This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

IP Address: 130.126.152.18

This content was downloaded on 09/09/2015 at 18:28

Please note that terms and conditions apply.

Disentangling climatic and anthropogenic controls on global terrestrial evapotranspiration

trends

View the table of contents for this issue, or go to the journal homepage for more

2015 Environ. Res. Lett. 10 094008

(http://iopscience.iop.org/1748-9326/10/9/094008)

Home Search Collections Journals About Contact us My IOPscience

Environ. Res. Lett. 10 (2015) 094008 doi:10.1088/1748-9326/10/9/094008

LETTER

Disentangling climatic and anthropogenic controls on globalterrestrial evapotranspiration trends

JiafuMao1,Wenting Fu2, Xiaoying Shi1, DanielMRicciuto1, Joshua BFisher3, Robert EDickinson2,YaxingWei1,Willis Shem1, Shilong Piao4, KaicunWang5, Christopher R Schwalm6,7, HanqinTian8,MingquanMu9, Altaf Arain10, PhilippeCiais11, Robert Cook1, YongjiuDai5, DanielHayes1,ForrestMHoffman12,MaoyiHuang13, SuoHuang10, DeborahNHuntzinger7,14, Akihiko Ito15, Atul Jain16,AnthonyWKing1, Huimin Lei17, Chaoqun Lu8, AnnaMMichalak18, Nicholas Parazoo3, Changhui Peng19,Shushi Peng11, Benjamin Poulter20, Kevin Schaefer21, Elchin Jafarov21, Peter EThornton1,WeileWang22,NingZeng23, ZhenzhongZeng4, FangZhao23, QiuanZhu19 andZaichunZhu24

1 Environmental SciencesDivision andClimate Change Science Institute, OakRidgeNational Laboratory, OakRidge, TN,USA2 Jackson School of Geosciences, theUniversity of Texas, Austin, TX,USA3 Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA,USA4 Sino-French Institute for Earth SystemScience, College of Urban and Environmental Sciences, PekingUniversity, Beijing 100871,

People’s Republic of China5 College ofGlobal Change and Earth System Science, BeijingNormal University, Beijing, People’s Republic of China6 Center for Ecosystem Science and Society, NorthernArizonaUniversity, Flagstaff, AZ 86011,USA7 School of Earth Sciences and Environmental Sustainability, Northern ArizonaUniversity, Flagstaff, AZ 86011,USA8 International Center for Climate andGlobal Change Research and School of Forestry andWildlife Sciences, AuburnUniversity, Auburn,

AL 36849,USA9 Department of Earth SystemScience, University of California, Irvine, CA,USA10 School of Geography and Earth Sciences andMcMaster Centre for Climate Change,McMaster University, Hamilton,Ontario, Canada11 Laboratoire des Sciences duClimat et de l’Environnement, LSCE, F-91191Gif sur Yvette, France12 Climate Change Science Institute andComputer Science andMathematics Division, OakRidgeNational Laboratory, Oak Ridge, TN

37831,USA13 Atmospheric Sciences andGlobal ChangeDivision, PacificNorthwest National Laboratory, Richland,WA99354,USA14 Department of Civil Engineering, ConstructionManagement, and Environmental Engineering, NorthernArizonaUniversity, Flagstaff,

AZ 86011,USA15 National Institute for Environmental Studies, Tsukuba, Ibaraki 305-8506, Japan16 Department of Atmospheric Sciences, University of Illinois, Urbana, IL 61801,USA17 Department ofHydraulic Engineering, TsinghuaUniversity, Beijing 100084, People’s Republic of China18 Department ofGlobal Ecology, Carnegie Institution for Science, Stanford, CA 94305,USA19 Institute of Environmental Sciences, University of Quebec atMontreal (UQAM), Case postale 8888, succCentre-Ville,Montreal, QC

H3C3P8, Canada20 Department of Ecology,Montana StateUniversity, Bozeman,MT59717,USA21 National Snow and IceDataCenter, Cooperative Institute for Research in Environmental Sciences, University of Colorado at Boulder,

Boulder, CO80309,USA22 Ames ResearchCenter, National Aeronautics and Space Administration,Moffett Field,MountainView, CA 94035,USA23 Department of Atmospheric andOceanic Science, University ofMaryland, College Park,MD20742,USA24 State Key Laboratory of Soil Erosion andDryland Farming on the Loess Plateau,Northwest A&FUniversity, Yangling 712100, People’s

Republic of China

E-mail:maoj@ornl.gov

Keywords: evapotranspiration, natural and anthropogenic controls, factorial analysis,MsTMIP

Supplementarymaterial for this article is available online

AbstractWeexamined natural and anthropogenic controls on terrestrial evapotranspiration (ET) changesfrom1982 to 2010 usingmultiple estimates from remote sensing-based datasets and process-orientedland surfacemodels. A significant increasing trend of ET in each hemisphere was consistently revealedby observationally-constrained data andmulti-model ensembles that considered historic natural andanthropogenic drivers. The climate impacts were simulated to determine the spatiotemporalvariations in ET.Globally, rising CO2 ranked second in thesemodels after the predominant climaticinfluences, and yielded decreasing trends in canopy transpiration and ET, especially for tropical forestsand high-latitude shrub land. Increasing nitrogen deposition slightly amplified global ET via enhanced

OPEN ACCESS

RECEIVED

16 June 2015

REVISED

17August 2015

ACCEPTED FOR PUBLICATION

18August 2015

PUBLISHED

8 September 2015

Content from this workmay be used under theterms of theCreativeCommonsAttribution 3.0licence.

Any further distribution ofthis workmustmaintainattribution to theauthor(s) and the title ofthework, journal citationandDOI.

© 2015 IOPPublishing Ltd

plant growth. Land-use-induced ET responses, albeit with substantial uncertainties across the factorialanalysis, wereminor globally, but pronounced locally, particularly over regions with intensive land-cover changes. Our study highlights the importance of employingmulti-streamET and ET-component estimates to quantify the strengthening anthropogenic fingerprint in the global hydrologiccycle.

1. Introduction

Intensified global hydrological cycle has been observedand modeled during the past few years (Hunting-ton 2006, Gerten et al 2008, Wang et al 2010, Duracket al 2012, Douville et al 2013, Sterling et al 2013, Wuet al 2013, Gedney et al 2014). Terrestrial evapotran-spiration (ET) is arguably the central component ofthis changing hydrologic cycle, and functions as a vitallink between energy, water and carbon cycles, therebyhaving important implications for the availability andusage of fresh water resources by humans andterrestrial ecosystems (Seneviratne et al 2006, Tren-berth et al 2009, Fisher et al 2011, Wang andDickinson 2012).

Natural environmental factors (e.g. precipitation,temperature, incident solar radiation, soil moisture,wind and atmospheric teleconnections) regulate ETand its variability across different terrestrial ecosys-tems (Teuling et al 2009, Jung et al 2010, Wanget al 2010, Vinukollu et al 2011, Zhang et al 2012, Mir-alles et al 2014). These natural controls and limita-tions/co-limitations of ET are scale-dependent. Theirmechanistic understanding is very important to pre-dict the tendency and variability of ET (Wang andDickinson 2012). Human-induced land use/landcover change, ground water withdrawals, and irriga-tion can directly alter the amount and timing of ET bymodifying the local water and energy balances (Piaoet al 2007, Gerten 2013, Leng et al 2013, 2014a, 2014b,Lo and Famiglietti 2013, Sterling et al 2013, Leiet al 2014c). Human activities that contribute to green-house gas emissions, atmospheric nitrogen deposition(NDE), and ozone pollution can also alter ET indir-ectly through changes in physiological, structural andcompositional responses of plants (Gedney et al 2006,Betts et al 2007, Sitch et al 2007, Cao et al 2009, Leakeyet al 2009). Discriminating these anthropogenic per-turbations from natural factors is expected to increasein importance as anthropogenic transformation of theEarth System becomes more pervasive (Seneviratneet al 2010, Gerten 2013).

