lable at ScienceDirect

Quaternary Science Reviews 87 (2014) 24e33

Contents lists avai

Quaternary Science Reviews

journal homepage: www.elsevier .com/locate/quascirev

Abundant C4 plants on the Tibetan Plateau during the Lateglacial andearly Holocene

Elizabeth K. Thomas a,*, Yongsong Huang a, Carrie Morrill b, Jiangtao Zhao c,Pamela Wegener a, Steven C. Clemens a, Steven M. Colman d, Li Gao a,e

aDepartment of Geological Sciences, Brown University, Box 1846, Providence, RI 02912, USAbCooperative Institute for Research in Environmental Sciences, University of Colorado and NOAA National Climatic Data Center, Boulder, CO 80305, USAc State Key Laboratory of Loess and Quaternary Geology, Institute of Earth Environment, Chinese Academy of Sciences, Xi’an 710075, Chinad Large Lake Observatory, University of Minnesota, Duluth, MN 55812, USAeGeoIsoChem Corporation, 738 Arrow Grand Circle, Covina, CA 91722, USA

a r t i c l e i n f o

Article history:Received 14 October 2013Received in revised form12 December 2013Accepted 17 December 2013Available online 24 January 2014

Keywords:Leaf waxCarbon isotopeC4 photosynthesisPaleoecologyTibetan PlateauMonsoon

* Corresponding author. Tel.: þ1 716 417 3062.E-mail address: elizabeth_thomas@brown.edu (E.K

0277-3791/$ e see front matter � 2013 Elsevier Ltd.http://dx.doi.org/10.1016/j.quascirev.2013.12.014

a b s t r a c t

Plants using the C4 (Hatch-Slack) photosynthetic pathway are key for global food production and accountfor ca 25% of terrestrial primary productivity, mostly in relatively warm, dry regions. The discovery ofmodern naturally-occurring C4 plant species at elevations up to 4500 m in Tibet and 3000 m in Africa andSouth America, however, suggests that C4 plants are present in a wider range of environments thanpreviously thought. Environmental conditions on the Tibetan Plateau, including high irradiance, rainfallfocused in summer, and saline soils, can favor C4 plants by offsetting the deleterious effects of lowgrowing season temperature. We present evidence based on leaf wax carbon isotope ratios from LakeQinghai that C4 plants accounted for 50% of terrestrial primary productivity on the northeastern TibetanPlateau throughout the Lateglacial and early Holocene. Despite cold conditions, C4 plants flourished dueto a combination of factors, including maximum summer insolation, pCO2 ca 250 ppmv, and sufficientsummer precipitation. The modern C3 plant-dominated ecosystem around Lake Qinghai was establishedca 6 thousand years ago as pCO2 increased and summer temperature and precipitation decreased. C4

plants were also intermittently abundant during the Last Glacial period; we propose that C4 plantscontributed a significant portion of local primary productivity by colonizing the exposed, saline QinghaiLake bed during low stands. Our results contrast with state-of-the-art ecosystem models that simulate<0.5% C4 plant abundance on the Tibetan Plateau in modern and past environments. The past abundanceof C4 plants on the Tibetan Plateau suggests a wider temperature range for C4 plants than can be inferredfrom modern distributions and model simulations, and provides paleoecological evidence to supportrecent findings that C4 plant evolution and distribution was determined by a combination of climatic andenvironmental factors (temperature, irradiance, precipitation amount and seasonality, and soil salinity).Moreover, this finding highlights the exceptional sensitivity of high-elevation ecosystems to environ-mental change, and provides critical benchmarks for ecosystem model validation.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Under present atmospheric CO2 concentrations, C4 plants areminor or absent in cool, high-elevation environments (Ehleringeret al., 1997; Collatz et al., 1998; Sage et al., 1999; Edwards andSmith, 2010). The modern distribution of C4 plants is likely aresult of preferential adoption of C4 photosynthesis by cladesadapted to warm, mesophytic conditions (>22 �C mean growing

. Thomas).

All rights reserved.

season temperature, ca 1200 mm yr�1 rainfall), where the C4photosynthetic pathway imparts a competitive advantage (Collatzet al., 1998; Edwards and Smith, 2010). Ecosystem models incor-porate a 22 �C minimum mean monthly temperature requirementfor C4 plants in addition to a 25 mm month�1 minimum precipi-tation threshold for all plants (Collatz et al., 1998; Still et al., 2003).Therefore, these models reconstruct few C4 plants at latitudes>50�

or elevations >2500 m asl (e.g., Fig. 1), which may result in un-derestimates of global primary production (Still et al., 2003).

The existence of>50 C4 plant species on the Tibetan Plateau andother high-elevation regions challenges this 22 �C threshold crite-rion (Fig. S1; Tieszen et al., 1979; Cavagnaro, 1988; Wang et al.,

Fig. 1. Location map. Lake Qinghai (black) and catchment (dark gray), Tibetan Plateau(light gray), modern boundary of 5% C4 plants (thick black line; Still et al., 2003), andsites mentioned in text (white dots; H, Huanxian; W, Weinan (Liu et al., 2005); X,Xunyi (Zhang et al., 2003); M, Mangshan (Peterse et al., 2011)). C4 plants abundance isvariable and heterogenous, ranging from 0 to 80% in Southeast Asia, the Philippines,and Indonesia (Still et al., 2003). Inset: C4 grass abundance at 9 ka simulated with theCCSM3 transient simulation (Z. Liu et al., 2009), showing no C4 grasses on the TibetanPlateau, but consistent minor C4 grass abundance in Southeast Asia, the Philippines,and Indonesia (white; see Fig. S5 for more time slices).

E.K. Thomas et al. / Quaternary Science Reviews 87 (2014) 24e33 25

2004) and indicates that C4 plants can successfully compete with C3plants at high-elevations despite cold conditions, given otherenvironmental conditions that favor C4 plants (Pyankov et al.,2000). Such conditions include 20% higher irradiance than at sealevel, >50% of rainfall falling in summer, and high salinity soils (upto 10 ppt; Baogen et al., 2013). Furthermore, the evolution of C4photosynthesis in cold-, arid-, salt-adapted clades, such as Cheno-podiaceae sensu stricto (s.s.), eudicotyledonous plants that areabundant in Asian deserts, may eliminate strict temperature limi-tations for some C4 plants (Pyankov et al., 2000; Kadereit et al.,2012).

The Tibetan Plateau and other high-elevation regions experi-ence amplified responses to global climatic change due to cloud-and snow-albedo feedbacks (X. Liu et al., 2009; Diffenbaugh et al.,2012). High-elevation plant ecosystems must adapt to amplifiedclimatic change under already harsh (cold, arid) conditions, and aretherefore bellwethers of climatic and environmental change. Re-constructions of plant ecosystem variability during climate regimesdifferent than today elucidate mechanisms that cause plantecosystem change, and provide information to improve ecosystemmodel accuracy in sensitive high-elevation regions.

We present a sub-millennial-resolution 32-ka-long reconstruc-tion of C4 plant abundance on the northeastern Tibetan Plateaubased on d13C of leaf wax C28 n-acids (d13Cwax) from Lake Qinghai(37 �N, 100 �E, 3194 m above sea level; Fig. 1). We augment ourfindings with a 16-ka-long pollen record from Lake Qinghai (Ji et al.,2005); a detailed interpretation of the 32-ka-long pollen record isin preparation (S. Xue, personal communication). Our work buildson research by An et al. (2012), who developed an age model andinitial climate reconstructions from the Lake Qinghai drill cores.Wecompare our C4 plant abundance reconstruction with existingclimate, environmental, and ecological records.We use output froma 21-ka-long transient simulation of the Community Climate Sys-tem Model 3 (CCSM3) to evaluate past seasonal temperature andprecipitation changes around Lake Qinghai.We also compare our C4

plant abundance reconstruction with the CCSM3 transient simu-lation of C4 plant abundance.

