Agriculture, Ecosystems and Environment 82 (2000) 139158Consequences of future climate change and changing climatevariability on maize yields in the midwestern United StatesJane Southwortha,, J.C. Randolpha, M. Habeckb, O.C. Doeringb,R.A. Pfeiferb, D.G. Raoc, J.J. JohnstonaaSchool of Public and Environmental Affairs, Indiana University, 1315 E. Tenth Street, Bloomington, IN 47405-1701, USAbDepartment of Agricultural Economics, Purdue University, West Lafayette, IN, USAcCentral Research Institute for Dryland Agriculture, Hyderabad, IndiaAbstractAny change in climate will have implications for climate-sensitive systems such as agriculture, forestry, and some othernatural resources. With respect to agriculture, changes in solar radiation, temperature, and precipitation will produce changesin crop yields, crop mix, cropping systems, scheduling of eld operations, grain moisture content at harvest, and hence, on theeconomics of agriculture including changes in farm protability. Such issues are addressed for 10 representative agriculturalareas across the midwestern Great Lakes region, a ve-state area including Indiana, Illinois, Ohio, Michigan, and Wisconsin.This region is one of the most productive and important agricultural regions in the world, with over 61% of the land usedevoted to agriculture.Individual crop growth processes are affected differently by climate change. Aseasonal rise in temperature will increase thedevelopmental rate of the crop, resulting in an earlier harvest. Heat stress may result in negative effects on crop production.Conversely, increased rainfall in drier areas may allowthe photosynthetic rate of the crop to increase, resulting in higher yields.Properly validated crop simulation models can be used to combine the environmental effects on crop physiological processesand to evaluate the consequences of such inuences. With existing hybrids, an overall pattern of decreasing crop productionunder scenarios of climate change was found, due primarily to intense heat during the main growth period. However, theresults changed with the hybrid of maize (Zea mays L.) being grown and the specic location in the study region. In general,crops grown in sites in northern states had increased yields under climate change, with those grown in sites in the southernstates of the region having decreased yields under climate change. Yields from long-season maize increased signicantly inthe northern part of the study region under future climate change. Across the study region, long-season maize performed mostsuccessfully under future climate scenarios compared to current yields, followed by medium-season and then short-seasonvarieties. This analysis highlights the spatial variability of crop responses to changed environmental conditions. In addition,scenarios of increased climate variability produced diverse yields on a year-to-year basis and had increased risk of a lowyield.Results indicate that potential future adaptations to climate change for maize yields would require either increased toleranceof maximum summer temperatures in existing maize varieties or a change in the maize varieties grown. 2000 ElsevierScience B.V. All rights reserved.Keywords: Climate change; Variability; CERES-maize; Maize yields; Agriculture; Midwestern United States Corresponding author. Tel: +1-812-855-4953; fax: +1-812-855-7547.E-mailaddress: jsouthwo@indiana.edu (J. Southworth).0167-8809/00/$ see front matter 2000 Elsevier Science B.V. All rights reserved.PII: S0167- 8809( 00) 00223- 1140 J.Southworthetal. / Agriculture,EcosystemsandEnvironment82(2000)1391581. IntroductionInterest intheconsequences of increasingatmo-spheric CO2 concentration and its role in inuencingclimatechangecanbetracedasfar backas1827,althoughmorecommonlyattributedtotheworkofArrhenius (1896) and Chamberlain (1897), as cited inChiotti and Johnston (1995). A century later, concernoverclimaticchangehasreachedglobaldimensionsand concerted international efforts have been initiatedinrecent years toaddress this problem(Intergov-ernmental Panel onClimate Change (IPCC, 1995,1990)). Future climate change could have signicantimpactsonagriculture, especiallythecombinedef-fectsofelevatedtemperatures, increasedprobabilityof droughts, anda reducedcrop-water availability(Chiotti and Johnston, 1995).Based on climate records, average global tem-peratures at the earths surface are rising. Sinceglobal records began in the mid-19th century, the vewarmest yearshaveoccurredduringthe1990sand10 of the 11 warmest years have occurred since 1980(Pearce, 1997). Basedonarangeofseveralcurrentclimate models (IPCC, 1995, 1990), the mean annualglobal surface temperature is projected to increase by1 to 3.5C by the year 2100 and there will be changesinthespatial andtemporal patternsofprecipitation(IPCC, 1995).Thevariabilityof theclimate, under current andfutureclimatescenarios, hasbeenatopicofrecentinterest foranumberofreasons. Theconsequencesof changes in variability may be as important as thosethat arise due tovariations inmeanclimatic vari-ables (Hulme et al., 1999; Carnell and Senior, 1998;Semenov and Barrow, 1997; Liang et al., 1995; Rind,1991; Mearns et al., 1984). Whilemost studies ofclimate change impacts on agriculture have analyzedeffects of mean changes of climatic variables on cropproduction, impacts of changes in climate variabilityhavebeenmuchlessstudied(Mearnset al., 1997;Mearns, 1995).Withinmorerecent historical records, therehavebeen signicant periods of climatic uctuations suchas the dustbowl conditions of the 1930s in the UnitedStates, when the dramatic and negative effects ofclimate variability on agriculture were realized. KatzandBrown(1992) showedthat for agivenclimatevariable a change in the variance has a larger effect onagriculturalcroppingsystemsthandoesachangeinthe mean. The effect of possible changes in climaticvariabilityremainsasignicant uncertaintythat de-servesadditional attentionwithinintegratedclimatechange assessments (Barrowet al., 1996; Mearnset al., 1996; Semenovet al., 1996). The studyofeconomic effects of climate change on agriculture isparticularlyimportant becauseagricultureisamongthe more climate sensitive sectors (Kane et al., 1992).Changes in climate will interact with adaptations toincrease agricultural production affecting crop yieldsandproductivityindifferentwaysdependingonthehybrids and cropping systems in a region. Importantdirect effects will be through changes in temperature,precipitation, length of growing season, and timing ofextremeorcritical thresholdeventsrelativetocropdevelopment (SaarikkoandCarter, 1996). Also, anincreased atmospheric CO2 concentration could havea benecial effect on the growth of some species.Indirect effects will include potentially detrimentalchanges in diseases, pests, and weeds, the effects ofwhichhavenotyetbeenquantiedinmoststudies.Evidence continues to support the ndings of the IPCCthat global agricultural productioncouldbemain-tained relative to baseline production for a growingpopulationunder2 CO2equilibriumclimatecon-ditions (Rosenzweig and Hillel, 1998, 1993). In mid-dle and high latitudes, climate change will extend thelength of the potential growing season, allowing earlierplanting of crops in the spring, earlier maturation andharvesting,andthepossibilityoftwoormorecrop-ping cycles during the same season. Climate changealso will modify rainfall, evaporation, runoff, and soilmoisture storage. Both changes in total seasonal pre-cipitation or in its pattern of variability are importantto agriculture. Moisture stress and/or extreme heat dur-ing owering, pollination, and grain lling is harmfultomost