Based on mechanistic and empirical algorithmsthat are driven by remotely sensed observations, avariety of globally gridded diagnostic ET productshave been compiled and evaluated in recent years(Willmott et al 1985, Fisher et al 2008, Jiménezet al 2009, Jung et al 2009, Sheffield et al 2010, Zhanget al 2010b, Miralles et al 2011, Mueller et al 2011,Vinukollu et al 2011, Zeng et al 2012, Schwalm

et al 2013). These gridded ET estimates offer crucialsources and benchmarks for quantitative investiga-tions of historical ET dynamics over the land surface.However, the accuracy of these observation-based ETproducts has yet to be reconciled due to limitations inunderlying hypotheses and errors in input datasets(Mueller et al 2011, 2013, Polhamus et al 2012). More-over, due to their reliance on the satellite observations,these datasets offer a limited historical temporalrecord that encompasses only a few decades (Badgleyet al 2015).

To predict future changes in ET patterns, process-based simulation and understanding of the magni-tudes, mechanisms and interactions that control his-torical ET dynamics will be required and should bewithin uncertainty of both historical and present-dayobservations. Mechanistic land surface models(LSMs), driven bymeasurement-based environmentalproperties, are useful tools for the detection and attri-bution of natural and anthropogenic effects on ETdynamics. For the past decade, global factorial LSMexperiments have been conducted and analyzed by dif-ferent modeling groups to investigate the separateeffects of environmental stresses on land surface andsubsurface runoff, river flow, ET and water use effi-ciency (Gedney et al 2006, 2014, Piao et al 2007, Shiet al 2011, 2013, Tian et al 2011, Liu et al 2012, Taoet al 2014). The role of climate impacts on these hydro-logic variables has been characterized predominantlyacross different regions of the globe. The relative roleof natural environmental change versus anthro-pogenic activities, however, wasmodeled to be hetero-geneous and geographically dependent. Nevertheless,due to large differences in initial model conditions,driver data, and complex parameterizations that gov-ern models, the simulated ET was demonstrated tovary in magnitudes and responses across models atboth temporal and spatial scales (Wang et al 2010).

To disentangle these differences in simulated ETpatterns and the relative role of model sensitivity andstructure, the experimental setup and boundary/initial datamust be similar among different participat-ing models. We leveraged the controlled factorialexperiments and model simulation protocol from theMulti-Scale Synthesis and Terrestrial Model Inter-comparison Project (MsTMIP) (Huntzingeret al 2013). Further, we synthesized a global ET timeseries (1982–2010) based on a diverse set of diagnosticET products (table 1), and the methodology reportedrecently inMueller et al (2013). The partitioning of ET

2

Environ. Res. Lett. 10 (2015) 094008 JMao et al

Table 1.Overview of the diagnostic ET datasets used for themerged ETof this study, and the simulated ET fromMsTMIPmodels. Factorial results of theMsTMIPmulti-model are ALL: the impact from all historical forcing factors, CLI: theimpact fromhistorical climate only, OTH: all anthropogenic impacts, CO2: the historical CO2 impact only, NDE: the historical nitrogen deposition impact only, LUC: the historical land use/land cover change impact only, Y: theavailability of ET simulation for the particular impact, andN: the non-availability of ET simulation for the particular impact.

Group Name Algorithm Spatial resolution Precipitation data Time period Citation

GLEAM Modified Priestley–Taylor 0.25°×0.25° GPCPCMORPH 1982–2010 Miralles et al (2011)CSIRO Modified Penman–Monteith 0.5°×0.5° SILO 1984–2005 Zhang et al (2010b)MPI Empirically derived fromFLUXNET 0.5°×0.5° GPCC 1982–2008 Jung et al (2009)NTSG Modified Penman–Monteith 0.5°×0.5° GPCC 1983–2006 Zhang et al (2010a)

Diagnostic ET PRUNI (3 setsof data)

Penman–Monteith/Priestley–Taylor

(ISCCP, AVHRR, SRB)0.25°×0.25° Sheffield

et al (2006)1984–2007 Sheffield et al (2010)

PT-JPL Modified Priestley–Taylor 0.5°×0.5° Not required 1984–2006 Fisher et al (2008)UDEL ModifiedThornthwaite water budget 0.5°×0.5° GHCN2 1980–2008 Willmott et al (1985)PUB Empiricalmethod (TWSA,CRU) 0.5°×0.5° GRACE 1982–2009 Zeng et al (2012)AWB Water balance 0.5°×0.5° GPCP 1990–2006 Mueller et al (2011)

Group Name Algorithm Spatial Resolution Time Period Citation CLI LUC CO2 NDE ALL OTH

CLM4 Modified Penman–Monteith 0.5°×0.5° CRUNCEP 1982–2010 Lawrence et al (2007),Mao

et al (2012)Y Y Y Y Y Y

DLEM Penman–Monteith 0.5°×0.5° CRUNCEP 1982–2010 Tian et al (2011, 2012) Y Y Y Y Y Y

BIOME-BGC Penman–Monteith 0.5°×0.5° CRUNCEP 1982–2010 Thornton et al (2002) Y N N N Y Y

CLASS-CTEM-N+ Modified Penman–Monteith 0.5°×0.5° CRUNCEP 1982–2010 Huang et al (2011), Bartlettet al (2006)

Y Y Y Y Y Y

CLM4-VIC Modified Penman–Monteith 0.5°×0.5° CRUNCEP 1982–2010 Lei et al (2014a) Y Y Y Y Y Y

ISAM Modified Penman–Monteith 0.5°×0.5° CRUNCEP 1982–2010 Jain et al (1996) Y Y Y Y Y Y

MsTMIPET LPJ-WSL Modified Penman–Monteith 0.5°×0.5° CRUNCEP 1982–2010 Sitch et al (2003) Y Y Y N Y Y

ORCHIDEE-LSCE Modified Penman–Monteith 0.5°×0.5° CRUNCEP 1982–2010 Krinner et al (2005) Y Y Y N Y Y

SiB3-JPL Penman–Monteith 0.5°×0.5° CRUNCEP 1982–2010 Baker et al (2008) Y Y Y N Y Y

SiBCASA Penman–Monteith 0.5°×0.5° CRUNCEP 1982–2010 Schaefer et al (2008, 2009) Y Y Y N Y Y

TRIPLEX-GHG Modified Penman–Monteith 0.5°×0.5° CRUNCEP 1982–2010 Peng et al (2011) N N Y Y Y N

VEGAS Bulk transfer formula 0.5°×0.5° CRUNCEP 1982–2010 Zeng et al (2005) Y Y Y N Y Y

VISIT Penman–Monteith 0.5°×0.5° CRUNCEP 1982–2010 Ito and Inatomi (2012) Y Y Y N Y Y

3

Environ.R

es.Lett.10(2015)094008

JMao

etal

(e.g canopy transpiration (Tr) and evaporation fromwet canopy and bare soil (ET–Tr)) and the variation ofthose ET components are poorly understood and lessconstrained by observations (Lawrence et al 2007,Jasechko et al 2013, Swenson and Lawrence 2014,Wang et al 2014). The MsTMIP modeling frameworkcan advance our understanding of trends in ET by pro-viding predictions of the individual ET components.In this study, we thus further investigated the con-tribution of individual influencing factors to the spa-tial and temporal characteristics of these ETconstituents.