1.1. Setting

Lake Qinghai is well suited for evaluating average landscapeproperties because æolian and fluvial processes deliver leaf waxesto the closed-basin lake from a large catchment (30,000 km2) athigh-elevations (3194 to >4000 m above sea level; An et al., 2012).Today, Lake Qinghai’s catchment contains temperate and alpinesteppe vegetation, dominated by C3 Poaceae and Artemisia (Chenand Peng, 1993; Herzschuh et al., 2009). Mean annual precipita-tion is 370 mm, with 60% falling in June through August; meanannual evaporation is >1000 mm (“China Meteorological DataSharing Service System,” 2012). The average temperature for July,the warmest month, is 16 �C.

In the Lake Qinghai catchment, May is the first month withaverage surface temperatures above 0 �C, so is the month when C3plants begin to grow, if there is enough precipitation and/or snowmelt. In contrast, C4 plants thrive during the summer, when tem-perature is highest, but only when there is sufficient moisture. Weuse these basic guidelines to interpret our record of C4 plantabundance.

1.2. Determining past C4 plant abundance

Different enzyme-mediated reactions in the C3 and C4 photo-synthetic pathways impart distinct carbon isotope compositions onplant biomass: C4 plants are ca 15& enriched in 13C relative to C3plants (O’Leary, 1981). These distinct d13C compositions are alsoreflected in leaf waxes (Chikaraishi and Naraoka, 2007; Rao et al.,2008). Leaf waxes are n-alkyl lipids that form a protective layeron leaves; plants in arid regions produce particularly high abun-dances of leaf waxes (Shepherd and Griffiths, 2006). Leaf waxes areabraded from the leaf surface and carried to a lake basin by windand/or water and are preserved well in Quaternary lake sediments(Castañeda and Schouten, 2011). We use leaf wax isotopic end-member values for C3 and C4 plants, discussed below, to calculateC4 plant abundance.

2. Methods

2.1. Lipid biomarker extraction and purification

131 samples were obtained from ICDP drill cores LQDP051A andLQDP05 1F for organics analysis. The stratigraphy, age model, andinitial climate reconstructions for these drill cores are described inprevious work (An et al., 2012). Free lipids were extracted fromfreeze-dried samples with an Accelerated Solvent Extractor 200(Dionex) using dichloromethane (DCM):methanol 9:1 (v:v). N-alkanoic acids were prepared by first separating the TLEs on solidphase Aminopropyl silica gel flash columns using three columnvolumes of DCM:Isopropanol 2:1 (v:v) and 4% acetic acid in diethylether. The acid fraction was methylated in anhydrous 5% HCl inmethanol at 60 �C overnight. Hydroxyl-carboxylic acids wereremoved from the methylated acid fractions using silica gel flashcolumns with DCM as the eluant, after eluting apolar compoundswith hexane.

2.2. Carbon isotope analysis

The carbon isotope ratio of fatty acid methyl esters (FAMEs) wasdetermined on an HP 6890 gas chromatograph interfaced to aThermo Finnigan Delta Vplus isotope ratio mass spectrometerthrough a combustion reactor at Brown University. Isotope values

Fig. 2. Surface temperature, precipitation, and moisture balance (precipitation minusevaporation) time series from the CCSM3 transient simulation (Z. Liu et al., 2009). Eachtime series was produced from the average of four grid cells surrounding Lake Qinghai(Figs. S2 and S3). Average winter (DecembereFebruary) variables plotted in black; May,June, July, and August variables plotted individually. Simulated modern temperaturesare cooler than modern observations (“China Meteorological Data Sharing ServiceSystem,” 2012), likely because modern observations are point measurements andthe model temperatures are averaged over a large area. Nonetheless, the magnitude ofglacialeinterglacial-scale modeled temperature change is similar to reconstructedtemperature change on the CLP.

E.K. Thomas et al. / Quaternary Science Reviews 87 (2014) 24e3326

are expressed in per mille (&) relative to Vienna Pee Dee Belemnite(VPDB). The FAME d13C values were corrected for the isotopiccontribution of the carbon in the methyl group added duringmethylation. The analytical error, calculated using the results ofduplicate analyses of each extract, is 0.4& (1s) and an externalFAME standard, injected once after every sixth sample injection,had an analytical error of 0.7& (1s).

2.3. Transient simulation

The transient simulation of the past 21,000 years was carried outusing the low-resolution version (T31�3) of the CommunityClimate System Model, version 3 (CCSM3; Z. Liu et al., 2009; He,2011; Liu et al., 2012). CCSM3 is a global, ocean-atmosphere-seaice-land surface climate model coupled without flux corrections.The atmospheric component is the Community Atmosphere Modelversion 3 at a horizontal resolution of T31 (approximately 3.75� inlatitude and longitude) and 26 hybrid vertical levels. The landmodel, which predicts plant functional types using a dynamicvegetation scheme, uses the same grid as the atmosphericcomponent. The ocean model is an implementation of the ParallelOcean Program model with longitudinal resolution of 3.6� andvariable latitudinal resolution, with finer resolution in the NorthAtlantic and tropics (w0.9�). The vertical resolution is 25 levelsextending to a depth of 5.5 km. The dynamicethermodynamic seaice model uses the same horizontal grid and landmask as the oceancomponent. The T31�3 present-day control simulation reproducesthe major features of global climate, with improved ocean circu-lation compared to earlier versions of this model (Yeager et al.,2006). Results from the higher-resolution (T42�1) version ofCCSM3 for Last Glacial Maximum and middle Holocene conditionscapture important paleoclimate patterns documented by proxyrecords (Otto-Bliesner et al., 2006).

Detailed information about the forcings used in the transientexperiment can be found in He (2011), Liu et al. (2012), and Z. Liuet al. (2009). Briefly, orbital forcing varied transiently (Berger,1978), and greenhouse gas (CO2, CH4, and N2O) concentrationswere based on ice core measurements (Joos and Spahni, 2008).Continental ice sheets were based on the ICE-5G reconstruction(Peltier, 2004) and altered approximately once per thousand years.Meltwater forcing (MWF) prior to 14.5 kawas the DGL-A scheme (Z.Liu et al., 2009), which includes an abrupt termination at theBolling-Allerod. Following this, major periods of MWF were pre-scribed from 14.4 to 13.9 ka to the North Atlantic and SouthernOcean and from 12.9 to 11.7 ka and from 9.0 to 8.2 ka to the NorthAtlantic (He, 2011; Liu et al., 2012).

To generate time series of climate variables from the transientsimulation for the Lake Qinghai area, we averaged the four closestgrid cells to the lake (Fig. 2). This choice was made to average outrandom fluctuations from individual grid cells while accounting forthe relatively low resolution of themodel and the large topographicgradients in this region. The transient evolution of temperature andprecipitation for these four grid cells is representative ofmonsoonal Asia as a whole (Figs. 2, S2 and S3).

2.4. CMIP5 model results

To provide an estimate of conditions that the Tibetan Plateauplant ecosystem will be experiencing in ca 2050 AD, we extractedtemperature and pCO2 data for the Tibetan Plateau from CoupledModel Intercomparison Project Phase 5 (CMIP5) model projections.Temperature and precipitation anomalies for the Lake Qinghai areaat ca 2050 ADwere calculated from CMIP5model projections, usingthe Representative Concentration Pathways (RCP) 2.6 and 8.5(years 2040e2059 AD) relative to historical climate simulations

(years 1986e2005 AD). We used all models in the CMIP5 archive asof May 2013 that provided the information necessary for ouranalysis (Table S1). For models providing ensembles for the RCPs,we analyzed just one ensemble member given the expectation thatdifferences among the models are typically larger than differencesbetween ensemble members. The grid resolutions vary among themodels and to ensure that we consistently averaged the samespatial area we first bilinearly interpolated results from all modelsto the CCSM4 grid. Then, we calculated the area average for a3� � 3� box centered on the location of Lake Qinghai (37 �N,100 �E).