crops, suchasmaize, soybeans, andwheat(RosenzweigandHillel, 1993), themost importantcommodity crops in the midwestern United States.Inworldmarkets, theUnitedStatesaccountsformore than 25% of the global trade in wheat, maize,soybeans, andcotton(USDA, 1997; Adams et al.,1999). An important question concerns the ability ofNorth American agriculture to adapt to changing cli-maticconditions. Warmerclimatescenarios(25Cincreases in North America) have yielded estimates ofnegative impacts in eastern, southeastern, and MaizeJ.Southworthetal. / Agriculture,EcosystemsandEnvironment82(2000)139158 141Belt regions and positive effects in northern plains andwesternregions. Moremoderatewarmingproducedestimatesof predominatelypositiveeffectsinsomewarm-season crops (IPCC, 1995).Agricultural systems are managed and farmersalways have a number of possible adaptations or op-tionsopentothem.Theseadaptationstrategiesmaypotentially lessen future yield losses fromclimatechange or may improve yields in regions where ben-ecial climatechanges occur (Kaiser et al., 1995).Thus, agricultural productionrespondsbothtophy-siological changes incrops due toclimate changeand also to changes in agricultural management prac-tices, crop prices, costs and availability of inputs, andgovernment policies (Adams et al., 1999).This research addresses the issues of changingmean climatic conditions and changes in the variabi-lityofclimatearoundthesemeans. Theoccurrenceand frequency of extreme events will become increas-inglyimportant. This studyaddresses these issuesfor 10representativeagricultural areas inthemid-westernUnitedStatesforthreehybridsofmaizeintermsoftheircurrentandfutureyields. Changesinclimatic patterns may result in spatial shifts of agri-cultural practices, thereby impacting current land usepatterns. Specically, this research addresses (1) howthe mean changes in future climate will affect maizeyields across the study region for two future climatescenarios, (2) how the changes in climate variability,in addition to changes in the mean, affect poten-tial future maize yields, (3) the implications of suchchanges spatially, examiningpotential future gainsandlosses, and(4)somepossiblefutureadaptationstrategies.2. Methods and materials2.1. Study regionThe midwestern Great Lakes region (Indiana, Illi-nois, Ohio, Michigan, andWisconsin) was dividedinto 10 agricultural areas based on climate, soils, landuse, and current agricultural practices. Representativefarms were created in each area based upon local char-acteristics and farm endowments (Fig. 1). This regionis one of the most productive and important agricul-tural regions in the world (Smith and Tirpak, 1989).2.2. Maize crop modelThe decisionsupport systemfor agrotechnologytransfer (DSSAT) softwareisaset of cropmodelsthatshareacommoninputoutputdataformat. TheDSSAT itself is a shell that allows the user to organizeand manipulate crops, soils, and weather data and torun crop models in various ways and analyze their out-puts (Thornton et al., 1997; Hoogenboom et al., 1995).The version of DSSAT used in this analysis was thatsupplied by ISBNAT, which is DSSAT 3.5.Crop growth was simulated using the CERES-maizemodel, with a daily time step from sowing to matu-rity, basedonphysiological processesthat describethecropsresponsetosoilandaerialenvironmentalconditions. Phasic development is quantied accord-ing to the plants physiological age. In CERES-maize,sub-modelstreat leaf areadevelopment, drymatterproduction, assimilatepartitioning, andtillergrowthanddevelopment. Potential growthisdependent onphotosyntheticallyactiveradiationandits intercep-tion,whereasactualbiomassproductiononanydayis constrained by sub-optimal temperatures, soil wa-ter decits, and nitrogen and phosphorus deciencies.The input data required to run the CERES-maizemodel includes daily weather information (maximumand minimum temperatures, rainfall, and solar radia-tion); soil characterization data (data by soil layer onextractable nitrogen and phosphorous and soil watercontent); aset ofgeneticcoefcientscharacterizingthe hybrid being grown (Table 1); and crop manage-ment information, such as emerged plant population,rowspacing, andseedingdepth, andfertilizer andirrigation schedules (Thornton et al., 1997). The soildatawereobtainedfromUSSoilConservationSer-vice (SCS; now known as the Natural Resources Con-servationService)forthe10farmsites. Themodelapportionstherainreceivedonanydayintorunoffandinltrationintothesoil, usingtherunoffcurvenumber technique. Arunoff curvenumber wasas-signed to each soil, based on the soil type, depth andtexture as obtained from the STATSGO databases.Concentrations of nitrogen and phosphorous in thesoil were not limiting to crop growth and these mod-ulesareturnedoffwithintheCERES-maizemodelruns.TheCERES-maizemodel includesthecapabilityto simulate the effects of increased atmospheric CO2142 J.Southworthetal. / Agriculture,EcosystemsandEnvironment82(2000)139158Fig. 1. Location of 10 representative agricultural regions in the midwestern Great Lakes states.J.Southworthetal. / Agriculture,EcosystemsandEnvironment82(2000)139158 143Table 1Genetic coefcients for CERES-maize for (a) east-central Indiana,south-central Michigan, eastern Wisconsin, south-west Wisconsin,andthe Michiganthumb, andfor (b) easternIllinois, southernIllinois, south-west Indiana, and western IllinoisDesignation P1aP2bP5cG2dG3ePHINTf(a)Long 320 0.52 940 620 6.0 38.9Medium 200 0.30 800 700 6.3 38.9Short 110 0.30 680 820 6.6 38.9(b)Long 320 0.52 990 620 6.0 38.9Medium 200 0.30 843 700 6.3 38.9Short 110 0.30 716 820 6.6 38.9aThermal time fromseedlingemergence tothe endof thejuvenile phase (expressed in degree days above a base temperatureof 8C) duringwhichtheplant isnot responsivetochangesinphotoperiod.bExtent to which development (expressed as days) is delayedfor each hour increase in photoperiod above the longest photope-riod at which development proceeds at a maximum rate (which isconsidered to be 12.5 h).cThermal time from silking to physiological maturity (expres-sed in degree days above a base temperature of 8C).dMaximum possible number of kernels per plant.eKernel llingrateduringthelinear grainllingstageandunder optimum conditions (mg/day).fPhylochron interval; the interval in thermal time (degree days)between successive leaf tip appearances.concentrations onphotosynthesis andwater usebythe crop. Daily potential transpiration calculationsaremodiedbytheCO2concentrations(Lal et al.,1998; Dhakhwaet al., 1997; Phillipset al., 1996).Hence, under current conditionsthemodel canrununder the current atmospheric CO2concentration,approximately360 ppmv. However, whenevaluatingcrop growth under changed climate scenarios, whichare based on the assumption of global warming dueto increased concentration of CO2(and other green-housegases), theCO2concentrationwasincreasedaccordingly. The atmospheric CO2concentration forthe future climate scenarios usedinthis research,based on the years 20502059, is 555 ppmv.CERES-maize does not have an algorithmthatkills maize as a result of spring freezes. Hence, inthisanalysis, whenever aminimumair temperatureof 2.0C or less occurred during the spring (JulianDays1180) andafter theemergencedate, growthwas terminated. This value of termination was basedon discussion with Dr. Joe Ritchie (personal commu-nicatin, 1999) as being a realistic number for freezeloss of maize.Theselectionof CERES-maizeasthemodel forthis research was based on (1) the daily time step ofthe model allows us to address the issue of changes inplanting dates, (2) plant growth dependence on bothmeandailytemperaturesandtheamplitudeofdailytemperature values was desired (not