2.Datasets andmethods

We created a merged diagnostic ET data (DIA) from11 long-term diagnostic datasets, all based on differentassumptions and constrained with extensive in situobservations or satellite retrievals or both (table 1).Weremapped the monthly raw datasets from their origi-nal spatial resolutions to the half-degree resolution ofthe model output from 1982 through 2010 based ondata availability. Following Mueller et al (2013), weapplied both physical and statistical constraints forquality control and bias corrections. For the physicalconstraint, we developed a dataset of seasonal netradiation maxima using the surface radiation budget(SRB3.0) datasets (Gupta 1983). We then excludedgrid points with values exceeding net radiation max-ima by more than 25%. The outliers were identified asvalues that exceed ±3 standard deviations (Wee-don 2011). Then the median values of these quality-controlled multiple ET estimates were treated as themerged product, and were comprehensively com-pared with the LSM results in this study. As shown infigure S1, the annual anomalies of the previouslysynthesized ET in Mueller et al (2013) are well withinthe spread of this newly-merged diagnostic dataproduct. This updated product however, provideslonger-term dynamics and is more amenable forstudies atmulti-decadal timescales.

To isolate the contributions of environmental dri-vers tomulti-year ET variations, we utilized the factor-ial ET simulations from the MsTMIP data archive.Driven by the same environmental forcing (climatevariability and trends, rising atmospheric CO2 con-centrations causing fertilization and reducing stomatalopening, nitrogen deposition, land use/land coverchange, and soil texture and vegetation types), thesestate-of-the-art LSMs were employed to identify theprincipal drivers of interannual variability and multi-decadal changes of ET. Because the evaporation com-ponent for canopy and soil, and the snow sublimation,were not separately archived in the standard modeloutputs in the MsTMIP I protocol (Huntzingeret al 2013, 2015), we included all relevant availableoutputs, namely the ET, Tr and the total evaporation(ET–Tr). Four model experiments: (1) SG1 (time

varying climate), (2) SG2 (time-varying climate andland use change history), (3) SG3 (time-varying cli-mate, land use, and atmospheric CO2), and (4) BG1(time-varying climate, land-use, atmospheric CO2 andnitrogen deposition), were analyzed to quantify theeffects of each environmental forcing factor on thestudy variables for the years 1982 through 2010. Thetransient simulations began in 1901, turning on onetime-varying driver at a time. Simulations BG1 or SG3were used to address the combined impacts from var-ious historical forcing agents formodels with (BG1) orwithout (SG3) an explicit nitrogen cycle. Simulationor simulation differencing was used to quantify thecontribution to ET and ET component changes fromclimate change (CLI) (derived from SG1), land use/land cover change (LUC) (derived fromSG2-SG1), ris-ing atmospheric CO2 (CO2) (derived from SG3-SG2),NDE (derived from BG1-SG3), or all forcing (ALL)(derived from BG1 or SG3) (table 1). To account forthe overall effects from human activity (OTH), wederived the human-induced ET to be the differencebetween the BG1 and SG1 or SG3 and SG1simulations.

Annual cropland area and total tree coverage infor-mation for the 1982–2010 period were derived fromthemerged product of the SYNergetic land coverMAP(Jung et al 2006) and the annual time series of the landuse harmonization data (Hurtt et al 2011). Additionaldetails on the aforementioned driver data and experi-mental design can be found inWei et al (2014a, 2014b)andHuntzinger et al (2013, 2015).

Growing season ET generally dominates theannual sum over the vegetated area of land (Wanget al 2007).We focused our analysis on growing seasonET for all observational and modeled data. Thedynamic annual growing season information, used tomask the monthly ET between 1982 and 2010, wasfirst determined from the global inventory modelingandmapping studies normalized difference vegetationindex (NDVI3g) dataset (Pinzon and Tucker 2014)using a Savitzky–Golay filter (Chen et al 2004, Jonssonand Eklundh 2004). It was then refined by excludingthe freeze period identified by the Freeze/Thaw EarthSystem Data Record (Kim et al 2011, 2012). In parti-cular, the growing season of tropical rainforests was setto 12months and it started in January.

3. Results

Across the globe, statistically significant increasingtrends of ET were recorded from 1982 to 2010 in theobservation-based ET estimates (DIA) (1.18 mm yr−2)and modeled ET from the ALL simulation(0.93±0.31 mm yr−2) (figures 1 and S2, and tableS1). Significantly positive annual correlations betweenthe simulated ALL ET and the observed ET wereobtained, particularly in the Northern hemisphere(NH) (Land: R2=0.58, p<0.01, NH: R2=0.72,

4

Environ. Res. Lett. 10 (2015) 094008 JMao et al

p<0.01, and the Southern hemisphere (SH):R2=0.46, p<0.01). The simulated multiyearincreasing trend and interannual variability of the ALLET were mainly explained by the CLI ET. In contrast,the overall human-induced OTH ET was predicted todecrease somewhat, and to exhibit relatively smallinterannual variations.

Spatial analysis of linear trends of ET for themerged observation product revealed remarkablyconsistent increasing tendency over most continents(figure 2(a)). Local hotspots of reduced ET were diag-nosed to occur in the arid regions of Western NorthAmerica, central Africa, Northern China and South-eastern Asia. By contrast, the modeled changes of ALLET underestimated the magnitude of ET changes in

EasternNorth America andWestern Europe, andmis-sed the ET decreases in central Africa. But the place-ment of increasing or decreasing trends in ALL ETlargely agreed favorably with those of the observed ETtrends, indicating the suitability of examining multi-year ET trends using the all-factor simulations.

Spatial patterns of ET changes that are consistentbetween the ALL and CLM estimates confirm thedominance of climate forcing in explaining annual ETtrends (figures 2(b), (c) and 3(a)). This dominant cli-matic response of ET trends was chiefly associatedwith concurrent annual precipitation changes (spatialR2=0.34 for ALL ET and precipitation trend, andspatial R2=0.30 for CLI ET and precipitation trends,respectively, P<0.01), and tended to show large

Figure 1.Time series of annual anomalies of growing season ET (mm yr−1) over (a) the globe, (b) theNH, and (c) the SH from1982 to2010. Solid lines are themedian values of themerged ET (ET_DIA, black),MsTMIP ETof ALL (ET_ALL, red), CLI (ET_CLI, blue),andOTH (ET_OTH, green). Shaded areas indicate the ET range of independentMsTMIPmodels.

5

Environ. Res. Lett. 10 (2015) 094008 JMao et al

spatial heterogeneities of sign and magnitude(figure 4(a)). The spatial dominance patterns based onthe partial correlations among the total growing sea-son ET, precipitation, temperature and incident solarradiation affirmed that for the MsTMIP models,annual precipitation drove not only the interannualvariability of ET, but primarily accounted for the mul-tiyear ET trends over most land areas (figures 4(e) and(f)). Combined anthropogenic effects tended todecrease ET, most notably in Northeastern NorthAmerica, Western Amazon, Northwestern Europeand tropical Asia (figure 2(d)). These effects were sub-ject primarily to the net physiological and structural

impacts of CO2 concentration on the growth of plantsin ecosystems (figures 2(e), 3(d), S2 and S3(a)).

