Pliocene outcrops in Ethiopia. Here, we measure the plant wax carbon, and hydrogen isotope
evidence for vegetation and hydroclimate at Woranso-Mille and test for the presence of
environmental signatures encoded in the GDGT compounds preserved in Woranso-Mille
182
lacustrine and fluvial sediments and paleosols. We find that the moisture sensitive ratio between
isoprenoid and branched GDGTs (Dang et al., 2016) and the temperature sensitive MBT'
5Me
index (De Jonge et al., 2014a) can inform our paleoenvironmental reconstructions. Additionally,
we compare organic (plant waxes) to inorganic (pedogenic carbonate) proxies to constrain
controls on these paleovegetation indicators in Pliocene outcrop sediments. We also compare the
Woranso-Mille plant wax record to the marine sedimentary records from DSDP Site 231 in the
Gulf of Aden, where previous pollen (Bonnefille, 2010) and plant wax (Liddy et al., 2016) have
indicated that a regional cooling, drying and C
4
plant increase occurred through the late Pliocene
in eastern Africa.
6.2 Regional Setting
Woranso-Mille is located on the western side of the Afar depression, Ethiopia (Fig. 1) at a mean
elevation of 600 m. Woranso-Mille is traversed by the Waki and Mille rivers, which are sourced
100 km east in the Ethiopian highlands. The local geology comprises approximately 200 m of
mid-Pliocene fluvial lacustrine sediments, volcanic sediments and basalt (Saylor et al., 2016;
183
Alene et al., 2017) overlain by Pleistocene gravels (
184
Fig. E.1). Pliocene strata are eastward dipping, leaving older rocks (4 – 3.57 Ma) exposed on the
western side of the basin and younger rocks (3.42 – 3.2 Ma) exposed on the eastern side (Saylor
et al., 2016). Modern vegetation in Woranso-Mille is defined as Acacia-Commiphora bushland
(Heslop-Harrison, 2011), containing small, sparsely distributed shrubs, herbs and trees, e.g.,
Acacia bussei, Commiphora alaticaulis. Adjacent to rivers in the basin, a riparian vegetation
corridor comprises woody vegetation with a higher canopy. Pollen analysis of Pliocene lake
deposits from Woranso-Mille suggests that the landscape was dominated by savannah type
vegetation (e.g., Poacea, Asteracea) with periods of wetland (Typha) and riparian vegetation
(Trilepisium madagascariense) (Saylor et al., 2019) all of which are indicative of a wetter
environment compared to the modern vegetation.
Today, the climate is hot and arid, with an average annual temperature of 26.6˚C and 150 mm of
annual precipitation (Nicholson, 2000). The region receives most of this rain between June, July,
and August (Kiremt rains), typically from precipitation sourced from terrestrial moisture in the
Congo basin (Levin et al., 2009). Precipitation occurs during the Masika rains in March, April
and May from moisture sourced from the Indian ocean (Levin et al., 2009; Bedaso et al., 2020).
We compare the terrestrial record to the marine sedimentary records from DSDP Site 231 in the
Gulf of Aden, 800 km east of Woranso-Mille (Fig 6.1).
185
Fig 6.1 Location of hominin fossil localities Woranso-Mille and Hadar, and DSDP Site 231 A)
regional DEM (ETOPO1, 1 minute resolution) and B) 3D digital elevation map (Amante and
Eakins, 2009) highlighting the modern Mille River drainage basin (color shading highlighting
elevation change).
186
6.3 Methods
6.3.1 Field sampling
We collected biomarker and pedogenic carbonate nodule samples from Woranso-Mille in 2012
(biomarkers = 27, pedogenic carbonate horizons = 9) and 2018 (biomarker = 127, pedogenic
carbonate = 22). Pits were excavated to expose fresh outcrop surface (digging ~1 m back from
the face) before samples were collected to avoid fractured surface and modern contamination.
Samples were generally collected below the modern plant rooting zone (~50 cm) to avoid
contamination with modern organic matter. We collected biomarker samples from both
fluvial/lacustrine (99 samples) as well as paleosols (53 samples) and kerogen (2 samples).
Pedogenic carbonate was sampled from <0.5 m below the soil surface and the BK horizon. We
collected paired samples of pedogenic carbonates and biomarkers from the same position in
paleosols to allow comparison between the proxies.
6.3.2 Laboratory methods
6.3.2.1 Biomarker extraction and purification
Samples were freeze-dried, ground, and then lipids were extracted by Accelerated Solvent
Extraction (Dionex) using 9:1 dichloromethane (DCM):methanol (MeOH) at 100ºC and 1500 psi
to yield the total lipid extract (TLE). Before extraction, quartz sand was added to samples in a
1:1 ratio to increase extraction efficacy. The TLEs were then separated into neutral and acid
fractions using columns packed with NH
2
coated silica gel. Neutral fractions were eluted using a
2:1 DCM:isopropanol and acid fractions were eluted with 4% formic acid in diethyl ether.
Carboxylic acids were methylated with 95:5 MeOH:hydrochloric acid at 70 °C for 12 hours.
Fatty acid methyl esters (FAMEs) were then separated from non-polar molecules using a 5%
deactivated silica gel column eluted with DCM. Saturated compounds were then removed from
187
samples by using a silver nitrate silica gel column, with FAMEs being eluted by DCM. The
neutral fraction was separated into n-alkanes and the glycerol dialkyl glycerol tetraethers
(GDGTs) fraction using columns packed with 5% deactivated silica gel. n-Alkanes were eluted
with hexane and the GDGTs were eluted with DCM and methanol. n-Alkanes extracted from
sediments were passed through a copper column to remove elemental Sulphur.
6.3.2.2 Plant wax analyses
Identification and molecular abundance distribution of n-alkanoic acid methyl esters were
determined using gas chromatography (Agilent 6890) mass spectrometry (Agilent 5973)
and flame ionization detection. Compound identification was based on mass spectra of target
peaks compared to library and retention time compared to standards. Quantifying FAMEs was
conducted by comparing sample peak areas in the FID chromatograms to peak areas of an in-
house standard containing known concentrations of n-alkanoic acids. We report data for the
C
22
to C
32
n-alkanoic acids. Summed concentrations were calculated for ΣC
22-32acid
in ng/g dry
sediment. Carbon preference index (CPI) and average chain length (ACL) were calculated to
represent chain length distributions using the following equations:
ACL = ∑ (n × [C
n
]) / ∑ [C
n
] (1)
CPI = 2[C
n
]/([C
n−1
] + [C
n+1
]) (2)
where the chain length (n) refers to C
26
to C
30
n-alkanoic acids.
The n-alkane fraction contains an uncharacterized complex mixture precluding analysis of the
plant wax n-alkanes. The n-alkanoic acids were instead targeted for compound specific C and H
isotopic analyses.
6.3.2.3 Compound specific isotopic analyses
Carbon and hydrogen isotopes of n-alkanoic acids were measured using a Thermo Scientific
188
Trace gas chromatograph equipped with a Rxi-5ms column (30 m × 0.25 mm; film thickness,
1 μm) with a PTV injector operated in solvent-split mode, which was coupled to a Delta V Plus
isotope ratio mass spectrometer using an Isolink combustion/pyrolysis furnace (1000/1400 °C).
Linearity of isotopic reference gas CO
2
was determined over a range of amplitudes (1–10 V) and
the standard deviation (0.1‰) was measured daily. H
3
+
factor was measured daily and averaged
7ppm mV
–1
.
A standard of known isotopic compositions (A6 mix supplied by A.
Schimmelmann, University of Indiana) was measured daily, allowing for normalization to the
Vienna Standard Mean Ocean Water/Standard Light Antarctic Precipitation and Vienna Pee Dee
Belemnite isotopic scales for δD and δ
13
C values, respectively. The root mean square error of the
replicate analyses of the external standard was 5‰ for δD and 0.1‰ for δ
13
C. FAME δ
13
C and
δD values were corrected using a mass balance for the addition of the methyl group of known
isotopic composition (Lee et al., 2017).
6.3.2.4 GDGT analyses
The neutral polar fractions were dissolved in hexane:isopropanol (99:1) and filtered through a
0.45 μm polytetrafluoroethylene filter. Samples were then injected into an Agilent 1260 High-
Performance Liquid Chromatograph (HPLC) coupled to an Agilent 6120 mass spectrometer at
the University of Arizona. Separation of the GDGTs was accomplished using two Ethylene
Bridged Hybrid (BEH) Hydrophilic Interaction (HILIC) silica columns (2.1 mm × 150 mm, 1.7
μm; Waters) following the method of Hopmans et al. (2016). Single Ion Monitoring (SIM) of the
protonated molecules (M + H
+
ions) was used to detect and quantify GDGTs with abundances
determined by comparison to a C
46
internal standard at m/z 744 (Huguet et al., 2006). We report
concentrations of summed branched (∑brGDGT) and isoprenoidal (∑isoGDGTs) GDGTs per
unit mass of sediment (ng/g) as well as the concentrations of individual compounds that are
189
further used in a range of index calculations. In order to assess the temperature information
contained in the brGDGT abundances, we calculate the MBT´
5Me
(De Jonge et al., 2014,
Hopmans et al., 2016) where
MBT
5Me
′
=
Ia+Ib+Ic
Ia+Ib+Ic+IIa+IIb+IIc+IIIa
(3)
We obtain the mean annual temperature from MBT´
5Me
using the BayMBT
0
model (Dearing
Crampton-Flood et al., 2020). We use a soil calibration (Dearing Crampton-Flood et al., 2020)
for paleosol samples and a lacustrine calibration (Martínez-Sosa et al., 2021) for
fluvial/lacustrine samples.
The ratio of isoGDGTs to brGDGTs in soils has been proposed to be an index for aridity (Xie et
al., 2012; Dang et al., 2016) where:
R
i/b
= ∑ isoGDGTs/∑ brGDGTs (4)
6.3.3 Pedogenic carbonate isotopic analyses
Pedogenic carbonate nodules were collected from soil Bk horizons at least 50 cm below the
paleosol surface. After digestion in 100% H
3
PO
4
at 90°C, the isotope compositions of powdered
carbonates (δ
13
C and δ
18
O) were analyzed on a Nu Perspective with an online Nu Carb
autosampler. Corrections are based on a two-point calibration using the NBS-18 and NBS-19
calcite standards and normalized to the VPDB isotopic scale.
190
6.4 Results
6.4.1 n-Alkanoic acid abundances
n-Alkanoic acids were generally present in low concentrations in these Pliocene samples (∑
acid
median = 184 ng/g, IQR = 348 ng/g, n=130) with the exception of an organic kerogen or tar with
∑
acid
= 88700 ng/g. We find systematic differences in concentrations between facies with higher
median ∑
acid
in paleosols (274 ng/g, IQR = 477ng/g, n = 50) and lower abundances in fluvial and
lacustrine sediments (153 ng/g, IQR = 173 ng/g, n = 79). Across facies, we found grain size
correlated with concentration, with clays having the highest median ∑
acid
of 221 ng/g (IQR = 658
ng/g, n = 12) compared to silt (149 ng/g, IQR = 152 ng/g, n= 57) and sand (108 ng/g, IQR = 110
ng/g, n = 9).
Carbon preference index (CPI) values had a median = 1.9 (IQR = 0.6, n = 134) and average
chain length had a median = 27.6 (IQR = 1.8, n = 137). Fluvial/lacustrine samples had a median
CPI of 2.0 (IQR = 1.4, n= 81), which is similar to the median paleosol CPI of 1.9 (IQR = 0.3, n =
52). The median average chain length of paleosol was 27.9 (IQR = 0.9, n= 53) compared to
median fluvial lacustrine sediments of 27.5 (IQR = 2, n = 83). As previously noted, the n-alkanes
were below detection in most samples; however, we revisited the paleosols (n = 16), which
showed relatively high n-alkanoic acid concentrations for n-alkanes. We found that almost all
samples had a significant uncharacterized mixture (UCM) in chromatograms peaking around C
23
n-alkane and extended across long chain alkanes, precluding accurate isotopic analyses of the n-
alkanes.
6.4.2 Carbon and hydrogen isotopic compositions of n-alkanoic acids
We obtained δ
13
C values for individual, even-chain length C
26
to C
32
carbon chain length n-
alkanoic acids for 127 samples (Fig 6.2). We find that the n-alkanoic acid δ
13
C values are
191
consistent across C
26
-C
30
, with medians of -21.7‰ (IQR = 3.6, n= 56), -23.1‰ (IQR = 4.2, n =
127), and -22.4‰ (IQR = 3.3, n = 93) (Fig 6.2a). C
32
has a more positive median δ
13
C of -19.3‰
(IQR = 2.9, n = 68), which could be driven by an over representation of long chain (>C
30
) C
4
grass lipid input relative to their abundance on the landscape, although lower peaks of C
32
alkanoic acid could have also contributed. We found no significant difference in δ
13
C between
dominant lithologies or facies (Fig 6.2c) with both paleosols and fluvial/lacustrine samples
having median δ
13
C
28acid
values of -22.5‰ (IQR = 4.4, n = 51) and -23.4‰ (IQR = 4, n = 75),
respectively. The kerogen had a distinct isotopic composition, with δ
13
C
28acid
of -30.3‰. We
found a strong correlation between δ
13
C measured in C
26-32
n-alkanoic acids (Fig 6.3), with a
mean r = 0.76, p < 0.01.
192
Fig 6.2 Violin plots showing distributions of a) δ
13
C
and b) δD
data measured from C
26-32
alkanoic
acids.
c) δ
13
C
28acid
and d) δD
28acid
of samples delineated by lithology.
Fig 6.3 Correlation coefficient between A) δ
13
C and B) δD measured in C
26-32
alkanoic acids.
The median δD measured in longer chain lengths (δD
30
median
= -145‰, IQR = 54‰, n = 36,
δD
32
median
= -149‰, IQR = 51‰, n = 24) was more negative than in the mid/long chain
lengths (δD
26
median
= -131‰, IQR = 42‰, n = 25, δD
28
median
= -138‰, IQR = 43‰, n = 61)
(Fig 6.2b). There were no significant differences between δD
C28
measured in different samples
lithologies or facies (Fig 6.2d), with paleosols having a median δD
C28
= -140‰ (IQR = 52‰, n =
30) and fluvial/lacustrine samples having a median δD
28
= -129‰ (IQR = 32‰, n = 30).
Likewise, both silts (δD
28
median = -131‰, IQR = 37‰, n = 19) and clays (δD
28
median = -
127‰, IQR = 35‰, n = 7) had similar δD
28
distributions (Fig 6.2d). Sands had a more negative
δD
28
(median = -115‰, IQR = 33‰, n = 3) although the very small sample size preludes
interpretations. δD measured across all n-alkanoic acids chain lengths had a very strong
correlation (Fig 6.3B) and ranged between 0.87 to 0.97.
193
6.4.3 GDGT distributions
We analysed a subset of 60 samples for GDGT concentrations in both paleosols (n = 40) and
fluvial lacustrine deposits (n = 20) targeting both a lacustrine section (MRD) as well as paleosls
which had generally higher concentrations of plant wax. Concentrations of both isoGDGTs
(median = 0.01 ng/g, IQR =0.03, n=53) and brGDGTs (median = 0.02 ng/g, IQR = 0.07, n=53)
were very low. Fluvial-lacustrine samples have greater abundances of both isoGDGTs (median =
0.02 ng/g, IQR = 0.05, n=20) and brGDGTs (median = 0.03 ng/g, IQR = 0.07, n = 20) compared
to paleosol isoGDGTs (median = 0.01 ng/g, IQR = 0.01, n = 32) and brGDGTs (median = 0.02
ng/g, IQR = 0.03, n = 32). Silts had a higher abundance of isoGDGTs (median = 0.03 ng/g, IQR
= 0.05, n = 14) and brGDGTs (median = 0.05 ng/g, IQR = 0.06, n = 14) than clay isoGDGTs
(median = 0.004 ng/g, IQR = 0.025, n = 5) and brGDGTs (median = 0.025 ng/g, IQR = 0.008, n
= 5). BIT was high but variable in paleosol (median = 0.66, IQR = 0.53, n = 35) and fluvial
lacustrine (median = 0.75, IQR = 0.4, n = 21) samples. Among fluvial lacustrine samples BIT
values were similar across lithologies with, clays (median = 0.8, IQR = 0.4, n = 5), silts (median
= 0.7, IQR = 0.4, n = 14) and sand (median = 0.95, n = 2) having similar distributions. The R
i/b
was similar in paleosols (median = 0.51, IQR = 1.21, n = 37), fluvial lacustrine sediments
(median = 0.44, IQR = 1.31, n = 21). R
i/b
was also consistent across different fluvial lacustrine
lithologies with clay (median = 0.2, IQR = 0.9, n = 5), silt (median = 0.6, IQR = 1.7, n = 14) and
sand (median = 0.2, n = 2) having similarly low values. We find MBT'
5Me
in paleosol samples
(median = 0.92, IQR = 0.05, n = 22) and fluvial lacustrine silt sediments (median = 0.94, IQR =
0.03, n = 6) all have similar values near the upper limit (1.00) of the index. However, several
brGDGT compounds (IIb, IIc) were below detection limits, which could lead to a warm bias.
194
6.5 Discussion
6.5.1 Biomarkers across facies
Biomarkers have the advantage that they can be extracted from multiple sedimentary facies, and
we first examine the results in that context. The outcrops include fluvial and lacustrine sediments
and paleosols. Both paleosols and fluvial/lacustrine samples have similar δ
13
C
28acid
distributions
(Fig 6.4A), suggesting that they record both C
3
and C
4
vegetation at Woranso-Mille during the
late Pliocene. Paleosols with pedogenic carbonates samples have similar median δ
13
C
28acid
to
paleosols without pedogenic carbonates (-22.1‰ vs. -22.2‰ respectively), although the
δ
13
C
28acid
range (-16‰ to -26‰) of pedogenic carbonate wax samples is indicative of pure C
4
to
mixed C
4
/C
3
vegetation. Fluvial/lacustrine samples integrate waxes over the watershed of a
river/lake and thus capture a larger wax source area than paleosols, whose wax inputs are
predominantly from overlying vegetation, although some fluvial input can occur in floodplain
settings. The δ
13
C
28acid
measured in fluvial lacustrine samples reflects landscape δ
13
C
28acid,
whereas δ
13
C
PC
from paleosols reflects more local inputs. From this, it might be suspected that
the IQR in δ
13
C
28acid
measured in fluvial/lacustrine samples would be smaller than the IQR in
paleosols, given the integration of a larger wax source area. However, the IQR in
fluvial/lacustrine δ
13
C
28acid
is larger than in paleosols (4.5 vs. 2.5, respectively) which suggests
that paleosols capture a smaller paleoenvironmental range than fluvial/lacustrine sediments,
possibly because paleosols only form under certain environmental conditions that allow
pedogenesis to keep pace with sediment input (Kraus, 1999).
The distributions of δD
28acid
across each facies (Fig 6.4b) suggest no significant facies biases
associated with this proxy. However, fluvial/lacustrine sample δD
28acid
is generally more positive
than the paleosol facies (-131‰ vs. -139‰, respectively). Interpreting this difference in δD
28acid
195
is challenging at Woranso-Mille, given the numerous controls on precipitation δD
,
including
amount effect and precipitation moisture source (Levin et al., 2009; Bedaso et al., 2020).
Different plant types also have different biosynthetic fractionations of D/H, with C
4
grasses
having more negative fractionations than C
3
shrubs (Sachse et al., 2012). At Woranso-Mille,
paleosols typically have a more positive δ
13
C
28acid
, indicative of more C
4
vegetation, which could
explain the more negative δD
28acid
.
Fig 6.4 Violin plots of A) δ
13
C
28acid
,
B) δD
28acid
, C) R
i/b
and
D) BayMBT
0
mean annual air
temperature, measured in fluvial/lacustrine, paleosol and paleosol with pedogenic carbonate
sediments.
Sampled paleosols include those with and without pedogenic carbonates. We find the R
i/b
is
lower in association with pedogenic carbonates (Fig 6.4c), consistent with these soils having had
196
a higher moisture content. Elsewhere in arid environments, lower R
i/b
has correlated with higher
soil moisture content (Dang et al., 2016). Based on the modern soil study (Dang et al., 2016),
these data would suggest that pedogenic carbonate precipitating soils had >20% soil moisture
content; however, 50% of paleosols without pedogenic carbonate have an R
i/b
>1, indicating a
soil moisture content <20%. Pedogenic carbonate only forms in seasonally dry conditions, and
the depth of carbonate formation is proportional to the mean annual precipitation (Zamanian et
al., 2016). In this study, pedogenic carbonate was only sampled at >0.5 m depth to ensure that
the impact of atmospheric CO
2
carbonate δ
13
C was minimized. This could explain why we see
lower R
i/b
in paleosols with pedogenic carbonate compared to paleosols without pedogenic
carbonates, with paleosols with pedogenic carbonate forming in wetter conditions. Given that
paleosols with pedogenic carbonates formed from soil with higher soil moisture contents, this
suggests pedogenic carbonate paleosols generally record wetter, more C
4
environments than
either paleosols without pedogenic carbonates and fluvial lacustrine samples, which record a
broader range of paleoenvironmental information.
Fluvial lacustrine samples have poor preservation of brGDGT compounds, and as such, we were
only able to calculate the temperature sensitive brGDGT index MBT'
5Me
on 6 samples. However,
the BayMBT
0
temperature estimates (MAT) from these samples were high (median = 25.9°C),
with no samples <24°C (Fig 6.4d). The median BayMBT
0
temperature
of paleosols is lower than
that of fluvial/lacustrine samples at 25.2°C, although the IQR is larger (1.9 vs. 0.9 respectively)
with a greater proportion of temperatures beneath the modern MAT of 27°C. In a modern
Savannah environment, MBT'
5Me
is reflective of the mean annual air temperature in the surface
soil but can be biased to higher (warmer) values beneath the soil surface due to salinity and high
pH (Peaple et al 2022). However, here we see that MBT'
5Me
and BayMBT
0
temperatures from
197
deep soil samples
are lower than fluvial/lacustrine samples suggesting that this “warm” bias is
either not present or reduced in these Woranso-Mille paleosols. MBT'
5Me
is typically higher in
soils for the same temperature than in lakes in modern global compilations (Dearing Crampton-
Flood et al., 2020; Martínez-Sosa et al., 2021). Thus, given that no fluvial lacustrine samples
have an MBT'
5Me
<0.9 suggests that lakes are reflecting higher temperature conditions (Fig 6.4d).
Closed vegetation cover can drastically reduce soil temperature relative to unvegetated soils by
up to 25°C (Cerling et al., 2011), possibly explaining the offset between paleosols and
fluvial/lacustrine samples we observe here. Overall, the temperatures reconstructed from these
Pliocene samples are similar to modern MAT (Nicholson, 2000). Previous reconstructions of
equatorial terrestrial Pliocene surface temperatures suggest that temperatures were similar to
modern (Passey et al., 2010; Zhang et al., 2019), which agrees with our Woranso-Mille
BayMBT
0
data. Locally, the increased water availability on the landscape (Curran and Haile-
Selassie, 2016; Saylor et al., 2019; Su and Haile-Selassie, 2022) would have reduced ground
surface temperatures (Lin et al., 2003), possibly offsetting the increase in mean temperatures
from higher atmospheric CO
2
concentrations (Stap et al., 2016).
6.5.2 Comparing plant wax and pedogenic carbonate proxies
Given that both δ
13
C
wax
and δ
13
C
pc
proxies are used to reconstruct plant δ
13
C, we sampled both
from the same position in paleosol Bk horizons to directly compare the proxies. We found no
correlation between paired pedogenic carbonate and plant wax plant δ
13
C (r=0.02). A lack of
correlation between plant wax and soil carbonate δ
13
C in Pleistocene deposits has also previously
been reported (Sarangi et al., 2021) from channel sands and ascribed to sparse vegetation cover
leading to reduced soil respiration rates.
198
At depth >0.5 m below the soil surface, pedogenic carbonate δ
13
C is influenced by the δ
13
C of
respired plant CO
2
(Cerling, 1984). The theory, therefore, suggests that the δ
13
C of total organic
carbon (δ
13
C
TOC
)
should be 14-16‰ more negative than δ
13
C
PC
(Zamanian et al., 2016)
due to the
fraction of CO
2
during diffusion in soil (4.4‰) and precipitation of carbonate (12-14‰). δ
13
C of
plant wax is 0–10‰ (mean = 7‰) more negative than equivalent soil organic matter due to a
biosynthetic fractionation in wax production (Diefendorf and Freimuth, 2017). As such, plant
waxes would be expected to be more negative than an equivalent δ
13
C
PC.
Fig 6.5 Violin plot showing distributions of δ
13
C
28acid
(left) and δ
13
C
PC
(right). δ
13
C
28acid
samples
are grouped by facies type.
In modern C
4
grassland soils in the Serengeti National Park, the mean δ
13
C
PC
- δ
13
C
28acid
(excluding surface samples) = 18.7‰ (Zhang et al., 2021). This is larger than the mean δ
13
C
PC
-
δ
13
C
28acid
measured in paired samples from Woranso-Mille (14.1‰, Fig 6.5). Although unpaired,
n-alkane and n-alkanoic acid samples from the late Pleistocene Nachukui formation in West
Turkana have more enriched δ
13
C relative to pedogenic carbonate δ
13
C from the same formation
(Uno et al., 2016b). The fact that we see a similar relative enrichment of plant wax δ
13
C relative
to pedogenic carbonates across two Plio-Pleistocene age sites could suggest an intrinsic
199
difference between these two proxies when sampled from ancient outcrop sediments. Identifying
the exact cause of this difference is not possible. However, it could reflect the combined effects
of degradation of plant waxes (Wu et al., 2019), recrystallizing of pedogenic carbonates
(Zamanian et al., 2016), differences in peak seasonal C
3
and C
4
plan respiration rates, seasonality
of pedogenic carbonate formation and differences between C
3
and C
4
plant rooting depths (Uno
et al., 2016b).
6.5.3 Comparing marine and terrestrial paleoenvironmental archives
To evaluate the environmental responses recorded from terrestrial and marine proxy archives, we
compare our Woranso-Mille δ
13
C
wax
record to the marine core DSDP Site 231 δ
13
C
wax
record.
Site DSDP Site 231 is located 800 km east of Woranso-Mille in the Gulf of Aden (Fig 6.1). It
receives aeolian transported plant wax from the Afar triangle and the Somali peninsula
(Prospero, 2002) and thus represents a regionally integrated record. Woranso-Mille plant wax is
currently sourced from within the smaller drainage basin of the Mille River, which flows from
the Ethiopian highlands 100 km to the east.
Between 3.8 to 3.7 Ma, there is a synchronous increase in δ
13
C
wax
at Woransno-Mille and marine
core site DSDP Site 231 (Liddy et al., 2016) (Fig 6.6B), reflecting an expansion of C
4
vegetation
and an opening of the landscape. From 3.7 to 3.4 Ma, δ
13
C
wax
decreases at both sites, likely
reflecting an increase in woody vegetation followed by an increase in δ
13
C
wax
between 3.4 to 3.2
Ma, which is interpreted to reflect an increase in C
4
shrubs (Liddy et al., 2016). This
synchronicity in the changes in δ
13
C
wax
at both sites is remarkable given the age, different
sedimentary archives, and distance between locations. This suggests that DSDP 231 δ
13
C
wax
record reflects vegetation changes also observed in smaller sedimentary basins on a timescale of
200
~100 kyr, demonstrating the utility of marine records. However, the range in δ
13
C
wax
values at
DSDP Site 231 (5‰) is smaller than measured at Woranso-Mille (14‰), likely due to greater
spatial and temporal integration in the marine δ
13
C
wax
.
Fig 6.6. A. Insolation at 20 degrees latitude North and eccentricity (Laskar et al., 2004). B.