just a daily meantemperature growth dependence as for EPIC), (3) themodel simulates crop response to major climate vari-ables,includingtheeffectsofsoilcharacteristicsonwater availability, and is physiologically oriented, (4)themodel isdevelopedwithcompatibledatastruc-turessothatthesamesoilandclimatedatasetscanbe used for all hybrids of crops which helps in com-parison (Adams et al., 1990), and (5) comprehensivevalidation has been done across a wide range of dif-ferent climateandsoil conditions, andfordifferentcrop hybrids (Semenov et al., 1996; Wolf et al., 1996;Hoogenboom et al., 1995).2.3. Crop model validationCERES-maize has been extensively validated atsites both in the United States and abroad (Dhakhwaet al., 1997; Hoogenboom et al., 1995). Mavromatisand Jones (1998) found that using the CERES-wheatmodel coupled with a weather generator (WGEN orSIMMETEO) to simulate daily weather from monthlymean values is an efcient method for assessing theimpactsofchangingclimateonagriculturalproduc-tion. Properlyvalidatedcropsimulationmodelscanbeusedtodeterminetheinuences of changes inenvironment, such as climate change, on crop growth(Peiris et al., 1996).Intensivesite-specicvalidationwas alsounder-taken to ensure the model could mirror reality in therepresentative agricultural areas. Detailed experimen-tal farmlevel datawereusedtoensurethat yieldsproduced by the model reected the actual yields ineachrepresentativeagricultural area. CERES-maizewas initiallyvalidatedusingexperimental data re-portedbyNafziger(1994)formaizeplantedattwolocations in Illinois, USA(Fig. 2a). Experimentaldatawereavailableforfourplantingdatesovertheperiod19871990, andmeasuredyieldswerecom-paredagainst yieldssimulatedusingCERES-maize.144 J.Southworthetal. / Agriculture,EcosystemsandEnvironment82(2000)139158Fig. 2. Validationof CERES-maizeresultsfor (a) 15-year simulationof maizeyieldtoplantingdatesat Dekalb, Illinois(19751990)and(b)long-seasonmaizeyieldsasapercent ofmaximumforeasternIllinois, southernIllinois, andeasternWisconsinasmodeledbyCERES-maize compared to agronomists predictions.J.Southworthetal. / Agriculture,EcosystemsandEnvironment82(2000)139158 145Simulated yields corresponded to within +/10% ofthe observed maize yields. Overall response of simu-lated yields to planting date variation compared veryfavorably with observed yields.In addition, each representative farm was validatedusing historical yield data and past daily climate in-formation. This validationwas usedtoensure themodel couldreplicatepast yields, at eachlocationbeingmodeled, forlonger-season(currentlygrown)maize varieties and that the medium and short-seasonmaizevarietiesshowedthecorrect trendsbasedonexpert opinions of agronomists (Fig. 2b).Inaddition, CERES-maizewas usedtosimulatethe relationships between planting date and yield (ex-pressed as a percentage of the highest yield observedfor any combination of hybrid and planting date) forlong, medium, andshort-seasonmaizevarietiesfor20 different planting dates. The relationship betweenplanting date and yield suggest that long-season maizehas higher yields than medium and short-season maizeforearlierplantingdates,medium-seasonmaizehasthe highest yields for the middle planting dates, andshort-season maize has the highest yields for lateplantingdates. Theseresultsareconsistentwithex-pectationsabout hybridperformanceinthisregion.Further, the close match between the long-season hy-bridperformanceandagronomistexpectationsgivesuscondencethatthemodelisabletodescribetherelationshipbetweenplantingdateandmaizeyieldfor these locations (Fig. 2b). Similar results have beenobtained for all 10 representative agricultural areas inthe study region.2.4. Current climate analysis: VEMAPTheVEMAPdatasetincludesdaily,monthly,andannual climate data for the conterminous UnitedStates including maximum, minimum, and mean tem-perature, precipitation, solarradiation, andhumidity(Kitteletal., 1996). TheVEMAPbaseline(30-yearhistorical mean) climatedatawasusedfor eachofthe10representativeagricultural areasinthestudyregion. The weather generator SIMMETEO (as usedinDSSATversion3.5) usedtheseclimatedatatostochasticallygeneratedailyweather datainmodelruns. This approach of using monthly data to generatedaily data allowed us to generate variability scenarios.The climate variables distribution patterns mimic thecurrent climate variables distributions. As there is nowaytoknowfuturedistributions, theyarebasedoncurrent patterns. Sensitivityanalysisdeterminedtheeffects of differing climatic conditions upon differingcombinations of planting and harvest dates.2.5. Future climate scenariosThis research used the Hadley Center modelHadCM2 from England for future climate scenariodata. This model was created from the Unied Model,which was modied slightly to produce a new, cou-pled ocean-atmosphere GCM, referred to as HadCM2.Thishasbeenusedinaseriesof transient climatechangeexperimentsusinghistoricandfuturegreen-house gas and sulfate aerosol forcing. These modelssimulate time-dependent climate change (Barrowet al., 1996). Transient model experiments are consi-dered more physically realistic and complex, andallow atmospheric concentrations of CO2 to rise grad-ually over time (Harrison and Buttereld, 1996). Theresults from this model have been validated at a num-ber of locations. HadCM2 is also one of the modelsused in the US National Assessment project.HadCM2hasaspatialresolutionof2.5 3.75(latitude by longitude) and the representation producesa grid box resolution of 9673 grid cells, which pro-duces a surface spatial resolution of about 417 km 278 km reducing to 295 km278 km at 45 north andsouth. The atmospheric component of HadCM2 has 19levels and the ocean component 20 levels. The equili-brium sensitivity of HadCM2, that is the global-meantemperature response to a doubling of effective CO2concentration, is2.5C, somewhat lower thanmostother GCMs (IPCC, 1990).The greenhouse-gas-only version, HadCM2-GHG,used the combined forcing of all the greenhouse gasesasanequivalent CO2concentration. HadCM2-SULused the combined equivalent CO2 concentration plusanegativeforcingfromsulfateaerosols. HadCM2-GHGsimulated the change in forcing of the cli-mate systemby greenhouse gases since the earlyindustrial period. There is a small amount of forcing(0.4 Wm2) prior to this simulation period represent-ing the small increase in greenhouse gases from 1765to 1860. The addition of the negative forcing effectsof sulfate aerosols represents the direct radiative for-cing due to anthropogenic sulfate aerosols by means146 J.Southworthetal. / Agriculture,EcosystemsandEnvironment82(2000)139158of an increase in clear-sky surface albedo proportionalto the local sulfate loading (Carnell and Senior, 1998).The indirect effects of aerosols were not simulated.Our researchusedthe periodof 20502059forclimatescenarios. However, usingasinglescenariohaslimitations, asit isnot possibletocapturetherange of uncertainties as described by the IPCC.This study used two model scenarios to represent thelikely upper and lower boundaries of future (2050s)climate change. The results from HadCM2-GHG andHadCM2-SULcannot be viewed as a forecast orprediction,butratherastwopossiblerealizationsofhow the climate system may respond to a given for-cing. A comparison of the main three climate datasets(Table 2) highlights the differences in projected meanclimate data for the studyregion. Also, a climatevariability analysis was conducted on these two sce-narios,thusincreasingthenumberoffutureclimatescenarios to six. Hence, a range of probable climatechangescenarioswereexaminedtodeterminetheirimpacts on maize growth.2.6. Climate variability analysisInorder toseparatecropresponsetochangesinclimatic means from its response to changes in climatevariability,itisnecessaryrsttomodeltheimpactsof mean temperature changes on crop growth. Then atime series of climate variables with changed variabi-lity can be constructed and added to the mean changescenarios. Hence, when the analysis is undertaken onfuturemeanandvariabilitychanges, itis, therefore,possible to infer what type of climate change causedchanges in yield (Mearns, 1995).The mean conditions for the period 20502059 foreach location, individual future years of climate data,and both GCM model runs were used in this analysis.The variance of the time series was changed for bothtemperature (maximumand minimum) and preci-pitation in time steps of one month. The variance ofeachmonthwasalteredseparately, accordingtothefollowing algorithm from Mearns (1995):X