3. C28 n-acid carbon isotopes as a proxy for C4 plantabundance

3.1. Leaf wax sources

The most abundant n-acid chain lengths in Lake Qinghai sedi-ments are C26 and C28. Terrestrial plants in the modern LakeQinghai catchment produce mainly mid- to long-chain n-acids,with a peak at C24; C26 and C28 n-acids compose ca 10% each of totalterrestrial plant leaf wax production (Wang and Liu, 2012). Incontrast, aquatic plants produce mainly C16 n-acids. C26 is <10%,and C28 is<5% of total aquatic plant leaf wax production (Wang and

E.K. Thomas et al. / Quaternary Science Reviews 87 (2014) 24e33 27

Liu, 2012). Aquatic pollen is a minor component (<1% average) ofthe total pollen flux throughout the entire 32-ka record, exceptfrom 20.5 to 15.5 ka, when aquatic pollen averages 5%, and brieflypeaks at 65% at ca 18 ka (Fig. 3; Ji et al., 2005; S. Xue has generatedaquatic pollen data for the past 32 ka in Lake Qinghai and re-constructs pollen abundance up to 65% at 20.5, 18, and 15.5 ka,personal communication). Terrestrial plants are therefore thedominant source of long-chain n-acids, with the possible exception

Fig. 3. Lake Qinghai d13Cwax compared to local precipitation and pollen reconstructions. A.and gray bars are 14C age constraints, shownwith 1s calibrated range; gray arrows and numb135-year reservoir correction; black stars indicate depths of 14C ages that are >20 ka older thabundance and uncertainty calculated using likely range of C3 and C4 plant end member 13Cw

ostracode d18O (Liu et al., 2007), age model adjusted to match age model from ref. (An et alanalytical error is �0.2&). D. Percent carbonate and flux of large grains in Lake Qinghai scipitation, the increase at 11.5 ka is likely due to a sedimentological transition as well as a prlikely greater winter precipitation. E. Percent Poaceae, Chenopodiaceae s.s., tree, and aquaticfrom Ji et al. (2005) adjusted to match age model for our samples by correlating percent totaEvents 1e3, defined by 18O-enriched intervals in the Chinese speleothem records (Wang et

of brief intervals between 20.5 and 15.5 ka. We report d13C of theC28 n-acid because aquatic plants produce fewer C28 n-acids, andare less likely to contaminate the terrestrial signal.

3.2. Carbon isotope end members

In order to convert d13Cwax to C4 plant abundance, we mustdetermine reasonable C3 and C4 plant C28 n-acid d13C values for the

Qinghai d13Cwax. Error bars on data points are 1s analytical precision. Horizontal blackers indicate reservoir correction applied to ages of the same shade, Holocene ages had aan their depositional age based on this age model (An et al., 2012). B. Inferred C4 plantax values (solid black line and gray range). Thin black line shows 0% C4. C. Lake Qinghai

., 2012). Negative values in the early Holocene indicate more summer precipitation (2sediments (An et al., 2012). Higher %CaCO3 in the Holocene likely indicates more pre-ecipitation increase. Larger grain size flux corresponds to stronger westerly winds, andpollen, and pollen concentration in Lake Qinghai sediments (Ji et al., 2005). Age modell organic carbon from the two records. Gray bars highlight Younger Dryas and Heinrichal., 2008).

E.K. Thomas et al. / Quaternary Science Reviews 87 (2014) 24e3328

Tibetan Plateau. We are not aware of any published modern C28 n-acid d13C values for this region. We use C28 n-acid d13C values for C3and C4 plants at lower elevations, adjusted for known effects ofelevation and aridity, to determine end member values for the Ti-betan Plateau.

3.2.1. C3 plant d13Cwax end memberThe d13C of C28 fatty acid in C3 trees and herbs from Thailand and

Japan, where mean annual precipitation is ca 1500 mm,is �36.9 � 1.7& (Chikaraishi and Naraoka, 2007). When C3 plantsare drought-stressed, as is the case for plants on the northeasternTibetan Plateau, their biomass becomes enriched in 13C by up to 3&relative to non drought-stressed C3 plants (Passey et al., 2002). Acompilation of carbon isotope fractionation during CO2 uptake andfixation (Dleaf) by trees throughout the world shows that trees indry regions have lower Dleaf, and therefore higher d13C, than trees inwet regions, even within a single genus (Diefendorf et al., 2010). C3plants at higher elevations also have lower Dleaf, perhaps due todecreased atmospheric pressure (Li et al., 2007; Diefendorf et al.,2010; Zhou et al., 2011). Assuming that this relationship appliesto grasses and herbs as well as trees, we apply an empiricallyderived equation for leaf d13Cwax enrichment: Dleaf ¼ 4.20(�0.26)*log10(MAP) � 0.06(�0.01)*Oaltitude þ 9.31(�0.90), MAP ¼ meanannual precipitation in mm, (Diefendorf et al., 2010) using anaverage catchmentMAP of 370mmand an elevation of 3500m. Thecombination of aridity and high-elevation at Lake Qinghai wouldcause up to 5.3 � 2.2& enrichment of C3 plant d13Cwax compared toplants in Japan and Thailand (Chikaraishi and Naraoka, 2007;Diefendorf et al., 2010). The modern landscape around Lake Qing-hai is dominated by C3 plants, and the minimum late Holocened13Cwax in Lake Qinghai sediments is �32.9&. This is 4& enrichedin 13C relative to d13Cwax of C3 trees and herbs in Japan and Thailand,close to the predicted 5.3& enrichment. We thereforeuse�33� 2& for the C3 plant d13Cwax endmember at Lake Qinghai.

The glacial period was drier than the Holocene at Lake Qinghai(Colman et al., 2007; An et al., 2012), likely causing C3 plant d13C tobe more enriched. Although we do not have quantitative estimatesof precipitation amount, a site with 100 mm growing season pre-cipitation would contain C3 plants 2& enriched relative to modernC3 plants at Lake Qinghai using the equation above (Diefendorfet al., 2010). Thus, the glacial C3 plant d13Cwax end member valuemay have been closer to �31&, which would decrease our calcu-lated C4 plant abundance during the glacial period by 5e10%(Fig. 3B). Note that the �31& C3 plant d13Cwax end member resultsin negative C4 plant abundance during the late Holocene, and istherefore likely only applicable to the glacial period.

3.2.2. C4 plant d13Cwax end memberThe d13C of C28 fatty acid in C4 plants from Thailand and Japan,

where mean annual precipitation is ca 1500 mm, is �21.4 � 2.1&(Chikaraishi and Naraoka, 2007). Unlike C3 plants, drought-stressedC4 plants do not exhibit changes in biomass d13C (Passey et al.,2002). Drier conditions on the Tibetan Plateau, or changes in pre-cipitation amount between the glacial and Holocene periods likelydo not affect the C4 plant d13Cwax end member. There is evidencethat C4 Chenopodiaceae s.s., a dominant form of C4 plants in centralAsian deserts, have lower Dleaf at high-elevations (0.1& per 100 m;Pyankov et al., 2000; Van de Water et al., 2002). At 3500 m on thenortheastern Tibetan Plateau, the C4 plant d13Cwax end membervalue would be �18&, ca 3& enriched in 13C relative to C4 plantd13Cwax at low elevations (Chikaraishi and Naraoka, 2007).

