PIK Report

No. 128No. 128

FOR

POTSDAM INSTITUTE

CLIMATE IMPACT RESEARCH (PIK)

THE IMPACT OF CLIMATE CHANGEON COSTS OF FOOD

AND PEOPLE EXPOSED TO HUNGERAT SUBNATIONAL SCALE

Anne Biewald, Hermann Lotze-Campen, Ilona Otto,Nils Brinckmann, Benjamin Bodirsky, Isabelle Weindl,

Alexander Popp, Hans Joachim Schellnhuber

Herausgeber:Potsdam-Institut für Klimafolgenforschung

Technische Ausführung:U. Werner

POTSDAM-INSTITUTFÜR KLIMAFOLGENFORSCHUNGTelegrafenbergPostfach 60 12 03, 14412 PotsdamGERMANYTel.: +49 (331) 288-2500Fax: +49 (331) 288-2600E-mail-Adresse:pik@pik-potsdam.de

Authors:Dr. Anne Biewald (corresponding author)Prof. Dr. Hermann Lotze-CampenDr. Ilona OttoDr. Benjamin BodirskyIsabelle WeindlDr. Alexander PoppProf. Dr. Hans Joachim SchellnhuberPotsdam Institute for Climate Impact ResearchP.O. Box 60 12 03, D-14412 Potsdam, GermanyE-Mail: biewald@pik-potsdam.de

Nils BrinckmannUniversity of Potsdam (former member of Potsdam Institute for Climate Impact Research)

POTSDAM, OKTOBER 2015ISSN 1436-0179

This study has been funded by the World Bank under the Selection number 1143413.

3

 

AbstractClimate change and socioeconomic developmentswill have a decisive impact on people exposed tohunger.Thisstudyanalysesclimatechangeimpactsonagricultureandpotential implicationsfortheoccurrenceofhungerunderdifferentsocioeconomicscenariosfor2030,focusingontheworldregionsmostaffectedbypovertytoday:theMiddleEastandNorthAfrica,SouthAsia,andSub‐SaharanAfrica.Weuseaspatiallyexplicit,agroeconomicland‐usemodeltoassessagriculturalvulnerabilitytoclimatechange.TheaimsofourstudyaretoprovidespatiallyexplicitprojectionsofclimatechangeimpactsonCostsofFood,andtocombinethemwithspatiallyexplicithungerprojectionsfortheyear2030,bothunderapoverty,aswellasaprosperityscenario.Ourmodelresultsindicatethatwhileaverageyieldsdecreasewithclimatechangeinallfocusregions,the impactontheCostsofFood isverydiverse.CostsofFoodincreasemost in theMiddleEastandNorthAfrica,whereavailableagriculturallandisalreadyfullyutilizedandoptionstoimportfoodarelimited. The increase is least in Sub‐Saharan Africa, since production there can be shifted to areaswhichareonlymarginallyaffectedbyclimatechangeandimportsfromotherregionsincrease.SouthAsiaandSub‐SaharanAfricacanpartlyadapttoclimatechange,inourmodel,bymodifyingtradeandexpanding agricultural land. In the Middle East and North Africa, almost the entire population isaffectedbyincreasingCostsofFood,buttheshareofpeoplevulnerabletohungerisrelativelylow,dueto relatively strong economic development in these projections. In Sub‐Saharan Africa, theVulnerabilitytoHungerwillpersist,butincreasesinCostsofFoodaremoderate.WhileinSouthAsiaahighshareofthepopulationsuffersfromincreasesinCostsofFoodandisexposedtohunger,onlyanegligiblenumberofpeoplewillbeexposedatextremelevels.Independentoftheregion,theimpactsofclimatechangearelesssevereinaricherandmoreglobalizedworld.Adverse climate impacts on the Costs of Food could be moderated by promoting technologicalprogress inagriculture. Improvingmarketaccesswouldbeadvantageous for farmers,providing theopportunitytoprofitablyincreaseproductionintheMiddleEastandNorthAfricaaswellasinSouthAsia,butmayleadtoincreasingCostsofFoodforconsumers.Inthelong‐termperspectiveuntil2080,theconsequencesofclimatechangewillbecomeevenmoresevere:while in203056%oftheglobalpopulationmayfaceincreasingCostsofFoodinapoorandfragmentedworld,in2080theproportionwillriseto73%.

4

TableofContentsTableofContents............................................................................................................................................................................4 1  Introduction............................................................................................................................................................................6 1.1  Otherstudiesmodellingclimatechangeimpactsonagriculturalproductionandpoverty........6 1.2  Scopeandlimitations................................................................................................................................................7 

2  Methods....................................................................................................................................................................................8 2.1  Models.............................................................................................................................................................................8 2.1.1  ThebiophysicalcropmodelLPJmL...........................................................................................................8 2.1.2  TheagroeconomiclandusemodelMAgPIE...........................................................................................9 

2.2  DevelopingtheAgriculturalVulnerabilityIndicator................................................................................11 2.2.1  SpatiallyexplicitVulnerabilitytoHungerIndex...............................................................................11 2.2.2  Theagriculturalindicators:Yields,productionandCostsofFood............................................13 2.2.3  Spatiallyexplicitpopulation......................................................................................................................13 