t = + 1/2(Xt ) (1)and =22(2)where X

tis the new value of climate variable Xt(e.g.monthlymeanmaximumFebruarytemperatureforyear t), the mean of the time series (e.g. the mean ofthemonthlymeanmaximumFebruarytemperaturesfor a series of years), the ratio of the new to the oldvarianceofthenewandoldtimeseries, Xttheoldvalueofclimatevariable(e.g. theoriginal monthlymeanFebruarytemperatureforyeart), 2thenewvariance, and2the old variance.To change the time series to have a new variance2, the variance and mean of the original time serieswascalculatedandthenanewratio()waschosen(e.g. halvingthevariance). Fromtheparameters ,, and the original time series, a new time series withvariance was calculated using equation one. This al-gorithm was used to change both maximum and min-imum temperatures and the precipitation time series.This simple method, as developed by Mearns (1995)wasuseddespitemorecomplexmethodologiesbe-ing developed, due to the comparisons of variabilitytechniques. Mearns et al., (1996, 1997) illustratedthe results obtained fromthe more computation-ally advanced, upper-level statistical techniques weresurprisinglysimilar toprior results fromthemorestatistically simple methodologies. If anything, theseresearchers noted a likelihood of underestimating thenegative impacts of climate variability on crop yieldsusingthesimpler techniques, but intheseanalysesthe differences were quite minor. The approach usedhere permits the incorporationof changes inboththemeanandthevariabilityof futureclimateinacomputationallyinexpensive, highlyconsistent, andreproducible manner.2.7. Model and data limitationsThisstudydoesnot attempt topredict futurecli-mate, but rather, isanevaluationofpossiblefuturechanges in agricultural production in the midwesternUnited States that might result from future changes inclimate. Suchpotentialchangesprovideinsightintopossible larger societal changes needed to control andreduce CO2 in the atmosphere and to help select appro-priate strategies to prepare for change (Adams et al.,1990).The CERES-maize model, as with all models, con-tains several assumptions. Weeds, insects andcropdiseaseshavenodetrimental effect onyield. Also,J.Southworthetal./Agriculture,EcosystemsandEnvironment82(2000)139158147148 J.Southworthetal. / Agriculture,EcosystemsandEnvironment82(2000)139158extremeclimate-relatedevents suchas droughts oroods are not taken into account by the model in termsof extreme crop losses resulting from such events.Other limitations relate to the simplied realityrepresentedbytherepresentativefarms, theuseofasinglesoil typeat eachlocation, andhence, thelossof thespatial variabilityof soils, althoughtheselected soil type was that predominant at each loca-tion. However, theextensivevalidationandanalysisat the farm level is in itself a more detailed analysisthan previously undertaken.Preparing agriculture for adaptation to climatechangerequiresadvanceknowledgeofhowclimatewill change and when. The direct physical effects onplants andtheindirect effects onsoils, water, andother biophysical factors alsomust be understood.Currently, such knowledge is not available for eitherthe direct or indirect effects of climate change. How-ever, guidance canbe obtainedfromanimprovedunderstandingof current climaticvulnerabilities ofagriculture and its resource base. This knowledge canbe obtained from the use of a realistic range of climatechange scenarios and from the inclusion of the com-plexity of current agricultural systems and the rangeofadaptationtechniquesandpoliciesnowavailableandlikelytobeavailableinthefuture(Rosenburg,1992).3. Results3.1. Consequences of changing mean climate andclimate variability on maize yieldsChanges in yield were evaluated by comparing thefuture maize yields to the current VEMAP yields, onthe same hybrids, and then stating the change as per-centagedifference. Decreasesinyieldweregreatestindoubledvariabilityscenarios. Decreasesinyieldweregreater for HadCM2-GHGscenarios thanforHadCM2-SULscenarios. Increasing the variabilityof the future climate scenarios increased the variabi-lity of the year-to-year crop yields obtained from theDSSAT model. The greater variance associated withthe HadCM2-GHGclimate scenario is due to themore extreme increases in mean temperatures associ-ated with this climate scenario, compared to the lesserincreases for the HadCM2-SUL climate scenario.Long-seasonmaizeinallfutureclimatescenarios(Fig. 3) clearlyshoweddecreases inyields inthesouthernandcentral locations. Thedecreasesrangefrom 0 to 45% as compared to yields estimated usingVEMAP data. The largest decreases in yield occurredinwesternIllinois for all futureclimatescenarios.Incontrast, thenorthernlocationsshowedincreasesinyieldsunderthesesamefutureclimatescenarios.In the four northernmost agricultural areas (east cen-tralMichigan, southwestMichigan, easternWiscon-sin, and western Wisconsin) increases in maize yieldranged from 0.1 to 45% as compared to yields usingVEMAPdata. AnexceptiontothispatternwasfortheHadCM2-GHGscenariowithdoubledvariance,where southwestern Michigan and western Wisconsinhad 0.1to 10%decreasesinyield. Thedoubledvariability runs of both scenarios resulted in extremedecreases in yield in the southern locations, and lesssignicant increases in yield in the northern locationsas compared to unchanged or halved variability runs.The greatest gains in yield occurred in the halved vari-ability scenarios because the occurrence of very loworzeroyieldsdecreasedsubstantiallyinthehalvedvariability scenarios.Medium-season maize yields (Fig. 4) showed dra-maticdecreasesinyieldsascomparedtoVEMAP,eveninthenorthernagricultural areasofthestudyregion. Under the HadCM2-GHG and HadCM2-SULscenarios all regions experienced decreases in yieldsof maize, ranging from 15.1 to 40%. In these sce-narios, nolocationsexperiencedincreasesinyield.Under the doubled and halved variability scena-rios, however, the northernmost locations showedlesser decreases in yield and even some increases. Val-ues ranged from 30 to 20% for the halved variabilityscenarios, and from 40 to 15% for the doubled vari-ability scenarios. It must be noted that the actual (sim-ulated) yields obtained for medium-season maize areusually higher than those obtained for longer-seasonmaizeforall climatescenarios(Table3). However,whencomparedtothecurrent VEMAPyieldsasapercentage change the yields frequently decrease.Short-season maize showed a different pattern withall yields decreasing, as comparedtoyields usingVEMAPdata, except in western Illinois (Fig. 5).Inthis agricultural area, yields increasedunder allHadCM2-SULscenarios andthe halvedvariabilityHadCM2-GHG scenario, with increases ranging fromJ.Southworthetal. / Agriculture,EcosystemsandEnvironment82(2000)139158 149Fig. 3. Percent