Increasing nitrogen deposition led to increasingleaf area index (LAI) (figures 4(b) and S3(b)), and con-sequently to enhanced terrestrial ET, particularly overSouth America, Africa and Southeastern China(figures 2(f) and S2). The areas undergoing strongincrease in forest fraction and decrease in croplandfraction, such as in central Eastern North America andcentral Europe, clearly showed increasing annual ET(figures 2(g), 4(c) and (d)). In contrast, regions withevident loss of trees, such as Eastern China and South-eastern South America, show a downtrend of annual

Figure 2. Spatial distribution of the linear trends in ETmedian values (mm yr−2) for (a)ET_DIA, (b)ET_ALL, (c)ET_CLI, (d)ET_OTH, (e)CO2 (ET_CO2), (f)NDE (ET_NDE), and (g) LUC (ET_LUC) from1982 to 2010. The stippled areas represent the trendsare statistically significant (P<0.05), and the insets show the frequency distribution of the corresponding change.

6

Environ. Res. Lett. 10 (2015) 094008 JMao et al

ET. Compared to the CO2 and nitrogen depositioneffects, however, the effect of LUC on land ET wasimportant locally. Relatively large uncertainties fromthe LUC were also found between individual models(figures S2 and S6).

Trends for the Tr and total ET–Tr were dominatedby the climatic changes across various continents. ForTr, 85.4% of the study area was impacted by the cli-matic changes, and 88.7% for ET–Tr (figures 3(b), (c),S4(a)–(f)). Congruent with the response of ET changesto rising CO2 (48.4 ppm during the period1982–2010), most areas, especially these regions cov-ered by tropical broadleaf evergreen trees and highlatitude shrubs, showed decreasing Tr. This is due tothe CO2-induced reduction in stomatal conductanceoverwhelming the LAI-induced increase of canopyevaporation and transpiration under elevated CO2

concentration (figures 3(e) and S4(j)). On the otherhand, CO2 fertilization would enhance canopy LAIthrough increasing photosynthate allocation to leaves,and caused more canopy transpiration and evapora-tion than the reduced transpiration by CO2 physiolo-gical effects, especially over dry areas with sparsevegetation (e.g. the Western North America, centralEurasia, and Australia) (figures S3(a) and S4(j)).Reversed ET–Tr trends in these arid regions imply thatdecreasing soil evaporationwas the dominant factor inchanging ET–Tr (figures S4(j)–(l)). Formost areas thatshowed decreasing Tr but increasing ET–Tr underCO2 enrichment, the augmented evaporation of inter-cepted rainfall and increasing soil evaporation mayhave been coincidental.

Increasing ET caused by nitrogen deposition wasdue to enhanced Tr (figures 2(f), S4(m) and S5). Adecrease of ET–Tr caused by the nitrogen depositioneffect, as seen in central North America and in Wes-tern Europe, was due to reduced soil evaporation(figures S4(n) and S5). The latter is a consequence ofthe increasing LAI providing more shade and so

reducing solar energy for soil evaporation. In addition,the increasing Tr further depleted soil water, whichreduced soil evaporation. In the evergreen broadleafforests of the Western Amazon and Congo basin,nitrogen deposition and higher LAI resulted inincreasing canopy evaporation. The increase in canopyevaporation more than offset the decrease in soil eva-poration and hence dominated the increasing ET–Trand even the nitrogen-induced increase in total ET(figures S4(m)–(o)).

LUC led to a decreasing trend in Tr across denselyinhabited regions that had experienced substantialland use perturbations (e.g. clearing trees for crops)during the study period. These occurred mainly inSoutheastern South America and the Eastern China(figures 4(c), (d), S4(p) and S5). Tr trends showed ageneral negative sign over central Eastern NorthAmerica and Western Europe, where croplands hadbeen replaced mainly by forests and woodlands. Thisreduction of Tr with reforestation implies that the treespecies that replaced the crops had lower stomatalconductance than the crop species, the younger andsmaller trees of the returning forests had lower LAIthan the croplands they replaced, or the available soilwater for plants decreased because of the removal ofirrigation. These aspects deserve further study.

4.Discussion

Between 1982 and 2010, the observation-based andsimulated ALL ET consistently showed a significantlyincreasing trend across the globe. These findings areconsistent with previous studies, which reported anintensified global hydrological cycle in response toglobal warming following the Clausius–Clapeyron law(the relationship between equilibrium water vaporpressure and temperature, about 7% per °C ofwarming) (Held and Soden 2006), as well as increasing

Figure 3. Spatial distribution of the dominant drivers for the ET, Tr andET–Tr changes for the period 1982–2010. (a)–(c)Dominantdrivers for the ET, Tr and ET–Tr trends of the ALL results, and (d)–(f) dominant drivers for the ET, Tr and ET–Tr trends of theOTHresults.

7

Environ. Res. Lett. 10 (2015) 094008 JMao et al

importance of the radiative component of ET (John-son and Sharma 2010). Climatic factors accounted formuch of the spatial and temporal variations interrestrial ET, Tr and ET–Tr. This supports previousstudies regarding the prevalent climatic mechanismscontrolling the long-term ET trends such as tempera-ture, precipitation, soil moisture, energy and internalclimate variability (Teuling et al 2009, Jung et al 2010,Wang et al 2010, Vinukollu et al 2011, Zhanget al 2012, Ukkola and Prentice 2013, Miralleset al 2014).

In our study, the rising atmospheric CO2 con-centration, as tested by model factorial experiments,induced an overall suppression of Tr and hence a gen-eral decreasing ET. Our results further suggest that thesign of change and regional pattern of these CO2 phy-siological effects on ET were moderated by changes inLAI. The overall response of ET was eventually deter-mined by the balance among the changes of Tr,canopy evaporation and soil evaporation. Theseresults are consistent with modeled and observedplant physiological responses to the increase of CO2

concentration in the atmosphere (Betts et al 2007,

Leakey et al 2009). They also reiterate previous find-ings that show the concurrent physiological and struc-tural responses of vegetation to rising CO2, andassociated hydrological effects (Gedney et al 2006,Leipprand and Gerten 2006, Ainsworth andRogers 2007, Betts et al 2007, Kurc and Small 2007,Piao et al 2007, Cao et al 2009, Leakey et al 2009, Leiet al 2014b).

Simulation experiments that consider NDEshowed enhanced global LAI as a result of increasingnutrient availability (figures 4(b) and S3(b)). Thenitrogen-induced enhancement of canopy Tr andcanopy evaporation, however, was regionally offset bydecreasing soil evaporation, and led to lower ET forthe nitrogen fertilization effect. Nonetheless, miner-alized nitrogen in the rooting system was governed bynot only the amount of deposited N, but also by leach-ing and denitrification, which are affected by environ-mental conditions (Hovenden et al 2014). Thishighlights the necessity of better understanding theinteractions among these environmental drivers, andthe underlying mechanisms responsible for biogeo-chemical and hydrologic cycles.

