Assessing agricultural risks of climate change in the21st century in a global gridded cropmodel intercomparisonCynthia Rosenzweiga,b,1, Joshua Elliottb,c, Delphine Deryngd, Alex C. Ruanea,b, Christoph Müllere, Almut Arnethf,Kenneth J. Booteg, Christian Folberthh, Michael Glotteri, Nikolay Khabarovj, Kathleen Neumannk,l, Franziska Pionteke,Thomas A. M. Pughf, Erwin Schmidm, Elke Stehfestk, Hong Yangh, and James W. Jonesg

aNational Aeronautics and Space Administration Goddard Institute for Space Studies, New York, NY 10025; bColumbia University Center for Climate SystemsResearch, New York, NY 10025; cUniversity of Chicago Computation Institute, Chicago, IL 60637; dTyndall Centre and School of Environmental Sciences,University of East Anglia, Norwich NR4 7TJ, UK; ePotsdam Institute for Climate Impact Research, 14473 Potsdam, Germany; fInstitute of Meteorology andClimate Research, Atmospheric Environmental Research, Karlsruhe Institute of Technology, 82467 Garmisch-Partenkirchen, Germany; gAgricultural andBiological Engineering Department, University of Florida, Gainesville, FL 32611; hEAWAG – Swiss Federal Institute of Aquatic Science and Technology 8600Dübendorf, Switzerland; iDepartment of the Geophysical Sciences, University of Chicago, Chicago, IL 60637; jEcosystems Services and Management Program(ESM), International Institute for Applied Systems Analysis (IIASA), Laxenburg A-2361, Austria; kPlanbureau voor de Leefomgeving (NetherlandsEnvironmental Assessment Agency), 3720 AH, Bilthoven, The Netherlands; lWageningen University, 6700 AK, Wageningen, The Netherlands; and mUniversityof Natural Resources and Life Sciences, 1180 Vienna, Austria

Edited by Hans Joachim Schellnhuber, Potsdam Institute for Climate Impact Research, Potsdam, Germany, and approved June 4, 2013 (received for reviewJanuary 31, 2013)

Here we present the results from an intercomparison of multipleglobal gridded crop models (GGCMs) within the framework of theAgricultural Model Intercomparison and Improvement Project andthe Inter-Sectoral Impacts Model Intercomparison Project. Resultsindicate strong negative effects of climate change, especially athigher levels of warming and at low latitudes; models that includeexplicit nitrogen stress project more severe impacts. Across sevenGGCMs, five global climate models, and four representative con-centration pathways, model agreement on direction of yieldchanges is found in many major agricultural regions at both lowand high latitudes; however, reducing uncertainty in sign of re-sponse in mid-latitude regions remains a challenge. Uncertaintiesrelated to the representation of carbon dioxide, nitrogen, and hightemperature effects demonstrated here show that further re-search is urgently needed to better understand effects of climatechange on agricultural production and to devise targeted adapta-tion strategies.

food security | AgMIP | ISI-MIP | climate impacts | agriculture

The magnitude, rate, and pattern of climate change impacts onagricultural productivity have been studied for approximately

two decades. To evaluate these impacts, researchers use bio-physical process-based models (e.g., refs. 1–5), agro-ecosystemmodels (e.g., ref. 6), and statistical analyses of historical data(e.g., refs. 7 and 8). Although these and other methods have beenwidely used to forecast potential impacts of climate change onfuture agricultural productivity, the protocols used in previousassessments have varied to such an extent that they constrain cross-study syntheses and limit the ability to devise relevant adaptationoptions (9, 10). In this project we have brought together sevenglobal gridded crop models (GGCMs) for a coordinated set ofsimulations of global crop yields under evolving climate conditions.This GGCM intercomparison was coordinated by the Agricul-

tural Model Intercomparison and Improvement Project (AgMIP;11) as part of the Inter-Sectoral Impact Model IntercomparisonProject (ISI-MIP; 12). In order to facilitate analyses across modelsand sectors, all global models are driven with consistent bias-corrected climate forcings derived from the Coupled ModelIntercomparison Project Phase 5 (CMIP5) archive (13). The objec-tives are to (i) establish the range of uncertainties of climatechange impacts on crop productivity worldwide, (ii) determine keydifferences in current approaches used by cropmodeling groups inglobal analyses, and (iii) propose improvements in GGCMs and inthe methodologies for future intercomparisons to produce morereliable assessments.

We examine the basic patterns of response to climate acrosscrops, latitudes, time periods, regional temperatures, and atmo-spheric carbon dioxide concentrations [CO2]. In anticipation ofthe wider scientific community using these model outputs and theexpanded application of GGCMs, we introduce these models andpresent guidelines for their practical application. Related studiesin this special issue focus on crop water demand and the fresh-water supply for irrigation (14), the application of the crop modelresults as part of wider intersectoral analyses (15), and the in-tegration of crop-climate impact assessments with agro-economicmodels (16).

1. Global Gridded Crop ModelsDetails of the seven global crop models used in this study areprovided in SI Appendix, Tables S1–S6. These include the En-vironmental Policy Integrated Climate Model [EPIC (17–20);originally the Erosion Productivity Impact Calculator (17)], theGeographic Information System (GIS)-based EnvironmentalPolicy Integrated Climate Model (GEPIC; 18–21), the GlobalAgroEcological Zone Model in the Integrated Model to Assessthe Global Environment (GAEZ-IMAGE; 22, 23), the Lund-Potsdam-Jena managed Land Dynamic Global Vegetation andWater Balance Model (LPJmL; 24–26), the Lund-Potsdam-Jena

Significance

Agriculture is arguably the sector most affected by climatechange, but assessments differ and are thus difficult to com-pare. We provide a globally consistent, protocol-based, multi-model climate change assessment for major crops with explicitcharacterization of uncertainty. Results with multimodel agree-ment indicate strong negative effects from climate change, es-pecially at higher levels of warming and at low latitudes wheredeveloping countries are concentrated. Simulations that con-sider explicit nitrogen stress result in much more severe impactsfrom climate change, with implications for adaptation planning.

Author contributions: C.R., J.E., D.D., A.C.R., C.M., K.J.B., M.G., F.P., and J.W.J. designedresearch; J.E., D.D., C.M., A.A., C.F., M.G., N.K., K.N., T.A.M.P., E. Schmid, E. Stehfest, andH.Y. performed research; J.E., D.D., A.C.R., and C.M. contributed analytic tools; C.R., J.E.,D.D., A.C.R., C.M., and K.J.B. analyzed data; and C.R., J.E., D.D., A.C.R., and C.M. wrotethe paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. E-mail: cynthia.rosenzweig@nasa.gov.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1222463110/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1222463110 PNAS Early Edition | 1 of 6

AGRICU

LTURA

LSC

IENCE

SSU

STAINABILITY

SCIENCE

SPEC

IALFEATU

RE

General Ecosystem Simulator with Managed Land (LPJ-GUESS;24, 27, 28), the parallel Decision Support System for Agro-tech-nology Transfer [pDSSAT; 29, 30; using the Crop EnvironmentResource Synthesis (CERES) models for maize, wheat, and riceand the Crop Template approach (CROPGRO) for soybean], andthe Predicting Ecosystem Goods And Services Using Scenariosmodel (PEGASUS; 31).These models differ in regard to model type, inclusion and

parameterization of soil and crop processes, management inputs,and outputs. These dissimilarities must be taken into account ininterpreting the results of the intercomparison and in the use ofresults in other analyses (SI Appendix, Table S1). Examples in-clude the biological and environmental stresses affecting crops ineach model and the treatment of how increasing [CO2] affectsplant growth and yield. GAEZ-IMAGE, LPJ-GUESS, andLPJmL focus on water and temperature responses, whereas theother four models also consider stresses related to nitrogen de-ficiency and severe heat during various stages of development. Inaddition to these, pDSSAT considers oxygen stress, PEGASUSconsiders phosphorus and potassium stresses, and EPIC andGEPIC both consider oxygen, phosphorus, bulk density, andaluminum stresses.

2. Comparison with Intergovernmental Panel on ClimateChange Fourth Assessment Report ResultsA relevant question is to what extent findings of this substantialeffort of coordinated GGCM modeling are different from whatwas reported in the Intergovernmental Panel on Climate ChangeFourth Assessment Report (IPCC AR4; 32) (Fig. 1). Cropmodeling results in the IPCC AR4 showed that small beneficialimpacts on rainfed crop yields may be found in mid- and high-latitude regions with moderate-to-medium local increases intemperature (1–3 °C) along with associated [CO2] increase andrainfall changes (figure 5.2 in ref. 32; reproduced as orange dotsand quadratic fit in Fig. 1). In low-latitude regions, even moderatetemperature increases (1 to 2 °C) were found to have negativeyield impacts for major cereals, because the climate of manytropical agricultural regions is already quite close to the high-temperature thresholds for suitable production of these cereals(33, 34). Furthermore, increases in tropical temperatures can leadto greater evaporative demand and thus water stress on crops.We find that general patterns of the GGCM results are simi-

lar, especially among those models that simulate nitrogen stresson crops and include fertilizer application rates based on ob-servational databases (red line in Fig. 1). GGCMs without ni-trogen stress tend to be more optimistic in yield response (greenline in Fig. 1). The 15–85% range of all GGCM results (indicatedby the shaded envelope) suggests that climate impacts on trop-ical croplands are generally more negative than the mid- andhigh-latitude impacts. There is considerable variation in responsewithin these broader latitudinal bands, with mid- and high-latitudecrop yields spanning both positive and negative responses, es-pecially at high levels of temperature change (which are alsoassociated with higher [CO2]). The GGCM results generally dis-play a wider range of uncertainty compared to the AR4 results,reflecting the much broader geographical coverage, projectedtemperature, and diversity of crop models represented in thecurrent study.

3. GGCM Structural DifferencesA major source of uncertainties in projected climate impactsacross the globe is the result of variations in GGCM approaches,assumptions, and structures. Documentation of these differencesis fundamental to at least partially constraining them and to im-proving analyses of ensemble crop projections.