δ
13
C
wax
records from marine core DSDP 231(Liddy et al., 2016) and Woranso-Mille. C. Soil
carbonate δ
13
C data from Afar compiled by Levin, (2013, references therein) and including new
δ
13
C data in this publication. D. Bovid enamel δ
13
C from Afar paleoanthropological sites
201
compiled by Paquette and Drapeau, (2021). E. Carbonate clumped isotope derived temperatures
from paleosols in the Turkana Basin, Kenya. F. Modelled atmospheric CO
2
concentration (Stap
et al., 2016).
What triggered these shifts in the paleoenvironment? Over millions of years, a decrease in
atmospheric CO
2
concentration (Ehleringer et al., 1991) and changes in oceanic gateways (Cane
and Molnar, 2001) have been identified as potential drivers of a long term increase in the
abundance of C
4
plants in east Africa. Additionally, over thousands of years, changes in
eccentricity modulated precession altered the seasonal distribution of insolation (Fig 6.6A), with
higher seasonal insolation changes leading to a greater proportion of C
4
vegetation at Olduvai
Gorge (Tanzania) (Magill et al., 2013). The DSDP 231 δ
13
C
wax
record shows both the
Milankovitch scale and long-term shifts and changes on intermediate timescales. For example,
between 3.8–3.6 Ma there is a (+4‰) shift to more positive and stable δ
13
C
wax
at DSDP 231,
implying an increase in C
4
vegetation, before returning to a C
3
environment at 3.5 Ma with a (–
3‰) shift in δ
13
C
wax
. This same C
4
excursion is recorded in the new Woranso-Mille δ
13
C
wax
record (Fig 6.6B), bovid tooth enamel δ
13
C (Fig 6.6D) (Paquette and Drapeau, 2021) and
tentatively in the Afar paleosol carbonate δ
13
C (Fig 6.6C) (Levin, 2013) record. Pollen data from
DSDP 231(Bonnefille, 2010) shows an increase in the dry adapted C
4
Amaranthace during this
period, implying a decrease in precipitation drove an increase in arid vegetation and an opening
up of the landscape.
Between 3.55–3.45 Ma at Woranso-Mille, we see a decrease in δ
13
C
wax
toward C
3
vegetation,
which we interpret as a brief return to a more closed woodland environment, suggesting an
increase in precipitation, before returning to more C
4
(drier) conditions from 3.4–3.2 Ma. Bovid-
assemblage-based paleoenvironmental reconstructions at the Hadar fossil site (Reed, 2008)
between 3.4-3.2 Ma also show similar trends to Woranso-Mille and DSDP 231, in addition to the
202
bovid tooth enamel (Paquette and Drapeau, 2021) (Fig 6.6D) and pedogenic carbonate records
(Levin, 2013) (Fig 6.6C).
Detecting a causal mechanism for such vegetation changes on intermediate (0.2 Ma) time scales
is challenging due to temporal uncertainty in the various records and the combination of tectonic,
oceanic, and climatic factors which are generally not well constrained (Levin, 2015). However,
we note the apparent correlation between periods of low CO
2
(Fig 6.6F ) (Stap et al., 2016) and
the increase in the proportion of C
4
plants in our Afar vegetation records (Fig 6.6B). Low
atmospheric CO
2
would benefit C
4
plants directly by increasing the efficiency of carbon fixation
vs. C
3
plants, although studies have suggested that atmospheric CO
2
variability is often not a
primary cause of vegetation change (Cotton et al., 2016). However, decreasing CO
2
would also
decrease global temperatures, impacting regional hydrological cycling.
The high-resolution nature of the sampling at Woranso-Mille also provides information on
shorter-term environmental changes. Lake sediments at the base and flood plain deposits at the
top of our record show high amplitude changes in δ
13
C
wax
over thousands of years (Fig 6.6B).
Interestingly, these high amplitude δ
13
C
wax
changes occur during periods of high eccentricity
variability (Fig 6.6A), which has also been observed from Pliocene lacustrine deposits in eastern
Africa (Lupien et al., 2020). Furthermore, insolation has been shown to negatively correlate with
δ
13
C
wax
from early Pleistocene east African lake sediments (Magill et al., 2013) on precessional
timescales, suggesting low insolation resulted in an expansion of C
4
environments. Our results
confirm this association and highlight the importance of eccentricity in modulating the pace of
regional paleoenvironment changes.
203
6.5.4 Hominin paleoenvironment reconstruction
Previous environmental reconstructions of early hominin habitats based on non-hominin
mammalian fossil assemblages (e.g., Behrensmeyer and Reed, 2013; Kimbel and Delezene,
2009; Reed, 2008) and pedogenic carbonate nodule proxy data (e.g., Bedaso et al., 2013; Cerling
et al., 2011; Levin, 2013; Levin et al., 2015; Quinn and Lepre, 2020) suggested a landscape
mosaic composed of grassland, woodland and closed-canopy vegetation (Reed, 2008; Kimbel
and Delezene, 2009; Behrensmeyer and Reed, 2013) across Pliocene eastern African hominin
sites. These landscape mosaics were also present at Woranso-Mille (Su and Haile-Selassie, 2022)
over the late Pliocene and were inhabited by Au. anamensis, Au. afarensis and uniquely at
Woranso-Mille, Au. deyiremeda. To understand the ecological context of Woranso-Mille
hominins, we binned Woranso-Mille hominin fossil sites by hominin species and then assigned a
species to each biomarker sample. We then plotted the distributions of biomarker ecological
indicators (Fig 6.7) to establish the relationship between hominin species and
paleoenvironmental indicator.
204
Fig 6.7 Violin plots of a) δ
13
C
28acid
b) δD
28acid
c) R
i/b
and d) BayMBT
0
mean annual air
temperature,
with samples delineated by the hominin species present at outcrop locality.
We find that Au. afarensis inhabited both C
3
and
C
4
environments (Fig 6.7a) as demonstrated by
the range in δ
13
C
28acid
from -29.6‰ to -16.8‰, although the median value of -21.5‰ suggests
that Au. afarensis was more commonly associated with C
4
vegetation. Previous studies
(Behrensmeyer and Reed, 2013; Su and Harrison, 2015; Su and Haile-Selassie, 2022) have
demonstrated that Au. afarensis was a cosmopolitan species that inhabited dry and relatively
open woodland, bushland, and shrubland at Laetoli in addition to the wetter and more riparian
environments of Woranso-Mille (Su and Haile-Selassie, 2022) and Hadar (Behrensmeyer and
Reed, 2013). As such, while Au. afarensis is more commonly present in C
4
environments at
Woranso-Mille; this does not necessarily imply a habitat preference as the median δ
13
C
28acid
205
measured in all Woranso-Mille samples is -21.7‰. Au. anamensis existed in comparatively more
C
3
environments (median δ
13
C
28acid =
-24.5‰) than Au. afarensis (median δ
13
C
28acid =
-21.5‰) (Fig
6.7a). Previously, pollen and plant wax identified an Au. anamensis fossil site to be close to C
3
open woodland and riparian forest vegetation (Saylor et al., 2019). We now demonstrate that this
association between Au. anamensis and C
3
vegetation extend throughout Woranso-Mille
sediments (Fig 6.7a). Both Au. anamensis and Au. afarensis adjacent sediments have similar
δD
28acid
(Fig 6.7b) and a median R
i/b
of <1 (Fig 6.7c) indicative of non-arid environments. Au.
anamensis has previously been linked to heterogeneous environments at multiple eastern African
hominin sites (Saylor et al., 2019; Bobe et al., 2020; Quinn and Lepre, 2020) and likely had a
preference for similarly mixed C
3
/C
4
vegetation at Woranso-Mille. The age range of Au.
anamensis of 4.2–3.8 Ma (Haile-Selassie et al., 2016a) places it in the most C
3
environment
present regionally anytime during the past 5 Ma according to δ
13
C
wax
measured from DSDP 231
(Liddy et al., 2016). As such, we have both regional and local paleoenvironmental information
pointing toward Au. anamensis having a stronger C
3
ecological preference than Au. afarensis.
Therefore, the fact that we see Au. anamensis disappearing from the fossil record prior to a
prolonged paleoenvironmental switch to C
4
vegetation at 3.6 Ma (Fig 6.6b) might suggest that
the paleo vegetation change could have contributed to the extinction of Au. anamensis. Perhaps
this ecological change favored the more cosmopolitan Au. afarensis, at the expense of the more
ecologically constrained Au. anamensis.
Both Au. anamensis and Au. afarensis adjacent sediments have similar δD
28acid
(Fig 6.7b) and
median R
i/b
of <1 indicative of non-arid environments, however, samples from Woranso-Mille
fossil sites including Mesgid Dora (MSD), Am-Ado (AMA) and Aralee Issie (ARI) contain
hominin fossils which cannot be placed confidently into either Au. anamensis or Au. afarensis
206
(Curran and Haile-Selassie, 2016; Su and Haile-Selassie, 2022). Paleoenvironmental
reconstruction of these sites indicates that were dominantly dry
(median R
i/b
= 2.2) mixed C
3
environments (median δ
13
C
28acid
= -25.9‰). These samples represent paleosol and
fluviolacustrine facies (Fig 6.8A) and cluster together in δ
13
C
28acid
vs. R
i/b
space (Fig 6.8B),
separate from Au. anamensis and Au. afarensis exclusive sites. Further fossil evidence will be
needed to ascertain which of either Au. anamensis or Au. afarensis was inhabiting this
paleoenvironment at Woranso-Mille.
Fig 6.8 Cross plots showing δ
13
C
28acid
versus R
i/b.
A) Color represents different sample facies. B)
Color represents samples from sites associated with a hominin species.
Au. deyiremeda existed simultaneously in Woranso-Mille (3.4 Ma) as Au. afarensis (Haile-
Selassie et al., 2015; Melillo et al., 2021) in localities that are only 5 km apart. Thus,
reconstructing the paleoenvironment that Au. deyiremeda inhabited is interesting in
understanding how Woranso-Mille could support two different hominin species nearby. We find
that Au. deyiremeda existed in a predominantly C
4
environment
,
signified by relatively positive
δ
13
C
28acid
(median = -22.2‰) (Fig 6.8B) which is similar to Au. afarensis (median = -21.5‰).
207
Directly comparing the median δ
13
C
28acid
between the Burtele site (Au. deyiremeda) and the
contemporaneous Leado Dido’a (Au. afarensis) site (-22.2‰ vs -21.5‰) shows that both sites
were C
4
dominated. Au. deyiremeda samples have a smaller δ
13
C
28acid
IQR than Au. afarensis
(1.8 vs 2.5), although this could be a consequence of the relatively fewer samples from the single
Au. deyiremeda fossil site (Burtele) (n = 18) compared to Au. afarensis sites (n = 44). In contrast
to the plant waxes, pedogenic carbonate nodule δ
13
C is very negative (mean = -13.7‰, 1 = 0.3)
(Levin et al., 2015), which is indicative of heavily wooded vegetation (Su and Haile-Selassie,
2022). Bovid community analysis indicates that alcelaphins (Wildebeest) were dominant
amongst bovids, which indicates the nearby presence of open grassland (Su and Haile-Selassie,
2022). Taken together, the proxies indicate that Au. deyiremeda existed in an open grassland
environment, close to the water bodies and riparian woodlands. This environment was likely
similar to that inhabited by Au. afarensis in Leado Dido’a, which was predominantly C
4
open
grassland inhabited by alcelaphins (wildebeest) and antelopes (antelopes), in addition to aquatic
species including crocodiles and hippopotamids (Su and Haile-Selassie, 2022). Root casts
indicate the presence of trees, possibly indicating the presence of riparian woodland (Su and
Haile-Selassie, 2022).
The Burtele foot (Haile-Selassie et al., 2012) was also discovered at the BRT site and had
arboreal adaptations similar to the older hominin Ardipithecus ramidus (4.5-4.3 Ma), yet appears
in the fossil record alongside the younger Au. afarenis (3.9-2.9 Ma), whose foot morphology is
more suited for bipedal locomotion. The Burtele foot specimen was recovered from sediments
with a relatively positive δ
13
C
28acid
(-21.4‰) indicative of an open C
4
environment.
Paleoenvironmental reconstructions of the arboreally-adapted Ar. ramidus from Aramis,
Ethiopia, suggests it was living in a similar environment defined as tree/bush savannah with 20-
208
10 % woody cover (Cerling et al., 2010), indicating that it is not unusual for arboreally adapted
hominins to exist in an environment without significant tree cover. Baboons and other land
primates live in open savannah habitats and forage for food on the ground during the day before
returning to the safety of trees at night, and the Burtele foot hominin may have existed in a
similar ecological niche.
The single kerogen sample from the Mesgid Dora (MSD) locality had a δ
13
C
28acid
of -30.3‰,
indicative of C
3
vegetation. We interpret this sample as being deposited in a pond environment,
where anoxic conditions and rapid burial led to the excellent preservation of organic material,
unique in the Woranso-Mille stratigraphic record. This sample was associated with fossil wood,
that of a C
3
palm (Barboni, personal communication), consistent with the δ
13
C
28acid
interpretation.
Palm springs have previously been identified as important environments for early hominins in
eastern Africa during the mid-Pliocene, providing water, shade, and food sources (Barboni et al.,
2019). This palm spring microhabitat at Woranso-Mille is in association with Au. anamensis/Au.
afarensis like hominins (Haile-Selassie, 2010) and a faunal assemblage based
paleoenvironmental reconstructions of a heterogeneous environment with dense riparian
vegetation and more open grass/shrubland (Curran and Haile-Selassie, 2016).
209
6.5.5 Hominin dietary expansion
Hominin dietary change can be tracked using carbon isotopes of tooth enamel, with the carbon
isotopic content of the vegetation consumed by a hominin imprinting on the tooth enamel. A
profound transition occurred in hominin diets between 4 Ma and 3.65 Ma, with an increase in the
proportion of C
4
plants consumed reflected in an increase in the δ
13
C of tooth enamel (Lee-Thorp
et al., 2012; Cerling et al., 2013; Wynn et al., 2013; Levin et al., 2015) (Fig 6.9). Due to low-
resolution proxy records, it has been difficult to link environmental changes to this dietary shift
(Levin et al., 2015). However, the transition to a prolonged and stable C
4
-dominated
environment between 3.8-3.6 Ma at Woranso-Mille coincides with the C
4
hominin dietary
expansion (Fig 6.9) as well as the appearance of the specialist C
4
consuming primate
Theropithecus oswaldi in the fossil record (Frost et al., 2014). A 200 kyr period of relative C4
vegetation dominance would likely have forced hominins inhabiting Woranso-Mille to
incorporate C
4
plants into their diets due to a scarcity of C
3
plants. δ
13
C of enamel indicates that
this wider dietary breadth was then maintained even when the environment returned to C
3
and
mixed C
3
/C
4
from 3.6–3.2 Ma, likely due to the competitive advantage this dietary breadth
conveyed.
Interestingly, the transition of hominin diets to include C
4
vegetation coincided with the
appearance of Au. afarensis and the disappearance of Au. anamensis from the fossil record at
Woranso-Mille (Haile-Selassie, 2010; Haile-Selassie et al., 2016b, 2019). The dietary change
could represent a transition between hominin species mediated by an environmental change. Au.
anamensis (C
3
diet preference) was associated with a C
3
woodland environment around 3.8 Ma
(Saylor et al., 2019) and Au. afarensis was discovered living in C
4
environments (with a mixed
210
C
3
/C
4
diet), so it is tempting to ascribe the switch in hominin diet to an Au. afarensis expansion,
although more species-specific hominin tooth enamel is needed to confirm this.
Fig 6.9 Woranso-Mille δ
13
C wax (small black circles), at individual sites (for colors, see legend)
with the sample most proximal to hominin fossil (symbol). b) Hominin tooth enamel δ
13
C for a
variety of species (see legend) collected from sites across eastern Africa (Lee-Thorp et al., 2012;
Cerling et al., 2013; Wynn et al., 2013; Levin et al., 2015).
211
6.6 Conclusions
We demonstrate that plant wax n-alkanoic acids and GDGT molecules can enable
paleoenvironmental reconstruction across multiple facies types in the hominin fossil locality of
Woranso-Mille, Ethiopia. We find that pedogenic carbonate δ
13
C reconstructs proportionally
more C
3
vegetation compared to paired δ
13
C plant wax samples, possibly caused by differences
in proxy formation seasonality, plant rooting depth, and time integration. We compare the time
series of δ
13
C
wax
from Woranso-Mille shows to the Gulf of Aden offshore marine core wax
record DSDP 231 and find that although Woranso-Mille δ
13
C
wax
shows larger amplitude changes,
the two records share similar long term changes in the proportion of C
3
and C
4
plants on the
landscape. This suggests that marine core records are useful in determining regional
paleoenvironmental changes on >100 kyr time scales, yet they are dampened in their amplitude
of change, likely due to averaging over large areas. We find that a shift in the dietary preference
of hominins to include C
4
vegetation at 3.6 Ma co-occurred with an increase in the proportion of
C
4
plants at Woranso-Mille, which indicates that hominins increased their dietary breadth in
response to a change in a decrease in landscape C
3
vegetation. We also show Au. anamensis
existed in more C
3
dominated environments than the more cosmopolitan Au. afarensis, which
existed in a range of both C
3
and C
4
environments. The ratio of isoprenoid: branched GDGT is
low (<1) in sediments associated with both hominins, suggesting high soil moisture and,
therefore, relatively wet environments. Au. deyiremeda was contemporaneous with Au. afarensis
at Woranso-Mille, and we show that both hominins existed in dominantly open C
4
environments,
although close to riparian vegetation zones, suggesting that both hominins had similar
environmental preferences.
212
Acknowledgments
This research was supported by funding from the WM Keck Foundation. The University of
Michigan and the University of Southern California supported the fieldwork participation of NL
and MP, respectively. USC supported the analytical work on organics; UM supported the
analytical work that focused on soil carbonates. Field research at Woranso-Mille
paleoanthropological study area was conducted under a permit from the Authority for Research
and Conservation of Cultural Heritage (ARCCH) of the Ministry of Culture and Tourism of
Ethiopia, the Afar Regional State, and all of its district administrations. We thank the Authority
for Research and Conservation of Cultural Heritage (ARCCH) for permission to conduct field
and laboratory work; the Afar people of Woranso-Mille and the Mille District administration for
their hospitality; the project’s fieldwork crew members for their tireless support of field
activities.
References
Alene, M., Hart, W.K., Saylor, B.Z., Deino, A., Mertzman, S., Haile-Selassie, Y., Gibert, L.B.,
2017. Geochemistry of Woranso–Mille Pliocene basalts from west-central Afar, Ethiopia:
Implications for mantle source characteristics and rift evolution. Lithos 282–283, 187–200.
Amante, C., Eakins, B.W., 2009. NOAA Technical Memorandum NESDIS NGDC-24 ETOPO1
1 Arc-minute global relief model: procedures, data sources and analysis.
Barboni, D., 2014. Vegetation of Northern Tanzania during the Plio-Pleistocene: A synthesis of
the paleobotanical evidences from Laetoli, Olduvai, and Peninj hominin sites. Quaternary
International 322–323, 264–276.
Barboni, D., Ashley, G.M., Bourel, B., Arráiz, H., Mazur, J.C., 2019. Springs, palm groves, and
the record of early hominins in Africa. Review of Palaeobotany and Palynology 266, 23–41.
213
Bedaso, Z.K., DeLuca, N.M., Levin, N.E., Zaitchik, B.F., Waugh, D.W., Wu, S.Y., Harman,
C.J., Shanko, D., 2020. Spatial and temporal variation in the isotopic composition of
Ethiopian precipitation. Journal of Hydrology 585, 124364.
Bedaso, Z.K., Wynn, J.G., Alemseged, Z., Geraads, D., 2013. Dietary and paleoenvironmental
reconstruction using stable isotopes of herbivore tooth enamel from middle Pliocene Dikika,
Ethiopia: Implication for Australopithecus afarensis habitat and food resources. Journal of
Human Evolution 64, 21–38.
Behrensmeyer, A.K., Reed, K.E., 2013. Reconstructing the Habitats of Australopithecus:
Paleoenvironments, Site Taphonomy, and Faunas, in: Vertebrate Paleobiology and
Paleoanthropology. Springer, pp. 41–60.
Berke, M.A., Johnson, T.C., Werne, J.P., Grice, K., Schouten, S., Sinninghe Damsté, J.S., 2012.
Molecular records of climate variability and vegetation response since the Late Pleistocene
in the Lake Victoria basin, East Africa. Quaternary Science Reviews 55, 59–74.
Bobe, R., Manthi, F.K., Ward, C. V., Plavcan, J.M., Carvalho, S., 2020. The ecology of
Australopithecus anamensis in the early Pliocene of Kanapoi, Kenya. Journal of Human
Evolution 140, 102717.
Bonnefille, R., 2010. Cenozoic vegetation, climate changes and hominid evolution in tropical
Africa. Global and Planetary Change 72, 390–411.
Cane, M.A., Molnar, P., 2001. Closing of the Indonesian seaway as a precursor to east African
aridification around 3-4 million years ago. Nature 411, 157–162.
Cerling, T.E., 1984. The stable isotopic composition of modern soil carbonate and its
relationship to climate. Earth and Planetary Science Letters 71, 229–240.
Cerling, T.E., Levin, N.E., Quade, J., Wynn, J.G., Fox, D.L., Kingston, J.D., Klein, R.G., Brown,
214
F.H., 2010. Comment on the paleoenvironment of ardipithecus ramidus. Science.
Cerling, T.E., Manthi, F.K., Mbua, E.N., Leakey, L.N., Leakey, M.G., Leakey, R.E., Brown,
F.H., Grine, F.E., Hart, J.A., Kaleme, P., Roche, H., Uno, K.T., Wood, B.A., 2013. Stable
isotope-based diet reconstructions of Turkana Basin hominins. Proceedings of the National
Academy of Sciences of the United States of America 110, 10501–10506.
Cerling, T.E., Wynn, J.G., Andanje, S.A., Bird, M.I., Korir, D.K., Levin, N.E., MacE, W.,
MacHaria, A.N., Quade, J., Remien, C.H., 2011. Woody cover and hominin environments
in the past 6-million years. Nature 476, 51–56.
Colcord, D.E., Shilling, A.M., Sauer, P.E., Freeman, K.H., Njau, J.K., Stanistreet, I.G.,
Stollhofen, H., Schick, K.D., Toth, N., Brassell, S.C., 2018. Sub-Milankovitch
paleoclimatic and paleoenvironmental variability in East Africa recorded by Pleistocene
lacustrine sediments from Olduvai Gorge, Tanzania. Palaeogeography, Palaeoclimatology,
Palaeoecology 495, 284–291.
Cotton, J.M., Cerling, T.E., Hoppe, K.A., Mosier, T.M., Still, C.J., 2016. Climate, CO2, and the
history of North American grasses since the Last Glacial Maximum. Science Advances 2,
e1501346.
Curran, S.C., Haile-Selassie, Y., 2016. Paleoecological reconstruction of hominin-bearing
middle Pliocene localities at Woranso-Mille, Ethiopia. Journal of Human Evolution 96, 97–
112.
Dang, X., Yang, H., Naafs, B.D.A., Pancost, R.D., Xie, S., 2016. Evidence of moisture control
on the methylation of branched glycerol dialkyl glycerol tetraethers in semi-arid and arid
soils. Geochimica et Cosmochimica Acta 189, 24–36.
De Jonge, C., Hopmans, E.C., Zell, C.I., Kim, J.H., Schouten, S., Sinninghe Damsté, J.S., 2014.
215
Occurrence and abundance of 6-methyl branched glycerol dialkyl glycerol tetraethers in
soils: Implications for palaeoclimate reconstruction. Geochimica et Cosmochimica Acta
141, 97–112.
Dearing Crampton-Flood, E., Tierney, J.E., Peterse, F., Kirkels, F.M.S.A., Sinninghe Damsté,
J.S., 2020. BayMBT: A Bayesian calibration model for branched glycerol dialkyl glycerol
tetraethers in soils and peats. Geochimica et Cosmochimica Acta 268, 142–159.
Diefendorf, A.F., Freimuth, E.J., 2017. Extracting the most from terrestrial plant-derived n-alkyl
lipids and their carbon isotopes from the sedimentary record: A review. Organic
Geochemistry.
Ehleringer, J.R., Sage, R.F., Flanagan, L.B., Pearcy, R.W., 1991. Climate change and the
evolution of C4 photosynthesis. Trends in Ecology and Evolution.
Feakins, S.J., Eglinton, T.I., deMenocal, P.B., 2007. A comparison of biomarker records of
northeast African vegetation from lacustrine and marine sediments (ca. 3.40 Ma). Organic
Geochemistry 38, 1607–1624.
Frost, S.R., Jablonski, N.G., Haile-Selassie, Y., 2014. Early Pliocene Cercopithecidae from
Woranso-Mille (Central Afar, Ethiopia) and the origins of the Theropithecus oswaldi
lineage. Journal of Human Evolution 76, 39–53.
Haile-Selassie, Y., 2010. Phylogeny of early Australopithecus: new fossil evidence from the
Woranso-Mille (central Afar, Ethiopia). Philosophical Transactions of the Royal Society B:
Biological Sciences 365, 3323–3331.
Haile-Selassie, Y., Gibert, L., Melillo, S.M., Ryan, T.M., Alene, M., Deino, A., Levin, N.E.,
Scott, G., Saylor, B.Z., 2015. New species from Ethiopia further expands Middle Pliocene
hominin diversity. Nature 521, 483–488.
216
Haile-Selassie, Y., Melillo, S.M., Su, D.F., 2016a. The Pliocene hominin diversity conundrum:
Do more fossils mean less clarity? Proceedings of the National Academy of Sciences
201521266.
Haile-Selassie, Y., Melillo, S.M., Vazzana, A., Benazzi, S., Ryan, T.M., 2019. A 3.8-million-
year-old hominin cranium from Woranso-Mille, Ethiopia. Nature 573, 214–219.
Haile-Selassie, Y., Ryan, T.M., Levin, N.E., Saylor, B.Z., Deino, A., Mundil, R., Scott, G.,
Alene, M., Gibert, L., 2016b. Dentognathic remains of Australopithecus afarensis from
Nefuraytu (Woranso-Mille, Ethiopia): Comparative description, geology, and
paleoecological context. Journal of Human Evolution 100, 35–53.
Haile-Selassie, Y., Saylor, B.Z., Deino, A., Levin, N.E., Alene, M., Latimer, B.M., 2012. A new
hominin foot from Ethiopia shows multiple Pliocene bipedal adaptations. Nature 483, 565–
569.
Heslop-Harrison, J.S. (Pat), 2011. Atlas of the potential vegetation of Ethiopia. Annals of Botany
107, vi–vii.
Kimbel, W.H., Delezene, L.K., 2009. “Lucy” redux: A review of research on Australopithecus
afarensis. American Journal of Physical Anthropology 140, 2–48.
Kraus, M.J., 1999. Paleosols in clastic sedimentary rocks: their geologic applications. Earth-
Science Reviews 47, 41–70.
Laskar, J., Robutel, P., Joutel, F., Gastineau, M., Correia, A.C.M., Levrard, B., 2004. A long-
term numerical solution for the insolation quantities of the Earth. Astronomy &
Astrophysics 428, 261–285.
Lee-Thorp, J., Likius, A., Mackaye, H.T., Vignaud, P., Sponheimer, M., Brunet, M., 2012.
Isotopic evidence for an early shift to C4 resources by Pliocene hominins in Chad.
217
Proceedings of the National Academy of Sciences of the United States of America 109,
20369–20372.
Levin, N.E., 2013. Compilation of East Africa soil carbonate stable isotope data. Integrated Earth
Data Applications 10, 75–89.
Levin, N.E., 2015. Environment and Climate of Early Human Evolution. Annual Review of
Earth and Planetary Sciences 43, 405–429.
Levin, N.E., Haile-Selassie, Y., Frost, S.R., Saylor, B.Z., 2015. Dietary change among hominins
and cercopithecids in Ethiopia during the early Pliocene. Proceedings of the National
Academy of Sciences of the United States of America 112, 12304–9.
