ContentslistsavailableatSciVerseScienceDirect
Field
Crops
Research
j o ur n a l ho me p a g e :w w w . e l s e v i e r . c o m / l o c a t e / f c r
Estimating
crop
yield
potential
at
regional
to
national
scales
Justin
van
Wart
a,∗, K.
Christian
Kersebaum
b, Shaobing
Peng
c, Maribeth
Milner
a,
Kenneth
G.
Cassman
aaDepartmentofAgronomyandHorticulture,UniversityofNebraska-Lincoln,202KeimHall,Lincoln,NE68583-0915,USA bLeibniz-CenterofAgriculturalLandscapeResearch(ZALF),EberswalderStraße84D-15374Müncheberg,Germany
cCropPhysiologyandProductionCenter(CPPC),NationalKeyLaboratoryofCropGeneticImprovement,MOAKeyLaboratoryofCropPhysiology,EcologyandCultivation(TheMiddle
ReachesofYangtzeRiver),HuazhongAgriculturalUniversity,Wuhan,Hubei430070,China
a
r
t
i
c
l
e
i
n
f
o
Articlehistory: Received18June2012 Receivedinrevisedform 24November2012 Accepted27November2012 Keywords: Yieldpotential Simulationmodel Foodsecurity Yieldgap Rice Maize Wheat
a
b
s
t
r
a
c
t
Worldpopulationwillincrease35%by2050,whichmayrequiredoublingcropyieldsonexistingfarmland tominimizeexpansionofagricultureintoremainingrainforests,wetlands,andgrasslands.Whetherthis ispossibledependsonclosingthegapbetweenyieldpotential(Yp,yieldwithoutpest,disease,nutrient orwaterstresses,orYwunderwater-limitedrainfedconditions)andcurrentaveragefarmyieldsinboth developedanddevelopingcountries.Quantifyingtheyieldgapisthereforeessentialtoinformpolicies andprioritizeresearchtoachievefoodsecuritywithoutenvironmentaldegradation.Previousattempts toestimateYpandYwatagloballevelhavebeentoocoarse,general,andopaque.Ourpurposewasto developaprotocoltoovercometheselimitationsbasedonexamplesforirrigatedriceinChina,irrigated andrainfedmaizeintheUSA,andrainfedwheatinGermany.SensitivityanalysisofsimulatedYporYw foundthatrobustestimatesrequiredspecificinformationoncropmanagement,+15yearsofobserved dailyclimatedatafromweatherstationsinmajorcropproductionzones,andcoverageof40–50%oftotal nationalproductionarea.NationalYpestimateswereweightedbypotentialproductionwithin100-km ofreferenceweatherstations.Thisprotocolisappropriateforcountriesinwhichcropsaremostlygrown inlandscapeswithrelativelyhomogenoustopography,suchasprairies,plains,largevalleys,deltasand lowlands,whichaccountforamajorityofglobalfoodcropproduction.Resultsareconsistentwiththe hypothesisthataveragefarmyieldsplateauwhentheyreach75–85%ofestimatednationalYp,which appearstooccurforriceinChinaandwheatinGermany.Predictionofwhenaveragecropyieldswill plateauinothercountriesisnowpossiblebasedontheestimatedYporYwceilingusingthisprotocol.
©2012ElsevierB.V.
1. Introduction
Worldpopulationisprojectedtoincrease35%by2050,which willrequirea 70–100% risein foodproduction givenprojected trends in diets, consumption, and income (Bruinsma, 2009; Rosegrantetal.,2009;UNFPA,2010).Increasedfoodproductioncan beachievedbyraisingcropyieldsonexistingfarmland,expanding cropproductionarea,orboth.Expansionofcroparea,however, comesat the expense of substantial greenhouse gasemissions (IPCC,2007;Searchingeretal.,2008),whichwouldcontributeto climatechange(KarlandTrenberth,2003).
Theextenttowhichincreasedfoodproductionrequires expan-sionofcultivatedareawillbedeterminedlargelybycropyield potential(Yp),whichisdefinedasthemaximumattainableyield per unit land area that can be achieved by a particular crop
∗ Correspondingauthor.
E-mailaddresses:Justin.vanwart@gmail.com(J.vanWart),ckersebalm@zalf.de
(K.C.Kersebaum),speng@mail.hzau.edu.cn(S.Peng),
Mmilner1@unl.edu(M.Milner),kcassman1@unl.edu(K.G.Cassman).
cultivarinanenvironmenttowhichitisadaptedwhenpestsand diseasesareeffectivelycontrolledandnutrientsarenon-limiting (Evans, 1993). In irrigated systems, Yp is determined by tem-peratureregimeand solarradiationduringthegrowingseason. Water-limitedyieldpotential(hereaftercalledwater-limitedyield; Yw)istherelevantmeasureofmaximumyieldattainableinrainfed systems.DespitetheimportanceofYpandYwtofoodproduction capacity,theyarenotexplicitlyconsideredinstudiesofindirect landusechangeasaffectedbypoliciesaboutbiofuels(Searchinger etal.,2008),conservationofbiodiversity(Phalanetal.,2011),or futurefoodsecurity(Godfrayetal.,2010).Accurateestimatesof YpandYwarealsoneededtointerpretyieldtrendsinregionsand countrieswhereaggregatedataindicateyieldstagnation(Cassman etal.,2003;Lobelletal.,2009).Forexample,riceyieldsappearto haveplateauedinJapanandChina;maizeyieldshavebeen stag-nantinChina,Italy,andFrance;andwheatyieldsarenotincreasing innorthernEuropeandIndia(Brissonetal.,2010;Cassmanetal., 2010).Yieldstagnationinthesemajorgrainproductionareasputs pressureonotherregionstoeitheraccelerateyieldgrowthratesor expandcultivatedareatomakeupthedifferencebetweenglobal supplyanddemand.Hence,understandingthecauseoftheseyield 0378-4290©2012ElsevierB.V.
http://dx.doi.org/10.1016/j.fcr.2012.11.018
Open access under CC BY-NC-ND license.
plateausiscriticaltodeterminingwhetheritispossibletoresume yieldadvanceorifthefocusshouldbeplacedonacceleratingyields inothergrainproducingregions.
Oneexplanationforyieldplateausisthataveragefarmyields haveapproachedaYporYwceiling.Thatis,plateausoccurbecause itis impossibleforaverageyieldin a regionor nationtoreach Yp orYw for two reasons: (1)100% of farmers cannot achieve the perfectionof crop and soil management required to reach YporYw, and(2) cropresponsetoadditionalinputs exhibitsa diminishingmarginalyieldbenefitasyieldapproachesthe ceil-ing,whichdecreasesthemarginalcost-benefitofadditionalinputs andreducesincentivestoexploitthesmallremaininggapbetween farmyieldlevelsandYporYw.Thisisnottosaythatplateausare duetoreducedefficienciesinresponsetoinputs.Onthecontrary, improvedcultivarswithimprovedresistancetobioticandabiotic stressresistanceachievehigheryieldsthantheoldercultivarsthey replaceatthesamelevelofinputs,whichmeansgreaterinputuse efficiencies.Instead,plateausmayreflectthefactthatasyieldsrise towardtheplateau,marginalreturntoadditionalinputsbecome smallerandthusfarmershavelessmotivationtotryandachieve highest possibleyield. Therefore, averageregionaland national yieldscanbepredictedtoplateauwhentheyreach70–90%ofYp orYw(Cassman,1999;Cassmanetal.,2003;Grassinietal.,2009). BecausethereislittleevidencethattheYpceilinghasincreased duringthepast30yearsinmaizeandrice(Cassman,1999;Duvick andCassman,1999;Pengetal.,1999)orwheat(Grayboschand Peterson,2010),accurateestimatesofceilingyieldlevelsare criti-caltodeterminewhetherplateauingcropyieldsresultfromlackof anexploitablegapbetweenaveragefarmyieldsandYpinirrigated systems,orYwinrainfedsystems.
AtissueiswhetheravailablemethodstoestimateYpandYware goodenoughtohelpinterpretyieldtrendsthatindicateaplateau ortoinformdevelopmentofpoliciesthatseektoreduceGHG emis-sionsfromagriculture,includingdirectandindirecteffectsofland usechange.CropsimulationmodelscanbeusedtoestimateYpand Ywbasedoncurrentmanagement,geneticfeaturesofthecrop, weatherandwatersupply.Butcropmodelsthatperformwellin evaluationofyieldatthefieldorfarmlevelsdonotgenerally per-formwellwhenscaleduptoregionalornationallevels(Witetal., 2005).Inlargepartthisperformanceproblemreflectsdifficultyin scalingweatherdatafrompointestimatesatground-basedstations tolargergeospatialscales.