Figure 4. Spatial distribution of trends in (a) precipitation (PRE,mm/yr−2), (b)nitrogen deposition (NDE,mg N m−2 yr−2), (c)fractional tree coverage (TREE,%/yr−2), and (d) fractional crop coverage (CROP,%/yr−2) over the period 1982–2010, and spatialdistribution of dominant climatic variable (precipitation, temperature (TEM) and incident solar radiation (RAD)) responsible for (e)ET variability, and (f) both variability and trend. For (e), the dominance was derived by comparing theR2 of the partial correlationsbetween detrended ET and individual climatic factor. For (f), the dominancewas derived by comparing theR2 of the partialcorrelations between un-detrended ET and individual climatic factor. Both (e) and (f) share the same color legend in (f).

8

Environ. Res. Lett. 10 (2015) 094008 JMao et al

Previousmodeling studies (Boisier et al 2012, 2014,Shi et al 2013, Sterling et al 2013, Tao et al 2014) agreewith our results that anthropogenic activities modifiedET and its components locally, and human-inducedLUC effects tended to counteract each other at a globalscale. We found large uncertainties associated withLUC impacts among the MsTMIP LSMs, particularlyover the NH and areas having marked land cover con-versions. Though based on the same merged LUCdataset, different LSM groups prescribed the dynamicevolution of plant functional types with model-spe-cific classifications (Wei et al 2014a, 2014b). The sensi-tivity of biophysical and biogeochemical processes tothe reconstructed historical scenario of LUC, more-over, varied considerably from model to model(Huntzinger et al 2013). For example, for the SIB3-JPLmodels, abnormally higher LUC ET was simulatedover the NH and global land compared to that of othermodels (figure S6). In SIB3-JPL, ET is a function ofstomatal conductance and is sensitive to changes inphotosynthetically active radiation (PAR). In LUCsimulations, plant functional type changes over time,but the PAR is prescribed from present day NDVI cli-matology and is thus fixed to modern vegetation. Thiscan lead to a bias in gross primary production in caseswhere grasslands are converted to forests, since theNDVI and resulting fraction of incident PAR absorbedby green leaving in the canopy (fPAR) are calculatedfrom a modern day forest ecosystem but used to esti-mate stomatal conductance and ET for the historicalgrassland it replaced. The sensitivity to land-usechange and cultivated ecosystems (e.g., irrigated crop-lands) reinforce the need for better LUC characteriza-tion, improved parameterization of ET in croplands,and the development of forcing datasets (e.g., PAR)that are not artificially dependent upon land cover.Improvements in these areasmay help reduce the largeinter-model spreads in the responses of ET to LUC.

Quantitative estimation of ET partitioning hasbeen refined recently, but information on long-termvariations and the precise drivers of each ET compo-nent are lacking (Jasechko et al 2013,Wang et al 2014).By using a multi-model ensemble, we assessed theannual trends of the Tr and ET–Tr over nearly threedecades, and further estimated their spatial-temporalresponses to various environmental stresses. Thesemodeled results, however, remain rather uncertainwithout observational constrains that are sufficientlylong and representative. Comprehensive synthesis oflong-term observation-constrained ET components isneeded to improve our understanding of the control-ling mechanisms, and to better characterize the parti-tioning schemes.

5. Conclusions

The relative contribution of climate and anthropo-genic activities to the spatio-temporal changes in ET

was quantitatively characterized with the newly-merged ET and multifactor ensemble simulationsfrom MsTMIP. In the LSMs, climate, CO2, nitrogendeposition, and land use impacts were separatedexperimentally to determine the ET variationsbetween 1982 and 2010. Climate, and in particular,changes in precipitation, was the dominant control ofmulti-year ET trends and variability. The overall CO2

physiological and structural effect induced decreasingplants transpiration and the total ET, especially inareas where vegetation was dense. Compared toclimate change and the elevated CO2 effects, theimpacts of nitrogen deposition and land use change onET were less important and acted locally. Otherdetailed explorations are needed, such as the imple-mentation of more compelling statistical techniquesand fully-coupled modeling systems (Douvilleet al 2013, Wu et al 2013, Gedney et al 2014) to detectand attribute the natural and anthropogenic effects onET with more certainty. ET-related feedback studiesare also required to account for land-atmosphereinteractions and anthropogenic impacts in the inte-grated earth system models (Seneviratne et al 2010,Bond-Lamberty et al 2014, Collins et al 2015) and tounderstand future trajectories of drought (Sheffieldet al 2012, Zarch et al 2015). Given that humanactivities continue to grow and intensify in theAnthropocene Epoch, we emphasize utilizing multi-stream datasets and multi-modeling frameworks tobetter diagnose and project anthropogenic influenceson terrestrial ET, hydrologic cycle and overall climatechange.

Acknowledgments

This researchwas supported partially by the TerrestrialEcosystem Science Scientific Focus Area (SFA), whichis sponsored by the Terrestrial Ecosystem Science(TES) Program in the Climate and EnvironmentalSciences Division (CESD) of the Biological andEnvironmental Research Program in the US Depart-ment of Energy Office of Science. This research wassupported partially by the Biogeochemistry–ClimateFeedbacks SFA and project under contract of DE-SC0012534, which are both sponsored by the Regionaland Global Climate Modeling (RGCM) Program inthe Climate and Environmental Sciences Division(CESD) of the Biological and Environmental ResearchProgram in the US Department of Energy Office ofScience. CRS was supported by National Aeronauticsand Space Administration (NASA) Grants#NNX12AP74G, #NNX10AG01A, and#NNX11AO08A. JBF carried out this research at theJet Propulsion Laboratory, California Institute ofTechnology, under a contract withNASA. Funding forthe Multi-scale synthesis and Terrestrial Model Inter-comparison Project (MsTMIP, http://nacp.ornl.gov/MsTMIP.shtml) activity was provided through NASA

9

Environ. Res. Lett. 10 (2015) 094008 JMao et al

ROSES Grant #NNX10AG01A. Data managementsupport for preparing, documenting, and distributingmodel driver and output data was performed by theModeling and Synthesis Thematic Data Center at OakRidge National Laboratory (ORNL, http://nacp.ornl.gov), with funding through NASA ROSES Grant#NNH10AN681. Finalized MsTMIP data productsare archived at the ORNL DAAC (http://daac.ornl.gov). This isMsTMIP contribution#5.

Acknowledgments for specific MsTMIP partici-patingmodels:

Biome-BGC: Biome-BGC code was provided bythe Numerical Terradynamic Simulation Group atUniversity of Montana. The computational facilitiesprovided by NASA Earth Exchange at NASA AmesResearchCenter.

CLASS-CTEM-N+: This research was funded bytheNatural Sciences and Engineering

Research Council (NSERC) of Canada Discoveryand Strategic grants. CLASS and CTEM models wereoriginally developed by the Climate Research Branchand Canadian Centre for Climate Modelling and Ana-lysis (CCCMa) of Environment Canada, respectively.

CLM: This research is supported in part by the USDepartment of Energy (DOE), Office of Science, Bio-logical and Environmental Research. Oak RidgeNational Laboratory is managed by UT-Battelle, LLCforDOEunder contractDE-AC05-00OR22725.

CLM4VIC: CLM4VIC simulations were sup-ported in part by theUSDepartment of Energy (DOE),Office of Science, Biological and EnvironmentalResearch (BER) through the Earth System Modelingprogram, and performed using the EnvironmentalMolecular Sciences Laboratory (EMSL), a national sci-entific user facility sponsored by the U.S.DOE-BERand located at Pacific Northwest National Laboratory(PNNL). Participation of M Huang in the MsTMIPsynthesis is supported by the U.S.DOE-BER throughthe Subsurface Biogeochemical Research Program(SBR) as part of the SBR Scientific Focus Area (SFA) atthe Pacific Northwest National Laboratory (PNNL).PNNL is operated for the US DOE by BATTELLEMemorial Institute under contract DE-AC05-76RLO1830.