Levin, N.E., Zipser, E.J., Cerling, T.E., 2009. Isotopic composition of waters from Ethiopia and
Kenya: Insights into moisture sources for eastern Africa. Journal of Geophysical Research
114, D23306.
Liddy, H.M., Feakins, S.J., Tierney, J.E., 2016. Cooling and drying in northeast Africa across the
Pliocene. Earth and Planetary Science Letters 449, 430–438.
Lin, Xiaohua, Smerdon, Jason E, England, Anthony W, Pollack, Henry N, Lin, X, Smerdon, J E,
England, A W, Pollack, H N, 2003. A model study of the effects of climatic precipitation
changes on ground temperatures. Journal of Geophysical Research: Atmospheres 108, 4230.
Lupien, R.L., Russell, J.M., Feibel, C., Beck, C., Castañeda, I., Deino, A., Cohen, A.S., 2018. A
leaf wax biomarker record of early Pleistocene hydroclimate from West Turkana, Kenya.
Quaternary Science Reviews 186, 225–235.
Lupien, R.L., Russell, J.M., Grove, M., Beck, C.C., Feibel, C.S., Cohen, A.S., 2020. Abrupt
climate change and its influences on hominin evolution during the early Pleistocene in the
Turkana Basin, Kenya. Quaternary Science Reviews 245, 106531.
218
Magill, C.R., Ashley, G.M., Freeman, K.H., 2013. Ecosystem variability and early human
habitats in eastern Africa. Proceedings of the National Academy of Sciences of the United
States of America 110, 1167–1174.
Martínez-Sosa, P., Tierney, J.E., Stefanescu, I.C., Dearing Crampton-Flood, E., Shuman, B.N.,
Routson, C., 2021. A global Bayesian temperature calibration for lacustrine brGDGTs.
Geochimica et Cosmochimica Acta 305, 87–105.
Melillo, S.M., Gibert, L., Saylor, B.Z., Deino, A., Alene, M., Ryan, T.M., Haile-Selassie, Y.,
2021. New Pliocene hominin remains from the Leado Dido’a area of Woranso-Mille,
Ethiopia. Journal of Human Evolution 153, 102956.
Nicholson, S.E., 2000. The nature of rainfall variability over Africa on time scales of decades to
millenia, Global and Planetary Change 26. 137-158
Paquette, J., Drapeau, M.S.M., 2021. Environmental comparisons of the Awash Valley, Turkana
Basin and lower Omo Valley from upper Miocene to Holocene as assessed from stable
carbon and oxygen isotopes of mammalian enamel. Palaeogeography, Palaeoclimatology,
Palaeoecology 562, 110099.
Passey, B.H., Levin, N.E., Cerling, T.E., Brown, F.H., Eiler, J.M., 2010. High-temperature
environments of human evolution in East Africa based on bond ordering in paleosol
carbonates. Proceedings of the National Academy of Sciences of the United States of
America 107, 11245–9.
Peaple, M.D., Beverly, E.J., Garza, B., Baker, S., Levin, N.E., Tierney, J.E., Häggi, C. and
Feakins, S.J., 2022. Identifying the drivers of GDGT distributions in alkaline soil profiles
within the Serengeti ecosystem. Organic Geochemistry, 169, 104433.
Peterse, F., Van Der Meer, M.T.J., Schouten, S., Jia, G., Ossebaar, J., Blokker, J., Sinninghe
219
Damsté, J.S., 2009. Assessment of soil n-alkane δd and branched tetraether membrane lipid
distributions as tools for paleoelevation reconstruction. Biogeosciences 6, 2799–2807.
Prospero, J.M., 2002. Environmental characterization of global sources of atmospheric soil dust
identified with the NIMBUS 7 Total Ozone Mapping Spectrometer (TOMS) absorbing
aerosol product. Reviews of Geophysics 40, 1002.
Quinn, R.L., Lepre, C.J., 2020. Revisiting the pedogenic carbonate isotopes and
paleoenvironmental interpretation of Kanapoi. Journal of Human Evolution 140, 102549.
Reed, K.E., 2008. Paleoecological patterns at the Hadar hominin site, Afar Regional State,
Ethiopia. Journal of Human Evolution 54, 743–768.
Sachse, D., Billault, I., Bowen, G.J., Chikaraishi, Y., Dawson, T.E., Feakins, S.J., Freeman,
K.H., Magill, C.R., Mcinerney, F.A., Van Der Meer, M.T.J.J., Polissar, P., Robins, R.J.,
Sachs, J.P., Schmidt, H.-L., Sessions, A.L., White, J.W.C., West, J.B., Kahmen, A., 2012.
Molecular Paleohydrology: Interpreting the Hydrogen-Isotopic Composition of Lipid
Biomarkers from Photosynthesizing Organisms 40. doi:10.1146/annurev-earth-042711-
105535
Sarangi, V., Agrawal, S., Sanyal, P., 2021. The disparity in the abundance of C4 plants estimated
using the carbon isotopic composition of paleosol components. Palaeogeography,
Palaeoclimatology, Palaeoecology 561, 110068.
Saylor, B.Z., Angelini, J., Deino, A., Alene, M., Fournelle, J.H., Haile-Selassie, Y., 2016.
Tephrostratigraphy of the Waki-Mille area of the Woranso-Mille paleoanthropological
research project, Afar, Ethiopia. Journal of Human Evolution 93, 25–45.
Saylor, B.Z., Gibert, L., Deino, A., Alene, M., Levin, N.E., Melillo, S.M., Peaple, M.D., Feakins,
S.J., Bourel, B., Barboni, D., Novello, A., Sylvestre, F., Mertzman, S.A., Haile-Selassie, Y.,
220
2019. Age and context of mid-Pliocene hominin cranium from Woranso-Mille, Ethiopia.
Nature 573. doi:10.1038/s41586-019-1514-7
Schouten, S., Hopmans, E.C., Sinninghe Damsté, J.S., 2013. The organic geochemistry of
glycerol dialkyl glycerol tetraether lipids: A review. Organic Geochemistry 54, 19–61.
Shilling, A.M., Colcord, D.E., Karty, J., Hansen, A., Freeman, K.H., Njau, J.K., Stanistreet, I.G.,
Stollhofen, H., Schick, K.D., Toth, N., Brassell, S.C., 2019. Biogeochemical evidence for
environmental changes of Pleistocene Lake Olduvai during the transitional sequence of
OGCP core 2A that encompasses Tuff IB (~1.848 Ma). Palaeogeography,
Palaeoclimatology, Palaeoecology 532, 109267.
Stap, L.B., de Boer, B., Ziegler, M., Bintanja, R., Lourens, L.J., van de Wal, R.S.W., 2016. CO2
over the past 5 million years: Continuous simulation and new δ11B-based proxy data. Earth
and Planetary Science Letters 439, 1–10.
Su, D.F., Haile-Selassie, Y., 2022. Mosaic habitats at Woranso-Mille (Ethiopia) during the
Pliocene and implications for Australopithecus paleoecology and taxonomic diversity.
Journal of Human Evolution 163, 103076.
Su, D.F., Harrison, T., 2015. The paleoecology of the Upper Laetolil Beds, Laetoli Tanzania: A
review and synthesis. Journal of African Earth Sciences 101, 405–419.
Uno, K.T., Polissar, P.J., Jackson, K.E., deMenocal, P.B., 2016a. Neogene biomarker record of
vegetation change in eastern Africa. Proceedings of the National Academy of Sciences of
the United States of America 113, 6355–63.
Uno, K.T., Polissar, P.J., Kahle, E., Feibel, C., Harmand, S., Roche, H., Demenocal, P.B., 2016b.
A Pleistocene palaeovegetation record from plant wax biomarkers from the Nachukui
Formation, West Turkana, Kenya. Philosophical Transactions of the Royal Society B:
221
Biological Sciences 371, 20150235.
Wu, M.S., West, A.J., Feakins, S.J., 2019. Tropical soil profiles reveal the fate of plant wax
biomarkers during soil storage. Organic Geochemistry 128, 1–15.
Wynn, J.G., Sponheimer, M., Kimbel, W.H., Alemseged, Z., Reed, K., Bedaso, Z.K., Wilson,
J.N., 2013. Diet of Australopithecus afarensis from the Pliocene Hadar Formation, Ethiopia.
Proceedings of the National Academy of Sciences of the United States of America.
Xie, S., Pancost, R.D., Chen, L., Evershed, R.P., Yang, H., Zhang, K., Huang, J., Xu, Y., 2012.
Microbial lipid records of highly alkaline deposits and enhanced aridity associated with
significant uplift of the Tibetan Plateau in the Late Miocene. Geology 40, 291–294.
Zamanian, K., Pustovoytov, K., Kuzyakov, Y., 2016. Pedogenic carbonates: Forms and
formation processes. Earth-Science Reviews 157, 1–17.
Zhang, D., Beverly, E.J., Levin, N.E., Vidal, E., Matia, Y., Feakins, S.J., 2021. Carbon isotopic
composition of plant waxes, bulk organics and carbonates from soils of the Serengeti
grasslands. Geochimica et Cosmochimica Acta. doi:10.1016/J.GCA.2021.07.005
Zhang, L., Hay, W.W., Wang, C., Gu, X., 2019. The evolution of latitudinal temperature
gradients from the latest Cretaceous through the Present. Earth-Science Reviews 189, 147–
158.
222
Chapter 7: Dissertation conclusions
At the start of this dissertation, I discussed the critical benefit of using organic biomarker
proxies: their ability to be applied to almost any sediment deposit, including lacustrine, marine,
and outcrops. Throughout this dissertation, I have applied, refined, and developed multiple
biomarker proxies in different environments to reconstruct past environments and climates from
the Plio-Pleistocene. In Chapter 2, I developed a new machine learning-based method for
vegetation reconstructions based upon the chain length distributions of plant waxes. This study
involved calibrating the model using modern vegetation growing around Southern California,
applying the model to fossil waxes, and then validating the model through comparison with an
independent lake salinity proxy. This work highlights the advantages of applying new data
science techniques to extract diagnostic information from previously underutilized data types. In
Chapter 3, I analyzed distributions of GDGT molecules in dry and alkaline soils across an aridity
gradient. I found that while brGDGT compounds respond to temperature in the surface soil, they
do not respond to the temperature at depth in the soil. This has implications for studies seeking to
reconstruct past temperatures from paleosols formed from similar soil types. I additionally
compared the brGDGT proxy to clumped isotopes measured in pedogenic carbonates sampled in
the same soil profiles as the GDGTs. While both proxies broadly reconstructed similar
temperatures, there was little correlation between the proxies. One reason for this was the
relatively small temperature range, which was present in our Serengeti study area, but sample
specific biases undoubtedly contributed to the difference. This highlights the need for multi
proxy studies to constrain temperature changes to separate proxy signals from proxy noise.
223
My thesis then moved on to applying proxies in different environments and times. Firstly, in
Chapter 4, I applied biomarker and pollen proxies to reconstruct changes in the
paleoenvironment of southwestern North America over the past 200 kyr, again applying
biomarker proxies to Searles Lake sediments. I discovered that n-alkane δD, a tracer for δD of
precipitation, shows strong glacial to interglacial transitions, which are strikingly like the
transitions measured in δ
18
O
calcite
from the nearby speleothem, Devils Hole.
The controls on
Devils Hole δ
18
O
calcite
have not been well constrained, and therefore the strong positive
correlation between these two records suggests that they are indeed both responding to the
isotopic composition of precipitation. GDGT evidence suggests that for the past 200 kyr, Searles
Lake was stratified, typified by high BIT and %isoGDGT-0 values indicating a low oxygen
water column. However, between 140-130 kyr Searles Lake is drastically different, characterized
by low BIT and isoGDGT-0%, indicating a well-oxygenated and overturning lake.
Additionally, ACE and IR
6+7me
salinity proxies indicate that Searles was a freshwater lake during
this period. This biomarker evidence for an overturning oxygenated water column matches high
shoreline tufa evidence indicating Searles was overflowing around this time, as well as lake
deposits in the adjacent Panamint and the terminal basin, Death Valley. So, this ~300 m deep,
fresh overflowing lake period was identified in the Searles Lake core record for the first time.
In Chapter 5, I continued examining Searles Lake deposits but went back to the Pliocene period,
when pCO2 was >400 ppm, and terrestrial temperatures were 2-6°C elevated relative to modern
large lakes that existed in the Mojave Desert. To understand why P-E was increased relative to
modern despite higher temperature conditions, I measured GDGTs and plant waxes extracted
224
from a commercial drill core (KM-3) currently owned by the USGS. I found that the δD of
precipitation was enriched more than both modern δD of precipitation and Pleistocene δD of
precipitation calculated in the previous chapter. This suggests that Searles Lake had a greater
input of precipitation originating from tropical moisture sources than at present and during the
past 200 kyr. Modeling using an isotopologue-enabled earth system model discovered that
atmospheric river-type winter precipitation and summer North American Monsoon precipitation
contributed to the relatively positive δD of precipitation measured downcore. This increase in
summer and winter precipitation also to the high lake levels reconstructed using both the ACE
and IR
6+7me
salinity proxies.
Finally, for Chapter 6, I stayed in the same late Pliocene period and arid environment but
switched locations to Woranso-Mille, Ethiopia. This chapter aimed to conduct a
paleoenvironmental reconstruction from early hominin-bearing sediments to understand the
ecology of early hominins better. To develop a paleoenvironmental reconstruction, we used
similar proxies as in previous chapters (plant waxes and GDGTs), but here we applied them to
ancient outcrop sediments instead of modern soils and lake sediment cores. As such, we first had
to compare our proxies between multiple facies environments and with other older proxies (soil
carbonates) previously applied at Woranso-Mille. I found that plant waxes preserved in paleosols
with pedogenic carbonates typically had more positive δ
13
C than those preserved in non-
pedogenic carbonate soils and fluvial/lacustrine samples.
Additionally, the moisture sensitive R
i/b
index indicated that pedogenic carbonate bearing soils
typically had higher moisture contents when forming. This suggested that pedogenic carbonate
bearing soils typically record wetter and more C
4
environments similar to the Serengeti today.
225
This result highlighted the utility of our biomarker proxies and showed that it is important to
reconstruct paleoenvironmental information from a broad range of facies. I then compared the
plant wax δ
13
C with paired pedogenic carbonate samples and found no correlation between their
isotopic values. I also discovered that plant wax δ
13
C was relatively more positive than an
equivalent pedogenic carbonate sample. The reasons for this proxy discrepancy are unclear,
although it could relate to diagenesis (both of plant waxes and carbonates), differences in
seasonality of proxy formation, and/or plant rooting depth. We then applied our biomarker
proxies to reconstruct the paleoenvironment of the 3 different hominins which existed at
Woranso-Mille over the late Pliocene. We found that Australopithecus afarensis existed in a
broad range of C
3
and C
4
paleoenvironments, whereas Australopithecus anamensis existed in
more C
3
dominated environments. Australopithecus deyiremeda was contemporaneous with Au.
afarensis at Woranso-Mille, and we show that both hominins existed in dominantly open C
4
environments, although close to riparian vegetation zones, suggesting that both hominins had
similar environmental preferences. I also demonstrated that long term trends in the plant wax
δ
13
C measured from Woranso-Mille were similar to those measured from the offshore marine
core DSDP 231, including the Afar Triangle in its modern wax source region. This, in effect,
ground truths that offshore marine record and confirms that changes observed in the terrestrial
environment are reflected in offshore wax records.
In conclusion, in this thesis, I have developed and applied biomarker proxies to reconstruct past
environments and climate states during the Plio-Pleistocene. This work demonstrates the utility
of multi proxy biomarker approaches to constrain paleoenvironmental signals across different
periods and facies environments.
226
References
Adam, D.P., Sarna-Wojcicki, A.M., Rieck, H.J., Bradbury, J.P., Dean, W.E., Forester, R.M.,
1989. Tulelake, California: The last 3 million years. Palaeogeography, Palaeoclimatology,
Palaeoecology 72, 89–103.
Agrawal, S., Galy, V., Sanyal, P., Eglinton, T., 2014. C4 plant expansion in the Ganga Plain
during the last glacial cycle: Insights from isotopic composition of vascular plant
biomarkers. Organic Geochemistry 67, 58–71.
Aichner, B., Herzschuh, U., Wilkes, H., 2010. Influence of aquatic macrophytes on the stable
carbon isotopic signatures of sedimentary organic matter in lakes on the Tibetan Plateau.
Organic Geochemistry 41, 706–718.
Aichner, B., Hilt, S., Périllon, C., Gillefalk, M., Sachse, D., 2017. Biosynthetic hydrogen
isotopic fractionation factors during lipid synthesis in submerged aquatic macrophytes:
Effect of groundwater discharge and salinity. Organic Geochemistry 113, 10–16.
Alene, M., Hart, W.K., Saylor, B.Z., Deino, A., Mertzman, S., Haile-Selassie, Y., Gibert, L.B.,
2017. Geochemistry of Woranso–Mille Pliocene basalts from west-central Afar, Ethiopia:
Implications for mantle source characteristics and rift evolution. Lithos 282–283, 187–200.
Amante, C., Eakins, B.W., 2009. NOAA Technical Memorandum NESDIS NGDC-24 ETOPO1
1 Arc-minute global relief model: procedures, data sources and analysis.
Anderson, V.J., Shanahan, T.M., Saylor, J.E., Horton, B.K., Mora, A.R., 2014. Sources of local
and regional variability in the MBT′/CBT paleotemperature proxy: Insights from a modern
elevation transect across the Eastern Cordillera of Colombia. Organic Geochemistry 69, 42–
51.
Bacon, S.N., Jayko, A.S., Owen, L.A., Lindvall, S.C., Rhodes, E.J., Schumer, R.A., Decker,
227
D.L., 2020. A 50,000-year record of lake-level variations and overflow from Owens Lake,
eastern California, USA. Quaternary Science Reviews 238, 106312.
Bai, Y., Chen, C., Xu, Q., Fang, X., 2018. Paleoaltimetry Potentiality of Branched GDGTs From
Southern Tibet. Geochemistry, Geophysics, Geosystems 19, 551–564.
Bajnai, D., Coplen, T.B., Methner, K., Löffler, N., Krsnik, E., Fiebig, J., 2021. Devils Hole
Calcite Was Precipitated at ±1°C Stable Aquifer Temperatures During the Last Half Million
Years. Geophysical Research Letters 48. doi:10.1029/2021GL093257
Ballog, R.A., Malloy, R.E., 1981. Neogene palynology from the southern California continental
borderland, Site 467, Deep Sea Drilling Project, Leg 63. Initial reports of the Deep Sea
Drilling Project, Leg 63, Long Beach, California to Mazatlan, Mexico, (U.S.Govt. Printing
Office; U.K. distributors, IPOD Committee, NERC, Swindon) 565–576.
Barboni, D., 2014. Vegetation of Northern Tanzania during the Plio-Pleistocene: A synthesis of
the paleobotanical evidences from Laetoli, Olduvai, and Peninj hominin sites. Quaternary
International 322–323, 264–276.
Barboni, D., Ashley, G.M., Bourel, B., Arráiz, H., Mazur, J.C., 2019. Springs, palm groves, and
the record of early hominins in Africa. Review of Palaeobotany and Palynology 266, 23–41.
Bartusek, S.T., Seo, H., Ummenhofer, C.C., Steffen, J., 2021. The role of nearshore air‐sea
interactions for landfalling atmospheric rivers on the US West Coast. Geophysical Research
Letters 48, e2020GL091388.
Bassett, R.L., Steinwand, A., Jorat, S., Petersen, C., Jackson, R., 2008. Forensic isotope analysis
to refine a hydrologic conceptual model. Ground Water 46, 372–383.
Baxter, A.J., van Bree, L.G.J., Peterse, F., Hopmans, E.C., Villanueva, L., Verschuren, D.,
Sinninghe Damsté, J.S., 2021. Seasonal and multi-annual variation in the abundance of
228
isoprenoid GDGT membrane lipids and their producers in the water column of a meromictic
equatorial crater lake (Lake Chala, East Africa). Quaternary Science Reviews 273, 107263.
Bechtel, A., Smittenberg, R.H., Bernasconi, S.M., Schubert, C.J., 2010. Distribution of branched
and isoprenoid tetraether lipids in an oligotrophic and a eutrophic Swiss lake: Insights into
sources and GDGT-based proxies. doi:10.1016/j.orggeochem.2010.04.022
Bedaso, Z.K., DeLuca, N.M., Levin, N.E., Zaitchik, B.F., Waugh, D.W., Wu, S.Y., Harman,
C.J., Shanko, D., 2020. Spatial and temporal variation in the isotopic composition of
Ethiopian precipitation. Journal of Hydrology 585, 124364.
Bedaso, Z.K., Wynn, J.G., Alemseged, Z., Geraads, D., 2013. Dietary and paleoenvironmental
reconstruction using stable isotopes of herbivore tooth enamel from middle Pliocene Dikika,
Ethiopia: Implication for Australopithecus afarensis habitat and food resources. Journal of
Human Evolution 64, 21–38.
Behrensmeyer, A.K., Reed, K.E., 2013. Reconstructing the Habitats of Australopithecus:
Paleoenvironments, Site Taphonomy, and Faunas, in: Vertebrate Paleobiology and
Paleoanthropology. Springer, pp. 41–60.
Berke, M.A., Johnson, T.C., Werne, J.P., Grice, K., Schouten, S., Sinninghe Damsté, J.S., 2012.
Molecular records of climate variability and vegetation response since the Late Pleistocene
in the Lake Victoria basin, East Africa. Quaternary Science Reviews 55, 59–74.
Berkelhammer, M., Stott, L., Yoshimura, K., Johnson, K., Sinha, A., 2012. Synoptic and
mesoscale controls on the isotopic composition of precipitation in the western United
States. Climate Dynamics 38, 433–454.
Beverly, E.J., Levin, N.E., Passey, B.H., Aron, P.G., Yarian, D.A., Page, M., Pelletier, E.M.,
2021. Triple oxygen and clumped isotopes in modern soil carbonate along an aridity
229
gradient in the Serengeti, Tanzania. Earth and Planetary Science Letters 567, 116952.
Beverly, E.J., Lukens, W.E., Stinchcomb, G.E., 2018. Paleopedology as a Tool for
Reconstructing Paleoenvironments and Paleoecology. Vertebrate Paleobiology and
Paleoanthropology 151–183.
Bhattacharya, T., Feng, R., Tierney, J.., Knapp, S., Burls, N.., Fu, M., 2022. Expansion and
intensification of the North American Monsoon during the Pliocene.
doi:https://doi.org/10.31223/X54W8N
Bhattacharya, T., Tierney, J.E., Addison, J.A., Murray, J.W., 2018. Ice-sheet modulation of
deglacial North American monsoon intensification. Nature Geoscience 1.
Bi, X., Sheng, G., Liu, X., Li, C., Fu, J., 2005. Molecular and carbon and hydrogen isotopic
composition of n-alkanes in plant leaf waxes. Organic Geochemistry 36, 1405–1417.
Bintanja, R., Van De Wal, R.S.W., 2008. North American ice-sheet dynamics and the onset of
100,000-year glacial cycles. Nature 2008 454:7206 454, 869–872.
Bischoff, J.L., Cummins, K., 2001. Wisconsin Glaciation of the Sierra Nevada (79,000-15,000 yr
B.P.) as recorded by rock flour in sediments of Owens Lake, California. Quaternary
Research 55, 14–24.
Bischoff, J.L., Rosenbauer, R.J., Smith, G.I., 1985. Uranium-series dating of sediments from
Searles Lake: Differences between continental and marine climate records. Science 227,
1222–1224.
Bobe, R., Manthi, F.K., Ward, C. V., Plavcan, J.M., Carvalho, S., 2020. The ecology of
Australopithecus anamensis in the early Pliocene of Kanapoi, Kenya. Journal of Human
Evolution 140, 102717.
Bonifacie, M., Calmels, D., Eiler, J.M., Horita, J., Chaduteau, C., Vasconcelos, C., Agrinier, P.,
230
Katz, A., Passey, B.H., Ferry, J.M., Bourrand, J.J., 2017. Calibration of the dolomite
clumped isotope thermometer from 25 to 350 °C, and implications for a universal
calibration for all (Ca, Mg, Fe)CO3 carbonates. Geochimica et Cosmochimica Acta 200,
255–279.
Bonnefille, R., 2010. Cenozoic vegetation, climate changes and hominid evolution in tropical
Africa. Global and Planetary Change 72, 390–411.
Boser, B.E., Guyon, I.M., Vapnik, V.N., 1992. 5th Annual ACM Workshop on COLT.
Bova, S., Rosenthal, Y., Liu, Z., Godad, S.P., Yan, M., 2021. Seasonal origin of the thermal
maxima at the Holocene and the last interglacial. Nature 2021 589:7843 589, 548–553.
Breiman, L., 2001. Random forests. Machine Learning 45, 5–32.
Brierley, C.M., Fedorov, A. V., Liu, Z., Herbert, T.D., Lawrence, K.T., LaRiviere, J.P., 2009.
Greatly expanded tropical warm pool and weakened Hadley circulation in the early
Pliocene. Science (New York, N.Y.) 323, 1714–1718.
Brittingham, A., Hren, M.T., Hartman, G., 2017. Microbial alteration of the hydrogen and carbon
isotopic composition of n-alkanes in sediments. Organic Geochemistry 107, 1–8.
Burls, N.J., Fedorov, A. V, 2017. Wetter subtropics in a warmer world: Contrasting past and
future hydrological cycles. Proceedings of the National Academy of Sciences 114, 12888–
12893.
Bush, R.T., Mcinerney, F.A., 2015. Influence of temperature and C 4 abundance on n-alkane
chain length distributions across the central USA. doi:10.1016/j.orggeochem.2014.12.003
Bush, R.T., McInerney, F.A., 2013. Leaf wax n-alkane distributions in and across modern plants:
Implications for paleoecology and chemotaxonomy. Geochimica et Cosmochimica Acta
117, 161–179.
231
Byrne, R., 1982. Preliminary pollen analysis of Deep Sea Drilling Project Leg 64 Hole 480
(Cores 1-11).
California Department of Forestry and Fire Protection, 2015. Vegetation (fveg) - CAL FIRE
FRAP [ds1327] .
Campbell, I.D., McDonald, K., Flannigan, M.D., Kringayark, J., 1999. Long-distance transport
of pollen into the Arctic. Nature 398, 29–30.
Cane, M.A., Molnar, P., 2001. Closing of the Indonesian seaway as a precursor to east African
aridification around 3-4 million years ago. Nature 411, 157–162.
Cerling, T.E., 1984. The stable isotopic composition of modern soil carbonate and its
relationship to climate. Earth and Planetary Science Letters 71, 229–240.
Cerling, T.E., Levin, N.E., Quade, J., Wynn, J.G., Fox, D.L., Kingston, J.D., Klein, R.G., Brown,
F.H., 2010. Comment on the paleoenvironment of ardipithecus ramidus. Science.
Cerling, T.E., Manthi, F.K., Mbua, E.N., Leakey, L.N., Leakey, M.G., Leakey, R.E., Brown,
F.H., Grine, F.E., Hart, J.A., Kaleme, P., Roche, H., Uno, K.T., Wood, B.A., 2013. Stable
isotope-based diet reconstructions of Turkana Basin hominins. Proceedings of the National
Academy of Sciences of the United States of America 110, 10501–10506.
Cerling, T.E., Wynn, J.G., Andanje, S.A., Bird, M.I., Korir, D.K., Levin, N.E., MacE, W.,
MacHaria, A.N., Quade, J., Remien, C.H., 2011. Woody cover and hominin environments
in the past 6-million years. Nature 476, 51–56.
Chabot, B.F., Billings, W.D., 1972. Origins and Ecology of the Sierran Alpine Flora and
Vegetation. Ecological Monographs 42, 163–199.
Channell, J.E.T., Singer, B.S., Jicha, B.R., 2020. Timing of Quaternary geomagnetic reversals
and excursions in volcanic and sedimentary archives. Quaternary Science Reviews 228,
232
106114.