3.2.3. Aquatic plant d13Cwax end member at Lake QinghaiAquatic plants are intermittently abundant in the Lake Qinghai

record (Fig. 3; Ji et al., 2005). Aquatic pollen in the Lake Qinghai

record is mainly Myriophyllum sp. with minor amounts of Pota-mogeton sp. and Sparganium sp. Carbon isotopes of bulk Potamo-geton sp. material from Lake Qinghai are �17.8 to �15.9& (W. Liuet al., 2013). Cladophora sp., aquatic phytoplankton that do notproduce pollen and produce only a small relative proportion of longchain n-acids, have bulk carbon isotope values of �33.6 to �28.6&(Wang and Liu, 2012; W. Liu et al., 2013). C28 n-acid carbon isotopestend to be depleted relative to bulk values, and therefore the bulkvalues cited here are maxima for aquatic d13Cwax end members. Thestudies of plants at Lake Qinghai do not provide informationregarding the concentration of n-acids produced by aquatic andterrestrial plants, although other research demonstrates thatterrestrial plants produce more n-acids than aquatic plants (Gaoet al., 2011). Aquatic plants are therefore unlikely to have contrib-uted significant amounts of n-acids to the Lake Qinghai sedimentsthroughout most of the record. One exception to this is the intervalat 18 ka, when aquatic pollen is >65% (S. Xue, personal commu-nication). Depleted d13Cwax values, ca �34&, also occur in this in-terval, and may be a result of increased aquatic plant productivity,particularly Cladophora sp., which has depleted bulk d13C values.

3.3. C4 plant abundance, error estimates, and alternateinterpretations

Lake Qinghai d13Cwax ranges from �32.9 to �21.9& (Fig. 3).d13Cwax was most 13C-depleted from 32 to 21 ka and 6 to 0 ka andmost 13C-enriched during the Lateglacial and early Holocene (13.7e8.3 ka; Fig. 3). This 10& range is most likely a result of changing C3and C4 plant biomass in the Lake Qinghai catchment. We thereforeuse the isotopic end members, �33 � 2& for C3 plants and a rangefrom �21 to �18& for C4 plants, and the Lake Qinghai d13Cwax re-cord to calculate C4 plant abundance from 32 ka to the Late Holo-cene (Fig. 3B; black line end members: �33& for C3 and �18& forC4 plants). C3 plants were dominant (>80%) during the glacialperiod, with two intervals of abrupt C4 plant expansion. During theLateglacial and early Holocene, C4 plants were abundant (ca 50%) inthe Lake Qinghai catchment. The modern C3-dominated ecosystemdeveloped ca 6 ka. Due to uncertainty in our end member esti-mates, we present a range of possible C4 plant abundance (Fig. 3B).

We stress that aridity-driven changes in C3 plant 13CO2discrimination would cause d13Cwax variations no greater thanca 3&. If aridity were the main cause of changes in Lake Qinghaid13Cwax, the dry late Holocene would be more enriched in 13C thanthe wet early Holocene (Passey et al., 2002; Wang et al., 2003;Diefendorf et al., 2010; An et al., 2012; Dietze et al., 2013; Hudsonand Quade, 2013). This is opposite the trend that we observe(Fig. 3). Furthermore, CAM plants, such as Crassulaceae and Isoetes,are rare in this part of the Tibetan Plateau, in both modern surveysand in the pollen record, and so would likely not influence d13Cwax(Chen and Peng, 1993; Ji et al., 2005; Herzschuh et al., 2010). Thus,we interpret the Lake Qinghai d13Cwax record as indicating changesin terrestrial plant ecology, specifically changes in C3 and C4 plantbiomass.

4. Discussion

We propose that two distinct sets of environmental conditionscaused C4 plants to flourish in the Lake Qinghai catchment: 1. From21 to 15 ka: C3 plants dominated, but arid conditions caused LakeQinghai level to drop, exposing the lake bed, which was colonizedby early-succession, cold- and salt-tolerant C4 Chenopodiaceae s.s.and, 2. During the Lateglacial and early Holocene: moderate pCO2and warm, wet summers caused C4 Chenopodiaceae s.s. and Poa-ceae to flourish throughout northeastern Tibet (Fig. 4). We usethese boundary conditions to incorporate precipitation into the

Fig. 4. Qinghai d13Cwax record compared to other climate and environmental reconstructions. A. Qinghai d13Cwax. Maximum summer insolation (gray; Laskar et al., 2004). B. Ages ofglacier maxima in the Dalijia Mountains (J. Wang et al., 2013). C. Atmospheric carbon dioxide concentrations from Antarctic ice cores (Jouzel et al., 2007). D. Mean annual airtemperature on the CLP (Peterse et al., 2011; see SI for calibration details; 1s analytical error is �0.2 �C). E. Total organic carbon (TOC) d13C fromWeinan and Huanxian loess sections(Liu et al., 2005) and from Lake Rutundu in Kenya (Wooller et al., 2003), and d13Cwax from Xunyi loess section (Zhang et al., 2003), note different y-axis scales. Inferred C4 abundancefrom the original publications (1s analytical error is �0.1e0.3&). F. Speleothem d18O from Sanbao and Hulu caves (Wang et al., 2008). Gray bars as in Fig. 3.

E.K. Thomas et al. / Quaternary Science Reviews 87 (2014) 24e33 29

model by Ehleringer et al. (1997) that predicts the CO2 and tem-perature conditions that favor C3 or C4 plants (Fig. 5, curvesrepresentative for the Tibetan Plateau). The boundary conditionsbetween C3 and C4 plants calculated by Ehleringer et al. (1997)cannot explain the changes in C4 plant abundance that we recon-struct at Lake Qinghai. We therefore constructed a schematic rep-resentation of the effects of precipitation (Fig. 5). Therein, isohyetsrepresent dry (ca 100 mm/growing season; Last Glacial), wet(ca 330 mm/growing season; early Holocene), and moderate pre-cipitation conditions (ca 220 mm/growing season; modern) andserve as boundaries between conditions that favor C3 and C4 plants.The isohyets were drawn to have slopes similar to those calculatedby Ehleringer et al. (1997), and were positioned to represent a bestfit to our %C4 plant results. Our results support recent findings thatthe conditions that favor C4 plant productivity shift not only inresponse to temperature and pCO2, but also in response to growing

season precipitation (Edwards and Still, 2008; Edwards and Smith,2010).

4.1. Plant ecosystems in the Lake Qinghai catchment during theglacial period

During the glacial period (32e14 ka), Lake Qinghai d13Cwax

ranged from �34.0 to �22.5&, indicating C3 plants were dominant(Fig. 3). Overall primary productivity was low from 32 to 15 ka, asindicated by low pollen counts (Fig. 3) and multiple bulk organicgeochemical proxies (X. Liu et al., 2013). Mean annual temperaturereconstructed on the Chinese Loess Plateau (CLP; Peterse et al.,2011) and summer temperature at Lake Qinghai in the transientsimulation of the CCSM3 (Z. Liu et al., 2009) were 3e6 �C lower thanpresent (Figs. 2e4, and S2; see SI). Precipitation, reconstructedusing carbonate in Lake Qinghai sediments and simulated by

Fig. 5. Schematic diagram illustrating environmental conditions (temperature, pCO2

and precipitation) controlling plant ecosystems on the Tibetan Plateau. Increasedprecipitation causes C4 plants to be favored under cooler and higher pCO2 conditions,as shown by isohyets based on data from this paper: wet isohyet (green), moderateisohyet (gray), dry isohyet (brown). Temperature and pCO2 axes and slopes of isohyetboundaries based on Ehleringer et al. (1997), and isohyets are drawn to yield a best fitto our %C4 plant results. “Moderate precipitation” corresponds to 220 mm in summer,based on modern instrumental data (“China Meteorological Data Sharing ServiceSystem,” 2012; all three conditions are arid: 50% more summer precipitation duringthe “wet” early Holocene would be ca 330 mm; 50% less summer precipitation duringthe “dry” LGM would be ca 110 mm; Fig. 2). Ellipses indicate conditions at time periodsin the Lake Qinghai d13Cwax record, outline color corresponds to precipitation condi-tions, fill corresponds to dominant photosynthetic pathway (C3, white; C4, yellow),ellipses cover the temperature range for that time period based on 20% temperatureamplification at Lake Qinghai relative to the CLP (X. Liu et al., 2009; Peterse et al., 2011;see SI). Black dots and bars are ranges of projected pCO2 and temperature at LakeQinghai in 2040e2059 AD from CMIP5 simulations using emissions scenarios RCP2.6and RCP8.5 (Taylor et al., 2011). (For interpretation of the references to colour in thisfigure legend, the reader is referred to the web version of this article.).