DLEM: The Dynamic Land Ecosystem Model(DLEM) developed in the International Center for Cli-mate and Global Change Research at Auburn Uni-versity has been supported by NASA InterdisciplinaryScience Program (IDS), NASA Land Cover/Land UseChange Program (LCLUC), NASA Terrestrial EcologyProgram, NASA Atmospheric CompositionModelingand Analysis Program (ACMAP); NSF Dynamics ofCoupled Natural-Human System Program (CNH),Decadal and Regional Climate Prediction using EarthSystem Models (EaSM); DOE National Institute forClimate Change Research; USDA AFRI Program andEPA STARProgram.

Integrated Science Assessment Model (ISAM)simulations were supported by the US National

Science Foundation (NSF-AGS-12-43071 and NSF-EFRI-083598), the USDA National Institute of Foodand Agriculture (NIFA) (2011-68002-30220), the USDepartment of Energy (DOE)Office of Science (DOE-DE-SC0006706) and the NASA Land cover and LandUse Change Program (NNX14AD94G). ISAM simula-tions were carried out at the National Energy ResearchScientific Computing Center (NERSC), which is sup-ported by the Office of Science of the US Departmentof Energy under contract DE-AC02-05CH11231, andat the Blue Waters sustained-petascale computing,University of Illinois at Urbana- Champaign, which issupported by the National Science Foundation(awards OCI-0725070 and ACI-1238993) and thestate of Illinois.

LPJ-wsl: This work was conducted at LSCE,France, using amodified version of the LPJ version 3.1model, originally made available by the Potsdam Insti-tute for Climate Impact Research.

ORCHIDEE-LSCE: ORCHIDEE is a global landsurface model developed at the IPSL institute inFrance. The simulations were performed with the sup-port of the GhG Europe FP7 grant with computingfacilities provided by LSCE (Laboratoire des Sciencesdu Climat et de l’Environnement) or TGCC (TresGrandCentre deCalcul).

VISIT: VISIT was developed at the National Insti-tute for Environmental Studies, Japan. This work wasmostly conducted during a visiting stay at Oak RidgeNational Laboratory.

This manuscript has been authored by UT-Bat-telle, LLC under Contract No. DE-AC05-00OR22725with the US Department of Energy. The United StatesGovernment retains and the publisher, by acceptingthe article for publication, acknowledges that the Uni-ted States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish orreproduce the published form of this manuscript, orallow others to do so, for United States Governmentpurposes. The Department of Energy will providepublic access to these results of federally sponsoredresearch in accordance with the DOE Public AccessPlan (http://energy.gov/downloads/doe-public-access-plan).

References

Ainsworth EA andRogers A 2007The response of photosynthesisand stomatal conductance to rising (CO2): mechanisms andenvironmental interactions Plant, Cell Environ. 30 258–70

BadgleyG, Fisher J B, Jiménez C, TuKP andVinukollu R 2015Onuncertainty in global evapotranspiration estimates fromchoice of input forcing datasets J. Hydrometeorol. 16 1449–55

Baker I T, Prihodko L,DenningA S, GouldenM,Miller S andDaRochaHR2008 Seasonal drought stress in the amazon:reconcilingmodels and observations J. Geophys. Res. Biogeo.114 1–10

Bartlett PA,MacKayMDandVerseghyDL 2006Modified snowalgorithms in theCanadian land surface scheme:model runsand sensitivity analysis at three boreal forest standsAtmos.Ocean 44 207–22

10

Environ. Res. Lett. 10 (2015) 094008 JMao et al

Betts RA et al 2007Projected increase in continental runoff due toplant responses to increasing carbon dioxideNature 4481037–U5

Boisier J P, deNoblet-DucoudréN, PitmanA J, Cruz F T,Delire C,vanDenHurk B J JM, van derMolenMK,Mueller C andVoldoire A 2012Attributing the impacts of land-coverchanges in temperate regions on surface temperature andheatfluxes to specific causes: results from the first LUCID setof simulations J. Geophys. Res. Atmos. 117D12116

Boisier J P, deNoblet-DucoudréN andCiais P 2014Historical land-use induced evapotranspiration changes estimated frompresent-day observations and reconstructed land-covermapsHydrol. Earth Syst. Sci. Discuss. 11 2045–89

Bond-Lamberty B et al 2014On linking an Earth systemmodel tothe equilibrium carbon representation of an economicallyoptimizing land usemodelGeosci.Model Devel. 7 2545–55

Cao L, BalaG, Caldeira K,Nemani R andBan-Weiss G 2009Climateresponse to physiological forcing of carbon dioxide simulatedby the coupled community atmospheremodel (CAM3.1) andcommunity landmodel (CLM3.0)Geophys. Res. Lett. 36 1–5

Chen J, Jonsson P, TamuraM,GuZH,Matsushita B and Eklundh L2004A simplemethod for reconstructing a high-qualityNDVI time-series data set based on the Savitzky–GolayfilterRemote Sens. Environ. 91 332–44

CollinsWD et al 2015The integrated Earth systemmodel (iESM):formulation and functionalityGeosci.Model Dev. Discuss. 8381–427

DouvilleH, Ribes A,DecharmeB, AlkamaR and Sheffield J 2013Anthropogenic influence onmultidecadal changes inreconstructed global evapotranspirationNat. Clim. Change 359–62

Durack P J,Wijffels S E andMatear R J 2012Ocean salinities revealstrong global water cycle intensification during 1950–2000Science 336 455–8

Fisher J B, TuKP andBaldocchiDD2008Global estimates of theland-atmosphere waterflux based onmonthly AVHRR andISLSCP-II data, validated at 16 FLUXNET sitesRemote Sens.Environ. 112 901–19

Fisher J B,Whittaker R andMalhi Y 2011 ET come home: potentialevapotranspiration in geographical ecologyGlob. EcologyBiogeography 20 1–18

GedneyN, Cox PM, Betts RA, BoucherO,Huntingford C andStott PA 2006Detection of a direct carbon dioxide effect incontinental river runoff recordsNature 439 835–8

GedneyN,HuntingfordC,WeedonGP, BellouinN, BoucherO andCoxPM2014Detection of solar dimming and brighteningeffects onNorthern hemisphere river flowNature Geosci. 7796–800

GertenD 2013A vital link: water and vegetation in theanthropoceneHydrol. Earth Syst. Sci. 17 3841–52

GertenD,Rost S, vonBlohWandLuchtW2008Causesof change in20th century global river dischargeGeophys. Res. Lett.351–5

Gupta SK 1983A radiative transfermodel for surface radiationbudget studies J. Quant. Spectrosc. Radiat. Trans. 29 419–27

Held IM and SodenB J 2006Robust responses of the hydrologicalcycle to global warming J. Clim. 19 5686–99

HovendenM J,Newton PCDandWills K E 2014 Seasonal notannual rainfall determines grassland biomass response tocarbon dioxideNature 511 583–6

Huang S, ArainMA,Arora VK, Yuan F, Brodeur J and PeichlM2011Analysis of nitrogen controls on carbon andwaterexchanges in a conifer forest using theCLASS-CTEMN+model Ecol.Modell. 222 3743–60

Huntington TG2006 Evidence for intensification of the globalwater cycle: review and synthesis J. Hydrol. 319 83–95