Chen, X., Liu, X., Wei, Y., Huang, Y., 2019. Production of long-chain n-alkyl lipids by
heterotrophic microbes: New evidence from Antarctic lakes. Organic Geochemistry 138,
103909.
Chou, C., Neelin, J.D., 2004. Mechanisms of global warming impacts on regional tropical
precipitation. Journal of climate 17, 2688–2701.
Clark, P.U., He, F., Golledge, N.R., Mitrovica, J.X., Dutton, A., Hoffman, J.S., Dendy, S., 2020.
Oceanic forcing of penultimate deglacial and last interglacial sea-level rise. Nature 2020
577:7792 577, 660–664.
Cleveland, D.M., Atchley, S.C., Nordt, L.C., 2007. Continental Sequence Stratigraphy of the
Upper Triassic (Norian–Rhaetian) Chinle Strata, Northern New Mexico, U.S.A.: Allocyclic
and Autocyclic Origins of Paleosol-Bearing Alluvial Successions. Journal of Sedimentary
Research 77, 909–924.
Coffinet, S., Huguet, A., Anquetil, C., Derenne, S., Pedentchouk, N., Bergonzini, L., Omuombo,
C., Williamson, D., Jones, M., Majule, A., Wagner, T., 2017. Evaluation of branched
GDGTs and leaf wax n-alkane δ2H as (paleo) environmental proxies in East Africa.
Geochimica et Cosmochimica Acta 198, 182–193.
Colcord, D.E., Shilling, A.M., Sauer, P.E., Freeman, K.H., Njau, J.K., Stanistreet, I.G.,
Stollhofen, H., Schick, K.D., Toth, N., Brassell, S.C., 2018. Sub-Milankovitch
paleoclimatic and paleoenvironmental variability in East Africa recorded by Pleistocene
lacustrine sediments from Olduvai Gorge, Tanzania. Palaeogeography, Palaeoclimatology,
Palaeoecology 495, 284–291.
Collister, J.W., Rieley, G., Stern, B., Eglinton, G., Fry, B., 1994. Compound-specific δ 13C
233
analyses of leaf lipids from plants with differing carbon dioxide metabolisms. Organic
geochemistry 21, 619–627.
Cotton, J.M., Cerling, T.E., Hoppe, K.A., Mosier, T.M., Still, C.J., 2016. Climate, CO2, and the
history of North American grasses since the Last Glacial Maximum. Science Advances 2,
e1501346.
Cracknell, M.J., Reading, A.M., 2014. Geological mapping using remote sensing data: A
comparison of five machine learning algorithms, their response to variations in the spatial
distribution of training data and the use of explicit spatial information. Computers and
Geosciences 63, 22–33.
Cross, M., McGee, D., Broecker, W.S., Quade, J., Shakun, J.D., Cheng, H., Lu, Y., Edwards,
R.L., 2015. Great Basin hydrology, paleoclimate, and connections with the North Atlantic:
A speleothem stable isotope and trace element record from Lehman Caves, NV. Quaternary
Science Reviews 127, 186–198.
Curran, S.C., Haile-Selassie, Y., 2016. Paleoecological reconstruction of hominin-bearing
middle Pliocene localities at Woranso-Mille, Ethiopia. Journal of Human Evolution 96, 97–
112.
Czop, M., Motyka, J., Sracek, O., Szuwarzyński, M., 2011. Geochemistry of the hyperalkaline
Gorka pit lake (pH>13) in the Chrzanow region, southern Poland. Water, Air, and Soil
Pollution 214, 423–434.
Dang, X., Yang, H., Naafs, B.D.A., Pancost, R.D., Xie, S., 2016. Evidence of moisture control
on the methylation of branched glycerol dialkyl glycerol tetraethers in semi-arid and arid
soils. Geochimica et Cosmochimica Acta 189, 24–36.
Danielson, J.J., Gesch, D.B., 2011. Global Multi-resolution Terrain Elevation Data 2010
234
(GMTED2010), U.S. Geological Survey Open-File Report 2011-1073.
Dansgaard, W., 1964. Stable isotopes in precipitation. Tellus 16, 436–468.
Davis, O.K., 1998. Palynological evidence for vegetation cycles in a 1.5 million year pollen
record from the Great Salt Lake, Utah, USA. Palaeogeography, Palaeoclimatology,
Palaeoecology 138, 175–185.
Davtian, N., Ménot, G., Bard, E., Poulenard, J., Podwojewski, P., 2016. Consideration of soil
types for the calibration of molecular proxies for soil pH and temperature using global soil
datasets and Vietnamese soil profiles. Organic Geochemistry 101, 140–153.
De Jonge, C., Hopmans, E.C., Zell, C.I., Kim, J.H., Schouten, S., Sinninghe Damsté, J.S., 2014a.
Occurrence and abundance of 6-methyl branched glycerol dialkyl glycerol tetraethers in
soils: Implications for palaeoclimate reconstruction. Geochimica et Cosmochimica Acta
141, 97–112.
De Jonge, C., Kuramae, E.E., Radujković, D., Weedon, J.T., Janssens, I.A., Peterse, F., 2021.
The influence of soil chemistry on branched tetraether lipids in mid- and high latitude soils:
Implications for brGDGT- based paleothermometry. Geochimica et Cosmochimica Acta
310, 95–112.
De Jonge, C., Radujković, D., Sigurdsson, B.D., Weedon, J.T., Janssens, I., Peterse, F., 2019.
Lipid biomarker temperature proxy responds to abrupt shift in the bacterial community
composition in geothermally heated soils. Organic Geochemistry 137, 103897.
De Jonge, C., Stadnitskaia, A., Hopmans, E.C., Cherkashov, G., Fedotov, A., Sinninghe Damsté,
J.S., 2014b. In situ produced branched glycerol dialkyl glycerol tetraethers in suspended
particulate matter from the Yenisei River, Eastern Siberia. Geochimica et Cosmochimica
Acta 125, 476–491.
235
Dearing Crampton-Flood, E., Peterse, F., Munsterman, D., Sinninghe Damsté, J.S., 2018. Using
tetraether lipids archived in North Sea Basin sediments to extract North Western European
Pliocene continental air temperatures. Earth and Planetary Science Letters 490, 193–205.
Dearing Crampton-Flood, E., Tierney, J.E., Peterse, F., Kirkels, F.M.S.A., Sinninghe Damsté,
J.S., 2020. BayMBT: A Bayesian calibration model for branched glycerol dialkyl glycerol
tetraethers in soils and peats. Geochimica et Cosmochimica Acta 268, 142–159.
Deino, A.L., Scott, G.R., Saylor, B., Alene, M., Angelini, J.D., Haile-Selassie, Y., 2010.
40Ar/39Ar dating, paleomagnetism, and tephrochemistry of Pliocene strata of the hominid-
bearing Woranso-Mille area, west-central Afar Rift, Ethiopia. Journal of Human Evolution
58, 111–126.
Dekens, P.S., Ravelo, A.C., McCarthy, M.D., 2007. Warm upwelling regions in the Pliocene
warm period. Paleoceanography 22. doi:10.1029/2006PA001394
Diefendorf, A.F., Freeman, K.H., Wing, S.L., Graham, H. V., 2011. Production of n-alkyl lipids
in living plants and implications for the geologic past. Geochimica et Cosmochimica Acta
75, 7472–7485.
Diefendorf, A.F., Freimuth, E.J., 2017. Extracting the most from terrestrial plant-derived n-alkyl
lipids and their carbon isotopes from the sedimentary record: A review. Organic
Geochemistry.
Diefendorf, A.F., Leslie, A.B., Wing, S.L., 2015a. Leaf wax composition and carbon isotopes
vary among major conifer groups. Geochimica et Cosmochimica Acta 170, 145–156.
Diefendorf, A.F., Mueller, K.E., Wing, S.L., Koch, P.L., Freeman, K.H., 2010. Global patterns in
leaf 13C discrimination and implications for studies of past and future climate. Proceedings
of the National Academy of Sciences of the United States of America 107, 5738–43.
236
Diefendorf, A.F., Sberna, D.T., Taylor, D.W., 2015b. Effect of thermal maturation on plant-
derived terpenoids and leaf wax n-alkyl components. Organic Geochemistry 89–90, 61–70.
Dion-Kirschner, H., McFarlin, J.M., Masterson, A.L., Axford, Y., Osburn, M.R., 2020. Modern
constraints on the sources and climate signals recorded by sedimentary plant waxes in west
Greenland. Geochimica et Cosmochimica Acta 286, 336–354.
Dirghangi, S.S., Pagani, M., Hren, M.T., Tipple, B.J., 2013. Distribution of glycerol dialkyl
glycerol tetraethers in soils from two environmental transects in the USA. Organic
Geochemistry 59, 49–60.
Ebisuzaki, W., 1997. A method to estimate the statistical significance of a correlation when the
data are serially correlated. Journal of Climate 10, 2147–2153.
Eglington, G., Hamilton, R.J., 1963. The distribution of Alkanes. Chemical Plant Taxonomy, T.
Swain, ed. Academic Press, London and New York. p.
Eglinton, G., Gonzalez, A.G., Hamilton, R.J., Raphael, R.A., 1962. Hydrocarbon constituents of
the wax coatings of plant leaves: A taxonomic survey. Phytochemistry 1, 89–102.
Eglinton, G., Hamilton, R.J., 1967. Leaf epicuticular waxes. Science 156, 1322–1335.
Ehleringer, J.R., Phillips, S.L., Comstock, J.P., 1992. Seasonal Variation in the Carbon Isotopic
Composition of Desert Plants. Functional Ecology 6, 396.
Ehleringer, J.R., Sage, R.F., Flanagan, L.B., Pearcy, R.W., 1991. Climate change and the
evolution of C4 photosynthesis. Trends in Ecology and Evolution.
Engle, M.A., Brunner, B., 2019. Considerations in the application of machine learning to
aqueous geochemistry: Origin of produced waters in the northern U.S. Gulf Coast Basin.
Applied Computing and Geosciences 3–4, 100012.
Faith, D.P., Minchin, P.R., Belbin, L., 1987. Compositional dissimilarity as a robust measure of
237
ecological distance. Vegetatio 1987 69:1 69, 57–68.
Fasham, M.J.R., 1977. A Comparison of Nonmetric Multidimensional Scaling, Principal
Components and Reciprocal Averaging for the Ordination of Simulated Coenoclines, and
Coenoplanes. Ecology 58, 551–561.
Feakins, S.J., 2013. Pollen-corrected leaf wax D/H reconstructions of northeast African
hydrological changes during the late Miocene. Palaeogeography, Palaeoclimatology,
Palaeoecology 374, 62–71.
Feakins, S.J., deMenocal, P.B., Eglinton, T.I., 2005. Biomarker records of late Neogene changes
in northeast African vegetation. Geology 33, 977–980.
Feakins, S.J., Eglinton, T.I., deMenocal, P.B., 2007. A comparison of biomarker records of
northeast African vegetation from lacustrine and marine sediments (ca. 3.40 Ma). Organic
Geochemistry 38, 1607–1624.
Feakins, S.J., Kirby, M.E., Cheetham, M.I., Ibarra, Y., Zimmerman, S.R.H., 2014. Fluctuation in
leaf wax D/H ratio from a southern California lake records significant variability in isotopes
in precipitation during the late Holocene. Organic Geochemistry 66, 48–59.
Feakins, S.J., Sessions, A.L., 2010. Controls on the D/H ratios of plant leaf waxes in an arid
ecosystem. Geochimica et Cosmochimica Acta 74, 2128–2141.
Feakins, S.J., Wu, M.S., Ponton, C., Galy, V., West, A.J., 2018. Dual isotope evidence for
sedimentary integration of plant wax biomarkers across an Andes-Amazon elevation
transect. Geochimica et Cosmochimica Acta 242, 64–81.
Feakins, S.J., Wu, M.S., Ponton, C., Tierney, J.E., 2019. Biomarkers reveal abrupt switches in
hydroclimate during the last glacial in southern California. Earth and Planetary Science
Letters 515, 164–172.
238
Feng, X., D’Andrea, W.J., Zhao, C., Xin, S., Zhang, C., Liu, W., 2019. Evaluation of leaf wax
δD and soil brGDGTs as tools for paleoaltimetry on the southeastern Tibetan Plateau.
Chemical Geology 523, 95–106.
Fick, S.E., Hijmans, R.J., 2017. WorldClim 2: new 1-km spatial resolution climate surfaces for
global land areas. International Journal of Climatology 37, 4302–4315.
Ficken, K.J., Li, B., Swain, D.L., Eglinton, G., 2000. An n-alkane proxy for the sedimentary
input of submerged/floating freshwater aquatic macrophytes, in: Organic Geochemistry.
Pergamon, pp. 745–749.
Finnegan, S., Bergmann, K., Eiler, J.M., Jones, D.S., Fike, D.A., Eisenman, I., Hughes, N.C.,
Tripati, A.K., Fischer, W.W., 2011. The Magnitude and Duration of Late Ordovician–Early
Silurian Glaciation. Science 331, 903–906.
Fleming, L.E., Tierney, J.E., 2016. An automated method for the determination of the TEX86
and U37K′ paleotemperature indices. Organic Geochemistry 92, 84–91.
Fontijn, K., Lachowycz, S.M., Rawson, H., Pyle, D.M., Mather, T.A., Naranjo, J.A., Moreno-
Roa, H., 2014. Late Quaternary tephrostratigraphy of southern Chile and Argentina.
Quaternary Science Reviews 89, 70–84.
Ford, H.L., Ravelo, A.C., Dekens, P.S., LaRiviere, J.P., Wara, M.W., 2015. The evolution of the
equatorial thermocline and the early Pliocene El Padre mean state. Geophysical Research
Letters 42, 4878–4887.
Forester, R.M., Lowenstein, T.K., Spencer, R.J., 2005. An ostracode based paleolimnologic and
paleohydrologic history of Death Valley: 200 to 0 ka. GSA Bulletin 117, 1379–1386.
Fornace, K.L., Whitney, B.S., Galy, V., Hughen, K.A., Mayle, F.E., 2016. Late Quaternary
environmental change in the interior South American tropics: new insight from leaf wax
239
stable isotopes. Earth and Planetary Science Letters 438, 75–85.
Fowler, A.M., Hennessy, K.J., 1995. Potential impacts of global warming on the frequency and
magnitude of heavy precipitation. Natural Hazards 11, 283–303.
Freimuth, E.J., Diefendorf, A.F., Lowell, T. V., 2017. Hydrogen isotopes of n-alkanes and n-
alkanoic acids as tracers of precipitation in a temperate forest and implications for
paleorecords. Geochimica et Cosmochimica Acta 206, 166–183.
Freimuth, E.J., Diefendorf, A.F., Lowell, T. V., Wiles, G.C., 2019. Sedimentary n-alkanes and n-
alkanoic acids in a temperate bog are biased toward woody plants. Organic Geochemistry
128, 94–107.
Friedman, I., Harris, J.M., Smith, G.I., Johnson, C.A., 2002. Stable isotope composition of
waters in the Great Basin, United States 1. Air-mass trajectories. Journal of Geophysical
Research: Atmospheres 107, ACL 14-1.
Friedman, I., Smith, G.I., Gleason, J.D., Warden, A., Harris, J.M., 1992. Stable isotope
composition of waters in southeastern California 1. Modern precipitation. Journal of
Geophysical Research 97, 5795.
Frost, S.R., Jablonski, N.G., Haile-Selassie, Y., 2014. Early Pliocene Cercopithecidae from
Woranso-Mille (Central Afar, Ethiopia) and the origins of the Theropithecus oswaldi
lineage. Journal of Human Evolution 76, 39–53.
Fu, M., Cane, M.A., Molnar, P., Tziperman, E., 2021. Wetter Subtropics Lead to Reduced
Pliocene Coastal Upwelling. Paleoceanography and Paleoclimatology 36, e2021PA004243.
Fu, M., Cane, M.A., Molnar, P., Tziperman, E., 2022. Warmer Pliocene Upwelling Site SST
Leads to Wetter Subtropical Coastal Areas: A Positive Feedback on SST. Paleoceanography
and Paleoclimatology 37. doi:10.1029/2021PA004357
240
Gallagher, T.M., Sheldon, N.D., 2016. Combining soil water balance and clumped isotopes to
understand the nature and timing of pedogenic carbonate formation. Chemical Geology 435,
79–91.
Galy, V., Eglinton, T., France-Lanord, C., Sylva, S., 2011. The provenance of vegetation and
environmental signatures encoded in vascular plant biomarkers carried by the Ganges-
Brahmaputra rivers. Earth and Planetary Science Letters 304, 1–12.
Gao, L., Hou, J., Toney, J., MacDonald, D., Huang, Y., 2011. Mathematical modeling of the
aquatic macrophyte inputs of mid-chain n-alkyl lipids to lake sediments: Implications for
interpreting compound specific hydrogen isotopic records. Geochimica et Cosmochimica
Acta 75, 3781–3791.
Ghosh, P., Adkins, J., Affek, H., Balta, B., Guo, W., Schauble, E.A., Schrag, D., Eiler, J.M.,
2006. 13C-18O bonds in carbonate minerals: A new kind of paleothermometer. Geochimica
et Cosmochimica Acta 70, 1439–1456.
Gies, H., Hagedorn, F., Lupker, M., Montluçon, D., Haghipour, N., Sophia Van Der Voort, T.,
Ian Eglinton, T., 2021. Millennial-age glycerol dialkyl glycerol tetraethers (GDGTs) in
forested mineral soils: 14C-based evidence for stabilization of microbial necromass.
Biogeosciences 18, 189–205.
Giorgi, F., Raffaele, F., Coppola, E., 2019. The response of precipitation characteristics to global
warming from climate projections. Earth System Dynamics 10, 73–89.
Glen, J.M.G., Liddicoat, J.C., Coe, R.S., 1999. A detailed record of paleomagnetic field change
from Searles Lake, California 1. Long-term secular variation bounding the
Gauss/Matuyama polarity reversal. Journal of Geophysical Research: Solid Earth 104,
12865–12882.
241
Glover, K.C., MacDonald, G.M., Kirby, M.E., Rhodes, E.J., Stevens, L., Silveira, E., Whitaker,
A., Lydon, S., 2017. Evidence for orbital and North Atlantic climate forcing in alpine
Southern California between 125 and 10 ka from multi-proxy analyses of Baldwin Lake.
Quaternary Science Reviews 167, 47–62.
Günther, F., Thiele, A., Gleixner, G., Xu, B., Yao, T., Schouten, S., 2014. Distribution of
bacterial and archaeal ether lipids in soils and surface sediments of Tibetan lakes:
Implications for GDGT-based proxies in saline high mountain lakes. Organic Geochemistry
67, 19–30.
Guo, J., Ma, T., Liu, N., Zhang, X., Hu, H., Ma, W., Wang, Z., Feng, X., Peterse, F., 2022. Soil
pH and aridity influence distributions of branched tetraether lipids in grassland soils along
an aridity transect. Organic Geochemistry 164, 104347.
Haile-Selassie, Y., 2010. Phylogeny of early Australopithecus: new fossil evidence from the
Woranso-Mille (central Afar, Ethiopia). Philosophical Transactions of the Royal Society B:
Biological Sciences 365, 3323–3331.
Haile-Selassie, Y., Gibert, L., Melillo, S.M., Ryan, T.M., Alene, M., Deino, A., Levin, N.E.,
Scott, G., Saylor, B.Z., 2015. New species from Ethiopia further expands Middle Pliocene
hominin diversity. Nature 521, 483–488.
Haile-Selassie, Y., Latimer, B.M., Alene, M., Deino, A.L., Gibert, L., Melillo, S.M., Saylor,
B.Z., Scott, G.R., Lovejoy, C.O., 2010. An early Australopithecus afarensis postcranium
from Woranso-Mille, Ethiopia. Proceedings of the National Academy of Sciences 107,
12121–12126.
Haile-Selassie, Y., Melillo, S.M., Su, D.F., 2016a. The Pliocene hominin diversity conundrum:
Do more fossils mean less clarity? Proceedings of the National Academy of Sciences
242
201521266.
Haile-Selassie, Y., Melillo, S.M., Vazzana, A., Benazzi, S., Ryan, T.M., 2019. A 3.8-million-
year-old hominin cranium from Woranso-Mille, Ethiopia. Nature 573, 214–219.
Haile-Selassie, Y., Ryan, T.M., Levin, N.E., Saylor, B.Z., Deino, A., Mundil, R., Scott, G.,
Alene, M., Gibert, L., 2016b. Dentognathic remains of Australopithecus afarensis from
Nefuraytu (Woranso-Mille, Ethiopia): Comparative description, geology, and
paleoecological context. Journal of Human Evolution 100, 35–53.
Haile-Selassie, Y., Saylor, B.Z., Deino, A., Levin, N.E., Alene, M., Latimer, B.M., 2012. A new
hominin foot from Ethiopia shows multiple Pliocene bipedal adaptations. Nature 483, 565–
569.
Halffman, R., Lembrechts, J., Radujković, D., De Gruyter, J., Nijs, I., De Jonge, C., 2022. Soil
chemistry, temperature and bacterial community composition drive brGDGT distributions
along a subarctic elevation gradient. Organic Geochemistry 163, 104346.
Hammer, U.T., Heseltine, J.M., 1988. Aquatic macrophytes in saline lakes of the Canadian
prairies. Hydrobiologia 158, 101–116.
Hammond, W.C., Blewitt, G., Li, Z., Plag, H.P., Kreemer, C., 2012. Contemporary uplift of the
Sierra Nevada, western United States, from GPS and inSAR measurements. Geology 40,
667–670.
Hargreaves, G.H., Samani, Z.A., 1985. Reference Crop Evapotranspiration from Temperature.
Applied Engineering in Agriculture 1, 96–99.
Harris, J.R., Grunsky, E.C., 2015. Predictive lithological mapping of Canada’s North using
Random Forest classification applied to geophysical and geochemical data. Computers and
Geosciences 80, 9–25.
243
Hay, R.L., Guldman, S.G., Matthews, J.C., Lander, R.H., Duffin, M.E., Kyser, T.K., 1991. Clay
mineral diagenesis in core KM-3 of Searles Lake, California. Clays and Clay Minerals 39,
84–96.
Haywood, A.M., Hill, D.J., Dolan, A.M., Otto-Bliesner, B.L., Bragg, F., Chan, W.-L., Chandler,
M.A., Contoux, C., Dowsett, H.J., Jost, A., Kamae, Y., Lohmann, G., Lunt, D.J., Abe-
Ouchi, A., Pickering, S.J., Ramstein, G., Rosenbloom, N.A., Salzmann, U., Sohl, L.,
Stepanek, C., Ueda, H., Yan, Q., Zhang, Z., 2013. Large-scale features of Pliocene climate:
results from the Pliocene Model Intercomparison Project. Climate of the Past 9, 191–209.
He, Y., Wang, H., Meng, B., Liu, H., Zhou, A., Song, M., Kolpakova, M., Krivonogov, S., Liu,
W., Liu, Z., 2020. Appraisal of alkenone- and archaeal ether-based salinity indicators in
mid-latitude Asian lakes. Earth and Planetary Science Letters 538, 116236.
Hendy, C.H., 1971. The isotopic geochemistry of speleothems—I. The calculation of the effects
of different modes of formation on the isotopic composition of speleothems and their
applicability as palaeoclimatic indicators. Geochimica et Cosmochimica Acta 35, 801–824.
Herbert, T.D., Schuffert, J.D., Andreasen, D., Heusser, L., Lyle, M., Mix, A., Ravelo, A.C.,
Stott, L.D., Herguera, J.C., 2001. Collapse of the California current during glacial maxima
linked to climate change on land. Science 293, 71–76.
Herbert, T.D., Yasuda, M., Burnett, C., 1995. Glacial-Interglacial Sea-Surface Temperature
Record Inferred from Alkenone Unsaturation indices, Site 893, Santa Barbara Basin, in:
Proceedings of the Ocean Drilling Program, 146 Part 2 Scientific Results.
doi:10.2973/odp.proc.sr.146-2.301.1995
Heslop-Harrison, J.S. (Pat), 2011. Atlas of the potential vegetation of Ethiopia. Annals of Botany
107, vi–vii.
244
Heusser, L.E., 2000. Rapid oscillations in western North America vegetation and climate during
oxygen isotope stage 5 inferred from pollen data from Santa Barbara Basin (Hole 893A),
Palaeogeography, Palaeoclimatology, Palaeoecology.
Heusser, L.E., Kirby, M.E., Nichols, J.E., 2015. Pollen-based evidence of extreme drought
during the last Glacial (32.6-9.0 ka) in coastal southern California. Quaternary Science
Reviews 126, 242–253.
Holmgren, C.A., Betancourt, J.L., Rylander, K.A., 2010. A long-term vegetation history of the
Mojave-Colorado desert ecotone at Joshua Tree National Park. Journal of Quaternary
Science 25, 222–236.
Hopmans, E.C., Schouten, S., Sinninghe Damsté, J.S., 2016. The effect of improved
chromatography on GDGT-based palaeoproxies. Organic Geochemistry 93, 1–6.
Hren, M.T., Pagani, M., Erwin, D.M., Brandon, M., 2010. Biomarker reconstruction of the early
Eocene paleotopography and paleoclimate of the northern Sierra Nevada. Geology 38, 7–
10.
Huang, X., Swain, D.L., Hall, A.D., 2020. Future precipitation increase from very high
resolution ensemble downscaling of extreme atmospheric river storms in California.
Science Advances 6, eaba1323.
Huguet, A., Fosse, C., Laggoun-Défarge, F., Toussaint, M.L., Derenne, S., 2010a. Occurrence
and distribution of glycerol dialkyl glycerol tetraethers in a French peat bog. Organic
Geochemistry 41, 559–572.
Huguet, A., Fosse, C., Metzger, P., Fritsch, E., Derenne, S., 2010b. Occurrence and distribution
of extractable glycerol dialkyl glycerol tetraethers in podzols. Organic Geochemistry 41,
291–301.
245
Huguet, A., Meador, T.B., Laggoun-Défarge, F., Könneke, M., Wu, W., Derenne, S., Hinrichs,
K.U., 2017. Production rates of bacterial tetraether lipids and fatty acids in peatland under
varying oxygen concentrations. Geochimica et Cosmochimica Acta 203, 103–116.
Huth, T.E., Passey, B.H., Cole, J.E., Lachniet, M.S., McGee, D., Denniston, R.F., Truebe, S.,
Levin, N.E., 2022. A framework for triple oxygen isotopes in speleothem paleoclimatology.
Geochimica et Cosmochimica Acta 319, 191–219.
Ibarra, D.E., Egger, A.E., Weaver, K.L., Harris, C.R., Maher, K., 2014. Rise and fall of late
Pleistocene pluvial lakes in response to reduced evaporation and precipitation: Evidence
from Lake Surprise, California. Geological Society of America Bulletin 126, 1387–1415.
Ibarra, D.E., Oster, J.L., Winnick, M.J., Caves Rugenstein, J.K., Byrne, M.P., Chamberlain, C.P.,
2018. Warm and cold wet states in the western United States during the Pliocene–
Pleistocene. Geology 46, 355–358.
Inglis, G.N., Carmichael, M.J., Farnsworth, A., Lunt, D.J., Pancost, R.D., 2020. A long-term,
high-latitude record of Eocene hydrological change in the Greenland region.
Palaeogeography, Palaeoclimatology, Palaeoecology 537, 109378.
Jacobel, A.W., McManus, J.F., Anderson, R.F., Winckler, G., 2016. Large deglacial shifts of the
Pacific intertropical convergence zone. Nature communications 7, 1–7.
Jacobel, A.W., McManus, J.F., Anderson, R.F., Winckler, G., 2017. Climate-related response of
dust flux to the central equatorial Pacific over the past 150 kyr. Earth and Planetary Science
Letters 457, 160–172.