3.1 Model Types. The seven GGCMs may be grouped into threetypes according to their original purpose, structure, and pro-cesses: site-based crop models (EPIC, GEPIC, and pDSSAT),agro-ecosystem models (LPJ-GUESS, LPJmL, and PEGASUS),and agro-ecological zone models (GAEZ-IMAGE) (SI Appen-dix, Fig. S1). A critical question is whether two models from thesame lineage, such as EPIC and GEPIC, and LPJ-GUESS andLPJmL are truly independent. For instance, in the case of EPICand GEPIC, the same model version is used (0810), but with dif-ferent parameterizations and assumptions about soil and manage-ment input data that are reflected in the variations in their results.Site-based models were developed to simulate processes at the

field scale, and include dynamic interactions among crop, soil,atmosphere, and management components (2, 20, 30). Thesemodels are often calibrated and validated with data from agro-nomic field experiments. The versions of the site-based modelsused in this study have been developed to run simulations onglobal grids, as has been done using DSSAT (29, 35–37).Agro-ecosystem models were primarily developed to simulate

carbon and nitrogen dynamics, surface energy balance, and soilwater balance. The LPJmL and LPJ-GUESS models are dy-namic global vegetation models that simulate the full globalcarbon and water cycles. Vegetation dynamics and agriculturalmodules were originally introduced to improve the simulations ofthese cycles. PEGASUS is a simple global vegetation modeldesigned to test how agroecosystems respond to climate changeand to evaluate potential benefits of various farming adaptationoptions at the global scale.

0 2 4 6 860

40

20

0

20

40

60

0 2 4 6 860

40

20

0

20

40

60

0 2 4 6 860

40

20

0

20

40

60

0 2 4 6 860

40

20

0

20

40

60

Regional mean temperature change °C

Maize, low latitude

Wheat, low latitude

Rice, low latitude

Soy, low latitude

0 2 4 6 860

40

20

0

20

40

60

Yie

ldch

ange

0 2 4 6 860

40

20

0

20

40

60

Yiel

dch

ange

0 2 4 6 860

40

20

0

20

40

60

Yiel

dch

ange

0 2 4 6 860

40

20

0

20

40

60

Regional mean temperature change °C

Yiel

dch

ange

Maize, mid- to high-latitude

Wheat, mid- to high-latitude

Rice, mid- to high-latitude

Soy, mid- to high-latitude

IPCC AR4GGCMs without explicit N stressGGCMs with explicit N stress

Fig. 1. Mean relative yield change (%) from reference period (1980–2010)compared to local mean temperature change (°C) in 20 top food-producingregions for each crop and latitudinal band. Results shown for the 7 GGCMs (6for rice) for all GCM combinations of RCP8.5 compared to results from IPCCAR4 (represented as orange dots and quadratic fit; 36). Quadratic least-squares fits are used to estimate the general response for the GGCMs withexplicit nitrogen stress (EPIC, GEPIC, pDSSAT, and PEGASUS; red line) and forthose without (GAEZ-IMAGE, LPJ-GUESS, and LPJmL; green line). The 15–85% range of all models for each ¼°C band is represented in gray. Limits oflocal temperature changes reflect differences in projected warming in cur-rent areas of cultivation.

2 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1222463110 Rosenzweig et al.

The agro-ecological zone methodology (used here by GAEZ-IMAGE) was developed to assess agricultural resources andpotential at regional and global scales and has been embeddedinto integrated assessment models for global environmentalchange (6, 23).

3.2 Model Processes. Crop processes simulated in all or some ofthe GGCMs include leaf area development, light interceptionand utilization, yield formation, crop phenology, root distribu-tion responsiveness to water availability at soil depth, water andheat stress, soil–crop–atmosphere water cycle dynamics, evapo-transpiration, soil carbon and nitrogen cycling, and the effect of[CO2] (SI Appendix, Table S1). All of the GGCMs explicitlysimulate the effects of temperature and water on crop growth;fewer models simulate, for example, the effects of specific heatstress at critical stages of crop development or the effects ofwater-logging on root function. GGCMs differ as to their simu-lation of some processes in individual crops, such as whichmodels simulate rice phenology as sensitive to day length as wellas temperature.Thus the GGCMs vary in their interactive responses to in-

creasing [CO2], rising temperature, and changes in water avail-ability, which are the core characteristics of projected climatechanges in agricultural regions around the world (32). How theGGCMs handle these factors and their interactions with nutrientavailability (especially N) has significant impacts on the results (41).This GGCM intercomparison focuses on long-term yield levels

affected by inputs (climate, [CO2], water, nutrients) rather thanon short-term shocks. The effects of pests and diseases are notincluded explicitly; pest vulnerability may be implicitly includedthrough calibration to observed yields in some of the models.LPJmL and PEGASUS, for instance, reflect the level of farmingintensification and technological inputs (such as the use ofpesticides). However this method does not allow for estimationof how the effects of pests and diseases may change underchanging climate conditions, an important area for future modeldevelopment.Climate change influences on short-term temperature extremes,

monsoon dynamics, and the frequency and intensity of pre-cipitation may also play a substantial role in the nature of futureagricultural impacts. GCMs do not fully resolve these features,and the representation of corresponding stresses remains an activearea of GGCM development.

3.3 Model Inputs. A key contrast among the GGCMs is in nutrientresponse in regard to underlying soil properties and to nutrientsapplied (nitrogen, phosphorus, and potassium), amount, andtiming. Disparities in the resulting nutrient stress may affect thesensitivity of yields to climate change because climate stresses andbenefits may also interact with (or be overwhelmed by) nutrientstresses. Alternate approaches in the GGCMs’ fertilization andnutrient schemes therefore need to be taken into account ininterpreting crop yield responses to [CO2] and other variables.GGCM differences in the simulation of water availability and

the application of irrigation also have a direct effect on climatesensitivity in irrigated regions. While the GGCMs deviate in howwater availability is determined, the effects of these deviationswere reduced by testing two irrigation scenarios: 1) no irrigation,and 2) full irrigation (assuming water is available to fully irrigatecrops) (see SI Appendix). In GEPIC, full irrigation was set as acomplete elimination of water stress of crops. In other GGCMs,full irrigation does not necessarily eliminate water stress com-pletely, as irrigation events are triggered by model-specific soilmoisture thresholds (rainfed and irrigated production responsesare shown in Fig. S5). In some cases, the ability of the crop plantto transpire water may not be sufficient to satisfy the atmosphericdemand (i.e., stomata may close despite full irrigation).

3.4 Model Procedures.An important disparity in GGCM outputs iswhether the models calculate actual or potential yields as theprimary output. The GAEZ-IMAGE and LPJ-GUESS results

represent potential yields, unlimited by nutrient or managementconstraints and without calibration of growth parameters to repro-duce historical yields. They are best suited to studies that are de-signed to advance scientific understanding of the plant-atmosphereprocesses being represented and their sensitivity to climatic stresses,rather than for economic forecasts or sensitivity to soil edaphicconditions. LPJmL is similar to LPJ-GUESS in that nitrogenstress is not explicitly represented; however, growth parametersin the model are calibrated so that simulations over the historicalperiod reproduce realistic average yield patterns (see SI Appendixfor details). GEPIC, PEGASUS, and pDSSAT used historicalpatterns of fertilizer application rates, while EPIC used stan-dardized low-, moderate-, and high-input management systemswith thresholds that trigger fertilizer and irrigation automatically.All four of these models explicitly represent nitrogen stress. Theissue of actual vs. potential yields is further complicated by thepresence of numerous other “yield gap” factors, including varia-tions in cultivars and farmer management, as well as soil char-acteristics, pests, diseases, and weeds (38).

4. Current and Future Yield Simulations4.1 Simulation of Current Crop Yields. The seven GGCMs largelyreproduce relative patterns of current crop yields (39) at multi-national regional scales but are dissimilar in the levels of theirbase yields (maize: Fig. 2; wheat, rice, and soybean results in SIAppendix, Figs. S2–S4). PEGASUS displayed the largest regionalvariation in simulated yields, whereas GAEZ-IMAGE displayedthe least. Each model has regions where crop yield simu-lations vary markedly from the patterns observed in the ref-erence period.LPJmL and LPJ-GUESS vary in reproducing current maize

yields, even though they both have a common base model, as doEPIC and GEPIC. Each of these two GGCM pairs vary in pa-rameter settings, assumptions, inputs (e.g., management, fertil-izer), processes (e.g., carbon allocation), and model procedures

EPIC

GEPIC

pDSSAT

M3-Observations

LPJ-GUESS

LPJmL

PEGASUS

IMAGE

Baseline Maize Yield (t/ha)

>1050

A B

C D

E F

G H

Fig. 2. Average reference period (1980–2010) GGCM maize yield (A–F, H),rescaled to a common global average to make the spatial patterns moreapparent, and historical yield M3 observation set (G) (39). Note that becausesome models are calibrated and others are not and because some modelssimulate potential rather than actual yields, it is not advisable to comparethe absolute yields in the ensemble with observations.

Rosenzweig et al. PNAS Early Edition | 3 of 6

AGRICU

LTURA

LSC

IENCE

SSU

STAINABILITY

SCIENCE

SPEC

IALFEATU

RE

(e.g., calibration) that are reflected in the wide variations in theirresults (SI Appendix).

4.2 Global Relative Yield Changes by Crop. Despite the differencesamong models in their assumed inputs and simulations of ab-solute yields, relative yield changes provide a more consistent setof results for comparison across models and with previously re-ported climate change impact results. When taken as a multi-GGCMandmulti-GCM ensemble, global results for relative changes in themajor crops under representative concentration pathway 8.5(RCP8.5; 42) with CO2 effects show broad agreement with resultsand regional patterns seen in previous studies (Fig. 3, Upper).End-of-century (2070–2099) maize yield changes with CO2

effects for RCP8.5 show substantial impacts and broad agree-ment among GGCMs, at least as to the sign of the effect. Resultsfor maize and wheat indicate high-latitude increases and low-lat-itude decreases with general agreement among models. However,the quality, depth, and hydraulic properties of soils for agriculturalproduction at high latitudes merit further investigation. Results forrice and soybean are consistent in the mid- and high-latituderegions showing yield increases, but show less agreement amongmodels in the tropical regions where median changes are small.Generally, the tropics are subject to more severe (or less benefi-cial) climate change impacts whereby CO2 fertilization does notcompensate for increases in water demand and shortening ofalready-short growing periods for annual C3 crops.

When the results are grouped by GGCMs with and withoutexplicit nitrogen fertilization (Lower Left and Lower Right inFig. 3; red and green lines in Fig. 1), results are substantiallymore negative with explicit nitrogen fertilization than without.The GGCMs with explicit nitrogen fertilization may captureenhanced dynamics of crop growth and yield interactions withCO2 fertilization; experiments show lower CO2 enhancement ofyield under nitrogen limitation (41). Further work is needed tounderstand how these interactions affect the GGCM results andidentify how variations in crop model parameter values alsoaffect simulated yields (e.g., ref. 43).