HuntzingerDN et al 2013TheNorth American carbon programmulti-scale synthesis and terrestrialmodel intercomparisonproject—:1. Overview and experimental designGeosci.ModelDev. 6 2121–33

HuntzingerDN et al 2015NACPMsTMIP: Global 0.5-degTerrestrial BiosphereModel Outputs (version 1) in StandardFormat (Tennessee, USA:OakRidgeNational Laboratory

Distributed Active ArchiveCenter, OakRidge) data set(http://daac.ornl.gov)

Hurtt GC et al 2011Harmonization of land-use scenarios for theperiod 1500–2100: 600 years of global gridded annual land-use transitions, wood harvest, and resulting secondary landsClim. Change 109 117–61

ItoA and InatomiM2012Use of a process-basedmodel forassessing themethane budgets of global terrestrial ecosystemsand evaluation of uncertaintyBiogeosci. 9 759–73

JainAK, KheshgiH S andWuebbles D J 1996A globally aggregatedreconstruction of cycles of carbon and its isotopesTellusB 48583–600

Jasechko S, Sharp ZD,Gibson J J, Birks S J, Yi Y and Fawcett P J 2013Terrestrial waterfluxes dominated by transpirationNature496 347–50

JiménezC, Prigent C andAires F 2009Toward an estimation ofglobal land surface heat fluxes frommultisatelliteobservations J. Geophys. Res. 114D06305

Jonsson P and Eklundh L 2004TIMESAT—a program for analyzingtime-series of satellite sensor dataComput. Geosci. 30833–45

Johhnson F and SharmaA 2010AComparison of Australian openwater body evaporation trends for current and futureclimates estimated from class A evaporation pans and generalcirculationmodels J. Hydrometeorol. 11 105–21

JungM,Henkel K,HeroldMandChurkinaG 2006 Exploitingsynergies of global land cover products for carbon cyclemodelingRemote Sens. Environ. 101 534–53

JungM, ReichsteinMandBondeauA 2009Towards globalempirical upscaling of FLUXNET eddy covarianceobservations: validation of amodel tree ensemble approachusing a biospheremodelBiogeosciences Discuss. 6 5271–304

JungM et al 2010Recent decline in the global landevapotranspiration trend due to limitedmoisture supplyNature 467 951–4

KimY,Kimball J S,McDonaldKC andGlassy J 2011Developing aglobal data record of daily landscape freeze/thaw status usingsatellite passivemicrowave remote sensing IEEETrans.Geosci. Remote 49 949–60

KimY,Kimball J S, ZhangK andMcDonaldKC2012 Satellitedetection of increasingNorthern hemisphere non-frozenseasons from1979 to 2008: implications for regionalvegetation growthRemote Sens. Environ. 121 472–87

KrinnerG, ViovyN, deNoblet-DucoudréN,Ogée J, Polcher J,Friedlingstein P, Ciais P, Sitch S andPrentice I C 2005Adynamic global vegetationmodel for studies of the coupledatmosphere–biosphere systemGlob. Biogeochem. Cycles 191–33

Kurc SA and Small E E 2007 Soilmoisture variations andecosystem-scalefluxes of water and carbon in semiaridgrassland and shrublandWater Resour. Res. 43W06416

LawrenceDM, Thornton P,OlesonKWandBonanGB 2007Thepartitioning of evapotranspiration into transpiration, soilevaporation, and canopy evaporation in a gcm: impacts onland–atmosphere interaction J. Hydrometeorol. 8 862–80

Leakey ADB, Ainsworth EA, Bernacchi C J, Rogers A, Long S P andOrtDR 2009 ElevatedCO2 effects on plant carbon, nitrogen,andwater relations: six important lessons fromFACE J. Exp.Bot. 60 2859–76

LeiH,HuangM, Leung LR, YangD, Shi X,Mao J,HayesD J,SchwalmC,Wei Y and Liu S 2014a Sensitivity of globalterrestrial gross primary production to hydrologic statessimulated by the community landmodel using two runoffparameterizations J. Adv.Model. Earth Syst. 6 658–79

LeiH, YangD andHuangM2014b Impacts of climate change andvegetation dynamics on runoff in themountainous region oftheHaihe River basin in the pastfive decades J. Hydrol. 511786–99

LeiH, YangD, YangH, YuanZ and LvH2014c Simulated impactsof irrigation on evapotranspiration in a strongly exploitedregion: a case study of theHaihe River Basin ChinaHydrol.Process. 29 2704–19

11

Environ. Res. Lett. 10 (2015) 094008 JMao et al

LeipprandA andGertenD 2006Content on evapotranspiration, soilmoisture and runoff under potential natural vegetationHydrol. Sci. J. 51 171–85

LengG,HuangM,TangQ, SacksW J, LeiH and Leung LYR 2013Modeling the effects of irrigation on land surfacefluxes andstates over the conterminousUnited States: sensitivity toinput data andmodel parameters J. Geophys. Res.-Atmos. 1189789–803

LengG,HuangM,TangQ,GaoH and Leung LYR 2014aModelingthe effects of groundwater-fed irrigation on terrestrialhydrology over the conterminousUnited StatesJ. Hydrometeorol. 15 957–72

LengG, TangQ,HuangMand Leung LR 2014bA comparativeanalysis of the impacts of climate change and irrigation onland surface and subsurface hydrology in theNorthChinaplainReg. Environ. Change 15 251–63

LiuM,TianH, LuC, XuX, ChenG andRenW2012 Effects ofmultiple environment stresses on evapotranspiration andrunoff over EasternChina J. Hydrol. 426–7 39–54

LoMHandFamiglietti J S 2013 Irrigation inCalifornia’s centralvalley strengthens the SouthwesternUSwater cycleGeophys.Res. Lett. 40 301–6

Mao J, Thornton PE, Shi X, ZhaoMandPostWM2012Remotesensing evaluation of CLM4GPP for the period 2000–09J. Clim. 25 5327–42

MirallesDG,Holmes TRH,De JeuRAM,Gash JH,Meesters AGCA andDolmanA J 2011Global land-surfaceevaporation estimated from satellite-based observationsHydrol. Earth Syst. Sci. 15 453–69

MirallesDG et al 2014 ElNiño–LaNiña cycle and recent trends incontinental evaporationNat. Clim. Change 4 122–6

Mueller B et al 2011 Evaluation of global observations-basedevapotranspiration datasets and IPCCAR4 simulationsGeophys. Res. Lett. 38 L06402

Mueller B et al 2013 Benchmark products for landevapotranspiration: LandFlux-EVALmulti-data set synthesisHydrol. Earth Syst. Sci. 17 3707–20

PengC,MaZ, Lei X, ZhuQ,ChenH,WangW, Liu S, LiW,FangX andZhouX 2011Adrought-induced pervasiveincrease in treemortality across Canada’s boreal forestsNat.Clim. Change 1 467–71

Piao S, Friedlingstein P, Ciais P, deNoblet-DucoudréN,LabatD andZaehle S 2007Changes in climate and land usehave a larger direct impact than rising CO2 on global riverrunoff trends Proc. Natl Acad. Sci. USA 104 15242–7

Pinzon J andTucker C 2014Anon-stationary 1981–2012AVHRRNDVI3g time seriesRemote Sens. 6 6929–60

PolhamusAM, Fisher J B andTuKP 2012What controls the errorstructure in evapotranspirationmodels?Agric. ForestMeteorol. 169 12–24