E.K. Thomas et al. / Quaternary Science Reviews 87 (2014) 24e3330

CCSM3, was low throughout the glacial period (Figs. 2 and 3, andS3; see SI; Z. Liu et al., 2009; An et al., 2012). Although these harshconditions were more favorable for C3 plants (Fig. 5), two distinctintervals exhibit rapid shifts to elevated C4 plant abundance: 21.0e20.0 ka (average 40%, range 20%e70%) and 17.5e15.8 ka (average35%, range 25%e60%). We interpret these peaks in d13Cwax as a localplant community response to changes in lake level during thecoldest and driest portion of the glacial period (23e15 ka; Figs. 3and 4).

4.1.1. C4 plants may have colonized dry, salty lake beds during theglacial period

The most likely reason for the large (up to 10&) intermittentincreases in d13Cwax from 21 to 15 ka (Fig. 3) is an increase in C4plant abundance. This is a surprising finding, however, because theconditions on the Tibetan Plateau during the glacial period wereextremely cold and dry, inhibiting most primary productivity, andcertainly not favorable to warm-adapted C4 plants (Figs. 3 and 5).We propose one possible interpretation of these brief d13Cwax in-creases: Lake Qinghai level dropped during the dry glacial period,so cold-, salt-, drought-tolerant C4 Chenopodiaceae s.s. colonizedthe dry, salty Lake Qinghai bed.

Summer monsoon precipitation was low during the glacialperiod, and westerly wind strength decreased starting ca 23 ka(Fig. 3; An et al., 2012), so total precipitation likely decreased,causing the Lake Qinghai level to drop. Modeled water balance(precipitation minus evaporation) was low during maximumglacial conditions, further evidence for enhanced aridity in the LakeQinghai catchment (Fig. 2). Submerged beach ridges and sandy-silt

layers that date to ca 19 ka indicate low levels at Lake Qinghai(Colman et al., 2007). Aquatic plant pollen is not readily transportedover long distances (Cook, 1988), so the presence of abundantaquatic plant pollen at the drill site from 21 to 15 ka suggests thataquatic plants were growing near the drill site (Fig. 3; Ji et al., 2005;S. Xue, personal communication). These plants generally root inwater that is <7 m deep (Sheldon and Boylen, 1977), so this pro-vides further evidence for lower Lake Qinghai levels. Other recordsat Lake Qinghai and elsewhere in Tibet indicate dry conditions andlower lake levels during the glacial period (Herzschuh et al., 2010;Mügler et al., 2010; An et al., 2012; X. Liu et al., 2013; Y. Wang et al.,2013). The Qinghai lake bed did not desiccate completely, however,as sedimentation was continuous with no evidence for subaerialexposure at the drill site, located at the depocenter of the lake (Anet al., 2012).

Glacier maxima in the Dalijia Mountains, 220 km southeast ofLake Qinghai, at 21.8 � 1.5 ka and 17.3 � 1.5 ka, are evidence ofmaximum cold conditions, corresponding to the lowest tempera-tures reconstructed on the CLP (Fig. 4; Peterse et al., 2011; J. Wanget al., 2013). Despite cold and arid conditions, low pCO2 would havefavored C4 plants (Figs. 4 and 5; Jouzel et al., 2007). Chenopodiaceaes.s. are early succession plants, and C4 photosynthesis is moreabundant in Chenopodiaceae s.s. lineages that exhibit salt-tolerance and succulence, both of which would have been criticaladaptations to surviving the arid glacial period on the TibetanPlateau (Pyankov et al., 2000; Wang, 2007; Kadereit et al., 2012).Chenopodiaceae s.s. pollen was abundant (20e40%) during theglacial d13Cwax peaks (Fig. 3; Ji et al., 2005; S. Xue, personalcommunication). We hypothesize that C4 Chenopodiaceae s.s.colonized the exposed, salty lake bed when Lake Qinghai levelsdropped during the glacial period. This increase in C4 Chenopo-diaceae s.s. would have been a local phenomenon, constrained toexposed lake beds (Fig. S1). This hypothesis can be tested in futurestudies: other desiccated lake beds may also show intermittentincreases in d13Cwax, whereas terrestrial sediment deposits (e.g.,loess) on the Tibetan Plateau would not contain evidence for thesed13Cwax increases.

4.2. Transition to the Lateglacial period

At ca 16 ka, the westerly winds strengthened, likely in responseto Heinrich Event 1 (H1; Sun et al., 2011; An et al., 2012). From 16 to14.5 ka, tree pollen increased in abundance and d13Cwax decreased,indicating that C3 plants dominated during H1 (Fig. 3; Ji et al.,2005). Immediately following H1, d13Cwax increased, but returnedto consistently low values (�32&) from 14.5 to 13.7 ka. Artemisiarepresented 80% of pollen from 14.5 to 13.7 ka, which wouldexplain the strong C3 plant signal in d13Cwax. Other proxies indicateincreasing temperature, increasing summer precipitation, andincreasing seasonal insolation contrast during this time period(Figs. 3 and 4, and S4), so it is unclear why the d13Cwax and pollensuggest stable ecological conditions from 14.5 to 13.7 ka.

4.3. C4 plants were abundant on the Tibetan Plateau during theLateglacial and early Holocene

During the Lateglacial and early Holocene (13.7e8.3 ka), LakeQinghai d13Cwax ranged from�29.5 to�21.8&, corresponding to anaverage of 50% C4 plants (Fig. 3). We hypothesize that this increasein C4 plant abundance was caused by a combination of environ-mental changes: high seasonal insolation contrast, moderate pCO2and warm conditions, and greater summer precipitation.

Winter insolation was low and summer insolation was high atLake Qinghai during the early Holocene, resulting in a seasonalinsolation contrast of ca 350 W m�2 (Fig. S4; Laskar et al., 2004).

E.K. Thomas et al. / Quaternary Science Reviews 87 (2014) 24e33 31

Cold winters and springs likely inhibited C3 plant growth, whereasshort, hot, wet summers favored C4 plant productivity (Ehleringeret al., 1997; Pyankov et al., 2000). Global pCO2 was 240e260 ppmv (Jouzel et al., 2007), values that favor C4 photosynthesisonly at higher temperatures (Fig. 5; Ehleringer et al., 1997). Late-glacial and early Holocene temperature was 1e3 �C higher thanpresent, according to a pollen record from Lake Luanhaizi in thenearby Qilian mountains, the CCSM3 transient simulation, and thetemperature reconstruction on the CLP (Figs. 2 and 4; Z. Liu et al.,2009; Herzschuh et al., 2010; Peterse et al., 2011). Precipitationincreased from 13.7 to 12 ka and peaked 12 to 8 ka, according tod18Oostracode from Lake Qinghai, the pollen record from Lake Luan-haizi, and the CCSM3 transient simulation (Figs. 2e4; Liu et al.,2007; Z. Liu et al., 2009; Herzschuh et al., 2010; An et al., 2012).Higher lake levels, suggesting wetter conditions, existedthroughout Tibet during the early Holocene, including at LakeDonggi Cona, 200 km southwest of Lake Qinghai (Dietze et al.,2013; Hudson and Quade, 2013).