4.3 Sensitivity of Yield Response to CO2. Projections of global rel-ative yield changes under RCP8.5 differ substantially amongGGCMs but also between simulations with and without CO2effects for maize, wheat, rice, and soybean (Fig. 4). By the end ofthe 21st century, most GGCMs show a range of approximately ±10% yield change across the five GCM scenarios when CO2effects are included (GCMs cause nearly double that range forPEGASUS and only half that range for GAEZ-IMAGE). Rel-ative global average model response to climate is more similarand much more negative across tropical and midlatitude bandsonce CO2 effects are removed, indicating that crop model pa-rameterization of CO2 effects remains a crucial area of research.Relative yield changes with and without CO2 effects are muchcloser in C4 maize than in the C3 crops.

Maize

Rice Soy

Wheat

Maize MaizeWheat Wheat

Rice RiceSoy Soy

All GGCMs

GGCMs with explicit N stress GGCMs without explicit N stress

<-50 >500

%

Fig. 3. Median yield changes (%) for RCP8.5 (2070–2099 in comparison to 1980–2010 baseline) with CO2 effects over all five GCMs x seven GGCMs (6 GGCMs for rice)for rainfed maize (35 ensemble members), wheat (35 ensemble members), rice (30 ensemble members), and soy (35 ensemble members). Hatching indicates areaswhere more than 70% of the ensemble members agree on the directionality of the impact factor. Gray areas indicate historical areas with little to no yield capacity.The bottom 8 panels show the corresponding yield change patterns over all five GCMs x four GGCMs with nitrogen stress (20 ensemble members from EPIC, GEPIC,pDSSAT, and PEGASUS; except for rice which has 15) (Left); and 3 GGCMs without nitrogen stress (15 ensemble members from GAEZ-IMAGE, LPJ-GUESS, and LPJmL).

4 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1222463110 Rosenzweig et al.

In near decades, relative yield changes display a lower range,both with and without CO2 effects, but after the 2050s that rangewidens considerably. LPJ-GUESS, a potential yield model thatallows for nutrient-unlimited yield increases, consistently dis-plays the highest relative changes with CO2 effects for all crops.The projected yield changes both with and without CO2 effects

for PEGASUS (an ecosystem model) are more negative than theLPJ ecosystem models (note that PEGASUS does not simulaterice), which is likely due to its utilization of radiation use effi-ciency (RUE) instead of leaf-level photosynthesis (40) for CO2effects and the inclusion of explicit heat stress. RUE-basedmodels simulate a universal saturating response to CO2 and af-fect water efficiency via adjustment of canopy conductance. Inthe leaf-level models, stomatal opening controls both photo-synthesis (CO2 availability) and transpiration. Recently, Free-AirCO2 Enrichment (FACE) experiment results (40) are being usedmore intensively to calibrate and test crop models in AgMIP.

4.4 Quantifying Uncertainty from GCMs and RCPs. GCMs and RCPscontribute substantially to the uncertainties of the results (Fig.5). Uncertainty is higher for soybean and rice than for maize andwheat, because they have more concentrated production areasand are therefore more sensitive to regional differences in GCMprojections. Uncertainties are greater in the later decades of thecentury, where GCM inputs and GGCM results can lead touncertainties several times larger in the highest RCP8.5 than inthe lowest RCP2.6. Uncertainty is higher for all crops when CO2effects are included, especially in soybean (which is not directlylimited by nitrogen) and in the end of the century when [CO2] ishighest. Note that the RCP nomenclature is misleading for earlierdecades, because RCP4.5 actually has slightly higher [CO2] thanRCP6.0 until ∼2060 (42).

5. Discussion and ConclusionsThe models used in this GGCM intercomparison are tools toanalyze the response of crops to climate change, and to betterunderstand risks and opportunities in regard to food productionand food security. For this information to be useful for decisionmakers, it needs to include analysis of sources of uncertainty dueto multiple greenhouse gas emissions pathways, climate models,and crop impact models (44). The work presented here begins tocharacterize the uncertainty cascade for GGCM simulations, in-cluding greenhouse gas emission scenarios, global climate simu-lations, variations in structure and implementation in crop models,and assumptions about agricultural management, in a frameworkthat can be compared across sectors.

Because of such variations in model structure, processes, inputs,assumptions, parameterizations, and outputs, the ensemble resultsfrom the GGCM intercomparison need to be used with care andmay not be appropriate for certain studies (see recommendationson data use in SI Appendix). Although the experimental designand climate change scenarios were meant to harmonize simu-lations to facilitate full comparability, several differences remainthat affect the GGCMs’ response to climate change and theirutility for different types of assessments, including economicanalyses. Particularly important are the parameterization of CO2effects, handling of fertilizer applications, simulation of actual vs.potential yields, and the extent of calibration. AgMIP is addressingthese in continuing work.Given these important caveats, we can conclude that the re-

sults from the GGCMs used in this study show general agree-ment with previous results, especially for those models that includenitrogen stress (e.g., 6, 32, 45). They indicate negative impacts onmajor crops in many agricultural regions at higher levels ofwarming. The inclusion of ecosystem-based models in this analysishas increased the range of uncertainty (previous analyses primarilyused site-based models). Relative global average model responseto climate is more similar once CO2 effects are removed, indicatingthat model parameterization of CO2 effects (on both photosyn-thesis and transpiration) remains a vital area of research.There is ample reason to be concerned in regard to climate

change and crop production. Many regions throughout the worldare projected to experience climate change-induced reductions

Fig. 4. Relative change (%) in RCP8.5 decadal mean production for eachGGCM (based on current agricultural lands and irrigation distribution) fromensemble median for all GCM combinations with (solid) and without (dashed)CO2 effects for maize, wheat, rice, and soy; bars show range of all GCM com-binations with CO2 effects. GEPIC, GAEZ-IMAGE, and LPJ-GUESS only contrib-uted one GCM without CO2 effects.

Maize Wheat Rice Soy

With CO2effects

Without CO2 effects

A

B

Fig. 5. Absolute deviation of decadal average production changes fromensemble median yield changes (as fraction of 1980–2010 reference periodmean production) for all GCM × GGCM combinations in RCP2.6 (dark blue),RCP4.5 (light blue), RCP6.0 (orange), and RCP8.5 (red) for maize, wheat, rice,and soy with (Upper) and without (Lower) CO2 effects. Simulations in A withCO2 effects included five GCMs and seven GGCMs (35 members), whereasGAEZ-IMAGE, GEPIC, and LPJ-GUESS ran only a single GCM without CO2

effects, resulting in 23 members in B.

Rosenzweig et al. PNAS Early Edition | 5 of 6

AGRICU

LTURA

LSC

IENCE

SSU

STAINABILITY

SCIENCE

SPEC

IALFEATU

RE

in crop yields in the climate scenario–crop model ensemble testedhere, and additional challenges are mounting (e.g., pests, watersupply, and soil degradation). The 2012 drought in the United Statesled to a reduction of maize yields of up to 25% (which is moderatecompared with the impacts projected here for some regions athigher levels of temperature increase), but US maize exportsdeclined by an even greater percentage (46). Although some high-latitude regions may become more climatically viable for crops,further study is needed to determine whether soil quality is suf-ficient for sustained agricultural production in these locations.AgMIP is dedicated to exploring the underlying mechanisms

behind GGCM differences and to quantifying uncertainties inclimate change impact assessments. AgMIP further endeavors toimprove agricultural models and expand the community of trans-disciplinary modelers, thus supporting effective adaptation andmitigationdecisions inagriculture at both global and regional scales.

Materials and MethodsCritical sources of uncertainty for climate change impacts on agricultural pro-ductivity are identified and characterized, including contrasts in results arisingfrom a range of global crop models, global climate models, and RCPs (42). SIAppendix provides a full description of materials and methods.

Simulations are driven using 20 climate scenarios from the Coupled ModelIntercomparison Project Phase 5 archive with five GCMs and four RCPs, each

bias-corrected at daily resolution based on the historical Water and GlobalChange forcing dataset derived from the European Centre for Medium-Range Weather Forecasts 40 Year Re-analysis (13). The reference period usedthroughout this analysis is 1980–2010. All models submitted simulations withCO2 effects for five GCMs for maize, wheat, and soybean (35 members). Allmodels except PEGASUS simulated five GCMs for rice (30 members). Allmodels simulated the Hadley Centre Global Environment Model (HadGEM)climate model without CO2 effects, but only LPJmL, pDSSAT, PEGASUS, andEPIC simulated the other four GCMs (23 members for maize and wheatwithout CO2 effects, and 18 members for rice).

ACKNOWLEDGMENTS. We thank the AgMIP research community and theISI-MIP team for their contributions to this effort; the US Department ofAgriculture and the United Kingdom Department for International De-velopment for their support of AgMIP; the World Climate ResearchProgramme’s Working Group on Coupled Modelling, which is responsiblefor CMIP; the climate modeling groups for producing and making availabletheir model output; and two anonymous reviewers for their helpful com-ments. At the Columbia Center for Climate Systems Research, we thank ErikMencos-Contreras and Shari Lifson for research and graphics assistance. ForCMIP the US Department of Energy’s Program for Climate Model Diagnosisand Intercomparison provides coordinating support and led development ofsoftware infrastructure in partnership with the Global Organization forEarth System Science Portals. The research leading to these results has re-ceived funding from the European Union’s Seventh Framework ProgrammeFP7/2007-2013 under Grant Agreement 266992.

1. Brisson N, et al. (2003) An overview of the crop model STICS. Eur J Agron 18(3-4):309–332.

2. Jones JW, et al. (2011) Use of crop models for climate-agricultural decisions. Hand-book of Climate Change and Agroecosystems. ICP Series on Climate Change Impacts,Adaptation, and Mitigation, eds Hillel D, Rosenzweig C (Imperial College Press, Lon-don), Vol 1, pp 131–157.

3. Keating BA, et al. (2003) An overview of APSIM, a model designed for farming sys-tems simulation. Eur J Agron 18(3-4):267–288.

4. van Ittersum MK, Donatelli M (2003) Modeling cropping systems: Science, softwareand applications. Eur J Agron 16:309–332.

5. Challinor AJ, Wheeler TR, Craufurd PQ, Slingo JM, Grimes DLF (2004) Design and optimi-zation of a large-area process-based model for annual crops. Agr For Meteorol 124:99–120.

6. Fischer G, Shah M, van Velthuizen H, Nachtergaele FO (2002) Global Agro-EcologicalAssessment for Agriculture in the 21st Century: Methodology and Results. (ResearchReport RR-02-02). (International Institute for Applied Systems Analysis and Food andAgriculture Organization of the United Nations, Laxenburg, Austria).