A common approach for estimating current or future crop yieldsata globallevelutilizesa weather databaseinterpolated to0.5◦×0.5◦ grid,or roughly 3100km2 at theequator(Fischer
etal., 2002;Lobell and Field,2007;Priya and Shibasaki,2001). Thestrengthofthisinterpolatedgridapproachisthatitprovides globalcoverageofterrestrialecosystems.Twoweaknessesof spa-tialinterpolationofdataare:(1)itreducesthedegreeofvariability intemperature,rainfall,andsolarradiationacrossalandscapedue tovariationintopographywithinthegridcell,and(2)the geospa-tialdistributionofcropareawithinagridisnotuniformandis typicallyconcentratedincertainzonesacrossthelandscape.The attenuationofvariabilityintemperature,rainfallandsolar radia-tioncanresultinoverorunder-estimationofyieldsforcropsthat relyonrainfallbyasmuchas10–50%(Baronetal.,2005). Further-more,thequalityofgeospatiallyinterpolatedweatherdataisnot uniformacrosstheglobebecausegeospatialdensityofweather sta-tionsisverylowinsomeregions.Anotherapproachistoassume highestyieldingfieldsforaparticularenvironmentasyield poten-tialyields,buttheseyieldsmaybetheresultofasinglegoodyear anddonotrepresentlong-termaverageyieldpotentialforagiven location(Lickeretal.,2010).Assumingallareasoftheglobecan behandledinthesamewayignorescomplexityinthegeospatial distributionofcultivatedlandduetodifferencesintopographyand weather.Useofactualweatherdataoveranumberofyearsfrom
groundstationsthatarespatiallycongruentwithgeospatial distri-butionofcropproductionavoidssuchweaknessesassociatedwith useofinterpolatedweatherdataor“averaged”cropproduction statisticsinestimatingYporYw.
Anotherissueisthemostappropriate time-stepforweather datausedtosimulatecropyields.Previousglobal,national,and regionalestimatesofYporYwaremostlybasedonweatherdata derivedfrommonthlymeansorsimulatedclimaticyearsbasedon historicalvariances(Andarzianetal.,2008;Deryngetal.,2011; Neumannetal.,2010;Nonhebel,1994;PriyaandShibasaki,2001; vanBusseletal.,2011).However,monthlymeansaretoocoarse andinterpolatingthesemeanstoderivedailyvaluesdoesnot cap-turewithinmonthvariabilityaddingadditionaluncertaintytothe weatherinputdata.AccuratesimulationofYpor Ywrequiresa dailytimesteptofullycapturetheimpactofcurrentcrop manage-mentpractices,oradaptivemanagementinresponsetochanges inclimateaswellasthehistoricalvariabilityofweather within thecourseofthemonth.BothYpandYwarehighlysensitiveto thedateofplantingand cultivarselection intermsofmaturity, which togetherdeterminethetiming ofkeygrowthstages and lengthofcrop-growingseason(Cassmanetal.,2010;Grassinietal., 2009;WangandConnor,1996;Yangetal.,2006).Suchsensitivityis especiallyimportanttoestimateYpandYwbysimulationofcrops grownintemperateagroecologicalzones,suchastheU.S.CornBelt, wherelengthofgrowingseasonisdeterminedbyexpecteddateof firstandlastfrost.Specificationofplantingandmaturitydatesalso areimportantinmultiplecroppingsystemsintropicaland semi-tropicalregionswheretwoorthreecropcyclesoccureachyearon thesamefield.
Inadditiontoweatherdatawithdailytime-step,an appropri-atesimulationmodelis neededtoestimateYpandYw. Models shouldbewelldocumentedandvalidatedagainstyieldsofcrops growninfieldswhere,apartfromweather,yield-limitingfactors havebeeneliminated(Kropffetal.,1993;Lobelletal.,2009;Yang etal.,2006).Whilesomepreviousstudieshaveusedgeneric, non-species-specificrelationshipsbetweenincidentsolarradiationand plant biomassproduction toestimatenet primary productivity (PenningdeVriesetal.,1997;DoorenbosandKassam,1979),such modelsarenotabletosimulatecropphenology,whichiscontrolled byspecies-specifictraitsessentialforaccuratesimulationofcrop maturityandgrainyield.
Basedonreviewoftheliterature,weconcludethatavailable methodsdo notgiverobust,reproducible, andtransparent esti-matesofcropYporYwatregionaltonationalscales.Giventhe needforaccurateestimatesofceilingyieldlevelsforinterpreting currentyieldtrendsandforstudiesoffuturefoodsecurityandland useunderchangingclimate,wesetouttodevelopan appropri-atemethodforestimatingtheseyieldbenchmarksatregionalto nationalscales.Todevelopsuchaprotocolrequiresaddressingthe followingissues:(1)whataretheminimumweatherdata require-mentsforaccuratesimulationofYporYwatagivenlocation,(2) whatlevelofspecificityincropmanagementpracticesisrequired, and(3)howbesttoscaleupestimatesofYpandYwfrom location-specificestimatestoregionalandnationalscales.Thesequestions wereexaminedforrainfedandirrigatedmaizeintheUSA(28Mha and 4Mha harvested area, respectively), irrigated rice in China (30Mha),andrainfedwheatinGermany(3Mha).Ourattemptsto answertothesequestionsledustoproposeaprotocolfor estimat-ingYpandYwatregionaltonationalscales.
2. Materialsandmethods
2.1. Geospatialdistributionofharvestedcroparea
Ageospatialdatabaseofharvestedcroparea(Portmannetal., 2010)wasusedtoidentifyregionswithlargecropproductionarea.
Fig.1. DistributionofNOAAweatherstationswith20+yearsofweatherdatasince1985.
Thisdatasetcontainstheharvestedareaof26cropsona0.5◦× 0.5◦
globalgrid,andforeachcrop,itdistinguishesirrigatedfromrainfed harvestedcroparea.Toourknowledge,itisthemostdetailedand comprehensivegeospatialdatabaseoncropareadistribution cur-rentlyavailable.Ofparticularnoteisthatthegeospatialdistribution ofcropareawasbasedonnationallyreporteddatacorroboratedby satelliteimagery.
2.2. Selectionofreferenceweatherstationsandqualitycontrol measures
Weatherdatabasesofsufficientgeospatialcoverageand qual-ity are essential in simulating crop yields at larger scales.We usedobservedweatherdatafromselectedstations,referredtoas referenceweatherstations(RWS),foundintheNationalOceanic andAtmosphericAssociation(NOAA),GlobalSummaryoftheDay (GSOD).Thisglobaldatasetincludesdailyvaluesforsurface max-imum and minimum temperature (Tmax, Tmin), precipitation, windspeed, and dewpoint temperature(NationalOceanic and AtmosphericAssociation,2010).Datafromstationsinthisdatabase undergo a number of quality control measures as described at: http://www.ncdc.noaa.gov/oa/climate/ghcn-daily/. Geospatial coverageisquitedenseinNorthAmerica,EuropeandEastAsia, andreasonablywelldistributedinpopulatedareasofsouthand southeastAsia,Africa,andLatinAmerica(Fig.1).Weatherstations aresparse,however,inareaswithlowpopulationdensityorlack ofinfrastructure.
WeatherstationswereonlyconsideredasapotentialRWSif they:(1)hadatleast20yearsofdatasince1980,(2)werelocatedin aprovinceorstatethatcontained>2%oftotalnationalproduction forthecropinquestion,and(3)hadfewerthan10%ofdata-daysand
fewerthan30consecutivedata-daysmissing.UsingthePortmann etal.(2010)geographicdistributionofcroparea,ArcGISwasused toiterativelysumtheharvestedareawithinabufferzoneoffixed radiusaroundeachstation,andstationswithinacountrywerethen ranked.Thestationwithgreatestharvestedareawithinthisbuffer zonewasselectedasareferenceweatherstation(RWS).Allother stationsneartheselectedRWS(withintwicethedistanceofthe bufferzoneradius)weretheneliminatedfromfurther consider-ationtoavoidoverlappingbufferzones,andthestationwiththe largestharvestedcropareaamongremainingstationswasthen selectedasasecondRWS.Theprocesswasrepeateduntil>50%of nationallyharvestedareafellwithin100kmofselectedRWS.In somecasescropproductionareaishighlyconcentratedsuchthat >50%ofnationallyharvestedareacannotbeachievedwithoutsome overlapamongRWS.Inthis case,5kmincrementaloverlapwas alloweduntil>50%ofnationallyharvestedareacouldbeachieved with25stationsorless.Thisselectionprocessavoidedsubjective selectionofRWS,limitedthenumberofstationsrequiredfor esti-mation,minimizedoverlapofbufferzonesusedforweighting,and providedgoodgeospatialcoverageofregionsthatcontributedmost tototal nationalproductionofthecropinquestion.Minimizing thenumberofRWSbecomesimportantbecauseinformationabout cropmanagementpractices withinbufferzonesisalsorequired foraccurateYpand Ywsimulation,anditrequiresconsiderable efforttoobtainthesedata(seeSection2.4below).Abufferzoneof 100kmwasusedforthisstudy.Zonesof50kmand150kmwere alsoconsidered,butdidnotallowfor50%coverageofharvested areabecauseof toomuchoverlap (150km)ornot enougharea covered(50km)(datanotshown).Preliminaryevaluationsof sta-bilityofYpandYwestimatesbasedonnumberofconsecutiveyears ofweatherdataandamountofcropareacoveredbyRWSbuffers
Table1
Influenceofdateoftransplantingordirectseeding(D.seed)plantingdatesandmaturitydate,orboth(dayofyear=DOY)onsimulatedYp(Mgha−1)usingNOAA-SR2for
selectprovincesandricecroppingsystems.