Jager, T., 1982. Soils of the Serengeti woodlands, Tanzania. Pudoc Wageningen.
Jannik, N.O., Phillips, F.M., Smith, G.I., Elmore, D., 1991. A 36Cl chronology of lacustrine
sedimentation in the Pleistocene Owens River system. Geological Society of America
246
Bulletin 103, 1146–1159.
Jansen, B., de Boer, E.J., Cleef, A.M., Hooghiemstra, H., Moscol-Olivera, M., Tonneijck, F.H.,
Verstraten, J.M., 2013. Reconstruction of late Holocene forest dynamics in northern
Ecuador from biomarkers and pollen in soil cores. Palaeogeography, Palaeoclimatology,
Palaeoecology 386, 607–619.
Jansen, B., van Loon, E.E., Hooghiemstra, H., Verstraten, J.M., 2010. Improved reconstruction
of palaeo-environments through unravelling of preserved vegetation biomarker patterns.
Palaeogeography, Palaeoclimatology, Palaeoecology 285, 119–130.
Jasechko, S., Lechler, A., Pausata, F.S.R., Fawcett, P.J., Gleeson, T., Cendon, D.I., Galewsky, J.,
LeGrande, A.N., Risi, C., Sharp, Z.D., Welker, J.M., Werner, M., Yoshimura, K., 2015.
Late-glacial to late-Holocene shifts in global precipitation δ18O. Climate of the Past 11,
1375–1393.
Jayko, A.S., Forester, R.M., Kaufman, D.S., Phillips, F.M., Yount, J.C., McGeehin, J., Mahan,
S.A., 2008. Late Pleistocene lakes and wetlands, Panamint Valley, Inyo County, California,
in: Special Paper of the Geological Society of America. pp. 151–184.
Jeng, W.L., 2006. Higher plant n-alkane average chain length as an indicator of petrogenic
hydrocarbon contamination in marine sediments. Marine Chemistry 102, 242–251.
Jouzel, J., Barkov, N.I., Barnola, J.M., Bender, M., Chappellaz, J., Genthon, C., Kotlyakov,
V.M., Lipenkov, V., Lorius, C., Petit, J.R., Raynaud, D., Raisbeck, G., Ritz, C., Sowers, T.,
Stievenard, M., Yiou, F., Yiou, P., 1993. Extending the Vostok ice-core record of
palaeoclimate to the penultimate glacial period. Nature 1993 364:6436 364, 407–412.
Kelson, J.R., Huntington, K.W., Breecker, D.O., Burgener, L.K., Gallagher, T.M., Hoke, G.D.,
Petersen, S. V., 2020. A proxy for all seasons? A synthesis of clumped isotope data from
247
Holocene soil carbonates. Quaternary Science Reviews 234, 106259.
Kelson, J.R., Huntington, K.W., Schauer, A.J., Saenger, C., Lechler, A.R., 2017. Toward a
universal carbonate clumped isotope calibration: Diverse synthesis and preparatory methods
suggest a single temperature relationship. Geochimica et Cosmochimica Acta 197, 104–
131.
Kemp, D.B., Robinson, S.A., Crame, J.A., Francis, J.E., Ineson, J., Whittle, R.J., Bowman, V.,
O’Brien, C., 2014. A cool temperate climate on the Antarctic Peninsula through the latest
Cretaceous to early Paleogene. Geology 42, 583–586.
Kim, J.H., Schouten, S., Hopmans, E.C., Donner, B., Sinninghe Damsté, J.S., 2008. Global
sediment core-top calibration of the TEX86 paleothermometer in the ocean. Geochimica et
Cosmochimica Acta 72, 1154–1173.
Kim, S.T., O’Neil, J.R., 1997. Equilibrium and nonequilibrium oxygen isotope effects in
synthetic carbonates. Geochimica et Cosmochimica Acta 61, 3461–3475.
Kimbel, W.H., Delezene, L.K., 2009. “Lucy” redux: A review of research on Australopithecus
afarensis. American Journal of Physical Anthropology 140, 2–48.
Knott*, J.R., Wan, E., Deino, A.L., Casteel, M., Reheis, M.C., Phillips, F.M., Walkup, L.,
McCarty†, K., Manoukian§, D.N., Jr.#, E.N., 2021. Lake Andrei: A Pliocene pluvial lake in
Eureka Valley, eastern California. From Saline to Freshwater: The Diversity of Western
Lakes in Space and Time 125–142.
Knott, J.R., Machette, M.N., Klinger, R.E., Sarna-Wojcicki, A.M., Liddicoat, J.C., Tinsley, J.C.,
David, B.T., Ebbs, V.M., 2008. Reconstructing late Pliocene to middle Pleistocene Death
Valley lakes and river systems as a test of pupfish (Cyprinodontidae) dispersal hypotheses,
in: Special Paper of the Geological Society of America. Geological Society of America, pp.
248
1–26.
Knott, J.R., Machette, M.N., Wan, E., Klinger, R.E., Liddicoat, J.C., Sarna-Wojcicki, A.M.,
Fleck, R.J., Deino, A.L., Geissman, J.W., Slate, J.L., Wahl, D.B., Wernicke, B.P., Wells,
S.G., Tinsley, J.C., Hathaway, J.C., Weamer, V.M., 2018. Late Neogene–Quaternary
tephrochronology, stratigraphy, and paleoclimate of Death Valley, California, USA. GSA
Bulletin 130, 1231–1255.
Koch, K., Bhushan, B., Barthlott, W., 2009. Multifunctional surface structures of plants: An
inspiration for biomimetics. Progress in Materials Science 54, 137–178.
Koehler, P.A., Anderson, R.S., 1994. Full-glacial shoreline vegetation during the maximum
highstand at Owens Lake, California. Great Basin Naturalist 54, 142–149.
Koehler, P.A., Anderson, R.S., Spaulding, W.G., 2005. Development of vegetation in the Central
Mojave Desert of California during the late Quaternary. Palaeogeography,
Palaeoclimatology, Palaeoecology 215, 297–311.
Körner, C., Farquhar, G.D., Wong, S.C., 1991. Carbon isotope discrimination by plants follows
latitudinal and altitudinal trends. Oecologia 88, 30–40.
Kraus, M.J., 1999. Paleosols in clastic sedimentary rocks: their geologic applications. Earth-
Science Reviews 47, 41–70.
Krull, E., Sachse, D., Mügler, I., Thiele, A., Gleixner, G., 2006. Compound-specific δ13C and
δ2H analyses of plant and soil organic matter: A preliminary assessment of the effects of
vegetation change on ecosystem hydrology. Soil Biology and Biochemistry 38, 3211–3221.
Kulongoski, J.T., Hilton, D.R., Izbicki, J.A., Belitz, K., 2009. Evidence for prolonged El Nino-
like conditions in the Pacific during the Late Pleistocene: a 43 ka noble gas record from
California groundwaters. Quaternary Science Reviews 28, 2465–2473.
249
Lachniet, M., Asmerom, Y., Polyak, V., Denniston, R., 2017. Arctic cryosphere and
Milankovitch forcing of Great Basin paleoclimate. Scientific Reports 2017 7:1 7, 1–10.
Lachniet, M.S., 2016. A Speleothem Record of Great Basin Paleoclimate: The Leviathan
Chronology, Nevada, in: Developments in Earth Surface Processes. Elsevier B.V., pp. 551–
569.
Lachniet, M.S., Denniston, R.F., Asmerom, Y., Polyak, V.J., 2014. Orbital control of western
North America atmospheric circulation and climate over two glacial cycles. Nature
Communications 5, 3805.
LaRiviere, J.P., Ravelo, A.C., Crimmins, A., Dekens, P.S., Ford, H.L., Lyle, M., Wara, M.W.,
2012. Late Miocene decoupling of oceanic warmth and atmospheric carbon dioxide forcing.
Nature 486, 97–100.
Laskar, J., Robutel, P., Joutel, F., Gastineau, M., Correia, A.C.M., Levrard, B., 2004. A long-
term numerical solution for the insolation quantities of the Earth. Astronomy &
Astrophysics 428, 261–285.
Lee-Thorp, J., Likius, A., Mackaye, H.T., Vignaud, P., Sponheimer, M., Brunet, M., 2012.
Isotopic evidence for an early shift to C4 resources by Pliocene hominins in Chad.
Proceedings of the National Academy of Sciences of the United States of America 109,
20369–20372.
Levin, N.E., 2013. Compilation of East Africa soil carbonate stable isotope data. Integrated Earth
Data Applications 10, 75–89.
Levin, N.E., 2015. Environment and Climate of Early Human Evolution. Annual Review of
Earth and Planetary Sciences 43, 405–429.
Levin, N.E., Haile-Selassie, Y., Frost, S.R., Saylor, B.Z., 2015. Dietary change among hominins
250
and cercopithecids in Ethiopia during the early Pliocene. Proceedings of the National
Academy of Sciences of the United States of America 112, 12304–9.
Levin, N.E., Zipser, E.J., Cerling, T.E., 2009. Isotopic composition of waters from Ethiopia and
Kenya: Insights into moisture sources for eastern Africa. Journal of Geophysical Research
114, D23306.
Li, C., Sessions, A.L., Kinnaman, F.S., Valentine, D.L., 2009. Hydrogen-isotopic variability in
lipids from Santa Barbara Basin sediments. Geochimica et Cosmochimica Acta 73, 4803–
4823.
Li, G., Li, L., Tarozo, R., Longo, W.M., Wang, K.J., Dong, H., Huang, Y., 2018. Microbial
production of long-chain n-alkanes: Implication for interpreting sedimentary leaf wax
signals. Organic Geochemistry 115, 24–31.
Li, J., Pancost, R.D., Naafs, B.D.A., Yang, H., Zhao, C., Xie, S., 2016. Distribution of glycerol
dialkyl glycerol tetraether (GDGT) lipids in a hypersaline lake system. Organic
Geochemistry 99, 113–124.
Li, Y., Zhao, S., Pei, H., Qian, S., Zang, J., Dang, X., Yang, H., 2018. Distribution of glycerol
dialkyl glycerol tetraethers in surface soils along an altitudinal transect at cold and humid
Mountain Changbai: Implications for the reconstruction of paleoaltimetry and paleoclimate.
Science China Earth Sciences 61, 925–939.
Liddicoat, J.C., Opdyke, N.D., Smith, G.I., 1980. Palaeomagnetic polarity in a 930-m core from
Searles Valley, California. Nature 286, 22–25.
Liddy, H.M., Feakins, S.J., Tierney, J.E., 2016. Cooling and drying in northeast Africa across the
Pliocene. Earth and Planetary Science Letters 449, 430–438.
Lin, J.C., Broecker, W.S., Hemming, S.R., Hajdas, I., Anderson, R.F., Smith, G.I., Kelley, M.,
251
Bonani, G., 1998. A Reassessment of U-Th and 14C Ages for Late-Glacial High-Frequency
Hydrological Events at Searles Lake, California. Quaternary Research 49, 11–23.
Lin, Xiaohua, Smerdon, Jason E, England, Anthony W, Pollack, Henry N, Lin, X, Smerdon, J E,
England, A W, Pollack, H N, 2003. A model study of the effects of climatic precipitation
changes on ground temperatures. Journal of Geophysical Research: Atmospheres 108, 4230.
Lisiecki, L.E., Raymo, M.E., 2005. A Pliocene-Pleistocene stack of 57 globally distributed
benthic δ 18O records. Paleoceanography 20, 1–17.
Litwin, R.J., Smoot, J.P., Durika, N.J., Smith, G.I., 1999. Calibrating Late Quaternary terrestrial
climate signals: Radiometrically dated pollen evidence from the southern Sierra Nevada,
USA. Quaternary Science Reviews. doi:10.1016/S0277-3791(98)00111-5
Liu, W., Yang, H., Li, L., 2006. Hydrogen isotopic compositions of n-alkanes from terrestrial
plants correlate with their ecological life forms. Oecologia 150, 330–338.
Lizama, C., Monteoliva-Sánchez, M., Prado, B., Ramos-Cormenzana, A., Weckesser, J.,
Campos, V., 2001. Taxonomic study of extreme halophilic archaea isolated from the “Salar
de Atacama”, Chile. Systematic and Applied Microbiology 24, 464–474.
Loomis, S.E., Russell, J.M., Heureux, A.M., D’Andrea, W.J., Sinninghe Damsté, J.S., 2014.
Seasonal variability of branched glycerol dialkyl glycerol tetraethers (brGDGTs) in a
temperate lake system. Geochimica et Cosmochimica Acta 144, 173–187.
Lu, H., Liu, W., Yang, H., Wang, H., Liu, Z., Leng, Q., Sun, Y., Zhou, W., An, Z., 2019. 800-
kyr land temperature variations modulated by vegetation changes on Chinese Loess Plateau.
Nature Communications 2019 10:1 10, 1–10.
Lupien, R.L., Russell, J.M., Feibel, C., Beck, C., Castañeda, I., Deino, A., Cohen, A.S., 2018. A
leaf wax biomarker record of early Pleistocene hydroclimate from West Turkana, Kenya.
252
Quaternary Science Reviews 186, 225–235.
Lupien, R.L., Russell, J.M., Grove, M., Beck, C.C., Feibel, C.S., Cohen, A.S., 2020. Abrupt
climate change and its influences on hominin evolution during the early Pleistocene in the
Turkana Basin, Kenya. Quaternary Science Reviews 245, 106531.
Lupien, R.L., Russell, J.M., Yost, C.L., Kingston, J.D., Deino, A.L., Logan, J., Schuh, A.,
Cohen, A.S., 2019. Vegetation change in the Baringo Basin, East Africa across the onset of
Northern Hemisphere glaciation 3.3–2.6 Ma. Palaeogeography, Palaeoclimatology,
Palaeoecology 109426.
Lüthi, D., Le Floch, M., Bereiter, B., Blunier, T., Barnola, J.-M., Siegenthaler, U., Raynaud, D.,
Jouzel, J., Fischer, H., Kawamura, K., Stocker, T.F., 2008. High-resolution carbon dioxide
concentration record 650,000–800,000 years before present. Nature 453, 379–382.
Lyle, M., Heusser, L., Ravelo, C., Andreasen, D., Olivarez Lyle, A., Diffenbaugh, N., 2010.
Pleistocene water cycle and eastern boundary current processes along the California
continental margin. Paleoceanography 25. doi:10.1029/2009PA001836
Magill, C.R., Ashley, G.M., Freeman, K.H., 2013. Ecosystem variability and early human
habitats in eastern Africa. Proceedings of the National Academy of Sciences of the United
States of America 110, 1167–1174.
Makou, M., Eglinton, T., McIntyre, C., Montluçon, D., Antheaume, I., Grossi, V., 2018. Plant
Wax n ‐Alkane and n ‐Alkanoic Acid Signatures Overprinted by Microbial Contributions
and Old Carbon in Meromictic Lake Sediments. Geophysical Research Letters 45, 1049–
1057.
Martínez-Botí, M.A., Foster, G.L., Chalk, T.B., Rohling, E.J., Sexton, P.F., Lunt, D.J., Pancost,
R.D., Badger, M.P.S., Schmidt, D.N., 2015. Plio-Pleistocene climate sensitivity evaluated
253
using high-resolution CO2 records. Nature 518, 49–54.
Martínez-Sosa, P., Tierney, J.E., 2019. Lacustrine brGDGT response to microcosm and
mesocosm incubations. Organic Geochemistry 127, 12–22.
Martínez-Sosa, P., Tierney, J.E., Stefanescu, I.C., Dearing Crampton-Flood, E., Shuman, B.N.,
Routson, C., 2021. A global Bayesian temperature calibration for lacustrine brGDGTs.
Geochimica et Cosmochimica Acta 305, 87–105.
McCallister, S.L., Del Giorgio, P.A., 2008. Direct measurement of the δ13C signature of carbon
respired by bacteria in lakes: Linkages to potential carbon sources, ecosystem baseline
metabolism, and CO2 fluxes. Limnology and Oceanography 53, 1204–1216.
McFarlin, J.M., Axford, Y., Masterson, A.L., Osburn, M.R., 2019. Calibration of modern
sedimentary δ2H plant wax-water relationships in Greenland lakes. Quaternary Science
Reviews 225, 105978.
McGee, D., 2020. Glacial-Interglacial Precipitation Changes. Annual Review of Marine Science.
McGee, D., Moreno-Chamarro, E., Marshall, J., Galbraith, E.D., 2018. Western U.S. lake
expansions during Heinrich stadials linked to Pacific Hadley circulation. Science Advances
4, eaav0118.
McPhillips, D., Brandon, M.T., 2010. Using tracer thermochronology to measure modern relief
change in the Sierra Nevada, California. Earth and Planetary Science Letters 296, 373–383.
Melillo, S.M., Gibert, L., Saylor, B.Z., Deino, A., Alene, M., Ryan, T.M., Haile-Selassie, Y.,
2021. New Pliocene hominin remains from the Leado Dido’a area of Woranso-Mille,
Ethiopia. Journal of Human Evolution 153, 102956.
Menges, J., Huguet, C., Alcañiz, J.M., Fietz, S., Sachse, D., Rosell-Melé, A., 2014. Influence of
water availability in the distributions of branched glycerol dialkyl glycerol tetraether in soils
254
of the Iberian Peninsula. Biogeosciences 11, 2571–2581.
Miller, M.D., Tziperman, E., 2017. The effect of changes in surface winds and ocean
stratification on coastal upwelling and sea surface temperatures in the Pliocene.
Paleoceanography 32, 371–383.
Mix, H.T., Caves Rugenstein, J.K., Reilly, S.P., Ritch, A.J., Winnick, M.J., Kukla, T.,
Chamberlain, C.P., 2019. Atmospheric flow deflection in the late Cenozoic Sierra Nevada.
Earth and Planetary Science Letters 518, 76–85.
Mix, H.T., Ibarra, D.E., Mulch, A., Graham, S.A., Page Chamberlain, C., 2016. A hot and high
Eocene Sierra Nevada. Bulletin of the Geological Society of America 128, 531–542.
Molnar, P., Cane, M.A., 2002. El Niño’s tropical climate and teleconnections as a blueprint for
pre-Ice Age climates. Paleoceanography 17, 11–1.
Molnar, P., Cane, M.A., 2007. Early Pliocene (pre-Ice age) El Niño-like global climate: Which
El Niño? Geosphere 3, 337–365.
Moore, J.G., Moring, B.C., 2013. Rangewide glaciation in the Sierra Nevada, California.
Geosphere 9, 1804–1818.
Moseley, G.E., Edwards, R.L., Wendt, K.A., Cheng, H., Dublyansky, Y., Lu, Y., Boch, R.,
Spötl, C., 2016. Reconciliation of the Devils Hole climate record with orbital forcing.
Science (New York, N.Y.) 351, 165–8.
Mulch, A., Sarna-Wojcicki, A.M., Perkins, M.E., Chamberlain, C.P., 2008. A Miocene to
Pleistocene climate and elevation record of the Sierra Nevada (California). Proceedings of
the National Academy of Sciences of the United States of America 105, 6819–6824.
Naafs, B.D.A., Gallego-Sala, A. V., Inglis, G.N., Pancost, R.D., 2017. Refining the global
branched glycerol dialkyl glycerol tetraether (brGDGT) soil temperature calibration.
255
Organic Geochemistry 106, 48–56.
Naafs, B.D.A., Oliveira, A.S.F., Mulholland, A.J., 2021. Molecular dynamics simulations
support the hypothesis that the brGDGT paleothermometer is based on homeoviscous
adaptation. Geochimica et Cosmochimica Acta 312, 44–56.
Nicholson, S.E., 2000. The nature of rainfall variability over Africa on time scales of decades to
millenia. Global and Planetary Change 26, 137–158.
Nobel, P.S., Bobich, E.G., 2002. Initial Net CO2 Uptake Responses and Root Growth for a CAM
Community Placed in a Closed Environment. Annals of Botany 90, 593–598.
Norton‐Griffiths, M., Herlocker, D., Pennycuick, L., 1975. The patterns of rainfall in the
Serengeti Ecosystem, Tanzania. African Journal of Ecology 13, 347–374.
Numelin, T., Kirby, E., Walker, J.D., Didericksen, B., 2007. Late Pleistocene slip on a low-angle
normal fault, Searles Valley, California. Geosphere 3, 163.
Olson, K.J., Lowenstein, T.K., 2021. Searles Lake evaporite sequences: Indicators of late
Pleistocene/Holocene lake temperatures, brine evolution, and pCO2. GSA Bulletin.
doi:10.1130/B35857.1
Oster, J.L., Ibarra, D.E., Winnick, M.J., Maher, K., 2015. Steering of westerly storms over
western North America at the Last Glacial Maximum. Nature Geoscience 8, 201–205.
Pancost, R.D., Hopmans, E.C., Sinninghe Damsté, J.S., 2001. Archaeal lipids in Mediterranean
cold seeps: molecular proxies for anaerobic methane oxidation. Geochimica et
Cosmochimica Acta 65, 1611–1627.
Paquette, J., Drapeau, M.S.M., 2021. Environmental comparisons of the Awash Valley, Turkana
Basin and lower Omo Valley from upper Miocene to Holocene as assessed from stable
carbon and oxygen isotopes of mammalian enamel. Palaeogeography, Palaeoclimatology,
256
Palaeoecology 562, 110099.
Passey, B.H., Levin, N.E., Cerling, T.E., Brown, F.H., Eiler, J.M., 2010. High-temperature
environments of human evolution in East Africa based on bond ordering in paleosol
carbonates. Proceedings of the National Academy of Sciences of the United States of
America 107, 11245–9.
Patten, D.T., Rouse, L., Stromberg, J.C., 2008. Isolated spring wetlands in the Great Basin and
Mojave deserts, USA: Potential response of vegetation to groundwater withdrawal.
Environmental Management 41, 398–413.
Peaple, M.D., Beverly, E.J., Garza, B., Baker, S., Levin, N.E., Tierney, J.E., Häggi, C., Feakins,
S.J., 2021a. Abundances and indices for soil microbial biomarkers (brGDGTs and
isoGDGTs) in eleven soil profiles across a Serengeti transect. Pangaea.
Peaple, M.D., Bhattacharya, T., Lowenstein, T.K., McGee, D., Olson, K.J., Stroup, J.S., Tierney,
J.E., Feakins, S.J., n.d. Biomarker and pollen evidence for late Pleistocene pluvials in the
Mojave Desert.
Peaple, M.D., Tierney, J.E., McGee, D., Lowenstein, T.K., Bhattacharya, T., Feakins, S.J.,
2021b. Identifying plant wax inputs in lake sediments using machine learning. Organic
Geochemistry 156, 104222.
Pedentchouk, N., Freeman, K.H., Harris, N.B., 2006. Different response of δD values of n-
alkanes, isoprenoids, and kerogen during thermal maturation. Geochimica et Cosmochimica
Acta 70, 2063–2072.
Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M.,
Prettenhofer, P., Weiss, R., Dubourg, V., Vanderplas, J., Passos, A., Cournapeau, D.,
Brucher, M., Perrot, M., Duchesnay, E., 2011. Scikit-learn: Machine Learning in Python,
257
Journal of Machine Learning Research.
Pei, H., Zhao, S., Yang, H., Xie, S., 2021. Variation of branched tetraethers with soil depth in
relation to non-temperature factors: Implications for paleoclimate reconstruction. Chemical
Geology 572, 120211.
PENNINGTON, W., 1979. THE ORIGIN OF POLLEN IN LAKE SEDIMENTS: AN
ENCLOSED LAKE COMPARED WITH ONE RECEIVING INFLOW STREAMS. New
Phytologist 83, 189–213.
Pérez-Angel, L.C., Sepúlveda, J., Molnar, P., Montes, C., Rajagopalan, B., Snell, K., Gonzalez-
Arango, C., Dildar, N., 2020. Soil and Air Temperature Calibrations Using Branched
GDGTs for the Tropical Andes of Colombia: Toward a Pan-Tropical Calibration.
Geochemistry, Geophysics, Geosystems 21, e2020GC008941.
Perol, T., Gharbi, M., Denolle, M., 2018. Convolutional neural network for earthquake detection
and location. Science Advances 4, e1700578.
Pester, M., Schleper, C., Wagner, M., 2011. The Thaumarchaeota: an emerging view of their
phylogeny and ecophysiology. Current Opinion in Microbiology 14, 300–306.
Peterse, Francien, Kim, J.-H., Schouten, S., Kristensen, D.K., Koç, N., Sinninghe Damsté, J.S.,
2009. Constraints on the application of the MBT/CBT palaeothermometer at high latitude
environments (Svalbard, Norway). Organic Geochemistry 40, 692–699.
Peterse, F., Martínez-García, A., Zhou, B., Beets, C.J., Prins, M.A., Zheng, H., Eglinton, T.I.,
2014. Molecular records of continental air temperature and monsoon precipitation
variability in East Asia spanning the past 130,000 years. Quaternary Science Reviews 83,
76–82.
Peterse, F., van der Meer, J., Schouten, S., Weijers, J.W.H., Fierer, N., Jackson, R.B., Kim, J.H.,
258
Sinninghe Damsté, J.S., 2012. Revised calibration of the MBT–CBT paleotemperature
proxy based on branched tetraether membrane lipids in surface soils. Geochimica et
Cosmochimica Acta 96, 215–229.
Peterse, F., Van Der Meer, M.T.J., Schouten, S., Jia, G., Ossebaar, J., Blokker, J., Sinninghe
Damsté, J.S., 2009. Assessment of soil n-alkane δd and branched tetraether membrane lipid
distributions as tools for paleoelevation reconstruction. Biogeosciences 6, 2799–2807.
Petersen, S. V., Defliese, W.F., Saenger, C., Daëron, M., Huntington, K.W., John, C.M., Kelson,
J.R., Bernasconi, S.M., Colman, A.S., Kluge, T., Olack, G.A., Schauer, A.J., Bajnai, D.,
Bonifacie, M., Breitenbach, S.F.M., Fiebig, J., Fernandez, A.B., Henkes, G.A., Hodell, D.,
Katz, A., Kele, S., Lohmann, K.C., Passey, B.H., Peral, M.Y., Petrizzo, D.A., Rosenheim,
B.E., Tripati, A., Venturelli, R., Young, E.D., Winkelstern, I.Z., 2019. Effects of Improved
17O Correction on Interlaboratory Agreement in Clumped Isotope Calibrations, Estimates
of Mineral-Specific Offsets, and Temperature Dependence of Acid Digestion Fractionation.
Geochemistry, Geophysics, Geosystems 20, 3495–3519.
Pierce, D.W., Cayan, D.R., Das, T., Maurer, E.P., Miller, N.L., Bao, Y., Kanamitsu, M.,
Yoshimura, K., Snyder, M.A., Sloan, L.C., Franco, G., Tyree, M., 2013. The Key Role of
Heavy Precipitation Events in Climate Model Disagreements of Future Annual Precipitation
Changes in California. Journal of Climate 26, 5879–5896.
Ponton, C., West, A.J., Feakins, S.J., Galy, V., 2014. Leaf wax biomarkers in transit record river
catchment composition. Geophysical Research Letters 41, 6420–6427.
Pound, M.J., Tindall, J., Pickering, S.J., Haywood, A.M., Dowsett, H.J., Salzmann, U., 2014a.
Late Pliocene lakes and soils: A global data set for the analysis of climate feedbacks in a
warmer world. Climate of the Past 10, 167–180.
259
Pound, M.J., Tindall, J., Pickering, S.J., Haywood, A.M., Dowsett, H.J., Salzmann, U., 2014b.
Late Pliocene lakes and soils: a global data set for the analysis of climate feedbacks in a
warmer world. Climate of the Past 10, 167–180.
Prospero, J.M., 2002. Environmental characterization of global sources of atmospheric soil dust
identified with the NIMBUS 7 Total Ozone Mapping Spectrometer (TOMS) absorbing
aerosol product. Reviews of Geophysics 40, 1002.
Pulli, J.J., Dysart, P.S., 1990. An experiment in the use of trained neural networks for regional
seismic event classification. Geophysical Research Letters 17, 977–980.