Schaefer K, Collatz G J, Tans P,DenningA S, Baker I, Berry J,Prihodko L, SuitsN and Philpott A 2008Combinedsimple biosphere/carnegie-ames-stanford approachterrestrial carbon cyclemodel J. Geophys. Res. Biogeo. 113G03034

Schaefer K, Zhang T, Slater AG, Lu L, Etringer A andBaker I 2009Improving simulated soil temperatures and soil freeze/thawat high-latitude regions in the simple biosphere/carnegieames-stanford approachmodel J. Geophys. Res. Earth Surf.114 F02021

SchwalmCR,HuntinzgerDN,Michalak AM, Fisher J B,Kimball J S,Mueller B, ZhangK andZhang Y 2013 Sensitivityof inferred climatemodel skill to evaluation decisions: a casestudy usingCMIP5 evapotranspiration Environ. Res. Lett. 8024–8

Seneviratne S I, LüthiD, LitschiM and Schär C 2006 Land-atmosphere coupling and climate change in EuropeNature443 205–9

Seneviratne S I, Corti T,Davin E L,HirschiM, Jaeger E B, Lehner I,Orlowsky B andTeuling A J 2010 Investigating soilmoisture–climate interactions in a changing climate: a review Earth-Science Rev. 99 125–61

Sheffield J, Goteti G andWoodE F 2006Development of a 50-yearhigh-resolution global dataset ofmeteorological forcings forland surfacemodeling J. Climate 19 3088–111

Sheffield J,WoodE F andMunoz-Arriola F 2010 Long-termregional estimates of evapotranspiration forMexico based ondownscaled ISCCPdata J. Hydrometeorol. 11 253–75

Sheffield J,WoodE F andRoderickML 2012 Little change in globaldrought over the past 60 yearsNature 491 435–8

Shi X,Mao J, Thornton PE,Hoffman FMandPostWM2011Theimpact of climate, CO2, nitrogen deposition and land usechange on simulated contemporary global river flowGeophys.Res. Lett. 38 L08704

Shi X,Mao J, Thornton PE andHuangM2013 Spatiotemporalpatterns of evapotranspiration in response tomultipleenvironmental factors simulated by the community landmodel Environ. Res. Lett. 8 024012

Sitch S et al 2003 Evaluation of ecosystem dynamics, plantgeography and terrestrial carbon cycling in the LPJ dynamicglobal vegetationmodelGlob. Change Biol. 9 161–85

Sitch S, Cox PM,CollinsW J andHuntingford C 2007 Indirectradiative forcing of climate change through ozone effects onthe land-carbon sinkNature 448 791–4

Sterling SM,Ducharne A and Polcher J 2013The impact of globalland-cover change on the terrestrial water cycleNat. Clim.Change 3 385–90

Swenson SC and LawrenceDM2014Assessing a dry surface layer-based soil resistance parameterization for the communitylandmodel usingGRACE and FLUXNET-MTE dataJ. Geophys. Res.-Atmos. 119 299–312

TaoB, TianH, RenW, Yang J, YangQ,HeR, CaiWand Lohrenz S2014 IncreasingMississippi river discharge throughout the21st century influenced by changes in climate, land use, andatmospheric CO2Geophys. Res. Lett. 41 4978–86

TianH,XuX, LuC, LiuM,RenW,ChenG,Melillo J and Liu J 2011Net exchanges of CO2, CH4, andN2ObetweenChina’sterrestrial ecosystems and the atmosphere and theircontributions to global climate warming J. Geophys. Res.Biogeosci. 116G02011

TianH et al 2012Century-scale responses of ecosystem carbonstorage and flux tomultiple environmental changes in theSouthernUnited States Ecosystems 15 674–94

TeulingA J et al 2009A regional perspective on trends in continentalevaporationGeophys. Res. Lett. 36 L02404

Thornton PE et al 2002Modeling andmeasuring the effects ofdisturbance history and climate on carbon andwater budgetsin evergreen needleleaf forestsAgric. For.Meteorol. 113185–222

TrenberthKE, Fasullo J T andKiehl J 2009 Earth’s global energybudgetBull. Am.Meteorol. Soc. 90 311–23

Ukkola AMandPrentice I C 2013Aworldwide analysis of trends inwater-balance evapotranspirationHydrol. Earth Syst. Sci. 174177–87

Vinukollu RK,Meynadier R, Sheffield J andWoodE F 2011Multi-model,multi-sensor estimates of global evapotranspiration:climatology, uncertainties and trendsHydrol. Process. 253993–4010

WangKC,Wang P, Li ZQ,CribbMand SparrowM2007A simplemethod to estimate actual evapotranspiration from acombination of net radiation, vegetation index, andtemperature J. Geophys. Res. 112D15107

WangKC,DickinsonRE,WildMand Liang S 2010 Evidence fordecadal variation in global terrestrial evapotranspirationbetween 1982 and 2002: 2 results J. Geophys. Res. 115D20113

WangKC andDickinsonRE 2012A review of global terrestrialevapotranspiration: observation,modelling, climatology, andclimatic variabilityRev. Geophys. 50RG2005

Wang L, Good S P andCaylor KK2014Global synthesis ofvegetation control on evapotranspiration partitioningGeophys. Res. Lett. 41 6753–7

Wei Y et al 2014a TheNorth American carbon programmulti-scalesynthesis and terrestrialmodel intercomparison project: II.Environmental driver dataGeosci.Model Dev. 7 2875–93

12

Environ. Res. Lett. 10 (2015) 094008 JMao et al

WeiY et al 2014bNACPMsTMIP: Global andNorth AmericanDriver Data forMulti-Model Intercomparison (Tennessee,USA:OakRidgeNational LaboratoryDistributed ActiveArchiveCenter, OakRidge) data set (http://daac.ornl.gov)

Willmott C J, RoweCMandMintz Y 1985Climatology of theseasonal terrestrial water cycle J. Climatol. 5 589–606

WuP,Christidis N and Stott P 2013Anthropogenic impact onEarth’s hydrological cycleNat. Clim. Change 3 807–10

ZarchMAA, Sivakumar B and SharmaA2015Droughts in awarming climate: a global assessment of standardisedprecipitation index (SPI) and reconnaissance drought index(RDI) J. Hydrol. 526 183–95

ZengN,Mariotti A andWetzel P 2005Terrestrialmechanisms ofinterannual CO2 variabilityGlob. Biogeochem. Cycles 19GB1016

ZengZ, Piao S, LinX, YinG, Peng S, Ciais P andMyneni RB 2012Global evapotranspiration over the past three decades:estimation based on thewater balance equation combinedwith empiricalmodels Environ. Res. Lett. 7 014026

ZhangK,Kimball J S,NemaniRRandRunningSW2010aAcontinuoussatellite-derivedglobal recordof land surface evapotranspirationfrom1983 to2006WaterResour.Res.46W09522

Zhang Y, Leuning R,Hutley L B, Beringer J,McHugh I andWalker J P 2010bUsing long-termwater balances toparameterize surface conductances and calculate evaporationat 0.05° spatial resolutionWater Resour. Res. 46W05512

Zhang Y, Leuning R, ChiewFHS,Wang E, Zhang L, LiuC, Sun F,PeelMC, ShenY and JungM2012Decadal trends inevaporation fromglobal energy andwater balancesJ. Hydrometeorol. 13 379–91

13

Environ. Res. Lett. 10 (2015) 094008 JMao et al