Province Season BaseYp(Mgha−1) TransplantDOY MaturityDOY TransplantandmaturityDOY
+7 −7 +7 −7 −7,+7 +7,−7 Sichuan Single 9.2 −4% 3% 7% −10% 10% −13% Anhui Early 3.3 −11% 11% 15% −15% 18% −23% Anhui Middle 8.6 −5% 3% 13% −14% 16% −18% Anhui D.Seed 6.1 −15% 12% −6% −8% 22% −24% Anhui Late 4.4 −10% 0% 4% −10% 13% −18%
indicatedthat20yearsand40–50%areacoveragewassufficient
toobtainstableestimates.Thiswasconfirmedbymorethorough
analysisasprovidedinthispaper.
UsingthisRWSselectionproceduremadeitpossibletoobtain
>50% coverage of total national harvested area for irrigated
rice in China (∼29Mha total harvested area, five-year average,
2004–2008)andirrigatedmaizeintheUSA(∼3Mhatotalharvested
area,2004–2008)withoutneedforoverlapamongRWS(Table2).
In contrast, rainfed maize production in the USA (∼28Mha, 2004–2008)ishighlyconcentratedinthecentralandeasternCorn Belt,asareweatherstationdata,suchthatupto25kmoverlapwas requiredbetweensomeoftheRWStocover14Mha(50%)of har-vestedareabasedonthePortmannetal.(2010)geospatialdata. Becauseofrelativelysmalllandareaandrelativelylargenumber ofweatherstationswithgoodqualityweatherdatainGermany, only6RWSwererequiredtocover>50%oftotalharvestedareaof rainfedwheat(∼3Mha).
WeatherdataforeachselectedRWSweresubjectedtoquality control(QC)measurestofillinmissingdataandidentifyandcorrect erroneousvaluesthatoccurduetotechnicalproblemscommonin weatherdataacquisition.Aspatialregressiontest(SRT)(Hubbard etal.,2005)wasusedtocheckandcorrectweatherdataatagiven RWSagainstdatafromnearbystationsbasedonthestrengthof cor-relationbetweennearbyandreferencestationdata.Developedfor useintheMidwestUSA,thisQCmethodwasfoundtooutperform otherQCapproachesinawidevarietyofclimate-zones(Hubbard etal.,2007;Youetal.,2008).Atleast2neareststationswereused withtheSRTtoidentifyandcorrectmissingandsuspiciousvalues forTmin,Tmax,dewpointtemperature,windspeed,and precipita-tion.Typicallyabout0.5%ofobservationswerecorrected,roughly 2daysperyear.FollowingHubbardetal.(2005),adailyvaluewas flaggedassuspiciousifitwasgreaterthan3standarddeviations (5for precipitation)fromtheSRTvalue,which is a regression-estimatedvaluebasedon15daysbeforeandafterthedailyvalue inquestion.Inrarecaseswhereasingledailyrecordwasmissing fromtheRWSandnearbystations(<0.01%ofallvalues),the aver-ageoftheprecedingandsucceedingdaywassubstitutedforthe missingvalue.
Evapotranspiration,relativehumidity,incidentsolarradiation, and vaporpressureare notmeasuredor reportedin theNOAA weatherdata,buttheyarerequiredbyoneormoreofthecrop simulationmodelsusedinthisstudy.Hence,evapotranspiration andrelativehumiditywereestimatedfollowingAllenetal.(1998). VaporpressurewasestimatedusingtheWobusmethod(Gerald
andWheatley,1984).Incidentsolarradiationwasobtainedinone oftwoways:(1)derived fromthesquarerootofthedifference betweendailyminimumandmaximum temperaturemultiplied byextraterrestrialsolarradiationandaconstant(Hargreavesand Samani,1982)or(2) obtainedfromNASAagroclimatologysolar radiationdata,whichareavailableona1◦by1◦globalgrid.These datawereobtainedfromtheNASALangleyResearchCenterPOWER ProjectfundedthroughtheNASAEarthScienceDirectorateApplied ScienceProgram.WethereforehadtwosourcesofNOAAweather dataforcropsimulation,bothusingactualdatafortemperatureand rainfallbutwithdifferentsourcesofdataforsolarradiation:either derived (Hargreaves andSamani, 1982),hereaftercalled NOAA-SR1,orbasedonsolarradiationfromtheNASA-POWERdatabase, hereaftercalledNOAA-SR2.
2.3. Soilproperties
Soiltextureandbulkdensityhavealargeinfluenceonwater holdingcapacity andarerequired forsimulation ofYw (Saxton etal.,1986).ForeachRWS,thedominantagriculturalsoilwithinthe 100kmbufferzonewasidentified.ForU.S.maizeproduction,soil texturewasidentifiedforthemostabundantsoiltypeassociated withthedensestmaizeproductionareawithineachRWSselected asaRWSforirrigatedorrainfedproduction.Thiswasachievedby evaluatingsoiltypesandareadistributionintheSSURGOdatabase (SoilSurveyStaff,2010)inrelationtothegeospatialdistribution of2009maizearea(NASS,2010a)withineach100kmRWSbuffer zone.Whenthereweretwoormoresoilsofsimilarextentand congruencewithmaizearea,thesoilwithhighestlandcapability class(KlingebielandMontgomery,1961)wasselectedasthemost representativesoilformaizeproductionintheRWSbufferzone.In GermanysoilswithineachRWSwereidentifiedusingadigitalsoil map(1:1.000.000)usingsoilprofiledescriptionsfordominantsoil typesfromHartwichetal.(1995).Forirrigatedrice,soilwater hold-ingcapacityisnotasensitivevariableforsimulationofYpbecause itisassumedthatfarmerscanapplyirrigationwheneverrainfall fallsbelowcropwaterrequirements.Therefore,simulationsofrice YpinChinadidnotrequirespecificationofsoilproperties. 2.4. Cropmanagement
Forthecropsandcountriesexaminedhere,farmershaveready accesstolatesttechnologiesand informationregardingplanting dates,seedingratesortransplantingpatterns,andcultivars.Indeed,
Table2
NationalestimatedYpandreported5-yearaverage(2004–2008)yields(takenfromIRRIforChina,FAOforGermanyandNASSfortheUS).
Country-crop Years Water
regime
Harvestedareain100km bufferzones(Mha)
Harvestedareaa
(Mha)
Coverage Ya(Mgha−1) Yp(Mgha−1) Ya/Yp(%)
China-rice 86–08 IR 15.0 29.12 51% 6.4 7.8 82%
US-maize 86–08 RF 14.0 27.7 50% 9.7 13.2 73%
US-maize 86–08 IR 1.9 3.5 54% 11.7 15.1 77%
German-wheat 86–08 RF 1.6 3.1 52% 7.6 9.5 80%
therearefewbarrierstoaltermanagementofthesepracticesif suchchangeswouldresultinhigheryieldsandprofit.Forthis rea-son,managementspecificationsforallsimulationswerebasedon currentaveragefarmerpracticesineachlocationwhereYporYw wassimulated.
Managementpracticesandtheextentofharvestedareaforrice systemsinChinawereobtainedfromagronomistsineachofthe majorrice-producingprovincesacrossChina.Dominantrice crop-pingsystemsrangedfromthree,two, oronericecropperyear onthesamepieceoflanddependingonwhethertheclimatewas warmenoughforyear-roundcropproduction.Morethan40 differ-entrice-basedcroppingsystemswereidentifiedin17provinces. Foreachricecrop,ineachofthedifferentcroppingsystems,crop managementdataincludedplantpopulationfordirect-seededrice orhillspacingintransplantedrice,dateofsowingortransplanting andtransplantseedlingage,dateoffloweringandmaturity,andthe mostwidelyusedcultivar.Emergencewasassumedtooccur7days afterdirectseeding.Cropphenology(seeding,flowering,and matu-ritydates)andtransplantingdatesreportedforeachricecropping systemwithinaprovincewereusedtoestimategenotype-specific coefficientsrequiredforsimulationofYpwithinRWSbufferzones locatedinthatprovinceusingcompanionsoftwareoftherice sim-ulationmodel(ORYZA2000).
Data for average U.S. maize sowing date by county were obtainedfromtheUSDA’sRiskManagementAgency(RMA),which requiresfarmersenrolledinUSDAinsuranceprogramstoreport theirplantingdatesbyfield.ForeachcountyinwhichaRWSwas located,theplantingdatewasconsideredtobethedateonwhich 50%ofthemaizeareawasplanted(meanvaluefor2003–2008). Seedingrateand growingdegreedaysrequired toreach matu-rityforthemostcommonhybridsusedwereobtainedfromfield researchers,seedcompanyagronomists,andfarmersfamiliarwith cropmanagementpracticesinbufferzoneareasaroundeachRWS. Iflong-termaverageyieldssimulatedbythemaizecropmodel(see Section2.5below)usingthereportedhybridmaturity(quantified bycumulativerelativematuritydays,calledCRM)hada>20%risk offrostoccurringbeforeendofgrainfilling,CRMwasadjustedafew daysearlieruntilriskoffrostwas<20%.Hybridmaturity,quantified incumulativerelativematuritydays(CRM),wasadjusteddown.