Quinn, R.L., Lepre, C.J., 2020. Revisiting the pedogenic carbonate isotopes and
paleoenvironmental interpretation of Kanapoi. Journal of Human Evolution 140, 102549.
Rasmussen, C.E., Williams, C.K.I., 2006. Gaussian processes for machine learning. MIT Press.
Reed, D.N., Anderson, T.M., Dempewolf, J., Metzger, K., Serneels, S., 2009. The spatial
distribution of vegetation types in the Serengeti ecosystem: the influence of rainfall and
topographic relief on vegetation patch characteristics. Journal of Biogeography 36, 770–
782.
Reed, K.E., 2008. Paleoecological patterns at the Hadar hominin site, Afar Regional State,
Ethiopia. Journal of Human Evolution 54, 743–768.
Reheis, M.C., Adams, K.D., Oviatt, C.G., Bacon, S.N., 2014. Pluvial lakes in the Great Basin of
the western United States—a view from the outcrop. Quaternary Science Reviews 97, 33–
57.
Reheis, M.C., Stine, S., Sarna-Wojcicki, A.M., 2002. Drainage reversals in Mono Basin during
the late Pliocene and Pleistocene. Geological Society of America Bulletin 114, 991–1006.
Remeika, P., Fischbein, I.W., Fischbein, S.A., 1988. Lower Pliocene petrified wood from the
260
Palm Spring Formation, Anza Borrego Desert State Park, California. Review of
Palaeobotany and Palynology 56, 183–198.
Reynolds, J., Reynolds, S., 1975. Aquatic angiosperms of some British Columbia saline lakes.
Syesis 8, 291–295.
Rimmer, A., Gal, G., Opher, T., Lechinsky, Y., Yacobi, Y.Z., 2011. Mechanisms of long-term
variations in the thermal structure of a warm lake. Limnology and Oceanography 56, 974–
988.
Roberts, S.M., Spencer, R.J., 1998. A desert responds to Pleistocene climate change: Saline
lacustrine sediments, Death Valley, California, USA, in: Quaternary Deserts and Climatic
Change. CRC Press, pp. 357–370.
Rommerskirchen, F., Plader, A., Eglinton, G., Chikaraishi, Y., Rullkötter, J., 2006.
Chemotaxonomic significance of distribution and stable carbon isotopic composition of
long-chain alkanes and alkan-1-ols in C4 grass waxes. Organic Geochemistry 37, 1303–
1332.
Rumelhart, D.E., Hinton, G.E., Williams, R.J., 1986. Learning representations by back-
propagating errors. Nature 323, 533–536.
Russell, J.M., Hopmans, E.C., Loomis, S.E., Liang, J., Sinninghe Damsté, J.S., 2018.
Distributions of 5- and 6-methyl branched glycerol dialkyl glycerol tetraethers (brGDGTs)
in East African lake sediment: Effects of temperature, pH, and new lacustrine
paleotemperature calibrations. Organic Geochemistry 117, 56–69.
Sachse, D., Billault, I., Bowen, G.J., Chikaraishi, Y., Dawson, T.E., Feakins, S.J., Freeman,
K.H., Magill, C.R., Mcinerney, F.A., Van Der Meer, M.T.J.J., Polissar, P., Robins, R.J.,
Sachs, J.P., Schmidt, H.-L., Sessions, A.L., White, J.W.C., West, J.B., Kahmen, A., 2012.
261
Molecular Paleohydrology: Interpreting the Hydrogen-Isotopic Composition of Lipid
Biomarkers from Photosynthesizing Organisms. Annual Review of Earth and Planetary
Sciences 40. doi:10.1146/annurev-earth-042711-105535
Sandquist, D.R., Ehleringer, J.R., 1995. Carbon isotope discrimination in the C₄ shrub Atriplex
confertifolia along a salinity gradient. The Great Basin Naturalist 135–141.
Sarangi, V., Agrawal, S., Sanyal, P., 2021. The disparity in the abundance of C4 plants estimated
using the carbon isotopic composition of paleosol components. Palaeogeography,
Palaeoclimatology, Palaeoecology 561, 110068.
Saylor, B.Z., Angelini, J., Deino, A., Alene, M., Fournelle, J.H., Haile-Selassie, Y., 2016.
Tephrostratigraphy of the Waki-Mille area of the Woranso-Mille paleoanthropological
research project, Afar, Ethiopia. Journal of Human Evolution 93, 25–45.
Saylor, B.Z., Gibert, L., Deino, A., Alene, M., Levin, N.E., Melillo, S.M., Peaple, M.D., Feakins,
S.J., Bourel, B., Barboni, D., Novello, A., Sylvestre, F., Mertzman, S.A., Haile-Selassie, Y.,
2019. Age and context of mid-Pliocene hominin cranium from Woranso-Mille, Ethiopia.
Nature 573. doi:10.1038/s41586-019-1514-7
Schauble, E.A., Ghosh, P., Eiler, J.M., 2006. Preferential formation of 13C-18O bonds in
carbonate minerals, estimated using first-principles lattice dynamics. Geochimica et
Cosmochimica Acta 70, 2510–2529.
Schmidt, D.F., Amaya, D.J., Grise, K.M., Miller, A.J., 2020. Impacts of Shifting Subtropical
Highs on the California Current and Canary Current Systems. Geophysical Research Letters
47, e2020GL088996.
Schouten, S., Hopmans, E.C., Schefuß, E., Sinninghe Damsté, J.S., 2002. Distributional
veriations in marine crenarchaeol membrane lipids: a new tool for reconstructing ancient
262
sea water temperatures? Earth and Planetary Science Letters 204, 265–274.
Schouten, S., Hopmans, E.C., Sinninghe Damsté, J.S., 2004. The effect of maturity and
depositional redox conditions on archaeal tetraether lipid palaeothermometry. Organic
Geochemistry 35, 567–571.
Schouten, S., Hopmans, E.C., Sinninghe Damsté, J.S., 2013. The organic geochemistry of
glycerol dialkyl glycerol tetraether lipids: A review. Organic Geochemistry 54, 19–61.
Schouten, S., Wakeham, S.G., Damsté, J.S.S., 2001. Evidence for anaerobic methane oxidation
by archaea in euxinic waters of the Black Sea. Organic Geochemistry 32, 1277–1281.
Seager, R., Neelin, D., Simpson, I., Liu, H., Henderson, N., Shaw, T., Kushnir, Y., Ting, M.,
Cook, B., 2014. Dynamical and thermodynamical causes of large-scale changes in the
hydrological cycle over North America in response to global warming. Journal of Climate
27, 7921–7948.
Seager, R., Ting, M., Held, I., Kushnir, Y., Lu, J., Vecchi, G., Huang, H.-P., Harnik, N.,
Leetmaa, A., Lau, N.-C., 2007. Model projections of an imminent transition to a more arid
climate in southwestern North America. Science 316, 1181–1184.
Seltzer, A.M., Ng, J., Aeschbach, W., Kipfer, R., Kulongoski, J.T., Severinghaus, J.P., Stute, M.,
2021. Widespread six degrees Celsius cooling on land during the Last Glacial Maximum.
Nature 593, 228–232.
Sessions, A.L., 2006. Isotope‐ratio detection for gas chromatography. Journal of separation
science 29, 1946–1961.
Setia, R., Gottschalk, P., Smith, P., Marschner, P., Baldock, J., Setia, D., Smith, J., 2013. Soil
salinity decreases global soil organic carbon stocks. Science of The Total Environment 465,
267–272.
263
Shankle, M.G., Burls, N.J., Fedorov, A. V., Thomas, M.D., Liu, W., Penman, D.E., Ford, H.L.,
Jacobs, P.H., Planavsky, N.J., Hull, P.M., 2021. Pliocene decoupling of equatorial Pacific
temperature and pH gradients. Nature 598, 457–461.
Shepherd, T., Wynne Griffiths, D., 2006. The effects of stress on plant cuticular waxes. New
Phytologist 171, 469–499.
Shilling, A.M., Colcord, D.E., Karty, J., Hansen, A., Freeman, K.H., Njau, J.K., Stanistreet, I.G.,
Stollhofen, H., Schick, K.D., Toth, N., Brassell, S.C., 2019. Biogeochemical evidence for
environmental changes of Pleistocene Lake Olduvai during the transitional sequence of
OGCP core 2A that encompasses Tuff IB (~1.848 Ma). Palaeogeography,
Palaeoclimatology, Palaeoecology 532, 109267.
Sinninghe Damsté, J.S., Ossebaar, J., Schouten, S., Verschuren, D., 2012a. Distribution of
tetraether lipids in the 25-ka sedimentary record of Lake Challa: extracting reliable TEX86
and MBT/CBT palaeotemperatures from an equatorial African lake. Quaternary Science
Reviews 50, 43–54.
Sinninghe Damsté, J.S., Rijpstra, W.I.C., Foesel, B.U., Huber, K.J., Overmann, J., Nakagawa, S.,
Kim, J.J., Dunfield, P.F., Dedysh, S.N., Villanueva, L., 2018. An overview of the
occurrence of ether- and ester-linked iso-diabolic acid membrane lipids in microbial
cultures of the Acidobacteria: Implications for brGDGT paleoproxies for temperature and
pH. Organic Geochemistry 124, 63–76.
Sinninghe Damsté, J.S., Rijpstra, W.I.C., Hopmans, E.C., Foesel, B.U., Wüst, P.K., Overmann,
J., Tank, M., Bryant, D.A., Dunfield, P.F., Houghton, K., Stott, M.B., 2014. Ether- and
ester-bound iso-diabolic acid and other lipids in members of Acidobacteria subdivision 4.
Applied and Environmental Microbiology 80, 5207–5218.
264
Sinninghe Damsté, J.S., Rijpstra, W.I.C., Hopmans, E.C., Jung, M.Y., Kim, J.G., Rhee, S.K.,
Stieglmeier, M., Schleper, C., 2012b. Intact polar and core glycerol dibiphytanyl glycerol
tetraether lipids of group I.1a and I.1b Thaumarchaeota in soil. Applied and Environmental
Microbiology 78, 6866–6874.
Sinninghe Damsté, J.S., Rijpstra, W.I.C., Hopmans, E.C., Weijers, J.W.H., Foesel, B.U.,
Overmann, J., Dedysh, S.N., 2011. 13,16-Dimethyl octacosanedioic acid (iso-Diabolic
Acid), a common membrane-spanning lipid of Acidobacteria subdivisions 1 and 3. Applied
and Environmental Microbiology 77, 4147–4154.
Sinninghe Damsté, J.S., Schouten, S., Hopmans, E.C., Van Duin, A.C.T., Geenevasen, J.A.J.,
2002. Crenarchaeol. Journal of Lipid Research 43, 1641–1651.
Smith, G.I., 2009. Late Cenozoic geology and lacustrine history of Searles Valley, inyo and San
Bernardino counties, California. US Geological Survey Professional Paper.
Smith, G.I., Barczak, V.J., Moulton, G.F., Liddicoat, J.C., 1983. Core KM-3, a surface-to-
bedrock record of late Cenozoic sedimentation in Searles Valley, California, Professional
Paper. doi:10.3133/PP1256
Spoor, F., Gunz, P., Neubauer, S., Stelzer, S., Scott, N., Kwekason, A., Dean, M.C., 2015.
Reconstructed Homo habilis type OH 7 suggests deep-rooted species diversity in early
Homo. Nature 2015 519:7541 519, 83–86.
Stap, L.B., de Boer, B., Ziegler, M., Bintanja, R., Lourens, L.J., van de Wal, R.S.W., 2016. CO2
over the past 5 million years: Continuous simulation and new δ11B-based proxy data. Earth
and Planetary Science Letters 439, 1–10.
Stiehl, T., Rullkötter, J., Nissenbaum, A., 2005. Molecular and isotopic characterization of lipids
in cultured halophilic microorganisms from the Dead Sea and comparison with the sediment
265
record of this hypersaline lake. Organic Geochemistry 36, 1242–1251.
Stock, G.M., Anderson, R.S., Finkel, R.C., 2004. Pace of landscape evolution in the Sierra
Nevada, California, revealed by cosmogenic dating of cave sediments. Geology 32, 193–
196.
Su, D.F., Haile-Selassie, Y., 2022. Mosaic habitats at Woranso-Mille (Ethiopia) during the
Pliocene and implications for Australopithecus paleoecology and taxonomic diversity.
Journal of Human Evolution 163, 103076.
Su, D.F., Harrison, T., 2015. The paleoecology of the Upper Laetolil Beds, Laetoli Tanzania: A
review and synthesis. Journal of African Earth Sciences 101, 405–419.
Tabor, C., Lofverstrom, M., Oster, J., Wortham, B., de Wet, C., Montañez, I., Rhoades, A.,
Zarzycki, C., He, C., Liu, Z., 2021. A mechanistic understanding of oxygen isotopic
changes in the Western United States at the Last Glacial Maximum. Quaternary Science
Reviews 274, 107255.
Tabor, N.J., Myers, T.S., 2015. Paleosols as indicators of paleoenvironment and paleoclimate.
Annual Review of Earth and Planetary Sciences 43, 333–361.
Tamalavage, A.E., van Hengstum, P.J., Louchouarn, P., Fall, P.L., Donnelly, J.P., Albury, N.A.,
Coats, S., Feakins, S.J., 2020. Plant wax evidence for precipitation and vegetation change
from a coastal sinkhole lake in the Bahamas spanning the last 3000 years. Organic
Geochemistry 150, 104120.
Thompson, R.S., Anderson, K.H., 2000. Biomes of western North America at 18,000, 6000 and
0 14C yr BP reconstructed from pollen and packrat midden data. Journal of Biogeography
27, 555–584.
Tierney, J.E., Haywood, A.M., Feng, R., Bhattacharya, T., Otto-Bliesner, B.L., 2019. Pliocene
266
Warmth Consistent With Greenhouse Gas Forcing. Geophysical Research Letters 46, 9136–
9144.
Tierney, J.E., Mayes, M.T., Meyer, N., Johnson, C., Swarzenski, P.W., Cohen, A.S., Russell,
J.M., 2010. Late-twentieth-century warming in Lake Tanganyika unprecedented since AD
500. Nature Geoscience 2010 3:6 3, 422–425.
Tierney, J.E., Russell, J.M., 2009. Distributions of branched GDGTs in a tropical lake system:
Implications for lacustrine application of the MBT/CBT paleoproxy. Organic Geochemistry
40, 1032–1036.
Tierney, J.E., Russell, J.M., Huang, Y., Damsté , J.S.S., Hopmans, E.C., Cohen, A.S., 2008.
Northern Hemisphere Controls on Tropical Southeast African Climate During the Past
60,000 Years. Science 322, 252–255.
Tipple, B.J., Pagani, M., 2010. A 35 Myr North American leaf-wax compound-specific carbon
and hydrogen isotope record: Implications for C4 grasslands and hydrologic cycle
dynamics. Earth and Planetary Science Letters 299, 250–262.
Tipple, B.J., Pagani, M., 2013. Environmental control on eastern broadleaf forest speciesâ€
TM
leaf wax distributions and D/H ratios. Geochimica et Cosmochimica Acta 111, 64–77.
Turich, C., Freeman, K.H., 2011. Archaeal lipids record paleosalinity in hypersaline systems.
Organic Geochemistry 42, 1147–1157.
Ueki, K., Hino, H., Kuwatani, T., 2018. Geochemical Discrimination and Characteristics of
Magmatic Tectonic Settings: A Machine-Learning-Based Approach. Geochemistry,
Geophysics, Geosystems 19, 1327–1347.
Uno, K.T., Polissar, P.J., Jackson, K.E., deMenocal, P.B., 2016a. Neogene biomarker record of
vegetation change in eastern Africa. Proceedings of the National Academy of Sciences of
267
the United States of America 113, 6355–63.
Uno, K.T., Polissar, P.J., Kahle, E., Feibel, C., Harmand, S., Roche, H., Demenocal, P.B., 2016b.
A Pleistocene palaeovegetation record from plant wax biomarkers from the Nachukui
Formation, West Turkana, Kenya. Philosophical Transactions of the Royal Society B:
Biological Sciences 371, 20150235.
Verschuren, D., Sinninghe Damsté, J.S., Moernaut, J., Kristen, I., Blaauw, M., Fagot, M., Haug,
G.H., 2009. Half-precessional dynamics of monsoon rainfall near the East African Equator.
Nature 2009 462:7273 462, 637–641.
Vos, M., Wolf, A.B., Jennings, S.J., Kowalchuk, G.A., 2013. Micro-scale determinants of
bacterial diversity in soil. FEMS Microbiology Reviews 37, 936–954.
Walker, J.D., Bidgoli, T.S., Didericksen, B.D., Stockli, D.F., Andrew, J.E., 2014. Middle
Miocene to recent exhumation of the Slate Range, eastern California, and implications for
the timing of extension and the transition to transtension. Geosphere 10, 276–291.
Wang, H., Liu, W., He, Y., Zhou, A., Zhao, H., Liu, H., Cao, Y., Hu, J., Meng, B., Jiang, J.,
Kolpakova, M., Krivonogov, S., Liu, Z., 2021. Salinity-controlled isomerization of
lacustrine brGDGTs impacts the associated MBT5ME’ terrestrial temperature index.
Geochimica et Cosmochimica Acta 305, 33–48.
Wang, H., Liu, W., Lu, H., Zhang, C., 2017. Potential degradation effect on paleo-moisture
proxies based on the relative abundance of archaeal vs. bacterial tetraethers in loess-
paleosol sequences on the Chinese Loess Plateau. Quaternary International 436, 173–180.
Wang, H., Liu, W., Zhang, C.L., Jiang, H., Dong, H., Lu, H., Wang, J., 2013. Assessing the ratio
of archaeol to caldarchaeol as a salinity proxy in highland lakes on the northeastern
Qinghai–Tibetan Plateau. Organic Geochemistry 54, 69–77.
268
Wang, L.X., 1992. A Neural Detector for Seismic Reflectivity Sequences. IEEE Transactions on
Neural Networks 3, 338–340.
Wang, M., Zheng, Z., Man, M., Hu, J., Gao, Q., 2017. Branched GDGT-based paleotemperature
reconstruction of the last 30,000 years in humid monsoon region of Southeast China.
Chemical Geology 463, 94–102.
Wang, Z., Zhang, F., Cao, Y., Hu, J., Wang, Huanye, Lu, H., Dong, J., Xing, M., Liu, H., Wang,
Hong, Liu, W., 2022. Linking sedimentary and speleothem precipitation isotope proxy
records to improve lacustrine and marine 14C chronologies. Quaternary Science Reviews
282, 107444.
Weijers, J.W.H., Panoto, E., van bleijswijk, J., Schouten, S., Rijpstra, W.I.C., Balk, M., Stams,
A.J.M., Damsté, J.S.S., 2009. Constraints on the Biological Source(s) of the Orphan
Branched Tetraether Membrane Lipids. http://dx.doi.org/10.1080/01490450902937293 26,
402–414.
Weijers, J.W.H., Schouten, S., van den Donker, J.C., Hopmans, E.C., Sinninghe Damsté, J.S.,
2007. Environmental controls on bacterial tetraether membrane lipid distribution in soils.
Geochimica et Cosmochimica Acta 71, 703–713.
Weijers, J.W.H., Wiesenberg, G.L.B., Bol, R., Hopmans, E.C., Pancost, R.D., 2010. Carbon
isotopic composition of branched tetraether membrane lipids in soils suggest a rapid
turnover and a heterotrophic life style of their source organism(s). Biogeosciences 7, 2959–
2973.
Weijers, J.W.H.H., Schouten, S., Hopmans, E.C., Geenevasen, J.A.J.J., David, O.R.P.P.,
Coleman, J.M., Pancost, R.D., Sinninghe Damste, J.S., Sinninghe Damsté, J.S., 2006.
Membrane lipids of mesophilic anaerobic bacteria thriving in peats have typical archaeal
269
traits. Environmental Microbiology 8, 648–657.
Wendt, K.A., Dublyansky, Y. V., Moseley, G.E., Edwards, R.L., Cheng, H., Spötl, C., 2018.
Moisture availability in the southwest United States over the last three glacial-interglacial
cycles. Science Advances 4, eaau1375.
Western Regional Climate Center, 2022. US COOP Station Map [WWW Document]. URL
https://wrcc.dri.edu/coopmap/ (accessed 10.31.18).
Western Regional Climate Center [WWW Document], 2020. URL https://wrcc.dri.edu/cgi-
bin/cliMAIN.pl?ca9035
Williams, A.P., Cook, E.R., Smerdon, J.E., Cook, B.I., Abatzoglou, J.T., Bolles, K., Baek, S.H.,
Badger, A.M., Livneh, B., 2020. Large contribution from anthropogenic warming to an
emerging North American megadrought. Science 368, 314–318.
Williams, C.K.I., Barber, D., 1998. Bayesian classification with gaussian processes. IEEE
Transactions on Pattern Analysis and Machine Intelligence 20, 1342–1351.
Williams, D.G., Ehleringer, J.R., 2000. Intra- and Interspecific Variation for Summer
Precipitation Use in Pinyon-Juniper Woodlands. Ecological Monographs 70, 517.
Willson, C.J., Manos, P.S., Jackson, R.B., 2008. Hydraulic traits are influenced by phylogenetic
history in the drought-resistant, invasive genus Juniperus (Cupressaceae). American journal
of botany 95, 299–314.
Windler, G., Tierney, J.E., Zhu, J., Poulsen, C.J., 2020. Unraveling Glacial Hydroclimate in the
Indo-Pacific Warm Pool: Perspectives From Water Isotopes. Paleoceanography and
Paleoclimatology 35, e2020PA003985.
Winnick, M.J., Welker, J.M., Chamberlain, C.P., 2013. Climate of the Past Geoscientific
Instrumentation Methods and Data Systems Stable isotopic evidence of El NiñoNi˜Niño-
270
like atmospheric circulation in the Pliocene western United States. Clim. Past 9, 903–912.
Winograd, I.J., Coplen, T.B., Landwehr, J.M., Riggs, A.C., Ludwig, K.R., Szabo, B.J., Kolesar,
P.T., Revesz, K.M., 1992. Continuous 500,000-year climate record from vein calcite in
Devils Hole, Nevada. Science 258, 255–260.
Winters, Y.D., Lowenstein, T.K., Timofeeff, M.N., 2013. Identification of Carotenoids in
Ancient Salt from Death Valley, Saline Valley, and Searles Lake, California, Using Laser
Raman Spectroscopy. Astrobiology 13, 1065–1080.
Wollheim, W.M., Lovvorn, J.R., 1996. Effects of macrophyte growth forms on invertebrate
communities in saline lakes of the Wyoming High Plains. Hydrobiologia 323, 83–96.
Wood, G.D., 2000. Pollen analysis of death valley sediments deposited between 166 and 114 ka.
Palynology 24, 49–61.
Woolfenden, W.B., 2003. A 180,000-year pollen record from Owens Lake, CA: Terrestrial
vegetation change on orbital sales. Quaternary Research 59, 430–444.
Wu, M.S., West, A.J., Feakins, S.J., 2019. Tropical soil profiles reveal the fate of plant wax
biomarkers during soil storage. Organic Geochemistry 128, 1–15.
Wynn, J.G., Sponheimer, M., Kimbel, W.H., Alemseged, Z., Reed, K., Bedaso, Z.K., Wilson,
J.N., 2013. Diet of Australopithecus afarensis from the Pliocene Hadar Formation, Ethiopia.
Proceedings of the National Academy of Sciences of the United States of America.
Xia, L., Cao, J., Stüeken, E.E., Zhi, D., Wang, T., Li, W., 2020. Unsynchronized evolution of
salinity and pH of a Permian alkaline lake influenced by hydrothermal fluids: A multi-proxy
geochemical study. Chemical Geology 541, 119581.
Xie, S., Pancost, R.D., Chen, L., Evershed, R.P., Yang, H., Zhang, K., Huang, J., Xu, Y., 2012.
Microbial lipid records of highly alkaline deposits and enhanced aridity associated with
271
significant uplift of the Tibetan Plateau in the Late Miocene. Geology 40, 291–294.
Yamamoto, Y., Ajioka, T., Yamamoto, M., 2016. Climate reconstruction based on GDGT-based
proxies in a paleosol sequence in Japan: Postdepositional effect on the estimation of air
temperature. Quaternary international 397, 380–391.
Yang, H., Lü, X., Ding, W., Lei, Y., Dang, X., Xie, S., 2015. The 6-methyl branched tetraethers
significantly affect the performance of the methylation index (MBT′) in soils from an
altitudinal transect at Mount Shennongjia. Organic Geochemistry 82, 42–53.
Yang, H., Pancost, R.D., Dang, X., Zhou, X., Evershed, R.P., Xiao, G., Tang, C., Gao, L., Guo,
Z., Xie, S., 2014. Correlations between microbial tetraether lipids and environmental
variables in Chinese soils: Optimizing the paleo-reconstructions in semi-arid and arid
regions. Geochimica et Cosmochimica Acta 126, 49–69.
Yu, K., D’odorico, Paolo, Collins, Scott L, Carr, David, Porporato, Amilcare, Anderegg, William
R L, Gilhooly Iii, W.P., Wang, Lixin, Bhattachan, Abinash, Bartlett, Mark, Hartzell,
Samantha, Yin, Jun, He, Yongli, Li, Wei, Tatlhego, Mokganedi, Fuentes, Jose D, Yu, C.:,
D’odorico, P, Collins, S L, Carr, D, Porporato, A, Anderegg, W R L, Gilhooly, W.P., Wang,
L, Bhattachan, A, Bartlett, M, Hartzell, S, Yin, J, He, Y, Li, W, Tatlhego, M, Fuentes, J D,
2019. The competitive advantage of a constitutive CAM species over a C4 grass species
under drought and CO2 enrichment. Ecosphere 10, e02721.
Zabihi, M., Pourghasemi, H.R., Pourtaghi, Z.S., Behzadfar, M., 2016. GIS-based multivariate
adaptive regression spline and random forest models for groundwater potential mapping in
Iran. Environmental Earth Sciences 75, 1–19.
Zamanian, K., Pustovoytov, K., Kuzyakov, Y., 2016. Pedogenic carbonates: Forms and
formation processes. Earth-Science Reviews 157, 1–17.
272
Zang, J., Lei, Y., Yang, H., 2018. Distribution of glycerol ethers in Turpan soils: implications for
use of GDGT-based proxies in hot and dry regions. Frontiers of Earth Science 2018 12:4
12, 862–876.
Zech, R., Gao, L., Tarozo, R., Huang, Y., 2012. Branched glycerol dialkyl glycerol tetraethers in
Pleistocene loess-paleosol sequences: Three case studies. Organic geochemistry 53, 38–44.
Zhang, D., Beverly, E.J., Levin, N.E., Vidal, E., Matia, Y., Feakins, S.J., 2021. Carbon isotopic
composition of plant waxes, bulk organics and carbonates from soils of the Serengeti
grasslands. Geochimica et Cosmochimica Acta. doi:10.1016/J.GCA.2021.07.005
Zhang, K., Wu, X., Niu, R., Yang, K., Zhao, L., 2017. The assessment of landslide susceptibility
mapping using random forest and decision tree methods in the Three Gorges Reservoir area,
China. Environmental Earth Sciences 76, 1–20.
Zhang, L., Hay, W.W., Wang, C., Gu, X., 2019. The evolution of latitudinal temperature
gradients from the latest Cretaceous through the Present. Earth-Science Reviews 189, 147–
158.
Zheng, Y., Heng, P., Conte, M.H., Vachula, R.S., Huang, Y., 2019. Systematic chemotaxonomic
profiling and novel paleotemperature indices based on alkenones and alkenoates: Potential
for disentangling mixed species input. Organic Geochemistry 128, 26–41.
Zomer, R.J., Bossio, D.A., Trabucco, A., Yuanjie, L., Gupta, D.C., Singh, V.P., 2007. Trees and
water: smallholder agroforestry on irrigated lands in Northern India. IWMI.