7. Schlenker W, Hanemann WM, Fisher AC (2006) The impact of global warming on U.S.agriculture: An econometric analysis of optimal growing conditions. Rev Econ Stat88(1):113–125.

8. Lobell DB, Burke MB (2010) On the use of statistical models to predict crop yield re-sponses to climate change. Agr For Meteorol 150(11):1443–1452.

9. White JW, Hoogenboom G, Kimball BA, Wall GW (2011) Methodologies for simu-lating impacts of climate change on crop production. Field Crops Res 124(3):357–368.

10. Rötter RP, Carter TR, Olesen JE, Porter JR (2011) Crop-climate models need an over-haul. Nat Climate Change 1(4):175–177.

11. Rosenzweig C, et al. (2013) The Agricultural Model Intercomparison and Improve-ment Project (AgMIP): Protocols and pilot studies. Agric For Meteorol 170:166–182.

12. Warszawski L, et al. (2013) The Inter-Sectoral Impact Model Intercomparison Project(ISI–MIP): Project framework. Proc Natl Acad Sci USA, 10.1073/pnas.1312330110.

13. Hempel S, Frieler K, Warszawski L, Shewe J, Piontek F (2013) A trend-preserving biascorrection – the ISI-MIP approach. Earth Syst Dynam 4:219–236.

14. Elliott J, et al. (2013) Constraints and potentials of future irrigation water availabilityon global agricultural production under climate change. Proc Natl Acad Sci USA,10.1073/pnas.1222474110.

15. Piontek F, et al. (2013) Multisectoral climate impact hotspots in a warming world. ProcNatl Acad Sci USA, 10.1073/pnas.1222471110.

16. Nelson GC, et al. (2013) Climate change effects on agriculture: Economic responses tobiophysical shocks. Proc Natl Acad Sci USA, 10.1073/pnas.1222465110.

17. Williams JR, et al. (1990) EPIC-Erosion/Productivity Impact Calculator. Technical Bul-letin Number 1768 (US Department of Agriculture Agricultural Research Service,Springfield, VA).

18. Williams JR (1995) The EPIC. Computer Models of Watershed Hydrology, Chapter 25,ed Singh VP (Water Resources Publications, Littleton, CO), pp 909–1000.

19. Gassman PW, et al. (2004) Historical development and applications of the EPIC and APEXmodels. ASAE/CSAE meeting paper no. 042097. Available at www.card.iastate.edu.

20. Izaurralde RC, et al. (2006) Simulating soil C dynamics with EPIC: Model descriptionand testing against long-term data. Ecol Modell 192(3-4):362–384.

21. Liu J, et al. (2007) GEPIC—modeling wheat yield and crop water productivity withhigh resolution on a global scale. Agr Syst 94(2):478–493.

22. Leemans R, Solomon AM (1993) Modeling the potential change in yield and distri-bution of the earth’s crops under a warmed climate. Clim Res 3:79–96.

23. Bouwman AF, Kram T, Klein T, Goldewijk K, eds (2006) Integrated Modelling ofGlobal Environmental Change. An Overview of IMAGE 2.4 (PBL Netherlands Envi-ronmental Assessment Agency, The Hague).

24. Bondeau A, et al. (2007) Modelling the role of agriculture for the 20th century globalterrestrial carbon balance. Glob Change Biol 13(3):679–706.

25. Fader M, et al. (2010) Virtual water content of temperate cereals and maize: Presentand potential future patterns. J Hydrol (Amst) 384(3-4):218–231.

26. Waha K, et al. (2012) Climate-driven simulation of global crop sowing dates. Glob EcolBiogeogr 21(2):247–259.

27. Smith B, et al. (2001) Representation of vegetation dynamics in the modeling ofterrestrial ecosystems: Comparing two contrasting approaches within European cli-mate space. Glob Ecol Biogeogr 10(6):621–637.

28. Lindeskog M, et al. (2013) Implications of accounting for land use in simulations of eco-system services and carbon cycling in Africa. Earth System Dynamics Discussions 4:235–278.

29. Elliott J, Glotter M, Best N, Kelly D, Wilde M, Foster I (2013) The parallel system forintegrating impact models and sectors (pSIMS). Proceedings of the Conference onExtreme Science and Engineering Discovery Environment: Gateway to Discovery(XSEDE ’13), Association for Computing Machinery, 21:1–8.

30. Jones JW, et al. (2003) The DSSAT cropping system model. Eur J Agron 18(3-4):235–265.31. Deryng D, SacksWJ, Barford CC, Ramankutty N (2011) Simulating the effects of climate and

agricultural management practices on global crop yield. Global Biogeochem Cycles 25(2).32. Easterling WE, et al. (2007) Food, fibre and forest products. Climate Change 2007:

Impacts, Adaptation and Vulnerability: Contribution of Working Group II to theFourth Assessment Report of the Intergovernmental Panel on Climate Change, edsParry ML, Canziani OF, Palutikof JP, van der Linden PJ, Hanson CE (Cambridge UnivPress, Cambridge, UK) pp 273–313.

33. Peng S, et al. (2004) Rice yields decline with higher night temperature from globalwarming. Proc Natl Acad Sci USA 101(27):9971–9975.

34. Hatfield JL, et al. (2011) Climate impacts on agriculture: Implications for crop pro-duction. Agron J 103(2):351–370.

35. Nelson GC, et al. (2009) Climate Change: Impact on Agriculture and Costs of Adap-tation (International Food Policy Research Institute, Washington, DC).

36. Nelson GC, et al. (2010) Food Security, Farming, and Climate Change to 2050: Scenarios,Results, Policy Options (International Food Policy Research Institute, Washington, DC).

37. Jones PG, Thornton PK (2003) The potential impacts of climate change on maizeproduction in Africa and Latin America in 2055. Glob Environ Change 13(1):51–59.

38. Lobell DB, Cassman KG, Field CB (2009) Crop yield gaps: Their importance, magni-tudes, and causes. Annu Rev Env Resour 34:179–204.

39. Monfreda C, Ramankutty N, Foley JA (2008) Farming the planet: 2. Geographic dis-tribution of crop areas, yields, physiological types, and net primary production in theyear 2000. Global Biogeochem Cycles 22(1):19.

40. Farquhar GD, Caemmerer S, Berry JA (1980) A biochemical model of photosyntheticCO2 assimilation in leaves of C3 species. Planta 149:78–90.

41. Kimball BA (2010) Lessons from FACE: CO2 Effects and Interactions with Water, Ni-trogen, and Temperature. The Handbook of Climate Change and Agroecosystems,eds Hillel D, Rosenzweig C (Imperial College Press, Singapore), pp 87–107.

42. Moss RH, et al. (2010) The next generation of scenarios for climate change researchand assessment. Nature 463(7282):747–756.

43. Iizumi T, et al. (2011) Probabilistic evaluation of climate change impacts on paddy riceproductivity in Japan. Clim Change 107:391–415.

44. Müller C (2011) Harvesting from uncertainties. Nature Climate Change 1:253–254.45. Rosenzweig C, Parry ML (1994) Potential impact of climate change on world food

supply. Nature 367:133–138.46. USDA (2012) Feed outlook. Capehart, Allen, and Bond. A Report from the Economic

Research Service. December 13, 2012. Available at http://www.ers.usda.gov/publications/fds-feed-outlook. Accessed January 15, 2013.

6 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1222463110 Rosenzweig et al.

1

Assessing agricultural risks of climate change in the 21st century

in a global gridded crop model intercomparison

Supplementary Appendix

Cynthia Rosenzweiga,b, Joshua Elliottb,c, Delphine Deryngd, Alex C. Ruanea,b, Christoph

Müllere, Almut Arnethf, Kenneth J. Booteg, Christian Folberthh, Michael Glotteri, Nikolay

Khabarovj, Kathleen Neumannk,l, Franziska Pionteke, Thomas A. M. Pughf, Erwin Schmidm,

Elke Stehfestk, Hong Yangh, and James W. Jonesg

aNational Aeronautics and Space Administration Goddard Institute for Space Studies, New

York, NY 10025; bColumbia University Center for Climate Systems Research, New York,

NY 10025; cUniversity of Chicago Computation Institute, Chicago, IL 60637; dTyndall

Centre and School of Environmental Sciences, University of East Anglia, Norwich NR4 7TJ,

UK; ePotsdam Institute for Climate Impact Research, 14473 Potsdam, Germany; fInstitute of

Meteorology and Climate Research, Atmospheric Environmental Research, Karlsruhe

Institute of Technology, 82467 Garmisch-Partenkirchen, Germany; gAgricultural and

Biological Engineering Department, University of Florida, Gainesville, FL 32611; hEAWAG

– Swiss Federal Institute of Aquatic Science and Technology 8600 Dübendorf, Switzerland; iDepartment of the Geophysical Sciences, University of Chicago, Chicago, IL 60637; jEcosystems Services and Management Program (ESM), International Institute for Applied

Systems Analysis (IIASA), Laxenburg A-2361, Austria; kPlanbureau voor de Leefomgeving

(Netherlands Environmental Assessment Agency), 3720 AH, Bilthoven, The Netherlands; lWageningen University, 6700 AK, Wageningen, The Netherlands; and mUniversity of

Natural Resources and Life Sciences, 1180 Vienna, Austria

PNAS - 2013

2

Table S1. Global Gridded Crop Models (GGCMs), participants, and references.

*DSSAT cropping system model includes CERES Maize, Rice, and Wheat and CROPGRO soybean.

Table S2: Summary of key characteristics and differences in GGCMs. EPIC GEPIC GAEZ-

IMAGE

LPJmL LPJ-

GUESS

pDSSAT PEGASUS

Type1 Site-based Site-based GAEZ Ecosystem Ecosystem Site-based Ecosystem

CO2 effects2 RUE, TE RUE, TE RUE LF, SC LF, SC RUE (for

wheat, rice, maize) and LF (for soybean)

RUE

Stresses3 W, T, H, A, N, P, BD, Al

W, T, H, A, N, P, BD, Al

W, T W, T W, T W, T, H, A, N

W, T, H, N, P, K

Fertilizer

application4

automatic N

input (max 200 kg ha-1 yr-1), PK (national stat. IFA), dynamic application

NP (national

stat: FertiSTAT), dynamic application

No

nutrient limitation

na na SPAM,

dynamic application

NPK

(national stat. IFA), annual application

Calibration5

Parameters

Site-specific

(EPIC 0810) na

Site-specific

and global F HIpot (for maize and rice)

na

na

Global

LAImax HI a

Uncalibrated

na

Site-specific

(DSSAT) na

Global

Outputs Actual yield & yield gap

Actual yield Potential yield

Actual yield Potential yield

Actual yield Actual yield

Notes (na = not applicable):

(1) Site-base crop model; GAEZ: Global agro-ecological zones; Ecosystem: Global ecosystem model

(2) Elevated CO2 effects: LF: Leaf-level photosynthesis (via rubisco or quantum-efficiency and leaf-

photosynthesis saturation; RUE: Radiation-use efficiency; TE: Transpiration efficiency; SC: stomatal

conductance

(3) W: water stress; T: temperature stress; H: specific-heat stress; A: oxygen stress; N: nitrogen stress; P:

phosphorus stress; K: potassium stress; BD: bulk density; Al: aluminum stress (based on pH and base

saturation)

(4) Fertilizer application, timing of application; NPK annual application of total NPK (nutrient-stress factor);

source of fertilizer application data; timing: annual or dynamic; IFA; FertiSTAT; SPAM

(5) F: fertilizer application rate; HIpot: Potential harvest index; LAImax: maximum LAI under unstressed conditions; HI: harvest index; αa: factor for scaling leaf-level photosynthesis to stand level; β: radiation-use

efficiency factor.