Warmer conditions were favorable for primary production, asindicated by increased pollen flux at ca 14.8 ka and multiple bulkorganic geochemical proxies from Lake Qinghai (Fig. 3; An et al.,2012; X. Liu et al., 2013). Poaceae and Chenopodiaceae s.s. eachcomposed 10e20% of total pollen flux during the Lateglacial andearly Holocene, Chenopodiaceae s.s. pollen increased slightly dur-ing the Younger Dryas (Fig. 3; Ji et al., 2005). Tree pollen decreasedto 10% during the Younger Dryas, and increased abruptly to 20% atthe end of the Younger Dryas, when total pollen counts alsoincreased. Although the existing data do not distinguish C3 from C4plant pollen, the 10& increase in d13Cwax cannot be explained bychanges in C3 plant water stress, and so C4 plants must haveflourished during the Lateglacial and early Holocene. Relativelywarm, wet conditions, occurring at a time of maximum seasonalcontrast during the Lateglacial and early Holocene, would befavorable to both C4 Poaceae and C4 Chenopodiaceae s.s. This in-crease in C4 plant abundance may have resulted from a few extanttaxa that proliferated on the QinghaieTibetan Plateau during thisperiod, or may have resulted from C4 plant species shifting tohigher elevations as climate became more conducive to theirgrowth (Fig. S1). In either case, the Lateglacial and early Holoceneincrease in C4 plant abundance was driven by regional and globalenvironmental conditions, and therefore C4 plants were likelydominant throughout the region around Lake Qinghai.

4.4. Mid-Holocene development of a C3-plant-dominatedecosystem

The mid-Holocene d13Cwax decrease is mirrored by decreasingbulk organic matter d13C (X. Liu et al., 2013). Aquatic plants exertonly a minimal influence on d13Cwax, so we can more confidentlyconclude that this decrease in carbon isotope values is due to adecrease in C4 plant biomass. The initial decline of C4 plants duringthe mid-Holocene (8e6 ka) coincided with high counts of treepollen, as well as decreasing precipitation and insolation, as well asdecreasing seasonal contrast (Figs. 2e4, and S4; Laskar et al., 2004;Ji et al., 2005; Peterse et al., 2011; An et al., 2012). C3 plantsdominated the Lake Qinghai catchment after ca 6 ka, coincidentwith high pCO2 and modeled increasing May temperatures andwinter precipitation, which favored C3 plant productivity (Figs. 2e4; Still et al., 2003). Whereas arboreal pollen declined during thelate Holocene, likely due to decreased precipitation (Herzschuhet al., 2009), Chenopodiaceae s.s. and Poaceae remained constant,suggesting that C3 plants within these clades became dominant,replacing C4 Chenopodiaceae s.s. and Poaceae that had dominatedduring the early Holocene (Fig. 3). Today, high pCO2 means thatmost C4 plant communities are near the edge of their ecological

niche, especially those in cold regions (Fig. 5). The predicted 2e3 �Ctemperature increase on the Tibetan Plateau by 2050 ADwill not besufficient to offset the deleterious effects of 450e500 ppmv CO2,likely resulting in the demise of C4 plants in cold, dry, high-elevation regions (Fig. 5; Taylor et al., 2011).

4.5. C4 plants in high-elevation ecosystems

In contrast to the northeastern Tibetan Plateau, high-elevationsites in equatorial Africa and South America hosted flourishing C4plant communities during the glacial period. The different timing ofmaximum C4 plant abundance is likely due to different tempera-tures and seasonal contrasts. The equatorial sites had higher meanannual temperatures than the northeastern Tibetan Plateau duringthe glacial period (Figs. 4 and S4). Seasonal insolation contrast isminimal at the equator, ranging from 40 to 60 W m�2, with mini-mum insolation contrast occurring during the glacial period whenthere were maximum C4 plants (Fig. S4; Berger et al., 2006). Thus,seasonal insolation contrast may not play a dominant role in plantecosystem variability at the equator. Instead C4 plants flourishedduring the glacial period likely due to sufficiently warm rainyseasons and low pCO2 (Figs. 4 and S4; Jouzel et al., 2007; Loomiset al., 2012).

At latitudes similar to Lake Qinghai but at lower elevation on theCLP, C4 plants rarely exceeded 40% abundance. Despite their rela-tively low abundance, C4 plants persisted on the CLP until the lateHolocene, perhaps due to warmer, wetter summers (Fig. 4; Zhanget al., 2003; Liu et al., 2005).

5. Conclusions

Our paleoecological data demonstrate that C4 plants can thriveon the Tibetan Plateau, and provide benchmarks for ecosystemmodels that infer <0.5% C4 plants on the Tibetan Plateau today. TheCCSM3 transient simulation infers 0% C4 grasses even during timeperiods when Lake Qinghai d13Cwax suggests up to 50% C4 plants(Figs. 1 and 3, and S5; Still et al., 2003; Z. Liu et al., 2009). Althoughwe have evidence for abundant C4 plant biomass around LakeQinghai during the glacial period, it will be most important tocorrectly model the increase in C4 plant abundance during theLateglacial and early Holocene, whichwas likely amorewidespreadphenomenon. Arid- and cold-adapted Chenopodiaceae s.s. likelyevolved C4 photosynthesis as a further adaptation to cold, dry, saltyconditions on the Tibetan Plateau (Kadereit et al., 2012), andtherefore do not fit with thewarm, wet thresholds for C4 plants thatare currently incorporated into ecosystem models (Collatz et al.,1998; Still et al., 2003). As the spatial resolution of climate andecosystem models increases, it will be necessary to account forunusual ecosystem dynamics, including cold-, arid-, salt-adaptedC4 Chenopodiaceae s.s., in spatially heterogenous, high-elevationregions.

The combination of saline soils, high irradiance, and summer-focused precipitation on the Tibetan Plateau inherently favor C4plants, whereas cool conditions and high pCO2 favor C3 plants. Adelicate ecological balance therefore exists on the Tibetan Plateauand in other high-elevation regions, where we document largepaleoecological transitions between C3 and C4 plants. Lake Qinghaid13Cwax provides a paleoecological perspective for studies of mod-ern distributions of C4 plants, which find that temperature, pre-cipitation, and pCO2 all play important roles (Ehleringer et al., 1997;Collatz et al., 1998; Sage et al., 1999; Edwards and Smith, 2010;Kadereit et al., 2012). Due to relatively high pCO2 today, C4 plantcommunities in cold, dry, high-elevation regions are at the edge oftheir ecological niche, and will likely decline further as pCO2 in-creases (Taylor et al., 2011).

E.K. Thomas et al. / Quaternary Science Reviews 87 (2014) 24e3332

Author contributions

YH designed the study; EKT, JZ, PW, and LG performed samplepreparation and analysis; EKT, YH, SMC, and SCC performed dataanalysis and interpretation; CM provided model output and anal-ysis; EKT wrote the main paper and the Supplementary Informa-tion, with input from YH, CM, SMC, and SCC.

Acknowledgments

We thank R. Tarozo for lab assistance and E. J. Edwards and oneanonymous reviewer for providing feedback on themanuscript. Wethank the International Continental Scientific Drilling Program, theUnited States National Science Foundation, the National NaturalScience Foundation of China and the Ministry of Science andTechnology of China for their support for the Lake Qinghai DrillingProject, whichwas led in part by A. Zhisheng.We thank Z. Liu and B.Otto-Bliesner for sharing CCSM3 output. The CCSM3 transientsimulation is supported by the P2C2 program at the US NationalScience Foundation, the Abrupt Change Program, EaSM program,and INCITE computing program at the US Department of Energy,and NCAR. We acknowledge the World Climate Research Pro-gramme’s Working Group on Coupled Modelling, which isresponsible for CMIP, and we thank the climate modeling groups(listed in Table S1) for producing and making available their modeloutput. For CMIP the U.S. Department of Energy’s Program forClimate Model Diagnosis and Intercomparison provides coordi-nating support and led development of software infrastructure inpartnership with the Global Organization for Earth System SciencePortals. This researchwas funded by NSF grant #ATM-0902805 to Y.Huang and a NSF Graduate Research Fellowship to EKT.

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.quascirev.2013.12.014.