Zomer, R.J., Trabucco, A., Bossio, D.A., Verchot, L. V., 2008. Climate change mitigation: A
spatial analysis of global land suitability for clean development mechanism afforestation
and reforestation. Agriculture, Ecosystems & Environment 126, 67–80.
273
Appendices
274
Appendix. A Supplementary information for Chapter 2
275
Fig. A.1.Chain length distributions of n-alkanes (blue) and n-alkanoic acids (green) in (A) Modern desert
plants. (B) Modern conifers. (C) Modern macrophytes. Columns A, B, and C show the relative proportion
of n-alkyl lipids. Error bars represent 1σ from the mean when multiple plants were analyzed.
276
Fig. A.2. Receiver operating characteristics (ROC) graphs of the five machine learning models. ROC
graphs plot the true positive rate (sensitivity) against the false positive rate (specificity) under different
classification threshold settings—models with higher discriminant capacity plot closer to the upper left-
hand corner of the plot. Left column shows the ROC curves of models tested against the original data.
Right column shows ROC curves of models tested against original data with uniformly distributed noise
added. Dashed black 1:1 line included for comparison.
277
Appendix. B Supplementary information for Chapter 3
Fig. B.1. Serengeti surface soil longitudinal transect showing: A) temperature, B) pH, C) TOC (%), D)
TOC δ
13
C, E) ∑brGDGTs (ng/g sediment), F) ∑isoGDGTs (ng/g sediment), G) ΛbrGDGTs (ng/g TOC),
H) ΛisoGDGTs (ng/g TOC).
278
Fig. B.2. Standard deviations of mean distributions were created from bootstrapping the BayMBT0 and
∆47 temperature data with different resampling sizes. As the resampling size increases, the variance
associated with the mean of resampled data decreases.
279
Appendix. C Supplementary information for Chapter 4
Extended palynology methods
An initial set of 115 pollen analyses at the U.S. Geological Survey’s Florence Bascom Science
Center in Reston, VA, by Ron Litwin and Nancy Durika. To extract pollen, 2–12 g of wet
sample were initially washed using deionized water to remove salts. Sediments were then
decalcified using HCl, siliciclastic material was removed using HF and heavy liquid separation
in ZnCl
2
, and mounted in glycerin jelly. One calibrated tablet of exotic Lycopodium clavatum
spores was added to each sample (62,712,2081 spores per tablet) to calculate absolute pollen
concentrations and pollen accumulation rates.
A subsequent set of 115 samples was processed at Syracuse University and the University of
California, Berkeley’s Quaternary Paleoecology Laboratory. We sampled 2 g of dry weight from
each sample, added a tablet of Lycopodium clavatum, used deionized water washes to dissolve
salts and disaggregate samples, then subjected the sample to HCl and HF acid digestion,
followed by KOH to remove organic acids, and acetolysis. Samples were stained with safranin
and suspended in silicone oil. We processed 2–3 samples using both methods and found no
systematic offsets in the pollen counts yielded by each method.
Modern plant study
To complement the downcore plant wax compound specific isotope work presented in the main
paper, we provide a supplementary study of representative taxa in modern vegetation. These
plant samples were sampled and lipid extractions were described previously (Peaple et al., 2021).
They are newly analyzed for compound specific carbon and hydrogen isotopic composition for
280
this study, using the same isotopic methods described in the main text. We studied representative
desert shrub vegetation in Searles Valley (SV), including C
3
creosote bush, Larrea tridentata (n
= 2) and rabbit brush, Ericameria nauseosa (n = 1) as well as C
4
saltbush (Atriplex hymenelytra,
Atriplex confertifoli and Atriplex canescens n = 4). From the valley slopes, we sampled the C
3
wild buckwheat, Eriogonum pusillum (n = 1) and spiny menodora, Menodora spinescens (n = 1)
as well as the Crassulacean Acid Metabolism (CAM) prickly pear, Opuntia chlorotica (n = 1).
We sampled conifers at 2150 m asl the San Bernardino Mountains (SBM), CA (165 km to the
south of Searles Lake) including: jeffrey pine, Pinus jeffreyi (n = 9); western juniper, Juniperus
occidentalis (n = 4), and white fir (Abies concolor) (n = 3).
281
Supplementary results
Modern plant survey
As context for the downcore plant wax isotope distributions (Fig. C.1a, b), we report δ
13
C
and δD
values for dominant homologs of the long chain n-alkanoic acids and n-alkanes in a survey of
living plants (Fig. C.1c, d). The availability of isotopic data varies as the molecular abundance
distributions differ between species.
Fig. C.1. Violin plots showing a) δ
13
C and b) δD values n-alkanoic acids (blue) and n-alkanes
(orange) in SLAPP-SL17 and c) δ
13
C and d) δD distributions for plants collected in the San
Bernadino Mountain (SBM, green) and Searles Valley (SV, pink).
Carbon isotopes
282
The desert plant community from Searles Valley has a large range in carbon isotopic values
indicating the use of C
3
, C
4
and CAM pathways (Nobel and Bobich, 2002; Yu et al., 2019).
Individual plants yielded δ
13
C
28acid
values ranging from -17.3‰ to 31.5‰ and δ
13
C
29alk
ranging
from -20.3‰ to -34.7‰ (Fig. C.1c, d). The most negative values were measured in C
3
shrubs
including Menodora spinescens (δ
13
C
29alk
= -32.5‰, δ
13
C
28acid
= -26.6‰, n = 1) and Larrea
tridentata (δ
13
C
29alk
= -31.6‰, δ
13
C
28acid
= -28.3‰, n = 1). Cacti including Opuntia chlorotica
(δ
13
C
28acid
= -19.7‰, δ
13
C
29alk
= -20‰, n = 1) and Cylindropuntia bigelovii (δ
13
C
28acid
= n.d.
δ
13
C
29alk
= -19.4‰, n = 1) have values denoting use of Crassulacean Acid Metabolism (CAM).
CAM plants are a minor component of the mixed shrub cover and are likely at their maximum
prevalence (arid conditions); as their carbon isotopic compositions are intermediate between the
C
3
and C
4
plants, they are not explicitly considered for sedimentary contributions. We detect the
significant and varying contribution of C
4
xerophytic and halophytic taxa in the pollen and δ
13
C
downcore. High δ
13
C values were measured in Atriplex (mean δ
13
C
28acid
= -18.2‰, δ
13
C
29alk
= -
21.7‰, n = 1) using the C
4
pathway.
In conifers in North America, n-alkanoic acids are usually more abundant than n-alkanes
(Diefendorf et al., 2015a), as corroborated by the sampling of pine and fir species present in the
San Bernardino Mountains today, where C
31
n-alkanes were rarely detectable; thus conifers are
unlikely to contribute to the n-alkane downcore record, except for junipers which have high
concentrations of C
31
-C
35
n-alkanes (Peaple et al., 2021). For the dominant n-alkane homologs:
Juniperus occidentalis δ
13
C
33alk
(mean -31.3‰, 1 = 0.8‰, n = 5), δ
13
C
29alk
, A. concolor (mean -
28.1‰, 1 = 0.98‰, n = 3) and δ
13
C
29alk
Pinus jeffreyi (mean -28.6‰, 1 = 0.7‰, n = 9). We
report data for the C
28
n-alkanoic acid homolog, often modal in plants, and reported downcore.
We find δ
13
C
28acid
ranges from -32.4 to -23.1‰. J. occidentalis had the lowest mean δ
13
C
28acid
of -
283
29.6‰ (1 = 3.2‰, n = 4); A. concolor had a mean of -24.7‰ (1 = 0.2‰, n = 3) and P. jeffreyi
a mean of -24.4‰ (1 = 1.1‰, n = 9). These conifers’ abundant fatty acids may dominate the
leaf wax transport during episodic montane runoff, as seen in the Lake Elsinore catchment
(Feakins et al., 2019). Our compound class comparison in modern species shows that the fatty
acids are more abundant in conifers as expected and more enriched. We show that the δ
13
C
acid
in
these species is very enriched for C
3
plants, with values of -24 to -25‰ measured in these C
3
conifers growing at 2 km (Fig. C.1c), values that could easily be misconstrued for a C
3
-C
4
mixture, when measured downstream in fluvial sediments or lakes.
Hydrogen isotopes
San Bernardino Mountain (SBM) conifers yielded δD
28acid
values from -155‰ to -190‰ with a
mean of -166‰. δD
28acid
mean values for individual species
were determined for J. occidentalis
(-160‰, n = 1), P. jeffreyi (-186‰, 1 = 5‰, n = 3) and A. concolor (-155‰, n = 1) and co-
occurring Artemisia tridentata (-142‰, 1 = 0.8‰, n = 2). Although the conifers do not have
abundant n-alkanes, the angiosperm shrub A. tridentata yields δD
29alk
(-152‰, 1 = 10‰, n = 3).
Desert plant taxa sampled in Searles Valley (SV) have δD
28acid
values ranging from -121‰ to -
217‰ (mean =
-163‰, 1 23‰, n = 15) and δD
29alk
ranging from -98‰ to -190‰ (mean =
-
155‰, 1 =25‰, n = 13).
We found no systematic offset in hydrogen isotopic compositions between photosynthetic
pathways (C
3
, C
4,
and CAM in order of prevalence on the landscape), suggesting an apparent
fractionation correction for the C
3
versus C
4
proportions is unlikely to be appropriate. The
conifers had a smaller range in δD
C28acid
than the desert plants (35‰ vs. 95‰), which could
reflect the greater diversity of plant types (i.e., shrubs, grasses and cacti) sampled in the Mojave
284
Desert. δD
C29alk
shows a similar distribution compared to δD
C28acid
in desert plants (Fig. C.1),
although the mean value is slightly more enriched ( -155‰ vs. -166‰). Among the SBM
vegetation we find a large range between δD
C31alk
measured in J. occidentalis compared to the
δD
C29
measured in A. tridentata (-77‰ vs -152‰). This might reflect different water uptake
strategies of the plants, as junipers uptake water throughout the year and respond quickly to even
small summer precipitation events (Williams and Ehleringer, 2000). However we do not see
similarly enriched δD
C28acid,
perhaps consistent with reports of seasonal offsets in the production
of n-alkanes and n-alkanoic acids in conifers (Freimuth et al., 2017).
Sediment core plant wax isotopic results
In the main text, we report two homologs, here we show results for additional compounds in the
homologous series as distributions (Fig. C.1) and downcore (Fig. C.2).
Carbon isotopes
For the n-alkanes (Fig. C.1a), δ
13
C
C27alk
ranges from -34.0‰ to -21.7‰ (mean -26.0‰), δ
13
C
29alk
ranges from -33.5‰ to -26.9‰ (mean -28.9‰) and δ
13
C
31alk
ranges from -32.5‰ to -27.0‰
(mean -29.4‰). δ
13
C
27alk
has a moderate positive
correlation with δ
13
C
C28acid
(r = 0.64), although
the correlations with the shorter chain δ
13
C
C26acid
and δ
13
C
24acid
are lower (r = 0.2, r = 0.38,
respectively). The correlations between δ
13
C
28acid
and the longer chain δ
13
C
29alk
and δ
13
C
31alk
are
moderate (r = 0.45 and r = 0.4, respectively). There is no correlation between the long chain
δ
13
C
31alk
and the shorter chain δ
13
C
26acid
and δ
13
C
24acid
(r = 0.12 and r = 0.15, respectively). There
is a strong correlation between δ
13
C
27alk
and δ
13
C
29alk
(r = 0.82) and between δ
13
C
29alk
and δ
13
C
31alk
(r = 0.78), although there is a weaker correlation between δ
13
C
27alk
and δ
13
C
31alk
(r = 0.64)
285
indicating a difference in sourcing. For the n-alkanoic acids (Fig. C.1a), δ
13
C
24acid
ranges from -
38.5‰ to -22‰, δ
13
C
26acid
ranges from -33.3‰ to -23.2‰ and δ
13
C
28acid
ranges from -32.0‰ to -
23.4‰ with a mean of -27.4‰.
Hydrogen isotopes
For the hydrogen isotopic composition of the n-alkanes, both δD
31alk
and δD
29alk
show very
similar trends downcore (Fig. C.2c), in addition to having similar means (-202‰ and -196‰,
respectively) and ranges (-223‰ to -170‰ and -217‰ to -164‰, respectively, Fig S1b). δD
27alk
is more enriched on average (mean = -190‰) but has a similar amplitude of variability (-223‰
to -173‰). For the n-alkanoic acids (Fig. C.1b), C
28
homolog, δD
C28acid
ranged from -227‰ to -
141‰ with a mean of -182‰. δD
26acid
has a larger range (-209‰ to -92‰) and is on average
more enriched (mean = -148‰) than δD
28acid
, as is δD
24acid,
which ranges from -192‰ to -128‰
with a mean of -167‰. There is a positive correlation between the hydrogen isotopic
compositions by n-alkanoic acid chain length, with the strongest correlation between δD
26acid
and
δD
28acid
(r = 0.6) (Fig. C.2d). There is no correlation between the C
31
n-alkanes and C
28
n-
alkanoic acids (r = 0), indicating different sources of these molecules.
286
Fig. C.2. Changes in the isotopic composition of n-alkanes and n-alkanoic acids with depth in SLAPP-
SL17 core a) δ13C measured from the C27, C29, and C31 alkane chain lengths. b) δ13C measured from
the C24, C26, and C28 alkanoic acid chain lengths. c) δD measured from the C27, C29, and C31 alkane
chain lengths. d) δD measured from the C24, C26, and C28 alkanoic acid chain lengths. e) Composite
287
core photo showing the presence of muds (dark) and salts (white) downcore. Thick salt accumulations
without biomarker sampling are indicated by pale pink shading.
Interpretation
Given the strong correlations between the C
29
and C
31
alkanes in both δD and δ
13
C (Fig. C.2a,
C), it is likely that they have a similar source. Even though bacteria and macrophytes have been
recorded as producing long chain alkanes (Aichner et al., 2010; Makou et al., 2018), terrestrial
plants are likely to be the dominant producers in most settings, and thus we interpret C
29alk
and
C
31alk
as having a terrestrial plant origin. The mean δ
13
C of both C
29alk
and C
31alk
and the small
range downcore (Fig. C.2a) suggests that these chain lengths have a C
3
source that is relatively
invariant through time. Previous pollen studies from Owens Lake (Litwin et al., 1999;
Woolfenden, 2003), in addition to pollen data presented here (Fig. C.3a), find Artemisia and
Juniper the dominant plants in SV and thus we assume that these plants dominate contributions
to C
29alk
and C
31alk
. The proportion of Amaranth pollen covaries with δ
13
C
29alk
and δ
13
C
31alk
(Fig.
C.3d),
suggesting that C
4
desert plants also contribute.
C
24acid
and C
26acid
likely had a different source to the long chain n-alkanes, given the lack of
correlation between their respective δ
13
C and δD. The very negative δ
13
C
24acid
between 56 m to
60 m (-38.5‰), combined with the inferred fresher water conditions (Fig. C.2d) during this
interval, could imply that there is C
24acid
algal input, as algae have been recorded as having lipid
δ
13
C as low as -40‰ (McCallister and Del Giorgio, 2008). No measured modern terrestrial
plants in this study have such low δ
13
C
acid
values (Fig. C.1a). The lack of such depleted δ
13
C
24acid
values in the rest of the core could suggest that the higher water salinities suppressed algae
production and thus decreased the relative contribution of algae C
24
acid. Throughout the rest of
the core, other macrophyte and terrestrial plant inputs likely contribute C
24
acid to Searles Lake;
as such, the source of C
24
is interpreted as changing downcore. The enriched values of δD
26acid
288
relative to other n-alkane and n-alkanoic acid chain lengths and the moderate positive correlation
between δD
26acid
ACE (r = 0.4) suggest that the source of C
26acid
was living in Searles
Lake/Owens Lake. Thus the δD
26acid
likely reflects both the δD of lake water and the salinity of
the lake water, which modulates the fractionation between water and lipid (Sachse et al., 2012).
Given that we see an increase in δD
26acid
in association with an increase in lake salinity, it implies
an algal source; as bacteria and algae, epsilon decreases with increases in salinity (Sachse et al.,
2012), whereas macrophytes fractionation increases with increasing salinity (Aichner et al.,
2017).
δ
13
C
28acid
and δ
13
C
27alk
as well as δD
28acid
and δD
27alk
covary in the core, and thus C
28acid
and C
27alk
likely share a similar source. Both δ
13
C
28acid
and δ
13
C
27alk
show a similar down core trend to the
longer chain (C
29alk
, C
31alk
) alkane δ
13
C, although they are more enriched by 1.5-6‰.
Additionally, while there is a strong correlation between δD
28acid
and δD
27alk
(r = 0.7) as well as a
strong correlation between δD
27alk
and both δD
29alk
and δD
31alk
, there is no correlation between
δD
C28acid
and either δD
29alk
or δD
31alk
. This suggests that the δD
27alk
is receiving lipid sourced
from both the producers of long chain n-alkanes (terrestrial plants) and aquatic/microbial
producers. C
28
n-alkanoic may include a mixture of aquatic and terrestrial plant sources, given
the moderate correlations between the δD and δ
13
C of C
28acid
and both mid chain length alkanes
(C
27alk
) and mid length chain alkanoic acids (C
26acid
). We, therefore, select the C
31
n-alkane as a
proxy for terrestrial vegetation and hydroclimate in the main text.
289
Fig. C.3. Downcore pollen and plant wax proxies from SLAPP-SRLS17. a) Proportion of pollen taxa. b)
Modelled vegetation types based on SVM machine learning of plant wax distributions in modern taxa
applied to the downcore record (Peaple et al., 2021b). c) Compare modeled desert plant types and sum
Amaranth and Asteraceae pollen. d) δ
13
C28acid and δ
13
C31alk compared to Amaranth pollen. e) δD28acid and
δD31alk. f) Precipitation δD estimated from the δD31alk with wax/p determined as i) “constant” -93‰ or
temporally varying as calculated based on the ii) “pollen,” i.e., the proportion of pollen taxa in the core,
290
iii), “C3 v. C4” proportion based on δ
13
C31alk, and iv) “ML” where “constant” is modified by mixing with
the SVM modeled desert plant proportions from machine learning. g) Composite core photo. Thick salt
accumulations without biomarker sampling (pale pink shading).
Modeling the apparent fractionation between plant wax and precipitation
The hydrogen isotopic composition of plant wax n-alkanes is of interest in reconstructing the
hydrogen isotopic composition of precipitation. However, that interpretation requires knowledge
of the apparent fractionation between precipitation and plant wax in the modern system and an
informed estimation of the past. The reconstructed precipitation isotopic composition is
dependent upon the choice of apparent fractionation. In order to quantify the implications of
interpretive choices, we assess the sensitivity of δD
precip
to changes in
wax/p
estimated using four
different methods (Fig. C.3f):
i. In the main text, we apply a “constant”
C3
of -93‰, based upon modern plant water and
plant wax n-alkane studies that found a constant average fractionation of -93‰ for
woody C
3
plants sampled across a W-E aridity gradient from the coastal and montane
woodlands to the Mojave Desert (Feakins and Sessions, 2010). Apparent fractionations
are relatively invariant in this range of climate and vegetation, supporting the most
straightforward approach to downcore reconstructions here.
ii. Pollen corrections have been applied in a few studies (Feakins, 2013; Inglis et al., 2020;
Tamalavage et al., 2020). We measured pollen and plant wax in parallel in the core; as
sampling depths differ, the data were first interpolated onto the same timestep using
linear interpolation with a resolution of 1.7 kyr. The “pollen” approach calculates
wax/p
291
using a mixing model (Fig. C.3f) is based on the relative proportion of pollen taxa, each
assigned an epsilon from the literature, with the following equation:
wax/p “pollen” =
f
pine
*
pine
+ f
artemisia
*
artemisia
+ f
asteraceae
*
asteraceae+
(1)
f
amaranthaceae
*
amaranthaceae
+ f
poaceae
*
poaceae
+ f
juniper
*
juniper
Where pine epsilon = -128‰ (Tipple and Pagani, 2013), artemisia epsilon = -60‰
(Feakins, unpublished), asteraceae epsilon = -112‰ (Sachse et al., 2012), amaranthaceae
epsilon = -81‰ (Sachse et al., 2012), poaceae epsilon = -134‰ (Sachse et al., 2012) and
juniper epsilon = -89 (Tipple and Pagani, 2013). We determine a mean
wax/p
“pollen” of -
113‰, which is close to the value assigned to the pine component (-128‰, which is the
dominant pollen Fig. C.3f) but reduced by the smaller apparent fractionations of desert
and juniper taxa. The downcore variation has a standard deviation of 6.5‰, a small
vegetation effect assessed by pollen.
iii. Carbon isotopic correction has been applied in tropical settings (Tipple and Pagani, 2010;
Fornace et al., 2016; Windler et al., 2020) to account for different apparent fractionation
for C
3
and C
4
vegetation types (Sachse et al., 2012), and this works well in the presence
of tropical C
4
grasses with more strongly differentiated fractionations. The approach is
tested here for xerophytic/halophytic C
4
woody shrubs. We assigned
C3
of -93‰,
denoting the apparent fractionation measured for C
3
woody vegetation surveyed across
Southern California (Feakins and Sessions, 2010), and estimated the
C4
component as -
81‰ based on values reported for chenono-amaranth (n = 10 samples, Bi et al., 2005;
292
Krull et al., 2006; Liu et al., 2006) from a global collation (Sachse et al., 2012). We
measured hydrogen and carbon isotopes on the same samples in the SLAPP-SRLS17
sediment core, thus used the δ
13
C
31alk
values to determine the f
C4,
assuming that the main
driver of δ
13
C
31alk
was the presence of C
4
plants, using a mixing model approach to
generate a carbon-isotope informed
wax/p
, “C
3
v C
4
” as follows:
wax/p “C3 v C4” =
f
1-C4
*
C3
+ f
C4
*
C4
(2)
This mixing model has a mean epsilon of -89‰ 1 = 1.0‰.
iv. Here we introduce a new approach to quantify the vegetation effect and use the machine
learning of plant wax n-alkane and n-alkanoic acid plant wax distributions in the
dominant plant taxa and the downcore record (Peaple et al., 2021). The machine learning
vegetation correction is based on the modeled proportion of vegetation using the SVM
algorithm (Peaple et al., 2021b). We assigned endmember fractionations for the desert
plants, macrophytes and conifers categories, identified in the machine learning approach,
as of -81‰ (Sachse et al., 2012), -86‰ (Aichner et al., 2010) and -109‰ (Tipple and
Pagani, 2013) respectively.
wax/p “ML” =
f
desert
*
desert
+ f
macro
*
macro
+ f
conifer
*
conifer
(3)
This results in a mean epsilon of -86.8‰ 1 =
293
All four Ɛ
wax/p
methods yield similar reconstructions (mean 1 = ) apart from the pollen
correction (ii), yielding δD
precip
estimates that are +20‰ offset compared to the other methods.
We discount this pollen-based estimate (ii) because a) the pollen record is dominated by pine,
which is likely overrepresented due to the wind-dispersed nature, and b) pines produce very little
C
31alk;
thus, a correction dominated by changes in pine is potentially spurious. We caution against
the pollen correction method in n-alkane records in similarly pine-pollen-dominated
assemblages. While we could weight to lower or eliminate the pine influence, our estimates by
the other approaches indicate that this would only shift the absolute value and have limited
influence on the downcore variability. We discount the mixing model approach (iii) due to the
weakness of its assumption that changes in δ
13
C
31alk
would be driven solely by changes in the
relative proportions of C
3
and C
4
plants. The approach is compromised here as both the changing
moisture variability and changing proportions of conifers would affect the C
3
“endmember.” We
also reject the machine learning approach (iv) as the plant groups reconstructed using this
method (conifers, desert plants, and macrophytes) have large intragroup ranges in epsilon which
cannot be constrained. We thus apply the regionally-determined, constant apparent fractionation
of -93‰ (i) to convert δD
31alk
into an estimate of the δD value of precipitation (δD
precip
) as
reported in the main text. We correct the changing composition of seawater (source water for
evaporation) due to changes in glacial ice volume using the benthic δ
18
O stack (Lisiecki and
Raymo, 2005) in order to evaluate changes in storm track and other regional climatic drivers of
precipitation isotopic variability.
294
Supplementary palynology results
Glacial to interglacial changes in pollen taxa composition occur downcore, with shrubs and pines
dominating glacial periods with Juniper and Artemsia expanding during cold glacial periods (Fig
S3a). A Nonmetric Multi-Dimensional Scaling analysis confirms that pollen taxa composition is
different during glacial and interglacial phases (Fig S4a). Changes in pollen taxa % could be
impacted by changes to pollen influx rates to lakes (PENNINGTON, 1979). We show that pollen
influx rates show similar trends to pollen % composition (Fig S3b, c, d), suggesting that pollen
taxa % reflects changes in catchment vegetation change.
295
Fig. C.4. Analysis of major pollen taxa in SLAPP-SRLS17 sediment core. a) Non-metric multi-
dimensional scaling (NMDS) analysis with glacial (blue diamond) and interglacial (red circle) samples
highlighted. b, c, d) Time series plots showing the pollen percentage and influx rate.
GDGT concentrations and indices
GDGT distributions in stratigraphic context
We report the concentration of the br and isoGDGTs and select compounds on the sediment
stratigraphic context (Fig. C.5). We find crenarchaeol has low concentrations (<1 ng/g)
throughout the core except during the interval 57 to 58.5 m (highlighted in blue), where
concentrations peak at 238 ng/g (Fig. C.5a). This depth interval also shows a brGDGT
concentration peak of 810 ng/g. Between 30 m to 76 m, brGDGT concentrations are variable
between 1 to 140 ng/g, and between 6 m to 30 m, brGDGT concentrations are lower and range
from 1 to 11 ng/g. ∑isoGDGTs have peak concentrations between 44 m to 51 m, 57 m to 58.5 m
and 65 m to 76 m, although unlike brGDGTs and crenarchaeol, ∑isoGDGTs have a maximum
concentration between 65 m to 75 m with a peak concentration of 1015 ng/g. Like ∑brGDGT,
∑isoGDGT also has lower concentrations in the upper section of the core between 6 to 30 m,
likely due to the higher salinity conditions as shown through the presence of salts in the SLAPP-
SRLS17 sediment core and high ACE values. Salts have higher accumulation rates and thus
greater dilution of organics and the environment is also more inhospitable to many organisms.
BIT and %GDGT-0 are uniformly high (>0.9) throughout much of the sediment sequence but
both shows decreases between 57 to 58.5 m. The low BIT and %GDGT-0 between 57 to 58.5 m
is driven primarily by the increase in crenarchaeol concentration (Fig. C.5b) and represents an
296
expansion in the amount of Thaumarcheota in the lake, likely in response to an increase in
oxygen and a decrease in salinity.
Caldarcheaol (isoGDGT-0, Figure 5c) makes up most of the ∑isoGDGT (Figure 5a) and thus
shows the same abundance trends with depth. Between 63 m to 76 m depth archaeol correlates
with caldarchaeol (r = 0.64), although there is a lack of correlation between 6 m and 63 m (r = 0).
There is a peak in the archeaol at 35 m where concentrations reach 909 ng/g and occur in muds
interbedded in salts, suggesting small ephemeral hypersaline lake conditions.
ACE (Fig. C.5d) is driven by changes in the relative proportion of archeaol and caldarchaeol
(isoGDGT-0, Fig. C.5c). ACE is variable with depth, ranging from 1 to 99, representing fresh
water to water salinities > 300 psu (Turich and Freeman, 2011). Low ACE values are recorded in
association with mud layers, with higher ACE seen from sections containing interbedded salts
(Fig. C.5g). This can be seen visually in the core; between 33 to 50 m ACE decreases from 95 to
varying between 40 to 2 and this matches the switching from small interbedded mud and salt
layers to massive mud deposits.