changeinmeanmaximumdecadal yieldfor long-seasonmaize, comparedtoVEMAPyields, for (a) halvedvariabilityHadCM2-GHG, (b)HadCM2-GHG, (c)doubledvariabilityHadCM2-GHG, (d)halvedvariabilityHadCM2-SUL, (e)HadCM2-SUL, and(f) doubled variability HadCM2-SUL.5.1 to 35%. This difference in pattern may be due todifferent factors than the changes in yield of mediumand long-season maize, perhaps related to the strongeast-west precipitation gradient or differences in soilsat this location.3.2. Mean maximum decadal yield versus plantingdatesThe impact of changing climate on planting datesin terms of the mean maximumdecadal yields(Figs. 35) illustrates that under future climate changescenarios later planting dates produced higher yields.In almost all cases (Table 3) the highest mean maxi-mumdecadal yieldoccurs at alater plantingdateunder future climate change, which explains whymedium-season maize varieties frequently have highertotal yieldsthanthelongerseasonmaize. Thelaterplantingdatesforall maizevarietieshavethemostbenecial impact on medium-season maize. The pro-ductivity induced shift to later planting dates is pastthecurrentoptimalplantingdatesforlongerseasonvarietiesandintothemedium-seasonmaizeoptimalplanting dates. Such delays in planting dates also havebeen found by other researchers under future climatechange scenarios (Jones et al., 1999).3.3. Spring freezesIn future climate scenarios, the frequency of maize-killing freeze events will decrease although under dou-bled variance scenarios the intensity of the freeze event150 J.Southworthetal. / Agriculture,EcosystemsandEnvironment82(2000)139158Fig. 4. Percent change in mean maximum decadal yield for medium-season maize, compared to VEMAP yields, for (a) halved variabilityHadCM2-GHG, (b)HadCM2-GHG, (c)doubledvariabilityHadCM2-GHG, (d)halvedvariabilityHadCM2-SUL, (e)HadCM2-SUL, and(f) doubled variability HadCM2-SUL.will increase (Fig. 6). This implies that increased frosttolerance is not an important issue for future climatechange and maize growth as initially expected.4. Discussion4.1. Crop yield changes by regionThefutureclimatescenarios, withincreasedtem-peraturesandprecipitation, resultedinsignicantlyalteredmaizeyields, at eachof the10agriculturalareas in the study region.Across the southern areas yields generallydecreasedduetothedailymaximumtemperaturesbecoming too high and hence, resulting in yielddecline. Western Illinois had yield decreases of 10to50%for long-season maize, 10 to40%for medium-seasonmaize, and 10to +40%forshort-seasonmaize. Forshort-seasonmaizewesternIllinois was theonlylocationwithyieldincreases,although these increases only occurred under theHadCM2-SULscenarios andthe halvedvariabilityHadCM2-GHGscenario, i.e. the less extreme cli-mates. Eastern Illinois had yield decreases of 10 to40% for long and medium-season maize, and 30to 50% decreases for short-season maize. SouthernIllinoisyielddecreasesrangedfrom0to 40%forlong-season maize, 10 to 40% for medium-seasonmaize, and 30to 40%for short-seasonmaize.J.Southworthetal. / Agriculture,EcosystemsandEnvironment82(2000)139158 151Table 3Meanmaximumdecadal yieldplantingdates(Juliandays) under VEMAPcurrent climateandfutureHadCM2GHG(halvedvariability(0.5)/unchangedvariability(1.0)/doubledvariability(2.0))andHadCM2SUL(halvedvariability(0.5)/unchangedvariability(1.0)/doubledvariability (2.0)) climate scenarios for maize varietiesSiteaVEMAPbHadCM2GHGcHadCM2SULd0.5 1.0 2.0 0.5 1.0 2.0ALSCe130 (260f) 186 (204) 186 (212) 179 (173) 179 (227) 186 (223) 179 (210)MSCg158 (261) 207 (218) 207 (185) 74 (165) 193 (215) 200 (194) 81 (178)SSCh179 (235) 195 (132) 207 (138) 218 (125) 200 (159) 207 (147) 200 (139)BLSC 123 (252) 186 (215) 179 (196) 158 (181) 172 (218) 172 (221) 158 (200)MSC 144 (272) 193 (234) 200 (201) 144 (218) 186 (255) 186 (212) 172 (248)SSC 158 (208) 144 (136) 207 (127) 151 (122) 207 (164) 207 (138) 144 (148)CLSC 116 (241) 179 (224) 172 (218) 165 (213) 158 (243) 158 (228) 144 (217)MSC 123 (267) 179 (262) 179 (205) 165 (268) 172 (312) 172 (217) 144 (292)SSC 158 (198) 137 (176) 137 (151) 165 (157) 172 (190) 172 (164) 158 (181)DLSC 137 (212) 165 (220) 172 (219) 114 (208) 137 (236) 151 (235) 144 (227)MSC 137 (274) 179 (272) 186 (222) 144 (269) 165 (303) 172 (232) 151 (299)SSC 158 (208) 130 (169) 137 (156) 151 (171) 165 (182) 172 (177) 172 (184)FLSC 151 (258) 200 (219) 186 (199) 193 (176) 186 (234) 186 (207) 186 (198)MSC 165 (258) 207 (181) 207 (167) 81 (160) 200 (215) 200 (195) 200 (174)SSC 179 (195) 130 (124) 102 (118) 130 (121) 207 (136) 207 (126) 116 (124)GLSC 151 (256) 186 (247) 186 (218) 186 (218) 179 (246) 165 (230) 172 (217)MSC 165 (255) 200 (232) 207 (207) 200 (192) 193 (250) 193 (209) 193 (212)SSC 151 (201) 207 (167) 207 (143) 207 (144) 207 (194) 207 (162) 200 (158)HLSC 123 (184) 179 (233) 158 (190) 172 (196) 165 (243) 151 (223) 151 (218)MSC 144 (252) 193 (265) 193 (189) 186 (238) 186 (299) 172 (214) 172 (280)SSC 158 (224) 193 (151) 144 (145) 179 (140) 193 (190) 179 (162) 186 (168)JLSC 123 (213) 172 (214) 179 (205) 172 (192) 165 (234) 158 (228) 144 (215)MSC 151 (272) 186 (264) 193 (200) 186 (232) 179 (284) 179 (217) 165 (277)SSC 165 (233) 193 (155) 207 (139) 186 (139) 207 (174) 186 (172) 172 (174)KLSC 137 (261) 74 (168) 186 (166) 74 (150) 179 (199) 186 (216) 88 (165)MSC 137 (219) 81 (179) 102 (149) 81 (155) 74 (188) 200 (180) 81 (185)SSC 137 (122) 109 (150) 102 (112) 109 (117) 95 (164) 109 (130) 116 (140)LLSC 137 (171) 179 (242) 165 (218) 158 (212) 151 (240) 130 (215) 158 (220)MSC 137 (269) 186 (300) 179 (207) 179 (262) 179 (307) 151 (213) 158 (286)SSC 137 (209) 200 (176) 200 (163) 193 (162) 193 (192) 186 (171) 186 (168)aA: eastern Illinois; B: east-central Indiana; C: north-west Ohio; D: south-central Michigan; F: southern Illinois; G: south-west Indiana;H: eastern Wisconsin; J: south-west Wisconsin; K: western Illinois; and L: Michigan thumb.bVEMAP dataset for current climate data.cHadCM2GHG is the Hadley Center data for 20502059 from the greenhouse gas only run.dHadCM2SUL is the Hadley Center data for 20502059 from the greenhouse gas and sulphate run.eLSC: long-season maize.fMaize yields are given in bu/ac.gMSC: medium-season maize.hSSC: short-season maize.152 J.Southworthetal. / Agriculture,EcosystemsandEnvironment82(2000)139158Fig. 5. Percent changeinmeanmaximumdecadal yieldforshort-seasonmaize, comparedtoVEMAPyields, for(a)halvedvariabilityHadCM2-GHG, (b)HadCM2-GHG, (c)doubledvariabilityHadCM2-GHG, (d)halvedvariabilityHadCM2-SUL, (e)HadCM2-SUL, and(f) doubled variability HadCM2-SUL.Fig. 6. Number of years across the decade of analysis, witha springfreeze event, for the sevenclimate scenarios: current climate(VEMAP), halved variability HadCM2-GHG (VARF), HadCM2-GHG (GHG), doubled variability HadCM2-GHG (VARH), halved variabilityHadCM2-SUL (VARR), HadCM2-SUL (SUL), doubled variability HadCM2-SUL (VART).J.Southworthetal. / Agriculture,EcosystemsandEnvironment82(2000)139158 153Resultswereveryconsistent acrossclimatescenar-ios. Southwest Indianayields decreasedbetween0and 20%forlong-seasonmaize, 0and 30%formedium-season and short-season maize. East-centralIndiana hadyielddecreases of 10to 30%forlong-seasonmaize, 0to 30%for medium-seasonmaize, and 20 to 50% for short-season maize.Agricultural areas in the northern states of the studyregiontypicallyexperiencedmore increasedyieldsunder the six future climate scenarios, especiallyfor long-season maize. Northwest Ohio had yieldchanges of +10 to 20% for long-season maize, +20to 30% for medium-season maize, and 0 to 30%for short-season maize. South-central Michigan yieldchanges ranged from +20 to 10% for long-seasonmaize, +20 to 20% for medium-season maize, and10 to 