PhenologicaldataforwheatinGermany(sowing,emergence, spike emergence, physiological maturity) were obtained from observationsoftheGermanWeatherService(DWD,www.dwd.de). Data of regional wheatarea, yields, and the mostwidely used cultivars indifferent regions wereobtainedfrom theliterature (SelingandLindhauer,2005;Selingetal.,2009), while informa-tiononmostwidelyusedseedingratewereobtainedfromwheat breedersandagronomists.Genotype-specificparametersforthe cultivarsusedwereobtainedfromtheGENCALCprogram,which iterativelychangesgenotypiccoefficientsuntilsimulationresults matchreporteddatesofphenologicalstages(Huntetal.,1993). 2.5. Cropsimulation
Crop Yp and Yw were simulated from 1990–2008 using ORYZA2000for ricein China(Boumanet al.,2001),1990–2008 usingHybridMaizeformaizeintheUS(Yangetal.,2004),andfrom 1983–1992using CERES-Wheat for wheat in Germany (Ritchie etal.,1985).Eachofthesemodelsrequiresdailyvaluesof maxi-mumandminimumtemperatureandsolarradiationtosimulate Yp,andalsorainfallforsimulationofYw.Grainyieldoutputsfrom themodelsarereportedatstandardmoisturecontentsof14,15.5, and13.5kgH2Okg−1grainforrice,maizeandwheat,respectively.
Eachofthesemodelshavebeenwelldocumentedandvalidatedin fieldexperimentswithoptimalmanagementtoallowexpressionof YporYwacrossawiderangeofenvironments(Bolingetal.,2010; BoumanandvanLaar,2006;Fengetal.,2007;Ghaffarietal.,2001;
Grassinietal.,2009;Liuetal.,2008;Singhetal.,2008;Timsina and Humphreys, 2006; Yang et al., 2006, 2004). ORYZA2000 requiresfourgenotype-specificcoefficientstodetermine pheno-logical development and final maturity, and it simulates daily canopyCO2 assimilationand total respiration.Dailynetcarbon
assimilationisestimatedbydifferenceandassimilateisallocated toroots,stems,leavesandgraindependingonstageof develop-ment(Boumanetal.,2001).HybridMaize issimilarinstructure toORYZA2000,butonlyrequiresasinglegenotype-specificinput parameter:growingdegreedaysuntilthecropreaches physiolog-icalmaturity(Yangetal.,2004).Mostofthemajorseedcompanies provide information on growing degree days to physiological maturity for theircommercial hybrids. Unlike ORYZA2000 and HybridMaize,whichsimulategrossphotosynthesisandrespiration separately,CERES-Wheatusestemperature-adjustedradiationuse efficiencytoconvertphotosyntheticallyactiveintercepted radia-tionintodrymatter(Jonesetal.,2003;Ritchieetal.,1985,1988). CERES-Wheatrequires6genotype-specificcoefficientstosimulate phenologicaldevelopmentinresponsetotemperature, photope-riod,andvernalizationrequirements.
2.6. EstimatingYporYwatregionalandnationalscales Regionalornationalestimatedyieldpotential(YP
R)isa
produc-tionweightedaveragedefinedas: YP R =
PP i HiforalliintheregionandforPP
i =YiP×Hi (1)
wherePP
i isthepotentialproduction,Hiistheharvestedareawithin
100km,andYP
i istheestimatedYporYwofRWSi.AnestimateofYp
orYwwithinaRWS100kmbufferzoneisderivedfromsimulations basedontheweatherdatafromtheRWSandcropmanagement practicesintheregionasdescribedabove.Forregionswheremore thanonecropisgrowneachyearonthesamepieceofland,suchas themultiplericecroppingsystemsinChina,YP
i isdefinedastotal
potentialproductionofeachcroppingsystem(early-season, late-season,etc.)dividedbythetotalareaplantedtoallricecropping systemssimulatedatthatRWS.Thesenational,long-term aver-ageYpandYwestimateswerecomparedagainstreported5-year averagenationalyields(Ya).TheseYadataarerepresentativeat thenationalscale,notataregionalscale.Whilethismethodof up-scalingworkswellforcountriesinwhichcropsaregrowninlarge, mostlyhomogenoustopographies,asisthecaseforthecropsand countriesexaminedinthispaper,modificationofup-scalingwill berequired,suchasuseofsmallerbufferzonesandagro-climatic zones,incountrieswherecropsaregrowninregionswithgreater heterogeneityintopography.
2.7. EvaluationofrequirementsforrobustYpandYwestimation SimulationsofYw forrainfedU.S.maize andGermanwheat were evaluated using three sources of weather data: NOAA-SR1andNOAA-SR2aspreviouslydescribed,andabenchmarkdata sourcethatprovidesdailymeasurementofallparametersrequired for cropsimulation. For maize, the benchmarkdatabases were obtainedfromtheHigh PlainsRegional ClimateCenter(HPRCC, 2011),whichisanetworkofweatherstationsinthewesternCorn Belt. For wheat,benchmark data were obtainedfrom the Ger-man WeatherService (DWD,2009).For irrigated rice in China, thebenchmarkweatherdatacamefromtheChinaMeteorological Association(CMA,2009).Ineachcountry,foursiteswereselected atwhichtherewerebothabenchmarkandNOAAweatherstation withat least10 years of weather data (1990–2008 for US and China,1983–1992forGermany).SitesincludedCedarRapids,IA, Lincoln,NE, McCook,NE and GrandIsland, NE in theUSA,Bad
Hersfeld,Braunschweig,Düsseldorf,andGeisenheiminGermany, andChengdu,Chongqing,Nanning,andGushiinChina.
Toexplorehow manyyears ofweather dataarerequiredto obtainarobustestimateoflong-termaverageYporYw,we sim-ulatedYwat23sitesacrosstheU.S.CornBelt from1986–2009 andthencalculatedtheaverageYwforeachandevery consecu-tiveintervalof2,3,4...and23yearswithinthesedatasets.This gaveus21observationsof2-yearaverages,20observationsof 3-yearaverages,etc.ateachsite.Themeanandstandarddeviationof allobservationsforeachintervalwerethencomputedandusedto calculatethecoefficientofvariation(CV).
ToevaluatesensitivityofYpandYwtospecificationof manage-mentpractices,simulationswereperformedusingreportedand modifiedmanagementpracticesandcropphenologyforselected sites.For selected simulationsof riceYp in China(inHoushan, AnhuiandChengdu,Sichuan),transplantingdate(orseedingdate fordirectlyseededsystems),maturitydate(representingan ear-lierorlatermaturingcultivar),andacombinationofthetwowere adjustedby ±7dayscompared toreportedvalues. Forselected rainfedUSmaizesimulations(GrandIsland,NE,CedarRapids,IA andGrissom,IN),plantingdatewasadjustedby±7and±14days, cumulativerelativematuritydayswereadjustedby±4days, seed-ingratewasadjustedby±12,000seedsha−1andacombinationof theseadjustmentswerecomparedwithsimulationsmadeusing thereportedmanagement.Thesesiteswereselectedforthis sensi-tivityanalysisbecausetheyrepresentawiderangeofricecropping systemsforirrigatedriceinChina,andforlargedifferencesin rain-fallacrosstheU.S.CornBelt.
TheinfluenceofharvestedareacoveredbyRWSbufferzones contributing to national estimates of Yp or Yw was examined throughcalculationsofYwforUS,YpforChina,andYwforGermany byincrementallyaddingastationtothenationalestimate, follow-ingthepreviouslydescribedprotocolforselectingstations,until 50%ormoreofnationallyreportedharvestedareawascoveredby 100kmbuffers.Atissuewashowmuchareacoveragewasneeded toachieveastableestimateofYporYw.
3. Results
3.1. Evaluationofweatherdatarequirements
ForYporYwsimulatedusingbenchmarkweatherdata,NOAA temperature and rainfall coupled with derived solar radiation (NOAA-SR1)largelyoverestimatedYpandYwcomparedtoNOAA datawithNASAobservedsolarradiation(NOAA-SR2)(Fig.2).The overestimationwasgreatestforirrigatedriceinChina(+41%), mod-erateforrainfedwheatinGermany(+12%),andrelativelysmallfor U.S.rainfedmaize(+4%).WithinChina,overestimationwasgreatest inSichuan,amountainousprovincewherericeisgrowninirrigated valleysasopposedtootherprovinceswheretopographyismostly flat.TopographyinmaizeandwheatproducingareasoftheUSA andGermanyisalsomostlyflat.IntheUS CornBelt,estimated solarradiationcloselyapproximatedobservedsolarradiation, per-hapsbecausethisareaissimilartotheoneinwhichtheestimation procedurewascalibrated(Hargreaves,1994).
AnnualestimatesofYporYwwereaveragedfrom1986–2009 (from 2 to 23-year averages) and the CV of these long-term, consecutive-yearaverageestimatescompared.Long-termaverage estimateswereconsideredrobustiftheyachievedaCVlessthan 0.05.ForrainfedmaizeintheU.S.CornBelt,thenumberof consec-utiveyearsofweatherdatarequiredtoobtainaYwestimatewitha CVof0.05wasassociatedwithmeanannualrainfallacrossa tran-sectofRWSintheU.S.CornBelt(Fig.3).Atlocationswhereannual rainfallwas>900mm(>−90◦longitude),only2–8years
consecu-tiveweatherdatawereneededwhereas11–15yearswererequired
Fig.2. ComparisonofYpsimulatedwithtwosourcesofNOAAweatherdata,either usingderivedsolarradiation(NOAASR1)orobservedsolarradiationfromthe NASA-POWERdatabase(NOAA-SR2)versusabenchmarkfromweatherstationsatwhich allrequiredweatherdataweredirectlymeasured.Valuesshownrepresent19-year meansforeachoffoursitesforriceandmaizeand10-yearmeansforeachoffour sitesinGermany.
atsiteswithannualrainfall<700mm(<−96◦longitude).Ina
sim-ilarexercisefor10RWSand32systemsofirrigatedriceinChina, aCVof0.05wasachievedwithin12consecutiveyearsforallsites analyzed.