References

An, Z., Colman, S.M., Zhou, W., Li, Xiaoqiang, Brown, E.T., Jull, A.J.T., Cai, Y., Huang, Y.,Lu, X., Chang, H., Song, Y., Sun, Y., Xu, H., Liu, W., Jin, Z., Liu, Xiaodong, Cheng, P.,Liu, Y., Ai, L., Li, Xiangzhong, Liu, Xiuju, Yan, L., Shi, Z., Wang, X., Wu, F., Qiang, X.,Dong, J., Lu, F., Xu, X., 2012. Interplay between the Westerlies and Asianmonsoon recorded in Lake Qinghai sediments since 32 ka. Sci. Rep. 2.

Baogen, S., Zhou, Q., Wang, H., Xu, J., 2013. Qinghai Lake basin vegetation com-munity recovery and study of soil salinity changes. Heilongjiang Anim. Husb.Vet. Med. 1, 83e84.

Berger, A., 1978. Long-term variations of caloric insolation resulting from the earth’sorbital elements. Quat. Res. 9, 139e167.

Berger, A., Loutre, M.F., Mélice, J.L., 2006. Equatorial insolation: from precessionharmonics to eccentricity frequencies. Clim. Past 2, 131e136.

Castañeda, I.S., Schouten, S., 2011. A review of molecular organic proxies forexamining modern and ancient lacustrine environments. Quat. Sci. Rev. 30,2851e2891.

Cavagnaro, J.B., 1988. Distribution of C3 and C4 grasses at different altitudes in atemperate arid region of Argentina. Oecologia 76, 273e277.

Chen, G.C., Peng, M., 1993. Types and distribution of vegetation in Qinghai LakeRegion. Phytoecol. Geobot. Sin. 17, 71e81.

Chikaraishi, Y., Naraoka, H., 2007. d13C and dD relationships among three n-alkylcompound classes (n-alkanoic acid, n-alkane and n-alkanol) of terrestrialhigher plants. Org. Geochem. 38, 198e215.

China Meteorological Data Sharing Service System [WWW Document], 2012. URLhttp://cdc.cma.gov.cn.

Collatz, G.J., Berry, J.A., Clark, J.S., 1998. Effects of climate and atmospheric CO2partial pressure on the global distribution of C4 grasses: present, past, andfuture. Oecologia 114, 441e454.

Colman, S.M., Yu, S.-Y., An, Z., Shen, J., Henderson, A.C.G., 2007. Late Cenozoicclimate changes in China’s western interior: a review of research on LakeQinghai and comparison with other records. Quat. Sci. Rev. 26, 2281e2300.

Cook, C.D.K., 1988. Wind pollination in aquatic angiosperms. Ann. Mo. Bot. Gard. 75,768e777.

Diefendorf, A.F., Mueller, K.E., Wing, S.L., Koch, P.L., Freeman, K.H., 2010. Globalpatterns in leaf 13C discrimination and implications for studies of past andfuture climate. Proc. Natl. Acad. Sci. 107, 5738e5743.

Dietze, E., Wünnemann, B., Hartmann, K., Diekmann, B., Jin, H., Stauch, G.,Yang, S., Lehmkuhl, F., 2013. Early to mid-Holocene lake high-stand sedi-ments at Lake Donggi Cona, northeastern Tibetan Plateau, China. Quat. Res.79, 325e336.

Diffenbaugh, N.S., Scherer, M., Ashfaq, M., 2012. Response of snow-dependent hy-drologic extremes to continued global warming. Nat. Clim. Change 3, 379e384.

Edwards, E.J., Smith, S.A., 2010. Phylogenetic analyses reveal the shady history of C4grasses. Proc. Natl Acad. Sci. 107, 2532e2537.

Edwards, E.J., Still, C.J., 2008. Climate, phylogeny and the ecological distribution ofC4 grasses. Ecol. Lett. 11, 266e276.

Ehleringer, J.R., Cerling, T.E., Helliker, B.R., 1997. C4 photosynthesis, atmospheric CO2,and climate. Oecologia 112, 285e299.

Gao, L., Hou, J., Toney, J., MacDonald, D., Huang, Y., 2011. Mathematical modeling ofthe aquatic macrophyte inputs of mid-chain n-alkyl lipids to lake sediments:implications for interpreting compound specific hydrogen isotopic records.Geochim. Cosmochim. Acta 75, 3781e3791.

He, F., 2011. Simulating Transient Climate Evolution of the Last Deglaciation withCCSM3. PhD thesis. University of Wisconsin, Madison.

Herzschuh, U., Birks, H.J.B., Ni, Jian, Zhao, Yan, Liu, Hongyan, Liu, Xingqi, Grosse, G.,2009. Holocene land-cover changes on the Tibetan Plateau. The Holocene 20,91e104.

Herzschuh, U., Birks, H.J.B., Mischke, S., Zhang, C., Böhner, J., 2010. A modern pol-leneclimate calibration set based on lake sediments from the Tibetan Plateauand its application to a Late Quaternary pollen record from the Qilian Moun-tains. J. Biogeogr. 37, 752e766.

Hudson, A.M., Quade, J., 2013. Long-term east-west asymmetry in monsoon rainfallon the Tibetan Plateau. Geology 41, 351e354.

Ji, S., Xingqi, L., Sumin, W., Matsumoto, R., 2005. Palaeoclimatic changes in theQinghai Lake area during the last 18,000 years. Quat. Int. 136, 131e140.

Joos, F., Spahni, R., 2008. Rates of change in natural and anthropogenic radiativeforcing over the past 20,000 years. PNAS 105, 1425e1430.

Jouzel, J., Masson-Delmotte, V., Cattani, O., Dreyfus, G., Falourd, S., Hoffmann, G.,Minster, B., Nouet, J., Barnola, J.M., Chappellaz, J., Fischer, H., Gallet, J.C.,Johnsen, S., Leuenberger, M., Loulergue, L., Luethi, D., Oerter, H., Parrenin, F.,Raisbeck, G., Raynaud, D., Schilt, A., Schwander, J., Selmo, E., Souchez, R.,Spahni, R., Stauffer, B., Steffensen, J.P., Stenni, B., Stocker, T.F., Tison, J.L.,Werner, M., Wolff, E.W., 2007. Orbital and millennial Antarctic climate vari-ability over the past 800,000 years. Science 317, 793e796.

Kadereit, G., Ackerly, D., Pirie, M.D., 2012. A broader model for C4 photosynthesisevolution in plants inferred from the goosefoot family (Chenopodiaceae ss).Proc. R. Soc. B Biol. Sci. 279, 3304e3311.

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.Astron. Astrophys. 428, 261e285.

Li, M., Liu, H., Li, L., Yi, X., Zhu, X., 2007. Carbon isotope composition of plants alongaltitudinal gradient and its relationship to environmental factors on the Qing-haieTibet Plateau. Pol. J. Ecol. 55, 67.

Liu, W., Huang, Y., An, Z., Clemens, S.C., Li, L., Prell, W.L., Ning, Y., 2005. Summermonsoon intensity controls C4/C3 plant abundance during the last 35 ka in theChinese Loess Plateau: carbon isotope evidence from bulk organic matter andindividual leaf waxes. Palaeogeogr. Palaeoclimatol. Palaeoecol. 220, 243e254.

Liu, W., Li, X., An, Z., Xu, L., Zhang, Q., 2013. Total organic carbon isotopes: a novelproxy of lake level from Lake Qinghai in the QinghaieTibet Plateau, China.Chem. Geol. 347, 153e160.

Liu, X., Cheng, Z., Yan, L., Yin, Z.-Y., 2009. Elevation dependency of recent and futureminimum surface air temperature trends in the Tibetan Plateau and its sur-roundings. Glob. Planet. Change 68, 164e174.

Liu, X., Colman, S.M., Brown, E.T., Henderson, A.C.G., Werne, Josef P., Holmes, J.A.,2013. Abrupt deglaciation on the northeastern Tibetan Plateaus: evidence fromlake Qinghai. J. Paleolimnol. in press.

Liu, X., Shen, J., Wang, S., Wang, Y., Liu, W., 2007. Southwest monsoon changesindicated by oxygen isotope of ostracode shells from sediments in Qinghai Lakesince the Lateglacial. Chin. Sci. Bull. 52, 539e544.