30% for short-season maize. The Michiganthumbareaexperiencedthegreatest yieldincreasesfor long-season maize, with +20 to +50% increasesabove current yields, for medium-season maize +20to 30% changes in yield, and 0 to 30% decreasesfor short-season maize. Southwest Wisconsin had+20 to 10% changes in yield for long-season maize,+10 to 30% for medium-season maize, and 20 to50% changes for short-season maize. Finally, east-ern Wisconsin, had yield changes of 0 to +40% forlong-season maize, +20 to 30% for medium-seasonmaize, and 10 to 40% for short-season maize.Theresults across all 10agricultural areas havesome signicant and consistent patterns. The twomain patterns are (1) short-season maize has lowyields compared to current yields under changed cli-mate scenarios except in western Illinois, and (2) thehalved variability climate scenarios produced both thehighest maize yield increases and some of the lowestdecreases in agricultural areas in the southern states,indicatingthat changesinfutureclimatevariability,producing more extreme climatic events, will be detri-mental to future agricultural production. Hence, as thisresearch illustrates, it is extremely important to modelboth changes in mean and variability of future climate.Our results indicate that the currently grown (pre-dominant) maize hybrid (long-season maize) willhaveincreasedorbetteryieldsunderfutureclimateconditions, compared to current yields, than willthemediumandshort-seasonmaizehybrids. How-ever, in terms of actual yield, medium-season maizeyields are frequently greater than those obtained fromlonger-seasonmaizefor thesameclimatescenariodue to the later planting dates under climate change.Short-seasonmaizedoesnot appeartobeasviableunder changedclimateconditions across thestudyregion.Spatially, resultsshowthat theagricultural areasinthenorthernstates(southwestWisconsin, easternWisconsin, south-central Michigan, northwest Ohio,andtheMichiganthumb) will experienceincreasesinmaizeyieldsasaresultofclimatechange,whilethose in the southern and central regions (western Illi-nois, eastern Illinois, southern Illinois, southwest In-diana, andeast-central Indiana)will showaclearlydecreasing trend. The more extreme climate scenario,asrepresentedbyHadCM2-GHG, resultsingreaterreductionof maizeyieldsthanHadCM2-SUL. TheHadCM2-GHGscenario produces mean monthly sum-mer temperatures that are 14Cwarmer thantheHadCM2-SULscenarios. Increasedsurfaceairtem-peratures result in a reduction in agricultural produc-tivityinmanycropsduetoearlieroweringandashortening of the grain-ll period. The shorter the cropduration, the lower yield per unit area (Lal et al., 1998),as is seen in results in central and southern locationsin the study region. However, in northern locations ofthe study area, where low temperatures currently limitthe grain ll period, increases in temperatures due toclimatechangewillresultinthegrainllingperiodlengthening and increased yields.4.2. Daily maximum temperaturesHigh temperatures affect agricultural productiondirectlythroughtheeffectsofheat stressat criticalphenological stagesinthecropsgrowth. Inmaize,hightemperaturesat thestagesofsilkingortassel-ingresultinsignicantdecreasesinyield. BoththeCERES-maize and another crop model, EPIC, use thetemperature-sumapproachtocalculatedevelopmen-tal time, where GDD = [(T max T min)/2] T base.In the case of the EPIC model, the only condition isthat GDD cannot be 35CMaximumtemperature>40CMaximumtemperature>35CMaximumtemperature>40CMaximumtemperature>35CMaximumtemperature>40C0.5 1.0 2.0 0.5 1.0 2.0 0.5 1.0 2.0 0.5 1.0 2.0A 5 0 92 80 95 13 28 28 44 45 37 0 5 0B 1 0 62 69 56 6 9 3 12 18 14 0 0 0C 1 0 20 34 35 0 1 1 6 5 6 0 0 0D 0 0 15 12 10 0 0 0 2 2 0 0 0 0F 9 0 89 83 93 22 28 30 32 44 44 0 1 0G 4 0 60 65 69 4 9 7 8 17 19 0 0 0H 0 0 43 35 44 1 4 0 2 4 4 0 0 0J 0 0 55 48 45 1 3 2 1 7 1 0 0 0K 3 0 87 80 91 30 28 39 44 45 47 0 5 0L 0 0 15 16 22 0 1 0 0 1 0 0 0 0aA: eastern Illinois; B: east-central Indiana; C: north-west Ohio; D: south-central Michigan; F: southern Illinois; G: south-west Indiana;H: eastern Wisconsin; J: south-west Wisconsin; K: western Illinois; and L: Michigan thumb.bVEMAP dataset for current climate data.cHadCM2GHG is the Hadley Center data for 20502059 from the greenhouse gas only run.dHadCM2SUL is the Hadley Center data for 20502059 from the greenhouse gas and sulphate run.atureisanimportantreasonfortheselectionoftheCERES-maize model.Using regression analyses, Rosenzweig (1993)found that daily maximum temperatures greater than33.3CinJulyandAugust were negativelycorre-lated with maize yield in the US Maize Belt and thatdaily maximum temperatures >37.7C caused severedamage to maize. The future climate scenarios usedhadmaximumdailytemperatures>35Conseveraldays duringJulyandAugust (Table4). Results inthistablematchcloselywithyieldchangeswithanincreasednumberofdayswithtemperaturesgreaterthan 35C within a given climate scenario, resultingin decreased yields.Hoogenboom et al. (1995), using the CERES-maizemodel, foundthat nomaizegrowthoccurredat airtemperatures below 8C; maximum crop growth andgrain ll occurred at daily temperatures of 34C; andgrowthwasreducedathighertemperaturesupuntil44C, above which no growth occurred. Due to thistemperaturesensitivity, maximumdailytemperaturewill become important in the future climate scenarios,as will the frequency and duration of such occurrences.J.Southworthetal. / Agriculture,EcosystemsandEnvironment82(2000)139158 155TheUSEnvironmental ProtectionAgency(EPA)foundadecreaseinmaizeyields under conditionsoffutureclimatechangeof442%duetotempera-tures rising above the range of tolerance for the maizecrops (EPA, 1998). Saarikko and Carter (1996) foundthat thethermal suitabilityforspringwheat inFin-land could shift northwards by 160180 km per 1Cincrease in mean annual temperature. In areas of cur-rent growth the timing of crop development under awarmer climate shifts to earlier in the year, thus short-eningthedevelopmentphase,resultingindecreasedyields. In northwest India, Lal et al. (1998) found a re-duction of 54%in wheat yields with a 4Crise in meandaily temperatures. Using a doubled atmospheric CO2concentration (720 ppmv) from present day and a 5Crise in mean daily temperatures, the decrease in yieldwas only 32% from current conditions. These resultsaresimilarintermsofpatternandtrendtothoseofthis research, with increasing summer maximum tem-peratures resulting in decreased yields.4.3. Impacts of CO2 fertilizationThis approach, using the CERES-maize model,enables us to model the predicted future climate, andCO2levels based on this future climate, and to eval-uatethecropresponse. ThisresearchusedafutureatmosphericCO2concentrationof 555 ppmv, com-pared to 360 ppmv for current conditions. For maize,a C4 crop, this response is not as important as for C3crops such as soybeans. C4 crops are more efcientphotosyntheticallythanC3plantsandshowlessre-sponse to increasing atmospheric CO2 concentration,which provides a future potential agricultural adapta-tion of C3 crops over C4 crops due to their enhancedgrowth functions with higher concentrations of CO2(RosenzweigandHillel, 1998; Rosenzweig, 1993).Assessment of both the effects increased atmosphericCO2concentrationsandclimaticchangeimpactsonagricultural productionisacrucial areaofresearchbecause the two factors occur together.4.4. Climate variability impacts on maize yieldsThehalvedvariabilityclimatescenariosproducedmaizeyieldswiththegreatestincreasesinyieldforlong-season maize. In addition, decreases in yield fortheagriculturalareasinthesouthernstates, andforthemedium-seasonandshort-seasonmaizevarietieswere much less extreme. These