3.2. SensitivityofYptochangesincropmanagement
Relativelysmallchangesincropmanagementspecificationsas inputtoyieldsimulationshadrelativelylargeeffectsonYpofrice (Table1).The impactof managementwasgreatestin crop sys-temspracticedinAnhui wherefarmersgrowatleasttwocrops annuallyonthesamepieceofland.Forexample,delayoradvance oftransplantingdatebysevendaysresultedinYpestimatesthat were−15to+12%greaterthanreportedtransplantingdates. Sim-ilarly,increasingordecreasingcropmaturitybysevendaysledto arangeof−15to+15%insimulatedYp.Andiffarmerscombined bothdelayedoradvancedtransplantingwithricecultivarswitha longerorshortermaturity,therangeinsimulatedYpvariedbyas
Fig.3.YearsofconsecutiveweatherdatarequiredtoobtainsimulatedYwformaize intheU.S.CornBeltwithaCVof0.05.TheeasternmostlocationisYoungstown,OH whilethewesternmostisGrandIsland,NE.Averageannualrainfallfortwolocations atextremesoftheregressionlineareprovidedasreferencepointsfortheinsert, whichshowstheeast–westpatternoftheannualrainfallgradient.
muchas46%(directseeded;D.Seed)comparedtocurrentpractices. InSichuanwithonlyonericecropperyear,Ypwassomewhatless sensitivetoeffectsofchangesinmanagementwitharangeof25% inYpduetocombinedeffectsoftransplantingdateandcultivar maturity.LikeSichuan,onlyonemaizecropisplantedeachyearin theU.S.CornBelt.Changingrelativematurityby±4days(roughly equivalentto52growingdegreedays,Celsius),resultedinmore than1tha−1differenceinYw.Andalthoughtheimpactof increas-ingrelativematurityby+4daystoachievealongergrowingseason hadapositiveeffectonsimulatedYw,itincreasedtheriskoffrost occurrencebeforegrainfillingends,whichincreasedvariabilityin Yw(datanotshown).
3.3. Geospatialcoveragenecessaryforaccuratecharacterization ofnationalYpandYw
Nationalestimates of Yw for both U.S. maize and wheat in Germanywererobustwithonlyasmallportionofnational pro-ductionareacoverage.Forexample,averagesimulatedYwofU.S. rainfedmaizerangedfrom13.4to14.3tha−1astheproportionof totalrainfedmaize areacoveredbyRWSbufferzonesincreased fromabout10%to40%.Incontrast,estimatednationalYpof irri-gatedriceinChinavariedfrom12tha−1toabout8.0tha−1asthe proportionofcoveredareaincreasedfrom5%to40%oftotalrice area(Fig.4).Thisisduetothreefactors:(i)selectionprotocolfor choosingRWS,(ii)spatialdistributionofriceyieldsinChina,and (iii)widediversityofricecroppingsystemsfoundinChina,where 2oreven3ricecropsareplantedeachseason,makingwindows foroptimalsowingandharvestingrelativelynarrow.Accordingto theprotocol,thefirstRWSselectedarethoseinareaswithgreatest riceproductionareadensitywithinthe100kmbufferzonesaround candidateweatherstations.InChina,theseareasalsohavethe high-estyieldpotential,meaningthataslower-yieldingproductionareas areaddedtotheweightednationalaverage,theestimatednational Ypdeclines.Itthereforerequiredatleast40%coverageofharvested cropareatoattainconsistentnationalyieldestimatesforirrigated riceinChinaduetothelargediversityofcroppingsystems.In con-trast,stableestimatesofYporYwwereobtainedwithrelatively lowcoverageofharvestedcropareafor USmaizeandwheatin Germanywheretopography, climate,and croppingsystems are relativelyhomogeneous.Inallthreecountries,coverageof40%of harvestedareawithin100kmofRWSprovidedrobustestimatesof YporYwandestimateswerenotimprovedbyaddingadditional RWStoincreasecoverageofharvestedarea.
3.4. NationalYpandYwestimates
NationalYpestimatesforirrigatedriceinChinaandmaizeinthe USA,andrainfedYwforU.SmaizeandrainfedwheatinGermany aregiveninTable2.EachestimateisbasedonselectedRWSthat includeatleast50%oftotalproductionareaforeachcrop,which required22RWSandnearly100simulationsforYpofriceinChina (includingthedifferentrice croppingsystems),24 RWSand an equivalentnumberofsimulationsforYwofUSrainfedmaize,4 RWSandsimulationsforYpofU.S.irrigatedmaize,and6RWSand simulationsforYwofrainfedwheatinGermany.InChina, multi-plecroppingsystemsandthelocationofsomeRWSbufferzones acrossneighboringprovincesrequiredalargenumberof simula-tionsbasedonthericecroppingsystemsineachprovince.
AveragefarmyieldsofriceinChinaandwheatinGermanywere 82and80%ofestimatednationalYpandYw,respectively.In con-trast,current averageYw ofrainfedU.S maizewasonly73% of estimatedYw,whichindicatesalargerexploitableyieldgapthan forriceinChinaorwheatinGermany.Likewise,rainfedUSmaize hasalargeryieldgapthanirrigatedmaize.
Fig.4. VariationinestimatednationalYporYwmaizeintheUS,riceinChinaand wheatinGermanyasinfluencedbythenumberofreferenceweatherstations(solid blackcircles)andassociatedproportionofharvestedtotalcropareausedtosimulate YporYw(openblackcircles).RangeofsimulatedYporYwatallreferenceweather stations(RWS)areshownbytheopenredcircles.ThenumberofRWSatwhich50% coverageofnationalharvestedareaoccursisshownbytheblackdottedline.
4. Discussion
Quantitative analyses of food security and environmental changeareoftenperformedtohelpguideformulationofregional andnationalpolicies.Accuracy,reproducibility,andtransparency areessentialattributesofsuchstudiesinordertoallowchallenge byothersandtoinstillconfidenceintheresults.Theseattributes aredifficulttoachieveinestimationsofYpandYwdueto limita-tionsonqualityofweatherdataatappropriategeospatialdensities thatarecongruentwithdistributionofharvestedcroparea,scarcity ofgeospatiallyexplicitinformationoncropmanagementpractices andsoils,andlackofappropriatecropmodels.Resultsfromthe currentstudy,however,indicateitispossibletoattainrobustand transparentestimatesofnationalYporYwifcarefulattentionis given tothemore sensitivecomponentsofYp and Yw estima-tion.Themethodologyexploredwithinthispaperismostsuitable forcountriesinwhichcropsaremostlygrowninprairies,plains, largevalleys,deltas,andlowlandswheretopographyisrelatively homogenous.Itisnoteworthythatalargemajorityofglobalfood cropproductionofthemajorcerealcropsoccursincountrieswhere
thesecropsaregrownonprairies,plains,largevalleys,deltas,and lowlands.Moreover,theapproachusedherewouldbeapplicable tocountriesinwhichcropsaregrowninmorecomplex topogra-phiesalthoughprotocolsmayrequiremodification.Forexample, bufferzonesof50km,100kmand150kmwereconsideredinthe currentstudy(datanotshown).While50kmzoneswereunableto capturemorethan40%ofharvestedareawithoutalargenumber ofRWS,especiallyintheUSandChina,150kmzonescausedtoo muchoverlapamongcandidateweatherstations.Forthesereasons the100kmbufferzoneswerechosenforthisanalysis.Incountries wherecropsaregrownin moreheterogeneouslandscapesthan thoseexaminedinthispaper,smallerbufferzonesmayprovemore appropriateanditmaybenecessarytoup-scaleestimatesofYpor YwfromRWStonationallevelsusingextrapolationbasedon agro-climaticzones.Inadditiontosmallerbufferzones,beneficialfuture workwouldalsoexaminevariabilityinsoilwaterholding capac-ityanditseffectonsimulationandupscalingresults.Forthetwo countrieswhereYwwassimulated,terrainisrelativelyflatandsoil typesonwhichthemajorityofthecropisgrownarerelatively uni-formandsoonlyasingle,dominantsoiltypewasusedasinputinto modelsimulations.
EstimatesofYpandYwaresignificantlyaffectedbyseveralkey factors.Theseincludeweatherdataqualityandestimation proce-dures,specificationofplantingdateandcultivarorhybridmaturity, thenumberofyearssimulatedtoestimatelong-termaverageYp orYw,andgeospatialweightingprocedurestoarriveataggregated estimatesatprovincialornationalscales.Eachofthesefactorsaffect estimatesofYporYwtoagreaterorlesserdegreedependingon thespecificcroppingsystems,location,anddistributionofavailable dataforthecropandcountryunderinvestigation.Forexamplein U.S.rainfedmaize,variabilityinestimatesofYwdependedon rain-fallwithhighrainfallareas(>850mmyear−1)requiringonly5–7 yearsofweatherdatatoachieveaCV<0.05whilemorethan10 yearswererequiredtoachieveasimilarlowCVforestimatesof Ywinlow rainfallareas(<700mmyear−1).Thusitseemslikely thatduration ofweather datarequiredforrobustsimulation of long-termYwincreasesasrainfalldecreasesorrainfallvariability increases.ForirrigatedriceinChina,however,nocleartrendwith rainfallwasobservedbecauseirrigationeliminatesyield-reducing impactofdryyears.