Liu, Z., Carlson, A.E., He, F., Brady, E.C., Otto-Bliesner, B.L., Briegleb, B.P.,Wehrenberg, M., Clark, P.U., Wu, S., Cheng, J., Zhang, J., Noone, D., Zhu, J., 2012.Younger Dryas cooling and the Greenland climate response to CO2. PNAS 109,11101e11104.

Liu, Z., Otto-Bliesner, B.L., He, F., Brady, E.C., Tomas, R., Clark, P.U., Carlson, A.E.,Lynch-Stieglitz, J., Curry, W., Brook, E., Erickson, D., Jacob, R., Kutzbach, J.,Cheng, J., 2009. Transient simulation of last deglaciation with a newmechanismfor Bolling-Allerod warming. Science 325, 310e314.

Loomis, S.E., Russell, J.M., Ladd, B., Street-Perrott, F.A., Sinninghe Damsté, J.S., 2012.Calibration and application of the branched GDGT temperature proxy on EastAfrican lake sediments. Earth Planet. Sci. Lett. 357e358, 277e288.

Mügler, I., Gleixner, G., Günther, F., Mäusbacher, R., Daut, G., Schütt, B., Berking, J.,Schwalb, A., Schwark, L., Xu, B., Yao, T., Zhu, L., Yi, C., 2010. A multi-proxyapproach to reconstruct hydrological changes and Holocene climate develop-ment of Nam Co, Central Tibet. J. Paleolimnol. 43, 625e648.

O’Leary, M.H., 1981. Carbon isotope fractionation in plants. Phytochemistry 20,553e567.

Otto-Bliesner, B.L., Brady, E.C., Clauzet, G., Tomas, R., Levis, S., Kothavala, Z., 2006.Last Glacial maximum and Holocene climate in CCSM3. J. Clim. 19, 2526e2544.

E.K. Thomas et al. / Quaternary Science Reviews 87 (2014) 24e33 33

Passey, B.H., Cerling, T.E., Perkins, M.E., Voorhies, M.R., Harris, J.M., Tucker, S.T.,2002. Environmental change in the Great Plains: an isotopic record from fossilhorses. J. Geol. 110, 123e140.

Peltier, W.R., 2004. Global glacial isostasy and the surface of the ice-age Earth: theICE-5G (VM2) model and GRACE. Annu. Rev. Earth Planet. Sci. 32, 111e149.

Peterse, F., Prins, M.A., Beets, C.J., Troelstra, S.R., Zheng, H., Gu, Z., Schouten, S.,Damsté, J.S.S., 2011. Decoupled warming and monsoon precipitation in East Asiaover the last deglaciation. Earth Planet. Sci. Lett. 301, 256e264.

Pyankov, V.I., Gunin, P.D., Tsoog, S., Black, C.C., 2000. C4 plants in the vegetation ofMongolia: their natural occurrence and geographical distribution in relation toclimate. Oecologia 123, 15e31.

Rao, Z., Jia, G., Zhu, Z., Wu, Y., Zhang, J., 2008. Comparison of the carbon isotopecomposition of total organic carbon and long-chain n-alkanes from surface soilsin eastern China and their significance. Chin. Sci. Bull. 53, 3921e3927.

Sage, R.F., Wedin, D.A., Li, M., 1999. The biogeography of C4 photosynthesis: patternsand controlling factors. In: Sage, Rowan F., Monson, Russell K. (Eds.), C4 PlantBiology. Academic Press, San Diego, p. 313. I.

Sheldon, R.B., Boylen, C.W., 1977. Maximum depth inhabited by aquatic vascularplants. Am. Midl. Nat. 97, 248e254.

Shepherd, T., Griffiths, D.W., 2006. The effects of stress on plant cuticular waxes.New Phytol. 171, 469e499.

Still, C.J., Berry, J.A., Collatz, G.J., DeFries, R.S., 2003. Global distribution of C3 and C4vegetation: carbon cycle implications. Glob. Biogeochem. Cycles 17, 6.1e6.14.

Sun, Y., Clemens, S.C., Morrill, C., Lin, X., Wang, X., An, Z., 2011. Influence of Atlanticmeridional overturning circulation on the East Asian winter monsoon. Nat.Geosci. 5, 46e49.

Taylor, K.E., Stouffer, R.J., Meehl, G.A., 2011. An overview of CMIP5 and the experi-ment design. Bull. Am. Meteorol. Soc. 93, 485e498.

Tieszen, L.L., Senyimba, M.M., Imbamba, S.K., Troughton, J.H., 1979. The distributionof C3 and C4 grasses and carbon isotope discrimination along an altitudinal andmoisture gradient in Kenya. Oecologia 37, 337e350.

Van de Water, P., Leavitt, S., Betancourt, J., 2002. Leaf 13C variability with elevation,slope aspect, and precipitation in the southwest United States. Oecologia 132,332e343.

Wang, G., Han, J., Liu, D., 2003. The carbon isotope composition of C3 herbaceousplants in loess area of northern China. Sci. China Ser. D Earth Sci. 46, 1069e1076.

Wang, J., Kassab, C., Harbor, J.M., Caffee, M.W., Cui, H., Zhang, G., 2013. Cosmogenicnuclide constraints on late Quaternary glacial chronology on the Dalijia Shan,northeastern Tibetan Plateau. Quat. Res. 79, 439e451.

Wang, L., Lü, H., Wu, N., Chu, D., Han, J., Wu, Y., Wu, H., Gu, Z., 2004. Discovery of C4species at high altitude in QinghaieTibetan Plateau. Ch. Sci. Bull. 49, 1392e1396.

Wang, R.Z., 2007. C4 plants in the deserts of China: occurrence of C4 photosynthesisand its morphological functional types. Photosynthetica 45, 167e171.

Wang, Y., Cheng, H., Edwards, R.L., Kong, X., Shao, X., Chen, S., Wu, J., Jiang, X.,Wang, X., An, Z., 2008. Millennial- and orbital-scale changes in the East Asianmonsoon over the past 224,000 years. Nature 451, 1090e1093.

Wang, Y., Herzschuh, U., Shumilovskikh, L.-S., Mischke, S., Birks, H.-J.-B.,Wischnewski, J., Böhner, J., Schlütz, F., Lehmkuhl, F., Diekmann, B.,Wünnemann, B., Zhang, C., 2013. Quantitative reconstruction of precipitationchanges on the NE Tibetan Plateau since the Last Glacial Maximum: extendingthe concept of pollen source-area to pollen-based climate reconstructions fromlarge lakes. Clim. Past Discuss. 9, 3563e3613.

Wang, Z., Liu, W., 2012. Carbon chain length distribution in n-alkyl lipids: a processfor evaluating source inputs to Lake Qinghai. Org. Geochem. 50, 36e43.

Wooller, M.J., Swain, D.L., Ficken, K.J., Agnew, A.D.Q., Street-Perrott, F.A., Eglinton, G.,2003. Late Quaternary vegetation changes around Lake Rutundu, Mount Kenya,East Africa: evidence from grass cuticles, pollen and stable carbon isotopes.J. Quat. Sci. 18, 3e15.

Yeager, S.G., Shields, C.A., Large, W.G., Hack, J.J., 2006. The low-resolution CCSM3.J. Clim. 19, 2545e2566.

Zhang, Z., Zhao, M., Lu, H., Faiia, A.M., 2003. Lower temperature as the main cause ofC4 plant declines during the glacial periods on the Chinese Loess Plateau. EarthPlanet. Sci. Lett. 214, 467e481.

Zhou, Y., Fan, J., Zhang, W., Harris, W., Zhong, H., Hu, Z., Song, L., 2011. Factorsinfluencing altitudinal patterns of C3 plant foliar carbon isotope composition ofgrasslands on the QinghaieTibet Plateau, China. Alp. Bot. 121, 79e90.