results were expectedfor the halved variability scenarios which resulted inless extreme events, e.g. fewer spring freezes (Fig. 6)and a decreased number of extreme temperature events(Table 4).The doubled climate variability scenarios repre-sent the most extreme climate scenarios modeled. Inaddition, adoublingof current or futurevariabilityconditions is probably at the maximum limit of likelychanges in climate variability. As such, these doubledvariability scenarios probably represent the most ex-tremevariabilitychangesthatmightoccurby2050.Underthedoubledvariabilityscenarios, particularlythedoubledvariabilityHadCM2-GHGscenario, thegreatest decreases inmaize yields for long-seasonmaize are found. Medium-seasonandshort-seasonmaize also experience large decreases in yield underthese scenarios. The doubled variability scenario alsoresultsinhighlyvariableyear-to-year variabilityinmaize yields across the 10 years modeled and studied.These results are in accordance with those found byother researchgroups, andarenot surprisinggiventhe high incidence of days with extreme temperatures,compared to the halved variability scenarios (Table 4).Whenevaluatingall sixfutureclimatescenariosinterms of their impacts on midwestern maize yields itis quite evident that the most detrimental agriculturalimpacts would arise from a future climate similar tothe HadCM2-GHG doubled variability scenario.Moreextremeweathereventsandincreasedvari-ability of the weather will result in lower maize yields.Phillips et al. (1996) studied climate impacts on cropyields in the United States and found that changes invariability affected mean yields less than changes inmean climate, but did affect changes in inter-annualyield variability. Mearns et al. (1996, 1997) found thatincreases in climate variability, on a scale of doublingcurrent variability, resulted in substantially decreasedcropyields for wheat. Inaddition, theyfoundde-creased variability to have little effect on mean yield.Wolfetal.(1996)andSemenovetal.(1996)foundthat highertemperaturesinRothamsted, UKandinSeville, Spain, resulted in lower grain yields for springwheat. In addition, a doubling of temperature variabil-ity resulted in an additional decrease in yield (usingthemodelsCERES, AFRCWHEATandNWHEAT,156 J.Southworthetal. / Agriculture,EcosystemsandEnvironment82(2000)139158but not for SIRIUS which had no change) across allsites, whichwasrelatedtoanincreasednumberofdays at sub-optimal temperatures. For the same loca-tionsSemenovandBarrow,(1997)foundthatwhenchanges in climate variability are included in a climatechange analysis the results found are quite different.For wheat yields, a decrease in yield of 20% was ob-served when variability was doubled. Again, these re-sults are similar to those reported in this research.4.5. Potential adaptations to climatic changeAnotherimportantareaofresearchconcernspos-sible adaptation strategies to climate change and theeffects of those strategies. The most obvious adapta-tionsidentiedinthisresearchare(1)thedevelop-ment ofamoreheat tolerant hybridoflong-seasonmaizeand(2)switchingfrommaize(aC4crop)tosoybeans (a C3 crop) to take advantage of increasedatmospheric CO2 concentrations promoting increasedgrowthandgreater tolerances for hot temperatures(although how realistic this may be will be dependenton market factors). In fact, increased heat tolerance inshort and medium-season maize varieties may providethe opportunity to manipulate planting dates of thesehybridsandprovideadaptationequal orsuperiortoadaptationoflongtermvarietiesundersomecondi-tions, which is illustrated by the use of shorter seasonmaizevarietiesratherthansorghuminrecent yearsinthesouthwesternUnitedStates. Under increasedclimate variability and increased extreme events, soilmoisture management will become more critical andwill require improved soil inltration and water hold-ing capacity. Tillage and cropping systems that yieldthese benets will increase in economic value tofarmers. Also, there will be increased concern aboutsoil erosion with more extreme rain events, especiallyif agricultural programstandards for conservationcompliance that limits erosion are tightened.5. ConclusionsOur primary conclusions are:A lengthened growing season, dominated by a cen-tral period of high maximum daily temperatures, isa critical inhibitor to maize yields. Late spring andearly fall frosts do not affect maize yields.Thenorth-southtemperaturegradient inthemid-western Great Lakes states is extremely importantin inuencing patterns of maize yield under futureclimate conditions.Climate variability is a signicant factor inuenc-ing maize yields because increased climate variabil-ity results in the largest decreases in future maizeyields.Understanding responses of individual farms tochanges in mean climate and changes in climate vari-abilityis essential tounderstandingtheimpacts ofclimate change on agriculture at a regional scale(Wassenaar et al., 1999). The research discussedhereis part of alarger project examiningpossiblefarm-level adaptationstothepotential changespre-dictedfromthecropmodeling. Continuingresearchwill incorporate crop modeling of soybeans (DSSATSOYGRO) and wheat (DSSAT CERES-wheat), bothin terms of the potential mean changes in future cli-mate and the potential changes in climate variability.These results will be used as inputs into the PurdueCropLinearProgram(PC/LP)modelforfarmleveldecision analysis. The results from the DSSAT mod-els (CERES-maize, CERES-wheat, and SOYGRO)owintoPC/LP, thenasmanagement/economicde-cisions change the type of production, results are fedbackintothecropmodel for further adjustment tocrop production modeling. This will allow the devel-opment of farm level strategies to be created and thentestedbyrunningbackthroughthemodelscenarioswith the adaptations incorporated.The approach taken in this research examines adap-tation at the farm level. Other research has examinedagricultural response to climate change primarily on aregional or national basis. Both are important. How-ever, at the local level, climate change research mustincludethefull spectrumofclimate, soils, biology,management, and economics if there is to be any linkbetweenanalysisandreality.Thisresearchhopestoprovide the basis for strategic planning and risk man-agement by farmers and the agricultural infrastructureto better adapt to changing conditions.AcknowledgementsThis research was funded by Grant No. R 824996-01-0fromtheSciencetoAchieveResults (STAR)J.Southworthetal. / Agriculture,EcosystemsandEnvironment82(2000)139158 157Programof the United States Environmental Pro-tectionAgency. WethankDr. DavidViner for hishelp and guidance with the use of the HadCM2 data,which was provided by the Climate Impacts LINK1Project (DETR Contract EPG 1/1/68) on behalf of theHadley Center and the United Kingdom Meteorologi-cal Ofce. We thank Dr. Joe Ritchie at Michigan StateUniversity for his help and advice on CERES-maizethroughout this project and Dr. Susan Riha at CornellUniversity for her comments on the temperature sen-sitivity of CERES-maize. We thank Dr. Timothy Kit-tel at the National Center for Atmospheric Research(NCAR), Boulder, Colorado for recommendations,help and guidance with VEMAP data acquisition anduse. We greatly appreciate the assistance of Dr. LindaMearns at NCAR in discussions on climate variabilityand for her comments regarding this manuscript andour research. We thank Michael R. Kohlhaas for hishelp with editing earlier versions of this manuscript.Finally, we thank the participants in the IPCC GCTEFocus3Conference, heldinReading, UK, Septem-ber 1999, for wonderful discussions andinsightfulcomments and questions.ReferencesAdams, R.A., Hurd, B.H., Reilly, J., 1999. Agricultural and GlobalClimate Change: A Review of the Impacts to US AgriculturalResources. On-Line PewReport at http://www.pewclimate.org/projects.Adams, R.M., Rosenzweig, C., Peart, R.M., Ritchie, J.T., McCarl,B.A., Glyer, J.D., Curry, R.B., Jones, J.W., Boote, K.J., AllenJr., L.H., 1990. Global climate change and USagriculture.Nature 345, 219224.Barrow, E., Hulme, M., Semenov, M., 1996. Effect of using diff-erent methods in the construction of climate change scenarios:examples from Europe. Clim. Res. 7, 195211.Carnell, R.E., Senior, C.A., 1998. Changesinmid-latitudevari-ability due to increasing greenhouse gases and sulphate aerosols.Clim. Dyn. 14, 369383.Chiotti, Q.P., Johnston, T., 1995. Extending the boundaries ofclimatechangeresearch: adiscussiononagriculture. J. RuralStud. 