Thesourceofsolarradiationdatawasalsoasensitivefactor insimulationofYporYw. Forirrigatedricein China,estimates ofsolarradiation performedpoorlyin simulatingYpin regions wheretopographywasmountainous.Althoughtheweather sta-tionsthemselvesarelocatedonflatterrain,temperatureinthese areasmaybemoreaffectedbymountainousclimate(thinnerair, rainshadowaffects,andtrappedair)ratherthanbycloudcoverand solarradiation,whichmakesderivationsofsolarradiationbased ondiurnaltemperaturerangeapoorproxyforincidentradiation (ThorntonandRunning,1999;Winslowetal.,2001).The overes-timationinnon-mountainousregionsmayresultfromparticulate airpollution,whichreducesincidentsolarradiationandincreases nighttimetemperatureandissubsequentlyhighinthe industri-alizedcentralandeasternChinawhereamajorityofriceisgrown (Menonetal.,2002).Incontrast,estimationofsolarradiationbased onthedifferenceinTminandTmaxappearstobemuchmore accu-rateforsimulationofcropYpandYwintheMidwestUSAbecause thealgorithmforthederivationwasbasedonresearchconducted inthisregion(Hargreaves,1994),andbecausethereisrelatively littleairpollution.Butgiventhelackofaccuracyinmodeledsolar radiationinareaswithvariabletopographyorwithparticulateair pollution,astandardmethodforestimatingYporYwwould prefer-ablyrelyonobservedorsatellitederivedsolarradiation(Baietal., 2010;Whiteetal.,2011).
Resultsfromthisstudyemphasizetheimportanceofspecifying currentcropmanagementpracticestoobtainrelevantestimatesof
YporYwatregionalornationalscales.Whilesomemayarguethe valueofsimulatingmaximumpossibleYporYwwithoutregardto currentmanagement practicesandcroppingsystems,such esti-mates do not account for the biophysical and socio-economic constraintsunderwhichfarmersmustoperate.Indeed,historically and globally farmers areefficientin allocatingtheirland, labor andcapitaltooptimizeprofitandreducerisk(HerdtandMandac, 1981;Hopper,1965;Sheriff,2005).Forthethreecountriesinwhich yieldgapwasevaluatedinthecurrentstudy,cropyieldsare rela-tivelyhighandfarmershaveaccesstorecommendationsaboutbest managementpractices.Therearetypicallynobarrierstofarmer adoptionofearlierorlatersowing,useoflongerorshortermaturity cultivars,differentseedingratesortransplantingpatternsif farm-ersbelievedthatsuchchangesmadeasignificantdifferenceinyield andprofit.Andinmultiplecroppingsystemsfarmersarenot seek-ingtooptimizeproductionofasinglecropbutratherofanentire systemthatincludesseveralcrops,suchasthecaseofirrigatedrice systemsinmuchofChina.Therefore,farmersarelikelytousethe mostappropriatecombinationofsowingdateandcropmaturity fortheircroppingsystemconsideringrisksand costsassociated withotheroptions.Andwhilethere mightbesmallyieldgains fromchangingplantpopulationsfromcurrentpractices,thereis nopublishedevidencedefining“optimal”populationsacrossthe widerangeofenvironmentsevaluatedinthisstudy.
Theimpactofusingappropriatespecificationofcrop manage-ment canbeseenin both localand regionalornational scales. Forexample,achangeofonly7daysinplantingdateforirrigated riceinChinaaffectedsimulatedYpbyasmuchasa1Mgha−1or more.GiventheharvestedareaofirrigatedriceinAnhuiProvince, adifferenceof1Mgha−1 represents2.7millionmetrictonstotal production,whichisequivalentto550,000haofhighqualityfarm landatcurrentaverageyieldlevelsortheannualrice consump-tionofover25millionpeopleatcurrentriceconsumptionlevels inChina(ZhaiandYang,2006).Giventhis sensitivitytosowing date,estimatesofYporYwbasedonmeanmonthlyweatherdata, orweatherdataderivedfrommonthlymeans, canleadtolarge biastowardover-orunderestimationdependingonthedifference betweenactualaveragesowingdateandtheassumed(orinferred) sowingdateatthebeginning,mid-monthorendofthemonth,to accommodateuseofmonthlymeanweatherdata.
Temperatureandcumulativeinterceptedsolarradiationduring the growing season have a large influence on Yp. Total inter-ceptedradiationis sensitivetoboth intensityofsolarradiation (MJ m−2d−1)and lengthofgrowing season.Lengthof growing seasonisgovernedbytemperatureregimeandcropmanagement withregardtoplantingdateandmaturityofthemostwidelyused cropcultivars.Densityofplantsperunitlandareaalsoinfluences amountofinterceptedsolarradiation.Therefore,estimationofYp andYwrelevanttodominantcroppingsystemsinaregionrequires specificationofplantingdate,cropcultivarmaturity,andplant den-sitytypicallyusedbyfarmersinthatregion.
Becauseofthedetailedinformationrequiredtoaccurately sim-ulatecropyieldineachregion,itisexpedienttolimitthenumber ofRWSrequiredforrobustandreproducibleestimatesofYpor Ywatanationalscale.Atthesametime,ifadequateareaisnot represented,YporYwestimatesmaynotprovideaccurate repre-sentationatanationalscale.Thisisespeciallytrueinmorecomplex croppingsystemssuchasforriceinChinawheretheinitialRWS selectedinChinawasfromaprovincewhereonlyasingle, high-yieldingricecropwasgrown,whichisnottypicalofthemajorityof riceproductioninthatcountry.Subsequentstationswerelocatedin provincesthatproducemorethanonecropperyearandforwhich theYpofeachofthesemultiplecropsismuchlower.Thenational YpestimateforChina,aswellaseveryothercropandcountry ana-lyzed,variedlittleafter40–50%ofharvestedareawaswithinthe bufferzonesofselectedRWS.Whilesizeofbufferzonesmayneed
tovarydependingonsizeofcountryforwhichnationalYporYw estimatesaredesired,achieving50%coverageofharvestedarea islikelytoproviderobustestimatesofcropproductionpotential basedonresultsfromthecurrentstudy,whichincludedcontrasting cropswitha widerangeofwaterregimesandcroppingsystem complexity.
AveragefarmyieldsofriceinChinaandwheatinGermanywere 82and80%ofestimatednationalYpandYw,respectively.In con-trast,current averageYw ofrainfedU.S maizewasonly73% of estimatedYw,whichindicatesalargerexploitableyieldgapthan forriceinChinaorwheatinGermany.BothChinaandGermany haveseenverylittleornogrowthinyieldsofriceandwheat, respec-tively,forthepastdecade,especiallycomparedtopreviousdecades (FAO,2010),despitecontinuedprogressinagriculturaltechnology andatrendtowardhigherprices.Similarly,yieldofirrigatedU.S. maizehasnotincreasedinthepastdecadeandinthisstudy cur-rentaverageyieldsareestimatedtobe77%ofYp.Theseestimates ofyieldgapforriceinChina,wheatinGermany,andirrigatedmaize intheUSAareconsistentwiththehypothesisthataveragenational yieldsbegintoplateauwhenaveragefarmyieldsreach75to85% ofYporYw(Cassmanetal.,2003;Lobelletal.,2009).Thisyield levelhasbeenproposedasthepracticallimitfornationalaverage farmyieldsbecauseitisneitherlogisticallyfeasiblenorprofitable for100%offarmerstoachieveyieldsequivalenttomaximum bio-physicalpotentialyields.Therefore,analysisoffutureglobalfood securityshouldrestrictcropproductionpotentialtosomefraction ofYpandYwtoavoid overestimatingnationaland globalfood productionpotential.
5. Conclusions
Resultsfromthisstudysuggestimprovedprotocolstoobtain robust,transparent,andreproducibleestimatesofYpandYwat local,regional,andnationalscalesforcountriesinwhichcropsare producedinareaswithrelativelyhomogeneoustopographies.Such aprotocolwouldhavethefollowingcomponents:
(1)userealweatherdatawithdailytimestepandavoidmonthly means,ordataderivedfrommonthlymeans,andderivedsolar radiation;
(2)15yearsofweatherdataforrainfedagriculture,while5years maybesufficientforestimatesofYpforfullyirrigated produc-tionsystems;
(3)50% coverageof harvested areausing a proceduretoselect regionswithgreatestcropproductiondensity;
(4)specificationofcurrentaveragesowingdate,cultivarorhybrid maturity,andplantpopulationthatgivesmaximumyieldwith thissowing dateandcultivar/hybridmaturity.Withoutwell documentedevidenceofwhatthisoptimalplantpopulation is,estimatesofYpandYwshouldbebasedoncurrent aver-ageplantpopulationusedbyfarmersinthetargetlocationand croppingsystem;
(5)anappropriateweightingproceduretoestimateYporYwat regionalornationallevelsfrompointestimatesatselected ref-erenceweatherstations(suchasup-scalingbasedonharvested areawithinbufferzonesaroundsimulationsites);and (6)anappropriatecropmodelthathasbeenvalidatedagainstfield
datainwhichcropshavebeengrowntoproduceyieldsthat approachYpandYw.