Model Version References for model

description and applications Institution Contact person/Web address

EPIC EPIC0810 1,2 BOKU, University of Natural Resources

and Life Sciences, Vienna

Erwin Schmid

erwin.schmid@boku.ac.at

GEPIC EAWAG 3,4

EAWAG

Swiss Federal Institute of Aquatic Science

and Technology

Christian Folberth/Hong Yang

Christian.folberth@eawag.ch

Hong.yang@eawag.ch

GAEZ in

IMAGE 2.4 5,6

Netherland Environmental Assessment

Agency (PBL)

Elke Stehfest/Kathleen Neumann

Elke.stehfest@pbl.nl

Kathleen.Neumann@pbl.nl

LPJmL - 7,8,9,10 Potsdam Institute for Climate Impact

Research

Christoph Müller

Christoph.Mueller@pik-potsdam.de www.pik-potsdam.de/lpj

LPJ-

GUESS

2.1 with crop

module 7,11,12

Lund University, Department for Physical

Geography and Ecosystem Science,

IMK-IFU, Karlsruhe Institute of

Technology, Garmisch-Partenkirchen,

Germany

Stefan Olin/Thomas Pugh

Stefan.Olin@nateko.lu.se

thomas.pugh@imk.fzk.de

pDSSAT pDSSAT1.0

(DSSAT4.0) 13,14*

University of Chicago

Computation Institute

Joshua Elliott,

jelliott@ci.uchicago.edu

PEGASUS V. 1.1 15

Tyndall Centre

University of East Anglia, UK/McGill

University, Canada

Delphine Deryng

d.deryng@uea.ac.uk

3

Table S3. GGCM biophysical processes.

Notes (na = Not applicable):

(1) D: Dynamic simulation based on development and growth processes; PS: Prescribed shape of LAI curve as function

of phenology, modified by water stress & low productivity

(2) S: Simple approach; D: Detailed approach

(3) RUE: Simple (descriptive) radiation use efficiency approach; P-R: Detailed (explanatory) gross photosynthesis-

respiration (for more details, see 16)

(4) Yield formation depending on: HI: Fixed harvest-index; HIws: HI modified by water stress; Prt: Partitioning during

reproductive stages; B: Total (above-ground) biomass; Gn: Number of grains and grain growth rate

(5) W: Water stress; T: Temperature stress; H: Specific heat stress; A: Oxygen stress; N: Nitrogen stress; P: Phosphorus

stress; K: Potassium stress; BD: Bulk density; Al: Aluminum stress (based on pH and base saturation) (6) V: Vegetative (source); R: Reproductive organ (sink); F: Number of grain (pod) set during the flowering period

(7) Crop phenology function of: T: Temperature; HU: Heat unit index; V: Vernalization; O: Other water/nutrient stress

effects considered; DL: photoperiod (day length)

(8) E: Ratio of supply to demand of water; S: Soil available water in root zone

(9) PM: Penman-Monteith; PT: Priestley-Taylor

(10) Number of soil layers

(11) Lin: Linear; W: Actual water depends on water availability in each soil layer; Exp: Exponential; Non: No roots-just

soil depth zone

(12) Carbon model; Nintrogen model; B(x): x number of microbial biomass pools; P(x): x number of organic matter

pools

(13) Elevated CO2 effects: RUE: Radiation use efficiency; TE: Transpiration efficiency; LF: Leaf-level photosynthesis

(via rubisco or quantum-efficiency and leaf-photosynthesis saturation; SC: Stomatal conductance

Mo

del

Lea

f a

rea

dev

elo

pm

en

t1

Lig

ht

inte

rcep

tio

n2

Lig

ht

uti

liza

tio

n3

Yie

ld

form

ati

on

4

Str

ess

es5

Ty

pe o

f h

ea

t

stress

6

Cro

p

ph

en

olo

gy

7

Ty

pe o

f w

ate

r

stress

8

Ev

ap

o-

tra

nsp

ira

tio

n9

So

il w

ate

r

dy

na

mic

s10

Ro

ot

dis

trib

uti

on

ov

er d

ep

th1

1

So

il p

ro

cess

es1

2

CO

2 e

ffects

13

EPIC D S RUE HIws

Prt B

W, T,

H, A,

N, P,

BD,

Al

V T(HU),

V, O

E PM

10 Lin, W C, N,

B(1),

P(6)

RUE, TE

GEPIC D S RUE HIws

Prt B

W, T,

H, A,

N, P,

BD,

Al

V T(HU),

V, O

E PM 5 Lin, W C, N,

B(1),

P(6)

RUE, TE

IMAGE D S RUE HI W, T,

BD

na T E PT 1 W na RUE

LPJmL PS S P-R HIws W, T na T, V S PT 5 Exp na LF, SC

LPJ-GUESS D S P-R HIws W, T na T, V S PT 2 Lin na LF, SC

pDSSAT D S;

Soy: D

RUE;

Soy: P-R

Gn W, T,

H, A,

N

V, R,

F

T, V,

DL, O

E PT 4 Exp C, N,

P(3)

RUE, TE,

Soy: LF,

TE

PEGASUS D S RUE Prt W, T,

H, N,

P, K

V, F T(HU) E PT 3 Non na RUE, TE

4

Table S4. GGCM inputs and agricultural management practices.

Model Spatial

scale

Temp-

oral

scale1

Climate

input

variables2

Soil input

data3 Spin up4 Planting

date5

Crop

cultivars6

Irri-

gation7,8

Fertilizer

application9

Crop

residue10

CO2

level11

EPIC

0.5° lon

x 0.5º

lat

D, H

Tmn,

Tmx, P,

Rad, RH,

WS

ISRIC-WISE

(17),

ROSETTA

(18),

AWC (19),

ALBEDO

(20),

HYD (21)

Soil OM,

C, NH3,

NO3,

H2O,

P(1)

S

(fraction

of

PHU),

fixed

planting

window

GDD

- fixed

90/100/50

0/50/208

maximum

applied

irrigation:

500 mm

yr-1

Automatic N

input (max 200

kg Ha-1 yr-1)

PK (national

stat. IFA)

dynamic

application

No, can be

simulated

380

ppm

(2005)

GEPIC

0.5° lon

x 0.5º

lat

D

Tmn,

Tmx, P,

Rad, RH,

WS

ISRIC-WISE

(17)

Soil OM,

C, NH3,

NO3,

H2O, P,

CR (20)

S

(fraction

of

PHU),

clim.

adapt

GDD, 2

cultivars

for mai

- fixed

90/100/20

00/1000/0

.018

NP (national

stat.

FertiSTAT),

dynamic

application

Yes, Crop-

specific

364

ppm

(2000)

IMAGE

0.5° lon

x 0.5º

lat

M, WG Ta, P

Soil

reduction

factor (22)

based on

FAO soil

map (23)

CR (210)

clim.

Adapt

(implicit

planting

date)

GDD +

clim.

adapt

na na

Yes, does

not affect

yield

370

ppm

(2000)

LPJmL

0.5° lon

x 0.5º

lat

D Ta, P, Cld

(or Rad)

HWSD (24),

STC HYD

(25),

THM (26)

H2O

(200)

S (9),

fixed

planting

day after

1951

GDD+V

(whe,

sunfl,

rapes);

BT (mai);

static

(others)

- fixed

300/90/10

0/varies7 na

Yes, does

not affect

yield

370

ppm

(2000)

LPJ-GUESS

0.5° lon

x 0.5º

lat

D Ta, P, Cld

(or Rad)

HWSD (24),

STC HYD

(25),

THM (26)

H2O (30)

S (9),

fixed

planting

window

GDD+V

(whe,

sunfl,

rapes);

BT (mai);

static

(others) +

clim. adap

200/90/10

0/1007 na

Yes, does

not affect

yield

379

ppm

(2005)

pDSSAT

0.5° lon

x 0.5º

lat

D

Tmn,

Tmx, P,

Rad

HWSD (24)

Soil OM,

C, NH3,

NO3,

H2O (1)

S (27),

fixed

planting

window

GDD

and/or

latitude,

2-3 for

each cell

- fixed

40/80/100

/757

ric:

30/50/100

/1007

SPAM (28),

dynamic

application

Yes, does

not affect

yield

330 ppm

(1975)

PEGASUS

0.5° lon

x 0.5º

lat

D

Ta, Tmn,

Tmx, P,

Cld (or

Sun)

AWC

(ISRIC-

WISE, 17)

H2O (4)

S (15),

clim.

adapt

GDD +

clim.

adapt

40/90/100

/1007

NPK (national

stat. IFA),

annual

application

na 369 ppm

(2000)

Notes (na = Not applicable):

(1) D: Daily time-step; H: Hourly time-step; M: Monthly time-step; WG: Monthly climate data interpolated to daily

using a weather-generator (2) Tmn: Minimum temperature, Tmx: Maximum temperature, P: Precipitation, Rad: Percentage of radiation, RH:

Relative humidity, WS: Wind speed, Ta: Average temperature, Cld: Percentage of cloud cover, Sun: Fraction of

sunshine hours

(3) Source of soil property inputs (i.e., source of basic soil properties), plus method for deriving parameters required by

models); AWC: Available water capacity; HYD: Hydraulic soil parameters; THM: Thermal parameters; HWSD:

Harmonized world soil database (24); STC: Soil texture classification based on USDA soil texture classification

(http://edis.ifas.ufl.edu/ss169); ISRIC-WISE (17); ROSETTA (18)