297
Fig. C.5. Downcore GDGT concentrations and indices in SLAPP-SRLS17, showing concentrations of a)
summed and selected GDGTs concentrations, b) BIT and %GDGT-0, c) caldarchaeol and isoGDGT-0, d)
ACE and IR6+7me index for salinity, e) pH, f) reconstructed MAF temperatures and g) composite core
stratigraphic column. Overflowing lake conditions with mud facies and low ACE salinity (blue shading),
298
only the T2 deep lake is associated with high productivity (a) and a well-mixed lake (b). Thick salt
accumulations without biomarker sampling (pale pink shading).
Evaluation of two salinity proxies
To determine the relationship between the ACE and IR
6+7me
salinity proxies, we compare their
time series (Fig. C.6a) cross wavelet spectra (Fig. C.6b) and frequency spectra (Fig. C.6c). Both
proxies strongly covary in the precession frequency band between 160-100 kyr, although they
show little covariation between 100-0 kyrs (Fig. C.6b). Both proxies share significant amplitude
in precessional frequencies, although, unlike IR
6+7me,
ACE has a large amplitude obliquity peak
(Fig. C.6c).
Fig. C.6. Time and frequency response of salinity proxies ACE (blue) and IR6+7me (orange). a) Time series
of ACE and IR6+7me b) Wavelet coherence between ACE and IR6+7me. c) Weighted wavelet Z transform
frequency spectrum of ACE (blue) and IR6+7me (orange).
299
Lake pH
We calculated pH using CBT′ following an Eastern African lake calibration (Russell et al.,
2018). Reconstructed pH is relatively invariant with a mean of 8.9 and a standard deviation of
0.6 (Fig. C.5e), similar to estimates from clay mineral diagenesis (Hay et al., 1991). However,
there is a spike in high pH conditions (pH 11) between 67 to 68.5 m. Searles Lake was relatively
fresh and overflowing during this period, so a higher pH is unusual. Usually, high pH and low
salinity conditions are caused by hydrothermal inputs into lakes (Xia et al., 2020), and the tufa
deposits located in the southwest of the lake at the Pleistocene highstand would indicate that
significant amounts of Ca
+
rich spring waters were being added to the lake during this depth
interval. However, one explanation is that when stratified, the surface water of Searles Lake had
a lower pH than deep anoxic water due to precipitation falling on the lake's surface and
dissolution of atmospheric CO
2
into the lake water. However, when Searles Lake was
overturning, the pH of the surface water would increase, reflecting the influence of deeper, more
alkaline water. A similar change in surface water pH occurs in the seasonally overturning
hyperalkaline Gorka Pit Lake, Poland (Czop et al., 2011).
Temperature
The BayMBT
0
temperature calibration (Martínez-Sosa et al., 2021) applied to the measured
MBT′
5Me
downcore yields MAF between 7 to 25°C downcore (Fig. C.5f). MBT′
5Me
decreased
between 76 to 57 m to a minimum of 7°C whereby Searles Lake is interpreted to be overflowing
into Lake Panamint. MBT′
5Me
then rapidly increased to higher temperatures (23-25°C) between
56 to 40 m. Temperatures then decreased between 40 to 30 m to 12-15°C in association with the
300
appearance of salts in the core. Temperatures ranged between 30 to 6 m and ranged from 15-
23°C, with a cooler period of 16-17 °C occurring between 22 to 26 m in association with mud
deposits. There is a positive association between BayMBT
0
temperature and the presence of
desert shrub pollen in the core which suggests that despite the hypersaline and hyperalkaline lake
chemistry, the MBT´
5Me
is primarily responding to changes in water temperature.
In order to interrogate the choice of calibration, we compared compilations of soil (Dearing
Crampton-Flood et al., 2020) and lake brGDGTs (Martínez-Sosa et al., 2021) filtered to only
consider those between 25 – 14 °C equivalent to the temperature range experiences in Searles
Valley (Fig. C.7). Searles Lake brGDGT distributions are more similar to the lake than soil
brGDGTs distributions, giving us confidence in the choice of lake calibration. This test also
indicates that the salinity variations do not yield a community of producers with a different
brGDGT distribution, which is reassuring as salinity can lead to warm biases in some lakes,
including nearby Mono Lake (Martinez Sosa et al., 2021).
301
Fig. C.7. Violin plots of brGDGT distributions from soil (Dearing Crampton-Flood et al., 2020) and lake
(Martínez-Sosa et al., 2021) compilations, selecting those entries with MAT of 14–25 °C, for comparison
to brGDGT distributions from Searles Lake SLAPP samples. Searles appears to have lake-like
distributions.
References
Aichner, B., Herzschuh, U., & Wilkes, H. (2010). Influence of aquatic macrophytes on the stable
carbon isotopic signatures of sedimentary organic matter in lakes on the Tibetan Plateau.
Organic Geochemistry, 41(7), 706–718. https://doi.org/10.1016/j.orggeochem.2010.02.002
Aichner, B., Hilt, S., Périllon, C., Gillefalk, M., & Sachse, D. (2017). Biosynthetic hydrogen
isotopic fractionation factors during lipid synthesis in submerged aquatic macrophytes:
Effect of groundwater discharge and salinity. Organic Geochemistry, 113, 10–16.
https://doi.org/10.1016/J.ORGGEOCHEM.2017.07.021
Bi, X., Sheng, G., Liu, X., Li, C., & Fu, J. (2005). Molecular and carbon and hydrogen isotopic
composition of n-alkanes in plant leaf waxes. Organic Geochemistry, 36(10), 1405–1417.
https://doi.org/10.1016/J.ORGGEOCHEM.2005.06.001
Czop, M., Motyka, J., Sracek, O., & Szuwarzyński, M. (2011). Geochemistry of the
hyperalkaline Gorka pit lake (pH>13) in the Chrzanow region, southern Poland. Water, Air,
and Soil Pollution, 214(1–4), 423–434. https://doi.org/10.1007/s11270-010-0433-x
Dearing Crampton-Flood, E., Tierney, J. E., Peterse, F., Kirkels, F. M. S. A., & Sinninghe
Damsté, J. S. (2020). BayMBT: A Bayesian calibration model for branched glycerol dialkyl
glycerol tetraethers in soils and peats. Geochimica et Cosmochimica Acta, 268, 142–159.
https://doi.org/10.1016/j.gca.2019.09.043
302
Diefendorf, A. F., Leslie, A. B., & Wing, S. L. (2015). Leaf wax composition and carbon
isotopes vary among major conifer groups. Geochimica et Cosmochimica Acta, 170, 145–
156. https://doi.org/10.1016/j.gca.2015.08.018
Feakins, S. J. (2013). Pollen-corrected leaf wax D/H reconstructions of northeast African
hydrological changes during the late Miocene. Palaeogeography, Palaeoclimatology,
Palaeoecology, 374, 62–71. https://doi.org/10.1016/J.PALAEO.2013.01.004
Feakins, S. J., & Sessions, A. L. (2010). Controls on the D/H ratios of plant leaf waxes in an arid
ecosystem. Geochimica et Cosmochimica Acta, 74(7), 2128–2141.
https://doi.org/10.1016/J.GCA.2010.01.016
Fornace, K. L., Whitney, B. S., Galy, V., Hughen, K. A., & Mayle, F. E. (2016). Late Quaternary
environmental change in the interior South American tropics: new insight from leaf wax
stable isotopes. Earth and Planetary Science Letters, 438, 75–85.
https://doi.org/10.1016/J.EPSL.2016.01.007
Freimuth, E. J., Diefendorf, A. F., & Lowell, T. V. (2017). Hydrogen isotopes of n-alkanes and
n-alkanoic acids as tracers of precipitation in a temperate forest and implications for
paleorecords. Geochimica et Cosmochimica Acta, 206, 166–183.
https://doi.org/10.1016/J.GCA.2017.02.027
Hay, R. L., Guldman, S. G., Matthews, J. C., Lander, R. H., Duffin, M. E., & Kyser, T. K.
(1991). Clay mineral diagenesis in core KM-3 of Searles Lake, California. Clays and Clay
Minerals, 39(1), 84–96. https://doi.org/10.1346/CCMN.1991.0390111
Inglis, G. N., Carmichael, M. J., Farnsworth, A., Lunt, D. J., & Pancost, R. D. (2020). A long-
term, high-latitude record of Eocene hydrological change in the Greenland region.
Palaeogeography, Palaeoclimatology, Palaeoecology, 537, 109378.
303
https://doi.org/10.1016/J.PALAEO.2019.109378
Krull, E., Sachse, D., Mügler, I., Thiele, A., & Gleixner, G. (2006). Compound-specific δ13C
and δ2H analyses of plant and soil organic matter: A preliminary assessment of the effects
of vegetation change on ecosystem hydrology. Soil Biology and Biochemistry, 38(11),
3211–3221. https://doi.org/10.1016/J.SOILBIO.2006.04.008
Lisiecki, L. E., & Raymo, M. E. (2005). A Pliocene-Pleistocene stack of 57 globally distributed
benthic δ18 O records. Paleoceanography, 20(1), n/a-n/a.
https://doi.org/10.1029/2004PA001071
Litwin, R. J., Smoot, J. P., Durika, N. J., & Smith, G. I. (1999). Calibrating Late Quaternary
terrestrial climate signals: Radiometrically dated pollen evidence from the southern Sierra
Nevada, USA. Quaternary Science Reviews. https://doi.org/10.1016/S0277-3791(98)00111-
5
Liu, W., Yang, H., & Li, L. (2006). Hydrogen isotopic compositions of n-alkanes from terrestrial
plants correlate with their ecological life forms. Oecologia, 150, 330–338.
https://doi.org/10.1007/s00442-006-0494-0
Makou, M., Eglinton, T., McIntyre, C., Montluçon, D., Antheaume, I., & Grossi, V. (2018).
Plant Wax n ‐Alkane and n ‐Alkanoic Acid Signatures Overprinted by Microbial
Contributions and Old Carbon in Meromictic Lake Sediments. Geophysical Research
Letters, 45(2), 1049–1057. https://doi.org/10.1002/2017GL076211
Martínez-Sosa, P., Tierney, J. E., Stefanescu, I. C., Dearing Crampton-Flood, E., Shuman, B. N.,
& Routson, C. (2021). A global Bayesian temperature calibration for lacustrine brGDGTs.
Geochimica et Cosmochimica Acta, 305, 87–105.
https://doi.org/10.1016/J.GCA.2021.04.038
304
McCallister, S. L., & Del Giorgio, P. A. (2008). Direct measurement of the δ13C signature of
carbon respired by bacteria in lakes: Linkages to potential carbon sources, ecosystem
baseline metabolism, and CO2 fluxes. Limnology and Oceanography, 53(4), 1204–1216.
https://doi.org/10.4319/lo.2008.53.4.1204
Nobel, P. S., & Bobich, E. G. (2002). Initial Net CO2 Uptake Responses and Root Growth for a
CAM Community Placed in a Closed Environment. Annals of Botany, 90(5), 593–598.
https://doi.org/10.1093/AOB/MCF229
Peaple, M. D., Tierney, J. E., McGee, D., Lowenstein, T. K., Bhattacharya, T., & Feakins, S. J.
(2021). Identifying plant wax inputs in lake sediments using machine learning. Organic
Geochemistry, 156, 104222. https://doi.org/10.1016/J.ORGGEOCHEM.2021.104222
Pennington, W. (1979). The origin of pollen in lakes sedimetns: an enclosed lake compared with
one recieiving inflow stream. New Phytologist, 83(1), 189–213.
https://doi.org/10.1111/j.1469-8137.1979.tb00741.x
Russell, J. M., Hopmans, E. C., Loomis, S. E., Liang, J., & Sinninghe Damsté, J. S. (2018).
Distributions of 5- and 6-methyl branched glycerol dialkyl glycerol tetraethers (brGDGTs)
in East African lake sediment: Effects of temperature, pH, and new lacustrine
paleotemperature calibrations. Organic Geochemistry, 117, 56–69.
https://doi.org/10.1016/J.ORGGEOCHEM.2017.12.003
Sachse, D., Billault, I., Bowen, G. J., Chikaraishi, Y., Dawson, T. E., Feakins, S. J., et al. (2012).
Molecular Paleohydrology: Interpreting the Hydrogen-Isotopic Composition of Lipid
Biomarkers from Photosynthesizing Organisms, 40(1). https://doi.org/10.1146/annurev-
earth-042711-105535
Tamalavage, A. E., van Hengstum, P. J., Louchouarn, P., Fall, P. L., Donnelly, J. P., Albury, N.
305
A., et al. (2020). Plant wax evidence for precipitation and vegetation change from a coastal
sinkhole lake in the Bahamas spanning the last 3000 years. Organic Geochemistry, 150,
104120. https://doi.org/10.1016/J.ORGGEOCHEM.2020.104120
Tipple, B. J., & Pagani, M. (2010). A 35 Myr North American leaf-wax compound-specific
carbon and hydrogen isotope record: Implications for C4 grasslands and hydrologic cycle
dynamics. Earth and Planetary Science Letters, 299(1–2), 250–262.
https://doi.org/10.1016/J.EPSL.2010.09.006
Tipple, B. J., & Pagani, M. (2013). Environmental control on eastern broadleaf forest
speciesâ€
TM
leaf wax distributions and D/H ratios. Geochimica et Cosmochimica Acta, 111,
64–77. https://doi.org/10.1016/j.gca.2012.10.042
Turich, C., & Freeman, K. H. (2011). Archaeal lipids record paleosalinity in hypersaline
systems. Organic Geochemistry, 42(9), 1147–1157.
https://doi.org/10.1016/J.ORGGEOCHEM.2011.06.002
Williams, D. G., & Ehleringer, J. R. (2000). Intra- and Interspecific Variation for Summer
Precipitation Use in Pinyon-Juniper Woodlands. Ecological Monographs, 70(4), 517.
https://doi.org/10.2307/2657185
Windler, G., Tierney, J. E., Zhu, J., & Poulsen, C. J. (2020). Unraveling Glacial Hydroclimate in
the Indo-Pacific Warm Pool: Perspectives From Water Isotopes. Paleoceanography and
Paleoclimatology, 35(12), e2020PA003985. https://doi.org/10.1029/2020PA003985
Woolfenden, W. B. (2003). A 180,000-year pollen record from Owens Lake, CA: Terrestrial
vegetation change on orbital sales. Quaternary Research, 59(3), 430–444.
https://doi.org/10.1016/S0033-5894(03)00033-4
Xia, L., Cao, J., Stüeken, E. E., Zhi, D., Wang, T., & Li, W. (2020). Unsynchronized evolution
306
of salinity and pH of a Permian alkaline lake influenced by hydrothermal fluids: A multi-
proxy geochemical study. Chemical Geology, 541, 119581.
https://doi.org/10.1016/J.CHEMGEO.2020.119581
Yu, K., D’odorico, P., Collins, S. L., Carr, D., Porporato, A., Anderegg, W. R. L., et al. (2019).
The competitive advantage of a constitutive CAM species over a C4 grass species under
drought and CO2 enrichment. Ecosphere, 10(5), e02721.
https://doi.org/10.1002/ECS2.2721
307
Appendix. D Supplementary information for Chapter 5
GDGT indices
We calculate the branched isoprenoid tetraether (BIT) index:
𝐵𝐼𝑇 =
𝐼𝑎 +𝐼𝐼𝑎 +𝐼𝐼 𝑎 ′
+𝐼𝐼𝐼𝑎 +𝐼𝐼𝐼 𝑎 ′
𝐼𝑎 +𝐼𝐼𝑎 +𝐼𝐼 𝑎 ′
+𝐼𝐼𝐼𝑎 +𝐼𝐼𝐼 𝑎 ′
+𝑐𝑟𝑒 (1)
where brGDGTs Ia IIa and IIIa, including both the 5´ and 6´ methyl isomers, are compared with
the abundance of crenarchaeol (Hopmans et al., 2004). In lakes, BIT has traditionally been
interpreted to represent the balance between soil inputs of brGDGTs and lake production of
crenarchaeol (e.g., Verschuren et al., 2009). However, interpretations may differ as bacterial
production may dominate in many lakes, and changes in oxycline depth may control the
abundance of creanarchaeol-producing Thaumarchaeota (Baxter et al., 2021). As an additional
measure of lake stratification, we calculate the methane index (MI), which records the relative
contribution of methanotrophic Euryarchaeota versus Thaumarchaeota (Zhang et al., 2011),
where:
The temperature sensitive MBT′
5Me
index is the relative methylation of the 5′ isomers of the
brGDGTs (De Jonge et al., 2014, Hopmans et al., 2016), is expressed as:
𝑀𝐵𝑇 ´
5𝑀𝐸
=
(𝐼𝑎 +𝐼𝑏 +𝐼𝑐 )
(𝐼𝑎 + 𝐼𝑏 + 𝐼𝑐 +𝐼𝐼𝑎 +𝐼𝐼𝑏 +𝐼𝐼𝑐 +𝐼𝐼𝐼𝑎 )
(2)
The Type I, II and III brGDGTs have four, five, and six methyl groups, respectively, and the
Type a, b, and c brGDGTs have zero, one, and two rings, respectively. Duplicate analyses and
analyses of an internal laboratory standard throughout the runs yielded an error of 0.009
MBT´
5Me
units (1 ). To convert MBT´
5Me
to temperature we use the Bayesian BayMBT
0
model
which was generated by calibrating MBT´
5Me
against the mean temperature of the months above
freezing from a global lake dataset (Martínez-Sosa et al., 2021), including lakes over a range of
pH (4.3 to 10), salinity (0–275 psu) and temperature (1.6 to 28.1°C).
We calculate IR
6+7Me
an index sensitive to changes in lake salinity (Wang et al., 2021):
308
𝐼𝑅
6+7𝑀𝑒
= [
𝐼𝐼 𝑎 ′
+𝐼𝐼 𝑏 ′
+𝐼𝐼 𝑐 ′
+𝐼𝐼𝐼 𝑎 ′
+𝐼𝐼𝐼 𝑏 ′
+𝐼𝐼𝐼 𝑐 ′
𝐼𝐼𝑎 +𝐼𝐼𝑏 +𝐼𝐼𝑐 +𝐼𝐼𝐼𝑎 +𝐼𝐼𝐼𝑏 +𝐼𝐼𝐼𝑐 +𝐼𝐼 𝑎 ′
+𝐼𝐼 𝑏 ′
+𝐼𝐼 𝑐 ′
+𝐼𝐼𝐼 𝑎 ′
+𝐼𝐼𝐼 𝑏 ′
+𝐼𝐼𝐼 𝑐 ′
+
𝐼𝐼𝐼 𝑎 ′′′
+𝐼𝐼 𝑎 ′′′
𝐼𝐼𝐼𝑎 +𝐼𝐼𝐼 𝑎 ′
+𝐼𝐼𝐼 𝑎 ′′′
+𝐼𝐼𝑎 +𝐼𝐼 𝑎 ′
+𝐼𝐼 𝑎 ′′′
] × 0.5
(3)
Fig. D.1 Lake proxy reconstructions for the Late Pliocene. A) Summary of lake depth from Lake
Andrei (Knott et al., 2021), Lake Manly (Knott et al., 2018), and Searles Lake (Smith et al.,
1983). Searles Lake proxy reconstructions from core KM3 (this study) including: B) ACE index
of salinity, C) IR
6+7me
index of salinity, D) BayMBT
0
temperature reconstruction of mean air
temperature for months above freezing, E) Branched and Isoprenoid Tetraether (BIT) index, F)
∑C
22
-
32
alkanoic acid concentration, G) δ
13
C value of C
30
alkanoic acid and H) δD value of
precipitation.
309
Appendix. E Supplementary information for Chapter 6
Fig. E.1. Location and stratigraphic columns of sites within Woranso-Mille. Numbers represent field
codes of biomarker samples, and black circles represent biomarker sample stratigraphic position. Sample
codes were assigned sequentially, with the suffix “WM2018”. Samples with an adjacent * indicate that a
310
paired pedogenic carbonate sample was collected from the same location. Bones represent hominin fossil
stratigraphic position. Y-axis represents depth (m) and x-axis represents lithology (c = clay, z = silt, s =
sand and g = gravel). Woranso-Mille sampling site codes are as follows: BRT = Burtele, NFR =
Nefuraytu, LDD = Leado Dido’a, KSD = Korsi Dora, MSD = Mesgid Dora, MRD = Miro Dora, LHG =
Lahaysule Gera, KSA = Kuserale Dora, GUG = Gugubsi, MRV = Mille River Valley. Sites illustrated
using published stratigraphic and paleontological information: MRD (Saylor et al., 2019), NFR (Haile-
Selassie et al., 2016b), LDD (Melillo et al., 2021), BRT (Haile-Selassie et al., 2012, 2015), KSD (Deino
et al., 2010).
Abstract (if available)
Abstract
Anthropogenic emissions of CO2 are leading to global warming, with Earth's surface temperatures rising to levels unseen since the Pliocene, 3 million years ago. This temperature rise will impact precipitation patterns and induce large-scale environmental changes that will challenge human societies. However, predicting the impacts of these changes is difficult and relies primarily on models. In some regions, including southern California, models disagree on whether the climate will become wetter or drier as a result of this increase in temperature. Another approach to determining future changes is to look back at past periods of Earth’s history, with similar atmospheric CO2 concentrations and Earth surface temperatures as we will likely experience soon. The previous interglacial (130 – 115 kyr) and the Pliocene (5.3 – 2.6 Ma) are two periods with such conditions. Paleoclimatic reconstructions of these periods could provide insight for predicting future climate change impacts. To reconstruct the paleoclimate during these periods, I used biomarker proxies sensitive to temperature changes, precipitation, and landscape vegetation composition. Chapter 2 introduces a machine learning approach to gain information about vegetation change encoded in the molecular distribution of plant waxes preserved in ancient lake sediments from Searles Lake, California. I compare my vegetation reconstruction to an independent proxy of lake salinity and find a correlation between landscape desert plant abundance and high lake salinity, thereby validating my vegetation modeling approach.
In Chapter 3 I aim to understand the controlling factors on the distribution of glycerol dialkyl glycerol tetraethers (GDGTs) in a suite of arid and alkaline soils from the Serengeti National Park, Tanzania. GDGTs are produced by bacteria (branched-GDGTs) and archaea (isoprenoid-GDGTs) and act as lipid biomarkers sensitive to temperature, pH, and moisture changes. As we wish to apply the GDGT proxy to understand past climates, we must fully understand the process governing the distribution of different GDGT compounds in these modern soils. We find that brGDGTs in the surface soil are sensitive to changes in temperature. However, brGDGTs sampled deeper in the soil profiles were found not responsive to temperature and thus most likely represent compounds produced by a different bacterial community than is present in the surface. This study thus can guide future paleoenvironment studies focused on temperature reconstructions from ancient soil (paleosols).
In Chapter 4 I present a paleoclimatic reconstruction of southwestern North America from an expanded suite of biomarkers and pollen preserved in Searles Lake sediments between 200 – 5 ka. We find that the hydrogen isotopic composition of precipitation (δD of precipitation) increased during interglacial periods and decreased during glacial periods, which agrees with other regional records of precipitation isotopes. We also find that Searles Lake showed the largest changes in lake depth in association with terminal glacial phases, with termination 2 being characterized by a very deep overturning lake that overflowed into the adjacent basin. Crucially, we find that the previous interglacial was wetter than the present Holocene and a permanent lake was present in Searles Valley under a warmer climate. Chapter 5 presents biomarker data from Pliocene sediments deposited in Searles Valley. Previous sedimentological work indicated that a deep lake environment existed between 2.7 – 3.2 Ma in Searles Valley. Recent modeling studies suggest that this deep lake existed due to higher southern California coastal sea surface temperatures driving an increase in the North American Monsoon, which increased summer season precipitation. We tested this hypothesis using the hydrogen isotopic composition of our plant waxes. Moisture from the North American Monsoon region typically has a more positive δD, whereas winter precipitation from the North Pacific has a more negative δD. We find that the plant waxes in the Pliocene Searles Lake sediments are relatively positive, which is indicative of an increase in North American Monsoon precipitation, supporting the sea surface temperature “warmer = wetter” hypothesis.
Chapter 6 aims to both reconstruct the local paleoenvironments of early hominin species who lived in Woranso-Mille, Ethiopia, and to assess whether the paleoenvironment of Woranso-Mille was following regional trends forced through changes in paleoclimate or whether it was more sensitive to local tectonic and volcanic influences. By studying the isotopic composition of plant waxes preserved in sediments, we found that the early hominins were inhabiting mixed landscapes, including both open savannah and closed woodland environments. Interestingly, there was a shift to more C4 grass/shrub vegetation between the 3.8 – 3.2 Ma in Woranso-Mille which parallels the changes reported from a marine core record of terrestrial plant waxes offshore in the Arabian Gulf. Additionally, we find that an expansion of C4 plants in Woranso-Mille which occurred 3.8 – 3.6 Ma occurs at the same time as Hominin tooth enamel carbon isotopes shift to more positive values indicating an expansion of dietary breadth. This suggests the change in hominin diet was forced by a change in their habitat’s vegetation composition.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Antarctic climate variability from greenhouse to icehouse world
PDF
From tree tops to river runoff: tracing plant wax biomarkers across the Peruvian Andes and Amazon
PDF
Evolution of the Indian Monsoon and rise of C₄ photosynthesis in the Miocene and Pliocene
PDF
Isotopic fractionations in plant biomarker molecules with application to paleoclimate
PDF
Late Pleistocene changes in winter moisture source in the coastal southwest United States
PDF
Recent variability in the hyrdological cycle of tropical Asia from oxygen isotopes of tree celulose
PDF
Flowstone ideograms: deciphering the climate messages of Asian speleothems
PDF
Taking the temperature of the Common Era: statistics, patterns and dynamical insights
PDF
The geobiology of fluvial, lacustrine, and marginal marine carbonate microbialites (Pleistocene, Miocene, and Late Triassic) and their environmental significance
PDF
Seeing the future through the lens of the past: fusing paleoclimate observations and models
PDF
Great Salt Lake ooids: insights into rate of formation, potential as paleoenvironmental archives, and biogenicity
PDF
Ages, origins and biogeochemical role of water across a tropical mountain to floodplain transition
PDF
Germanium and silicon isotope geochemistry in terrestrial and marine low-temperature environments
PDF
Stromatolites as biosignatures and paleoenvironmental records: experiments with modern mats and examples from the Eocene Green River Formation
PDF
Green River formation stromatolites as a paleoclimate indicator: an investigation of the early Eocene climatic optimum through mass spectrometry, micro-X-ray fluorescence spectroscopy, and petrography
PDF
A multi-proxy high resolution record of paleohydrology from Zaca Lake, CA
PDF
Lithium isotopes as a carbon cycle proxy: examining the effects of high-temperature weathering, understanding brachiopod archives, and validating Cenozoic records
PDF
Critical zone response to perturbation: from mountain building to wildfire
PDF
Signs of photosynthesis on a neoproterozoic snowball Earth
PDF
Altitude effect on tree wood carbon isotopic composition in humid tropical forests
Asset Metadata
Creator
Peaple, Mark Donald
(author)
Core Title
From deep lakes to deserts: Plio-Pleistocene paleoenvironmental transitions
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Geological Sciences
Degree Conferral Date
2022-08
Publication Date
08/01/2022
Defense Date
05/06/2022
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
biomarkers,OAI-PMH Harvest,paleoclimate,Pleistocene,Pliocene
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Feakins, Sarah (
committee chair
), Carlson, Kristian (
committee member
), Corsetti, Frank (
committee member
)
Creator Email
05mpeaple@usc.edu,peaple@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-oUC111375877
Unique identifier
UC111375877
Legacy Identifier
etd-PeapleMark-11058
Document Type
Dissertation
Format
application/pdf (imt)
Rights
Peaple, Mark Donald
Type
texts
Source
20220801-usctheses-batch-965
(batch),
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Repository Email
cisadmin@lib.usc.edu
Tags
biomarkers
paleoclimate
Pleistocene
Pliocene