Acknowledgments
Completingtheworkreportedinthispaperwouldnothavebeen possiblewithoutinputfrom:JansBobertattheLeibnizCenterfor AgriculturalLandscapeResearchforprovidinguswithappropriate
genotypecoefficientsforWheatinGermany;theCentral Agrom-eteorologicalResearchStationoftheGermanWeatherServicein BraunschweigforprovidingdataforweatherstationsinGermany; RebeccaDavisattheUSDARMA forprovidingUSmaize county levelplantingdatedata;JinshunBaiandChinaAgricultural Uni-versitygenerouslyprovideduswithobservedsolardatacollected atspecificlocationsbytheChinaMeteorologicalAdministration; PatricioGrassiniandMartinvanIttersumfortheirsuggestionsfor themanuscript;andtheChineseagronomistswhoprovidedcrop managementdataforirrigatedriceinChina.
AppendixA. Supplementarydata
Supplementarydataassociatedwiththisarticlecanbefound,in theonlineversion,athttp://dx.doi.org/10.1016/j.fcr.2012.11.018.
References
Allen,R.G., Pereira,L.S., Raes, D.,Smith, M.,1998. CropEvapotranspiration– GuidelinesforComputingCropWaterRequirements.FoodandAgriculture Orga-nizationoftheUnitedNations,Rome,Reportnr56.
Andarzian,B.,Bakhshandeh,A.M.,Bannayan,M.,Emam,Y.,Fathi,G.,Alami,S.,2008. WheatPot:asimplemodelforspringwheatYpusingmonthlyweatherdata. Biosyst.Eng.99,487–495.
Bai,J.,Chen,X.,Dobermann,A.,Yang,H.,Cassman,K.G.,Zhang,F.,2010.Evaluation ofNASAsatellite-andmodel-derivedweatherdataforsimulationofmaizeyield potentialinChina.Agron.J.102,9–16.
Baron,C.,Balme,B.S.,Sarr,M.,Traoré,B.,Lebel,S.B.,Janicot,T.,Dingkuhn,S.M., 2005.FromGCMgridcelltoagriculturalplot:scaleissuesaffectingmodelling ofclimateimpact.Philos.Trans.R.Soc.B360,2095–2108.
Boling,A.A.,Pascual,Z.P.,Eusebio,W.H.,Gregorio,G.B.,Bouman,B.A.M.,Tuong,T.P., Redona,E.D.,2010.Progressandchallengesinidentifyingadaptivemanagement strategiesinriceproductionfortheprevailingclimate.PhilippineJ.CropSci.35, 28.
Bouman,B.A.M.,vanLaar,H.H.,2006.Descriptionandevaluationofthericegrowth modelORYZA2000undernitrogen-limitedconditions.Agric.Syst.87,249–273. Bouman,B.A.M.,Kropfff,M.J.,Tuong,T.P.,Wopereis,M.C.S.,TenBerge,H.F.M.,van Laar,H.H.,2001.Oryza2000:ModelingLowlandRice.InternationalRiceResearch Institute,LosBa ˜nos,Philippines.
Brisson,N.,Gate,P.,Gouache,D.,Charmet,G.,Oury,F.X.,Huard,F.,2010.Whyare wheatyieldsstagnatinginEurope?AcomprehensivedataanalysisforFrance. FieldCropsRes.119,201–212.
Bruinsma,J.2009.Theresourceoutlookto2050:byhowmuchdoland,waterand cropyieldsneedtoincreaseby2050?FAOexpertmeetingonhowtofeedthe worldin2050;2009/06.
Cassman,K.G.,1999.Ecologicalintensificationofcerealproductionsystems:Yp,soil quality,andprecisionagriculture.In:ProceedingsoftheNationalAcademyof Sciences(USA),pp.5952–5959.
Cassman,K.G.,Grassini,P.,vanWart,J.,2010.Cropyieldpotential,yieldtrendsand globalfoodsecurityinachangingclimate.In:Rosenzweig,C.,Hillel,D.(Eds.), HandbookofClimateChangeandAgroecosystems.ImperialCollegePress, Lon-don,p.37.
Cassman,K.G.,Dobermann,A.D.,Walters,D.,Yang,H.,2003.Meetingcerealdemand whileprotectingnaturalresourcesandimprovingenvironmentalquality.Ann. Rev.Environ.Res.28,315–358.
CMA,2009.Tmin,TmaxandIncidentSolarRadiation(1990–2008).China Meteoro-logicalAssociation,Beijing,China.
Deryng,D.,Sacks,W.J.,Barford,C.C.,Ramankutty,N.,2011.Simulatingtheeffects ofclimateandagriculturalmanagementpracticesonglobalcropyield.Global Biog.Cyc.25,18.
Doorenbos,J.,Kassam,A.H.,1979.YieldResponsetoWater.FoodandAgriculture OrganizationoftheUnitedNations,Rome,Italy,Reportnr33.
Duvick,D.N.,Cassman,K.G.,1999.Post-green-revolutiontrendsinYpoftemperate maizeinthenorth-centralUnitedStates.CropSci.39,1622–1630.
DWD,2009.Tmax,Tmin,IncidentSolarRadiation,RelativeHumidity, Precipita-tionandWindspeed(1983–1992).DeutscherWetterdienst,OffenbachamMain, Germany.
Evans,L.T.,1993.CropEvolution,AdaptationandYield.CambridgeUniversityPress, Cambridge.
FAO.2010.FAOSTAT.http://faostat.fao.org/default.aspx
Feng,L.P.,Bouman,B.A.M.,Tuong,T.P.,Cabangon,R.J.,Li,Y.L.,Lu,G.A.,Feng,Y.H., 2007.ExploringoptionstogrowriceusinglesswaterinnorthernChinausing amodellingapproach.I.Fieldexperimentsandmodelevaluation.Agric.Water Manage.88,1–13.
Fischer,G.,vanVelthuizen,H.,Shah,M.,Nachtergaele,F.O.,2002.Global Agroecolog-icalAssessmentforAgricultureinthe21stCentury:MethodologyandResults. FAOandIIASA,Laxenburg,Austria,ReportnrRR-02-02.
Gerald,C.F.,Wheatley,P.O.,1984.AppliedNumericalAnalysis.Addison-WesleyPub. Co,Reading,Mass.
Ghaffari,A.,Cook,H.F.,Lee,H.C.,2001.Simulatingwinterwheatyieldsunder tem-perateconditions:exploringdifferentmanagementscenarios.Eur.J.Agron.15, 231–240.
Godfray,H.C.J.,Beddington,J.R.,Crute,I.R.,Haddad,L.,Lawrence,D.,Muir,J.F.,Pretty, J.,Robinson,S.,Thomas,S.M.,Toulmin,C.,2010.Foodsecurity:thechallengeof feeding9billionpeople.Science327,812-818.
Grassini,P.,Yang,H.S.,Cassman,K.G.,2009.LimitstomaizeproductivityinWestern Corn-Belt:asimulationanalysisforfullyirrigatedandrainfedconditions.Agric. For.Meteorol.149,1254–1265.
Graybosch,R.A.,Peterson,C.J.,2010.Geneticimprovementinwinterwheatyields intheGreatPlainsofNorthAmerica,1959–2008.CropSci.50,1882–1890. Hargreaves,G.H.1994.Simplifiedcoefficientsforestimatingmonthlysolarradiation
inNorthAmericaandEurope.Logan,Utah:UtahStateUniversity.Reportnr DepartmentalPaper,DepartmentofBiologyandIrrigationEngineering. Hargreaves,G.H.,Samani,Z.A.,1982.Estimatingpotentialevapotranspiration.J.Irrig.
Drain.Eng.108,223–230.
Hartwich, R.,Behrens, J., Eckelmann,W., Haase, G.,Richter, A., Roeschmann, G.,Schmidt,R.,1995.BodenübersichtskartederBundesrepublikDeutschland 1:1.000.000.In:Erläuterungsheft,,KartemitErläuterungen,Textlegendeund Leitprofilen.FederalInstituteforGeosciencesandNaturalResources,Hannover, Germany.
Herdt,R.W.,Mandac,A.M.,1981.Moderntechnologyandeconomic-efficiencyof Philippinericefarmers.Econ.Dev.CulturalChange29,375–399.
Hopper,W.D.,1965.AllocationefficiencyinatraditionalIndianagriculture.J.Farm Econ.47,611–624.
HPRCC. 2011 HPRCC classic online services weather data retrieval system. <http://hprcc1.unl.edu/cgi-hpcc/home.cgi>(accessed2010August).
Hubbard,K.G.,Guttman,n.B.,You,J.,Chen,Z.,2007.AnimprovedQCprocessfor temperatureinthedailycooperativeweatherobservations.J.Atmos.Ocean. Technol.24,206–213.
Hubbard,K.G.,Goddard,S.,Sorensen,W.D.,Wells,N.,Osugi,T.T.,2005.Performance ofqualityassuranceproceduresforanappliedclimateinformationsystem.J. Atmos.Ocean.Technol.22,105–112.
Hunt,L.A.,Pararajasingham,S.,Jones,J.W.,Hoogenboom,G.,Imamura,D.T.,Ogoshi, R.M.,1993.GENCALC-softwaretofacilitatetheuseofcropmodelsforanalyzing fieldexperiments.Agron.J.85,1090–1094.