(4) Number years for spin up (x); OM: Organic matter, C: Carbon; NH3: Ammonia; NO3: Nitrate; H2O: Soil water; P:

Phosphorus; CR: Crop residue

(5) S: Simulation of planting dates according to climatic conditions; F: Fixed planting dates; source of planting date

data if applicable; PHU: Potential heat unit; Fixed planting window: Does not allow for adaptation to climate change;

clim. adapt: Dynamic planting window: adaptats to climate change (6) GDD: Simulates crop Growing Degree Days (GDDs) requirement according to estimated annual GDDs from daily

temperature; Number of cultivars; GDD+V GDD requirements and vernalization requirements computed based on past

climate experience; BT: Base Temperature computed based on past climate; fixed: Static GDD requirement (no

adaptation); clim. adapt: Dynamic GDD requirement (adaptation to climate change)

5

(7) Irrigation rules: IMDEP(cm): Depth of soil moisture measured; ITHRL(%): Critical lower soil moisture threshold to

trigger irrigation event; ITHRU(%): Upper soil moisture threshold to stop irrigation; IREFF(%): Irrigation application

efficiency

(8) Irrigation rules: EPIC and GEPIC models: BIR(%): Water stress in crop to trigger automatic irrigation; EFI(%):

Irrigation efficiency - runoff from irrigation water; VIMX(mm): Maximum of annual irrigation volume; ARMX(mm):

Maximum of single irrigation volume allowed; ARMN(mm): Minimum of single irrigation volume allowed

(9) Fertilizer application, timing of application; NPK annual application of total NPK (nutrient-stress factor); Source of

fertilizer application data; Timing: Annual or dynamic

(10) Remove residue or not (Yes/No) (11) CO2 concentration baseline for “no CO2” simulations (corresponding year)

6

Table S5: Model calibration and validation.

Notes (na = Not applicable):

(1) Site-based crop model; GAEZ: Global agro-ecological zones; Ecosystem: Global ecosystem model

(2) F: Fertilizer application rate; HIpot: Potential harvest index; LAImax: Maximum LAI under unstressed conditions; HI:

Harvest index; αa: Factor for scaling leaf-level photosynthesis to stand level; β: Radiation use efficiency factor

(3) FE: Field experiments; FAO: FAOSTAT national yield statistics; M3: Gridded dataset of crop-specific yields and

harvested areas for the year 2000 (29)

(4) Wilmott: Maximize Wilmott index of agreement (d) and RMSEu>RMSEs (RMSE: Root-mean-square error;

RMSEu: Unsystematic RMSE; RMSEs: Systematic RMSE) (30)

* GEPIC: Default parameters from field scale model EPIC0810 are mostly used. Potential HI has been adjusted for

maize cultivars and rice based on literature (i.e., field trials). Fertilizer application rates have been modified for a few countries that report very high yields and low fertilizer use, whereas most of these countries are known for their

intensive use of manure.

Model Model origin1 Calibration

method

Parameters

for

calibration2

Output variable

and dataset for

calibration3

Spatial scale of

calibration

Temporal scale

of calibration

Method for model

evaluation4

EPIC Site-based Site-specific

(EPIC 0810) na

Yield (FE &

FAO)

Field-scale &

National Various na

GEPIC Site-based

Site-specific

(EPIC 0810) &

Global*

F, HIpot

(maize, rice)

Yield (FE &

FAO) National

Average for

1997-2003 R2

IMAGE GAEZ NA na Potential Yield National Average for

1970-2005 na

LPJmL Ecosystem Global LAImax, HI,

a Yield (FAO) National

Average for

1998-2003 Wilmott

LPJ-

GUESS Ecosystem Uncalibrated na NA na na na

pDSSAT Site-based Site-specific

(DSSAT) na Yield (FE) Field-scale Various na

PEGASUS Ecosystem Global Yield (M3)

Gridcell level

(0.5ºlon x0.5ºlat

resolution)

Average for

1997-2004 Wilmott

7

Table S6: Simulation experiments and GGCMs outputs.

GGCMs GCMs-RCPs-CO2 CROP OUTPUT1

EPIC

HADGEM2-ES + 4RCPs-CO2 + 4RCPs-noCO2

IPSL-CM5A-LR + 4RCPs-CO2 + RCP8.5-noCO2

MIROC-ESM-CHEM + 4RCPs-CO2 + RCP8.5-noCO2

GFDL-ESM2M + 4RCPs-CO2 + RCP8.5-noCO2

NorESM1-M + 4RCPs-CO2 + RCP8.5-noCO2

Maize, wheat, soybean, rice,

barley, managed grass, millet,

rapeseed, sorghum, sugarcane,

drybean, cassava, cotton,

sunflower, groundnut

YIELD, PIRRWW, AET

GEPIC2

HADGEM2-ES + 4RCPs-CO2 + 4RCPs-noCO2

IPSL-CM5A-LR + 4RCPs-CO2

MIROC-ESM-CHEM + 4RCPs-CO2

GFDL-ESM2M + 4RCPs-CO2

NorESM1-M + 4RCPs-CO2

Maize, wheat, soybean, rice YIELD, PIRRWW, AET

IMAGE

HADGEM2-ES + 4RCPs-CO2 + 4RCPs-noCO2

IPSL-CM5A-LR + 4RCPs-CO2

MIROC-ESM-CHEM + 4RCPs-CO2

GFDL-ESM2M + 4RCPs-CO2

NorESM1-M + 4RCPs-CO2

Maize, wheat, soybean, rice YIELD

LPJmL

HADGEM2-ES + 4RCPs-CO2 + 4RCPs-noCO2

IPSL-CM5A-LR + 4RCPs-CO2 + 4RCPs-noCO2

MIROC-ESM-CHEM + 4RCPs-CO2 + 4RCPs-noCO2

GFDL-ESM2M + 4RCPs-CO2 + 4RCPs-noCO2

NorESM1-M + 4RCPs-CO2 + 4RCPs-noCO2

Maize, wheat, soybean, rice,

millet, cassava, sugar beet,

field pea, rapeseed, sunflower,

groundnut, sugarcane

YIELD, PIRRWW, AET,

PLANT-DAY, MATY-

DAY, BIOM, GSPRCP,

GSRSDS, SUMT

LPJ-GUESS

HADGEM2-ES + 4RCPs-CO2 + 4RCPs-noCO2

IPSL-CM5A-LR + 4RCPs-CO2

MIROC-ESM-CHEM + 4RCPs-CO2

GFDL-ESM2M + 4RCPs-CO2

NorESM1-M + 4RCPs-CO2

Maize, wheat, soybean, rice YIELD, PIRRWW, AET

pDSSAT

HADGEM2-ES + 4RCPs-CO2 + 4RCPs-noCO2

IPSL-CM5A-LR + 4RCPs-CO2 + 4RCPs-noCO2

MIROC-ESM-CHEM + 4RCPs-CO2 + 4RCPs-noCO2

GFDL-ESM2M + 4RCPs-CO2 + 4RCPs-noCO2

NorESM1-M + 4RCPs-CO2 + 4RCPs-noCO2

Maize, wheat, soybean, rice YIELD, PIRRWW, AET,

GSPRCP

PEGASUS3

HADGEM2-ES + 4RCPs-CO2 + 4RCPs-noCO2

IPSL-CM5A-LR + 4RCPs-CO2 + RCP8.5-noCO2

MIROC-ESM-CHEM + 4RCPs-CO2 + RCP8.5-noCO2

GFDL-ESM2M + 4RCPs-CO2 + RCP8.5-noCO2

NorESM1-M + 4RCPs-CO2 + RCP8.5-noCO2

Maize, wheat, soybean

YIELD, PIRRWW, AET,

PLANT-DAY, ANTH-DAY,

MATY-DAY, INITR,

ONITR, BIOM, LEACH,

GSPRCP, GSRSDS, SUMT

Outputs:

(1) YIELD (ton ha-1 yr-1): Dry matter; PIRRWW (mm yr-1): Potential irrigation water withdrawal; AET (mm yr-1):

Actual growing season evapotranspiration; PLANT_DAY (julian day): Planting date; ANTH-DAY (day from planting):

Date of anthesis; MATY-DAY (day from planting): Maturity date; INITR (ton ha-1 yr-1): Inorganic nitrogen application

rate; ONITR (ton ha-1 yr-1): Organic nitrogen application rate; BIOM (ton ha-1 yr-1): Total above ground biomass yield;

LEACH (ton ha-1 yr-1): Nitrogen leached; GSPRCP (mm yr-1): Growing season precipitation; GSRSDS (W m-2 yr-1):

Growing season incoming solar radiation; SUMT (Cº-day yr-1): Sum of daily mean temperature over growing season

(2) GEPIC: All GEPIC outputs for HadGEM2-ES have been shifted by one year in the period 2005-2030. Note as of January 21st, 2013, data have been updated on the server.

(3) PEGASUS: Outputs for NorESM1-M+RCP4.5 wheat are not available

8

S1. Model Processes

The geneaology of GGCMs used in this study is presented in Figure S1. Key characteristics of each GGCM are

provided in Tables S1-S6.

Figure S1: Crop model genealogy for site-based, ecosystem, and agro-ecologial zone (AEZ) models. The models

examined in this study are marked in red boxes.

S1.1. Differences between similar model versions

LPJ-GUESS simulates potential yield, while LPJmL simulates actual yields, and their main difference is the allocation

scheme to the different crop organs, in which the leaf area index (LAI) development is either a function of phenology and management intensity (LPJmL) or a direct feedback of daily net primary production and leaf area index (LPJ-

GUESS). EPIC and GEPIC both use an automatic N fertilization and irrigation schedule constrained by upper limits

(200 N kg/ha/a and 500 mm/a, respectively in EPIC; FertiSTAT values and 2000 mm/a, respectively, in GEPIC). In

addition, GEPIC simulations were run for each decade separately with a 20-year spin-up in order to equilibrate soil

processes while preventing total soil nutrient depletion (31).

S1.2 CO2 effects

GGCMs differ in whether and how they include the potentially beneficial effects on crops of elevated [CO2] from

greenhouse gas emissions and related carbon cycle feedbacks. Effects on crop growth are simulated in LPJmL, LPJ-

GUESS, and DSSAT-Soybean with detailed leaf-level biochemistry photosynthesis (via rubisco or quantum efficiency,

QE, and light-saturated photosynthesis, Amax; 32) and in PEGASUS, EPIC, GEPIC, and CERES maize, wheat, and

rice models in DSSAT through increased radiation use efficiency (RUE). Some models include high [CO2] effects on canopy conductance (LPJmL, LPJ-GUESS, and PDSSAT). The site-based crop models (EPIC, GEPIC and pDSSAT)

include interactions with nitrogen in the CO2 responses, reducing positive effects under low-nitrogen conditions.