11, 335350.Dhakhwa, G.B., Campbell, C.L., LeDuc, S.K., Cooter, E.J., 1997.Corn growth: assessing the effects of global warming and CO2fertilizationwithcropmodels. Agric. For. Meteorol. 87, 253272.Environmental ProtectionAgency(EPA), 1998. ClimateChangeand Indiana. Ofce of Policy, EPA 236-F-98-007g.1LINK homepage at http://www.cru.uea.uk/link/.Harrison, P.A., Buttereld, R.E., 1996. Effectsofclimatechangeon Europe-wide winter wheat and sunower productivity. Clim.Res. 7, 225241.Hoogenboom G., Tsuji, G.Y., Pickering, N.B., Curry, R.B., Jones,J.J., Singh, U., Godwin, D.C., 1995. Decision support system tostudyclimatechangeimpactsoncropproduction.In:ClimateChange and Agriculture: Analysis of Potential InternationalImpacts. ASA Special Publication No. 59, Madison, WI, USA,pp. 5175.Hulme, M., Barrow, E.M., Arnell, N.W., Harrison, P.A., Johns,T.C., Downing, T.E., 1999. Relative impacts of human-inducedclimate change and natural climate variability. Nature 397, 688691.IPCC(Intergovernmental Panel onClimateChange), 1990. Firstassessment report. In: Houghton, J.T., Jenkins, G.J., Ephraums,J.J. (Eds.), Scientic Assessment of Climate Change Reportof Working Group I. Cambridge University Press, UK.IPCC (Intergovernmental Panel on Climate Change), 1995. Secondassessment report climate change. In: Houghton, J.T.,Meira Filho, L.G., Callender, B.A., Harris, N., Kattenburg, A.,Maskell, K. (Eds.), The Science of Climate Change. CambridgeUniversity Press, UK.Jones, J.W., Jagtap, S.S., Boote, K.J., 1999. Climate change:implications for soybean yield and management in the USA. In:Kaufman, H.E. (Ed.), ProceedingofWorldSoybeanResearchConference VI, August 47, 1999, Chicago. University ofIllinois, Urbana-Champaign, IL. pp. 209222.Kaiser, H.M., Riha, S.J., Wilks, D.S., Sampath, R., 1995. Adap-tationtoglobal climatechangeat thefarmlevel. In: Kaiser,H.M., Drennen, T.E. (Eds.), Agricultural Dimensions of GlobalChange. St. Lucie Press, FL, USA, pp. 136152.Kane, S., Reilly, J., Tobey, J., 1992. Anempirical studyof theeconomic effects of climate change on world agriculture. Clim.Change 21, 1735.Katz, R.W., Brown, B.G., 1992. Extremeevents inachangingclimate: varibility is more important than averages. Clim.Change 21, 289302.Kittel, T.G.F., Rosenbloom, N.A., Painter, T.H., Schimel, D.S.,Fisher, H.H., Grimsdell, A., VEMAPParticipants, Daly, C.,Hunt Jr., E.R., 1996. The VEMAP phase I database: anintegrated input dataset for ecosystem and vegetation modelingfor the conterminous United States. CDROM and World WideWeb (URL=http://www.cgd.ucar.edu/vemap/).Lal, M., Singh, K.K., Rathore, L.S., Srinivasan, G., Saseendran,S.A., 1998. Vulnerability of rice and wheat yields in NW Indiato future changes in climate. Agric. For. Meteorol. 89, 101114.Liang, X., Wang, W., Dudek, M.P., 1995. Interannual variabilityof regional climate and its change due to the greenhouse effect.Glob. Plan. Change 10, 217238.Mavromatis, T., Jones, P.D., 1998. Comparison of climate changescenario construction methodologies for impact assessmentstudies. Agric. For. Meteorol. 91, 5167.Mearns, L.O., 1995. Researchissues indeterminingtheeffectsof changing climate variability on crop yields. In: ClimateChange and Agriculture: Analysis of Potential InternationalImpacts. ASA Special Publication No. 59, Madison, WI, USA,pp. 123143.158 J.Southworthetal. / Agriculture,EcosystemsandEnvironment82(2000)139158Mearns, L.O., Rosenzweig, C., Goldberg, R., 1997. Meanandvariance change in climate scenarios: methods, agriculturalapplications, andmeasures of uncertainty. Clim. Change35,367396.Mearns, L.O., Rosenzweig, C., Goldberg, R., 1996. The effectof changes in daily and inter-annual climatic variability onCeres-wheat: a sensitivity study. Clim. Change 32, 257292.Mearns, L.O., Katz, R.W., Schneider, S.H., 1984. Extremehigh-temperature events: changes in their probabilities with changesin mean temperature. J. Clim. Appl. Meteorol. 23, 16011613.Nafziger,E.D.,1994.Cornplantingdateandplantpopulation.J.Prod. Agric. 7, 5962.Pearce, F., 1997. State of the Climate ATime for Action.WWF Report.Peiris, D.R., Crawford, J.W., Grashoff, C., Jeffries, R.A., Porter,J.R., Marshall, B., 1996. Asimulationstudyof cropgrowthanddevelopment underclimatechange. Agric. For. Meteorol.79, 271287.Phillips, D.L., Lee, J.J., Dodson, R.F., 1996. Sensitivityof theUS corn belt to climate change and elevated CO2: I. Corn andsoybean yields. Agric. Syst. 52, 481502.Rind, D., 1991. Climate variability and climate change. In:Schlesinger, M.E. (Ed.), Greenhouse-Gas-Induced ClimateChange: A Critical Appraisal of Simulations and Observations.Development in Atmospheric Science 19, Elsevier, New York.Rosenburg, N.J., 1992. Adaptation of agriculture to climate change.Clim. Change 21, 385405.Rosenzweig, C., 1993. Modeling crop responses to environmentalchange. In: Solomon, A.M., Shugart, H.H. (Eds.), VegetationDynamics and Global Change. Chapman and Hall, NY, USA.Rosenzweig C., Hillel, D., 1998. Climate Change and theGlobal Harvest: Potential Impacts of the Greenhouse Effect onAgriculture. Oxford University Press, New York.Rosenzweig, C., Hillel, D., 1993. Agriculture in a greenhouseworld. Natl. Geograph. Res. Explor. 9, 201221.Saarikko, R.A., Carter, T.R., 1996. Estimating the development andregional thermal suitabilityof springwheat inFinlandunderclimatic warming. Clim. Res. 7, 243252.Semenov, M.A., Barrow, E.M., 1997. Use of a stochastic weathergenerator in the development of climate change scenarios. Clim.Change 35, 397414.Semenov, M.A., Wolf, J., Evans, L.G., Eckersten, H., Iglesias, A.,1996. Comparisonof wheat simulationmodelsunder climatechange. II. Application of climate change scenarios. Clim. Res.7, 271281.Smith, J.B., Tirpak, D.A. (Eds.), Appendix C: Agriculture in ThePotential Effects of Global Climate Change on the UnitedStates. US Environmental Protection Agency, Washington, DC,1989.Thornton, P.K., Bowen, W.T., Ravelo, A.C., Wilkens, P.W., Farmer,G., Brock, J., Brink, J.E., 1997. Estimatingmillet productionfor famine earlywarning: anapplicationof cropsimulationmodeling using satellite and ground based data in Burkina Faso.Agric. For. Meteorol. 83, 95112.UnitedStatesDepartment of Agriculture(USDA), 1997. Intera-gency agricultural projections committee. Agricultural BaselineProjections to 2005, Reecting the 1996 Farm Act. Staff ReportWAOB-97-1. World Agricultural Outlook Board, Ofce of theChief Economist, Washington, DC.Wassenaar, T., Lagacherie, P., Legros, J.P., Rounsevell, M.D.A.,1999. Modelling wheat yield responses to soil and climatevariability at the regional scale. Clim. Res. 11, 209220.Wolf, J., Evans, L.G., Semenov, M.A., Eckersten, H., Iglesias, A.,1996. Comparisonof wheat simulationmodelsunder climatechange. I. Model calibration and sensitivity analyses. Clim.Res. 7, 253270.