IPCC,2007.ClimateChange2007.CambridgeUniversityPress,NewYork. Jones,J.W.,Hoogenboom,G.,Porter,C.H.,Boote,K.J.,Batchelor,W.D.,Hunt,L.A.,
Wilkens,P.W.,Singh,U.,Gijsman,A.J.,Ritchie,J.T.,2003.TheDSSATcropping systemmodel.Eur.J.Agron.18,235–265.
Karl, T.R.,Trenberth, K.E., 2003. Modern global climate change. Science 302, 1719–1723.
Klingebiel,A.A.,Montgomery,P.H.,1961.Land-capabilityclassification.In: Agricul-turalHandbook.USDA,SoilConservationService,Washington,D.C,210p. Kropff,M.J.,Cassman,K.G.,vanLaar,H.H.,Peng,S.,1993.Nitrogenandyieldpotential
ofirrigatedrice.PlantSoil155–156,391–394.
Licker, R., Johnston, M., Foley, J.A., Barford, C., Kucharik, C.J., Monfreda, C., Ramankutty,N.,2010.Mindthegap:howdoclimateandagricultural manage-mentexplainthe‘yieldgap’ofcroplandsaroundtheworld.GlobalEcol.Biogeogr. 19,769–782.
Liu,Y.,Li,S.Q.,Chen,X.P.,Bai,J.S.,2008.AdaptabilityofHybrid-Maizemodeland potentialproductivityestimationofspringmaizeondryhighlandloessplateau. Trans.Chin.Soc.Agric.Eng.24,302–308.
Lobell,D.B.,Field,C.B.,2007.Globalscaleclimate-cropyieldrelationshipsandthe impactsofrecentwarming.Environ.Res.Lett.2,1–7.
Lobell,D.B.,Cassman,K.G.,Field,C.B.,2009.Cropyieldgaps:theirimportance, mag-nitudesandcauses.Annu.Rev.Environ.Resour.34,4.
Menon,S.,Hansen,J.,Nazarenko,L.,Luo,Y.,2002.Climateeffectsofblackcarbon aerosolsinChinaandIndia.Science297,2250–2253.
National Agricultural Statistical Service (NASS). 2010a. US irrigated and rainfed maize yield, harvested area and production 2004–2008. NASS Marketing and Information Services Office, Washington D.C. <http://http://www.nass.usda.gov/>(accessed2010August).
NationalAgriculturalStatisticsService(NASS).2010b.2009 USCroplandData Layer. NASS Marketing and Information Services Office, WashingtonD.C. <http://www.nass.usda.gov/research/Cropland/SARS1a.htm> (accessed 2010 August).
NationalOceanicandAtmosphericAssociation.2010.Globalsurfacesummaryof thedaydataset.<ftp://ftp.ncdc.noaa.gov/pub/data/gsod/>.
Neumann,K.,Verberg,P.H.,Stehfest,E.,Muller,C.,2010.Theyieldgapofglobalgrain production:aspatialanalysis.Agric.Syst.103,316–326.
Nonhebel,S.,1994.Theeffectsofuseofaverageinsteadofdailyweatherdatain cropgrowthsimulation-models.Agric.Syst.44,377–396.
Peng,S.,Cassman,K.G.,Virmani,S.S.,Sheehy,J.,Khush,G.S.,1999.Yieldpotential trendsoftropicalricesincethereleaseofIR8andthechallengeofincreasing riceyieldpotential.CropSci.39,1552–1559.
PenningdeVries,W.T.,Rabbinge,R.,Groot,J.J.R.,1997.Potentialandattainablefood productionandfoodsecurityindifferentregions.Philos.Trans.R.Soc.B360, 2067–2083.
Phalan,B.,Balmford,A.,Green,R.E.,Scharlemann,J.P.W.,2011.Minimizingtheharm tobiodiversityofproducingmorefoodglobally.FoodPolicy36,S62–S71. Portmann,F.T.,Siebert,S.,Döll,P.,2010.MIRCA2000–globalmonthlyirrigatedand
rainfedcropareasaroundtheyear2000:anewhigh-resolutiondatasetfor agriculturalandhydrologicalmodeling.GlobalBiog.Cyc.24,GB1011. Priya,S.,Shibasaki,R.,2001.NationalspatialcropyieldsimulationusingGIS-based
cropproductionmodel.Ecol.Model.135,113–129.
Ritchie,J.T.,Godwin,D.C.,Otter-Nacke,S.,1988.CERES-Wheat.UniversityofTexas Press,Austin,TX.
Ritchie,J.T.,Godwin,D.C.,Otter-Nacke,S.,1985.CERES-Wheat:AUserOriented WheatYieldModel:PreliminaryDocumentation.MichiganStateUniversity,MI, ReportnrAGRISTARSPublicationNo.YM-U3-04442-JSC-18892.
Rosegrant,M.W.,Ringler,C.,Zhu,T.J.,2009.Waterforagriculture:maintainingfood securityundergrowingscarcity.Ann.Rev.Env.Resour.34,205–222. Saxton,K.E.,Rawls,W.J.,Romberger,J.S.,Papendick,R.I.,1986.Estimating
general-izedsoil–watercharacteristicsfromtexture.SoilSci.Soc.Am.J.50,1031–1036. Searchinger,T.,Heimlich,R.,Houghton,R.A.,Dong,F.,Elobeid,A.,Fabiosa,J.,Tokgoz, S.,Hayes,D.,Yu,T.H.,2008.UseofU.S.croplandsforbiofuelsincreases green-housegasesthroughemissionsfromland-usechange.Science319,1238–1240. Seling,S.,Lindhauer,M.G.,2005.DieQualitätderdeutschenWeizenernte2005. PartI:quantitativesundqualitativesErgebnisinBundundLändern.Mühle+ Mischfutter146,663–671.
Seling,S.,Schwake-Anduschus,C.,Lindhauer,M.G.,2009.DieQualitätderdeutschen Weizenernte2009.PartI:quantitativesundqualitativesErgebnisinBundund Ländern.Mühle+Mischfutter146,672–681.
Sheriff,G.,2005.Efficientwaste?Whyfarmersover-applynutrientsandthe impli-cationsforpolicydesign.Rev.Agric.Econ.27,542–557.
Singh,A.K.,Tripathy,R.,Chopra,U.K.,2008.EvaluationofCERES-WheatandCropSyst modelsforwater–nitrogeninteractionsinwheatcrop.Agric.WaterManage.95, 776–786.
SoilSurveyStaff.2010.Soilsurveygeographic(SSURGO)databaseforthecornbelt. NaturalResourcesConservationService.UnitedStatesDepartmentof Agricul-tureAvailablefrom:Availableonlineathttp://soildatamart.nrcs.usda.gov
Thornton,P.E.,Running,S.W.,1999.Animprovedalgorithmforestimating inci-dentdailysolarradiationfrommeasurementsoftemperature,humidity,and precipitation.Agric.For.Meteorol.93,211–228.
Timsina,J.,Humphreys,E.,2006.PerformanceofCERES-riceandCERES-wheat mod-elsinrice-wheatsystems:areview.Agric.Syst.90,5–31.
UNFPA.2010.Stateofworldpopulation2010-fromconflictandcrisistorenewal: generations of change. United Nations Population Fund Available from:
http://www.unfpa.org/swp/2010/web/en/pdf/ENSOWP10.pdf
vanBussel,L.G.J.,Muller,C.,vanKeulen,H.,Ewert,F.,Leffelaar,P.A.,2011.Theeffect oftemporalaggregationofweatherinputdataoncropgrowthmodels’results. Agric.For.Meteorol.151,607–619.
Wang,Y.P.,Connor,D.J.,1996.Simulationofoptimaldevelopmentforspringwheat attwolocationsinsouthernAustraliaunderpresentandchangedclimate con-ditions.Agric.For.Meteorol.79,9–28.
White,J.W.,Hoogenboom,G.,Wilkens,P.W.,StackhouseJr.,P.W.,Hoell,J.M.,2011. Evaluationofsatellite-based,modeled-deriveddailysolarradiationdataforthe continentalUnitedStates.Agron.J.103,1242–1251.
Winslow,J.C.,Hunt,J.,Piper,S.C.,2001.Agloballyapplicablemodelofdailysolar irradianceestimatedfromairtemperatureandprecipitationdata.Ecol.Model. 143,227–243.
Wit,A.J.W.,Boogaard,H.L.,Diepen,C.A.,2005.Spatialresolutionofprecipitationand radiation:theeffectonregionalcropyieldforecasts.Agric.For.Meteorol.135, 156–168.
Yang,H.,Dobermann,A.,Cassman,K.G.,Walters,D.T.,2006.Features,applications, andlimitationsofthehybrid-maizesimulationmodel.Agron.J.98,737–748. Yang,H.S.,Doberman,A.,Lindquist,J.L.,Walters,D.T.,Arkebauer,T.J.,Cassman,K.G.,
2004.Hybrid-maize-amaizesimulationmodelthatcombinestwocrop model-ingapproaches.FieldCropsRes.87,131–154.
You,J.,Hubbard,K.G.,Goddard,S.,2008.Comparisonofmethodsforspatially esti-matingstationtemperaturesinaqualitycontrolsystem.Int.J.Climatol.28, 777–787.
Zhai,F.Y.,Yang,X.G.,2006.ReportontheSituationofNutritionandHealthofChinese residents:II.TheStatusofDietandNutrientUptakein2002.People’sMedical PublishingHouse,Beijing.