Furthermore, slightly different constant [CO2] values, ranging from 330 to 380 ppm, are used by each model in

experiments where [CO2] was held constant although these differences probably do not play a large role in the results.

All the GGCMs used daily climate inputs, which limits their ability to explicitly resolve the diurnal cycles of carbon

fluxes as do some carbon and ecosystem models (some processes in EPIC and pDSSAT are simulated on an hourly

timestep).

S1.3 Temperature and heat effects

The characteristics of GGCMs’ sensitivity to temperature changes and acute heat stress could produce substantially

different responses to both mean climate change and the interacting interannual and intraseasonal variability. In all the

GGCMs, crop phenology is a function of temperature, via accumulated growing degree days. In some GGCMs,

phenology also responds to photoperiod and water and nutrient stresses. Some models include vernalization of winter

9

varieties, and some but not all the GGCMs respond to specific heat stress, such as heat stress at anthesis and the effects

of high temperature on grain growth during the crop’s reproductive phase.

S1.4 Evapotranspiration

Differences in the parameterizations used to simulate evapotranspiration could substantially affect the regions and

severity of water stress under future climate change. Different GGCMs utilize Penman (33; available in EPIC and

GEPIC), Penman-Monteith (34; available in EPIC, GEPIC, and pDSSAT), Priestley-Taylor (35; available in EPIC,

GEPIC, and pDSSAT, and used in GAEZ-IMAGE, LPJmL, and PEGASUS), Hargreaves (36; available in EPIC and

GEPIC), and Baier-Robertson (37; available in EPIC and GEPIC) methods to simulate evapotranspiration. For this study, EPIC used Penman-Monteith for potential evaporation. EPIC, GEPIC, and pDSSAT utilize crop-specific

coefficients for calculation of actual evapotranspiration.

S1.5 Pests

EPIC includes a pest damage function, which was not activated in this analysis. Pest damage does not apply to those

models that simulate potential rather than actual yields (GAEZ-IMAGE and LPJ-GUESS).

S2. Model Configuration

There are key similarities and differences in GGCM inputs and management practices that may affect both the specific

farming systems represented and their initial yield patterns even before they respond to projected climate changes (See

Tables S1-S5 above).

S2.1 Soil properties

Sources of soil properties and methods for deriving GGCM inputs, such as available water capacity, include the

Harmonized World Soil Database (24), the USDA soil texture classification (http://edis.ifas.ufl.edu/ss169), ISRIC-

WISE (17), and ROSETTA (18). Hydraulic and thermal soil parameters are included in some GGCMs. EPIC simulates

soil degradation processes and runs each 30-year time slice independently rather than simulating a continuous time

series.

S2.2 Crops

While some of the models include a wide range of crops and crop types, all the GGCMs simulate wheat, maize, and

soybean, and all but PEGASUS simulate rice. Results presented here focus on those four crops, which are the top four global agricultural food commodities

S2.3 Land use and agricultural systems

All the models use a 0.5o grid, but there are differences in the grid cells simulated to represent agricultural land. While

some models simulated all land areas, others simulated only potential suitable cropland area according to evolving

climatic conditions and others utilized historical harvested areas in the year 2000 according to various data sources

(e.g., the Spatial Production Allocation Model, SPAM; 28). The MIRCA2000 land use database (38) is used for all

models to identify the location of irrigated and non-irrigated areas. There are differences in handling the fraction of

grid-cell area covered by the crops and row spacing/planting density within the cropping areas. Some (but not all)

models have mixed cropland, rotations, and multiple growing seasons (e.g., for aus, aman, and boro rice in

Bangladesh).

S2.4 Planting date Models differed in how planting and harvesting dates were handled in the intercomparison. All models simulated exact

planting dates according to climatic conditions, but some allowed for dynamic planting windows (PEGASUS, GEPIC,

and IMAGE), while others utilized fixed planting windows to historical values based on literature; pDSSAT based on

(27); LPJ-GUESS and LPJ-mL (9). As an example of determining planting dates, the EPIC crop model that underlies

EPIC and GEPIC GGCMs uses automatic adjustments of planting and harvesting dates due to annual weather

conditions. These are based on fractions of crop and regional-specific total heat units. Whenever the fraction of total

heat units for planting and harvesting is reached, planting and harvesting is triggered; the assumption in this analysis is

that total heat units remain constant over time.

S2.5 Climate data

All the GGCMs use daily climate inputs except LPJ-GUESS, which uses monthly climate data interpolated to daily

values. GAEZ-IMAGE, LPJmL, and LPJ-GUESS use daily average temperature, while EPIC, GEPIC, pDSSAT, and

PEGASUS use daily minimum and maximum temperature. For solar radiation, models use either direct surface

insolation or convert this quantity to the percentage of cloud cover or the fraction of sunshine hours. EPIC and GEPIC

additionally use relative humidity and windspeed in potential evapotranspiration calculations.

10

S2.6 Fertilizer application Since [CO2] effects tend to be reduced in low nitrogen-fertility conditions (39), it is important to know whether the

GGCMs responses to [CO2] depend on nitrogen status (as do GEPIC and pDSSAT). The EPIC model simulated high

and low input systems with respect to fertilization and irrigation by using various thresholds that trigger automatic

application (see Table S4). GEPIC applied nitrogen and phosphorous according to FertiSTAT data. PEGASUS applied

nitrogen, phosphorous, and potassium annually according to national statistics. pDSSAT applied nitrogen according to

(40), country averages from FertiSTAT, and crop-specific management intensities from SPAM (28). LPJmL, LPJ-

GUESS and GAEZ-IMAGE did not explicitly simulate nutrient limitations.

S2.7 Model calibration, spin-up, and outputs

GGCM differences regarding model calibration (i.e., adjustment of parameters), spin-up, and outputs may also affect

analysis of the projected climate response. For example, yields from some models are reported according to the year

containing the harvest date. Thus, if the harvest date falls near the start/end of the calendar year, no yields may be reported in some years but other years can report total yields for two harvests. Additionally, models with more

statistical calibration procedures may be affected by the implicit assumption of stationarity in climate statistics that can

change over the coming century.

S2.7.1 Calibration and spin-up

The GGCM simulations differed in calibration and spin-up procedures that can also affect projected climate impacts as

future climates further differentiate themselves from the historical period. When calibration was used, both variables

and data sources differed. LPJmL developed calibration procedures to observed FAO average yield around the year

2000 by adjusting maximum LAI, harvest index, and a scale factor for scaling leaf-level photosynthesis to the crop

stand level (8). PEGASUS calibration procedures tuned model results to M3 observed yield around 2000 (29) by

adjusting one global parameter representing a radiation-biomass conversion factor. In the case of EPIC, crop growth parameters were not adjusted to match simulated and reported yields. Simulated yields are compared to national

averages after a spin-up (nutrient-mining) period of 20 years, which has been found earlier to be adequate for

representing low soil nutrient status in low-input regions like sub-Saharan Africa (31). Other models (EPIC, pDSSAT)

relied only on previous underlying site-based calibration across broad regions, while others had no calibration

procedure (LPJ-GUESS) or contain a post-processing calibration procedure (GAEZ-IMAGE). However, only

uncalibrated yields from GAEZ-IMAGE are utilized in this study.

Crop models may be described as including more or fewer yield-constraining factors and processes (e.g., water,

nutrients, and heat stress). These yield gap issues have important implications for calibration, validation, and eventual

adaptation testing (41).

S3. Additional Results and Recommended Guidelines for Future Work

Average reference period (1980-2010) wheat, rice, and soy yields are presented in Figures S2-S4. Figure S5 displays

globally-aggregated production changes with CO2 effects separated by areas that are currently rainfed and areas with

irrigation. Differences in production changes between rainfed and irrigated areas in any given model are generally

smaller than the differences between simulations with and without CO2 effects.

S3.1 General recommendations for the use of GGCM ensemble results from Phase 1

The seven GGCMs that provided data to the AgMIP/ISI-MIP Phase 1 archive differ in model type, implemented mechanisms, model calibration, and implicit and explicit assumptions. These differences have implications for the use

and interpretation of data in analyses and assessments. The AgMIP/ISI-MIP publications of Phase 1 provide a good

orientation on GGCMs’ performance relative to the total range of results, which should be considered in the

interpretation of the data. We here point out a few general caveats but request that any researcher using these data

carefully check the suitability of the data for the intended analysis. If in doubt, please contact the individual GGCM

modelers.

Some of the models have been calibrated to national or grid cell yield observations. This implies that absolute yield data

are closer to observations, but it does not indicate their skill in simulating observed yield levels. Similarly, some of the

GGCMs may have been applied in specific regions more than in others and may thus have implicit assumptions that suit

cropping systems in these regions better than in others. Even though some GGCMs do not capture current yield patterns

well (e.g. due to lack of calibration data), the simulated relative yield trends may constitute valuable information for some applications such as economic assessments, if superimposed on observed yield patterns.

11

Many aspects, such as sensitivity to weather extremes and year-to-year variability have not been tested in detail and

analyses on these aspects need to evaluate the models’ skill before proceeding.

GGCM differences in model types, processes, inputs, and procedures imply ways that the results should be used.

Relative yield changes should be used rather than absolute yield values since models differ in their calibration

procedures as well as fertility inputs. For reference period yields, it is advisable to use an observation set such as the

M3 data (29) adjusted to represent future yields using relative yield change factors calculated from the GGCMs.

Furthermore, multi-year averages of yield results should be used because for some models yields are reported according

to the year containing the harvest date and some years may not have reported yields or may have two harvests.

Great care should be used in interpreting regional (i.e., continental, sub-continental, national, and sub-national) results,

since the objective of this study was to conduct a global-scale intercomparison, and regional input data, model settings,

and results have not been vetted. See for example the discussion of accumulation of uncertainties in (42). We

recommend that detailed validation be done at national and sub-national scales as a first step to use of these results at

finer-than-global scales. Work is continuing to attribute climate and [CO2] sensitivity differences to disparities in

GGCM properties and configurations.

S3.2 AgMIP GGCM Intercomparison Phase II

In the second phase of the AgMIP GGCM intercomparison, we are conducting a rigorous validation study and

designing protocols that provide further information relevant to policymakers. The next phase includes updated versions

of the models described here as well as a broader range of global gridded crop models (such as DayCent (43); GLAM (44, 45); MCWLA (46); Orchidee-Mil (47)). For example, the GGCM intercomparison Phase II protocol includes

simulations without nutrient limitation and with harmonized planting dates. Since economic growth is likely to spur

greater fertilizer applications in current low-input regions and improve management, this will improve comparability

across models and adaptation planning.

12

Figure S2: As in main text Figure 2, but for wheat.

13

Figure S3: As in main text Figure 2, but for rice. Note that PEGASUS does not simulate rice.

14

Figure S4: As in main text Figure 2, but for soybean.

15

Figure S5: As in main text Figure 4, but with rainfed and irrigated areas separated in simulations, with CO2 effects.

16

References for Supplementary Information 1. Williams, J.R. (1995) The EPIC. In: Singh, V.P. (Ed.). Computer Models of Watershed Hydrology. Water Resources

Publications. Littleton, CO. pp. 909–1000 (Chapter 25).

2. Izaurralde, R.C., et al. (2006) Simulating soil C dynamics with EPIC: Model description and testing against long-term data. Ecological Modeling, 192(3-4):362-384.

3. Williams, J.R., et al. (1990) EPIC-Erosion/Productivity Impact Calculator. United States Department of Agriculture

Agricultural Research Service. Technical Bulletin Number 1768. Springfield, VA.

4. Liu, J. and Williams, J and Zehnder, A.J.B. and Yang, H., (2007) GEPIC–modelling wheat yield and crop water

productivity with high resolution on a global scale, Agricultural Systems, 94 (2), pp. 478–493

5. Leemans, R. and A.M. Solomon (1993) Modeling the potential change in yield and distribution of the earth’s crops

under a warmed climate. Climate Research, 3:79-96.

6. Bouwman A.F., T. Kram, T. Klein, and K. Goldewijk (Eds.). (2006) Integrated Modelling of Global Environmental

Change. An Overview of IMAGE 2.4. PBL Netherlands Environmental Assessment Agency, The Hague.

7. Bondeau, A., et al. (2007) Modelling the role of agriculture for the 20th century global terrestrial carbon balance.

Global Change Biology 13:679–706, doi:10.1111/j.1365-2486.2006.01305.x

8. Fader, M., et al. (2010), Virtual water content of temperate cereals and maize: Present and potential future patterns, Journal of Hydrology, 384, 218-231.

9. Waha, K., et al. (2012) Climate-driven simulation of global crop sowing dates. Global Ecology and Biogeography

21:247-259.

10. Schaphoff, S., U. Heyder, S. Ostberg, D. Gerten, J. Heinke, and W. Lucht (2013) Contribution of permafrost soils to

the global carbon budget, Environmental Research Letters. in press.

11. Smith, B., et al. (2001) Representation of vegetation dynamics in the modelling of terrestrial ecosystems: comparing

two contrasting approaches within European climate space. Global Ecology and Biogeography 10:621-637

12. Lindeskog M., et al. (in review) Implications of accounting for land use in simulations of ecosystem services and

carbon cycling in Africa. (in review at Earth Systems Dynamics)

13. Elliott, J., M. Glotter, N. Best, D. Kelly, M. Wilde, and I. Foster (2013). The Parallel System for Integrating Impact

Models and Sectors (pSIMS). Accepted for publication in the Proceedings of the 2013 XSEDE Conference. 14. Jones, J.W., et al. (2003) The DSSAT cropping system model. Eur. J. Agron. 18:235-265.

15. Deryng, D., Sacks, W. J., Barford, C. C., & Ramankutty, N. (2011). Simulating the effects of climate and

agricultural management practices on global crop yield. Global Biogeochemical Cycles, 25(2).

16. Adam, Z., D. Charuvi, O. Tsabari, R.R. Knopf, and Z. Reich (2011) Biogenesis of thylakoid networks in

angiosperms: knowns and unknowns. Plant Molecular Biology 76(3-5), 221-234. doi:10.1007/s11103-010-

9693-5

17. Batjes, N.H. (2006) ISRIC-WISE Derived Soil Properties on a 5 by 5 Arc-minutes Global Grid. ISRIC – World Soil

Information, Wageningen, Netherlands. (Available at http://www.isric.org.)

18. Schaap, M.G., and W. Bouten (1996). Modeling water retention curves of sandy soils using neural networks. Water

Resources Research, 32(10), 3033-3040.

19. van Genuchten, M.T., F. Kaveh, and W.B. Russell, W.B. (1988) Direct and indirect methods for estimating the

hydraulic properties of unsaturated soils \ Land qualities in space and time : proceedings of a symposium Wageningen, the Netherlands, 22 - 26 August 1988.

20. Dobos, E. (2006) Albedo. In Encyclopedia of Soil Science, Second Edition (pp. 64-66).

21. USDA/NRCS (2012): Soil Survey Staff, Natural Resources Conservation Service, Soil Survey Geographic

(SSURGO) Database for [Survey Area, State]. Available online at http://soildatamart.nrcs.usda.gov. Accessed

Feb 2012.

22. Wood, S.R. and F.J. Dent (1983) LECS : a land evaluation computer system methodology. Part of: Manual / Food

and Agriculture Organization Nr. no. 5, version 1

23. FAO (1991) The Digitized Soil Map of the World (Release 1.0). Food and Agriculture Organization of the United

Nations)

24. FAO/IIASA/ISRIC/ISSCAS/JRC(2012) Harmonized World Soil Database (version 1.2), FAO, Rome, Italy and

IIASA, Laxenburg, Austria. 25. Cosby, B.J., G.M. Hornberger, R.B. Clapp, and T.R. Ginn, T. R. (1984) A statistical exploration of the relationships

of soil moisture characteristics to the physical properties of soils. Water Resources Research, 20(6), 682-690.

26. Lawrence, D.M., and A.G. Slater (2008) Incorporating organic soil into a global climate model. Climate Dynamics,

30(2-3), 145-160.

27. Sacks, W.J., D. Deryng, J.A. Foley, and N. Ramankutty (2010) Crop planting dates: an analysis of global patterns.

Global Ecology and Biogeography 19:607-620.

28. You, L., et al. (2000) Spatial Production Allocation Model (SPAM) 2000 Version 3 Release 1.

http://MapSPAM.info. (Accessed Feb, 2012).

29. Monfreda, C., Ramankutty, N. & Foley, J. A. (2008) Farming the planet: 2. Geographic distribution of crop areas,

yields, physiological types, and net primary production in the year 2000. GBC 22, 19.Monteith, J.L. 1965.

17

Evaporation and environment. Symposia of the Society for Experimental Biology 19: 205–224.

PMID 5321565.

30. Willmott, C. J., et al. (1985) Statistics for the evaluation and comparison of models, J. Geophys. Res., 90(C5),

8995–9005.

31. Folberth, C., R. Gaiser, K.C. Abbaspour, R. Schulin, H. Yang (2012) Regionalization of a large-scale crop growth

model for sub-Saharan Africa: Model setup, evaluation, and estimation of maize yields. Agriculture

Ecosystems & Environment 151:21-33.

32. Farquhar, G. D., S. Caemmerer, J.A. Berry (1980) A biochemical model of photosynthetic CO2 assimilation in

leaves of C3 species. Planta 149. 33. Penman, H.L. (1948) Natural Evaporation from Open Water, Bare Soil and G5 Grass. Proc. Roy. Soc. London

A(194), S. 120-145.

34. Monteith, J. (1965) Evaporation and environment. Symp. Soc. Exp. Biol. 19:205-234.

35. Priestley, C.H.B. and R.J. Taylor (1972) On the assessment of surface heat flux and evaporation using large-scale

parameters. Monthly Weather Review 100 (2): 81–82. Bibcode 1972MWRv..100...81P. doi:10.1175/1520-

0493(1972)100<0081:OTAOSH>2.3.CO;2.

36. Hargreaves, G.H. and Z.A. Samani (1985) Reference crop evapotranspiration from temperature. Transactions of

ASAE 1(2):96-99.

37. Baier, W. and G. W. Robertson (1965) Estimation of latent evaporation from simple weather observations. Can. J.

Plant Sci. 45:276-284.

38. Portmann, F.T., S. Siebert, P. Doll (2010) MIRCA2000 – global monthly irrigated and rainfed crop areas around the

year 2000: a new high-resolution data set for agricultural and hydrological modelling. Global Biogeochemical Cycles 24:GB1011.

39. Kimball, B.A. 2011. Lessons from FACE: CO2 effects and interactions with water, nitrogen, and temperature. In

Hillel, D. and Rosenzweig, C. (Eds.). Handbook of Climate Change and Agroecosystems: Impacts,

Adaptation, and Mitigation. ICP Series on Climate Change Impacts, Adaptation, and Mitigation Vol 1.

Imperial College Press. London. pp. 87-107.

40. Potter, Philip, Navin Ramankutty, Elena M. Bennett, Simon D. Donner (2010) Characterizing the Spatial Patterns of

Global Fertilizer Application and Manure Production. Earth Interact., 14, 1–22.

41. Lobell, D.B., K.G. Cassman, and C.B. Field (2009) Crop Yield Gaps: Their Importance, Magnitudes, and Causes.

Annu. Rev. Environ. Resour. 2009. 34:179–204

42. Roudier, P., B. Sultan, P. Quirion, C. Baron, A. Alhassane, S.B. Traore, and B. Muller (2011) The impact of future

climate change on West African crop yields: What does the recent literature say? Global Environmental Change 21:1073-1083.

43. Del Grosso, et al. (Eds.) (2001) Modeling Carbon and Nitrogen Dynamics for Soil

Management. CRC Press, Boca Raton, Florida, pp. 303-332.

44. Challinor, A.J., T.R. Wheeler, P.Q. Craufurd, J.M. Slingo, and D.I.F. Grimes (2004) Design and optimisation of a

large-area process-based model for annual crops. Agric. For. Meteorol. 124, 99–120.

45. Osborne, T., G. Rose, and T. Wheeler (2013): Variation in the global-scale impacts of climate change on crop

productivity due to climate model uncertainty and adaptation. Agric. For. Meteorol. 170, 183-194.

doi:10.1016/j.agrformet.2012.07.006

46. Tao, F., M. Yokozawa, and Z. Zhang (2009) Modelling the impacts of weather and climate variability on crop

productivity over a large area: A new process-based model development, optimization, and uncertainties

analysis. Agric. For. Meteorol. 149, 831–850 47. Berg, A., N. de Noblet-Ducoudré, N., B. Sultan, M Lengaigne, and M. Guimberteau (2013) Projections of climate

change impacts on potential C4 crop productivity over tropical regions. Agric. For. Meteorol. 170, 89-102.

doi:10.1016/j.agrformet.2011.12.003