2.3  Scenarios.....................................................................................................................................................................14 2.3.1  Socioeconomicscenarios............................................................................................................................14 2.3.2  Climatescenarios...........................................................................................................................................16 2.3.3  Marketaccessscenarios..............................................................................................................................18 2.3.4  Agriculturaltechnologicalprogressscenarios...................................................................................19 2.3.5  Referencescenarios......................................................................................................................................19 

3  Resultsanddiscussion....................................................................................................................................................20 3.1  Theimpactofclimatechangeonbiophysicalindicators.........................................................................20 3.2  Theimpactofclimatechangeonagroeconomicindicators...................................................................22 3.2.1  Climate‐inducedchangesinagriculturalproduction......................................................................22 3.2.2  Adaptationstrategiesaredifferentforthefocusregions.............................................................23 3.2.3  Climate‐inducedchangesinCostsofFood..........................................................................................25 3.2.4  ClimatechangedoesnotnecessarilyleadtoincreasingCostsofFood...................................26 

3.3  Assessingagriculturalvulnerabilityunderclimatechange...................................................................27 3.3.1  Regionalandnationalresults:AgriculturalVulnerabilityIndicator.........................................29 3.3.2  Regionalandnationalresults:Numberofpeopleaffected...........................................................30 

3.4  Adaptationoptions:Technologicalprogressandimprovedmarketaccess....................................32 3.4.1  Results:Technologicalprogress..............................................................................................................32 3.4.2  Fasttechnicalprogresscanalleviatenegativeclimatechangeimpacts..................................33 3.4.3  Results:Marketaccess.................................................................................................................................34 3.4.4  Bettermarketaccesshelpsproducersbutmighthurtlocalconsumers.................................34 

3.5  Theglobalperspective...........................................................................................................................................35 3.5.1  GlobalimpactofclimatechangeonCostsofFood...........................................................................35 3.5.2  Assessingglobalagriculturalvulnerability.........................................................................................37 

5

3.6  Lookingbeyond2030.............................................................................................................................................38 3.6.1  Until2080CostsofFoodwillincreaseinallthreefocusregions..............................................38 3.6.2  Thenumberofpeoplenegativelyimpactedwillincreaseintheverylongterm................41 

4  Conclusions..........................................................................................................................................................................41 4.1  Mainfindings.............................................................................................................................................................41 4.2  Policyimplicationsofthisstudy........................................................................................................................43 4.3  Limitationsofthemodelandthestudyapproach.....................................................................................44 4.4  Comparisontotheliterature...............................................................................................................................45 

5  References............................................................................................................................................................................46 6  Appendix...............................................................................................................................................................................50 6.1  TranslationofSSPindicatorsintomodelparametersforthenon‐focusregions.........................50 6.2  RegressionforthespatiallyexplicitVulnerabilitytoHungerIndex...................................................51 6.3  Differenceinprojectionsofthespatiallyexplicitpopulation...............................................................52 6.4  Comparisonofresultsforthedifferentgeneralcirculationmodelsfor2080,withandwithoutCO2‐fertilization......................................................................................................................................................................53 6.5  Additionalindicators:Irrigatedandrainfedyieldsforthefocusregions........................................54 6.6  Regionalchangesinproduction,area,tradeandcosts............................................................................55 6.7  Spatiallyexplicitchangesinproduction,CostsofFoodandtheAgriculturalVulnerabilityIndicatorforthetechnicalprogressscenarios,forthethreefocusregions..................................................59 6.8  Spatiallyexplicitchangesinproduction,CostsofFoodandtheAgriculturalVulnerabilityIndicatorforthemarketaccessscenarios,forthethreefocusregions...........................................................62 6.9  Theimpactofclimatechangeonproductionandyieldsin2080comparedto2030forthethreefocusregions................................................................................................................................................................65 6.10  Globalimpactsofclimatechangeonyields,wateravailability,production,CostsofFoodandpeopleexposedtohunger...................................................................................................................................................67 6.11  Landandwatershadowprices2030for10worldregions...................................................................71 

6

1 IntroductionA comprehensiveunderstandingof climate change impacts requires extending the researchbeyondthephysicalpropertiesoftheclimatesystemtohumanimpactsandthegeographicandsocioeconomicfactors that influencethem(Wheeler,2011).Severalstudieswarnthatclimate‐relatedreductions inglobal food production in combinationwith increasing food demanddue to population and incomegrowthwill lead to increasedprices of food (Nelson et al., 2009;Willennockel, 2011). For example,Parryetal.(2005)projectsthatglobalcerealpricescouldincreaseby30‐70%by2050andmorethan160%by2080.Suchrisingpricesalsoincreasetheriskofhunger,asfoodconsumptionbecomesmoreexpensive.While rising prices also have the potential to increase the income of peopleworking inagriculture,theneteffectsaremostoftennegative.Thepoorest20%ofthepopulationoftenreceiveparts of their income from agriculture, but are still net food‐buyers, while net food‐producers canratherbefoundinmiddleincomesegments(FAO,2011;IvanicandMartin,2008).(Parryetal.,2009)estimatethatacombinationofclimatechange,highpopulationgrowthandincreasingregionalincomedisparitiescouldincreasethenumberofpeopleatriskofhungerworldwidebyupto20%by2050.However,itremainsuncertain,howtheregionalandlocaldistributionoftheseeffectsmaybe.Theaimsofourstudyaretherefore:toprovidespatiallyexplicitprojectionsofclimatechangeimpactsonCosts of Food, an indicatordefined as the averageproductionCosts of Food and feed crops in aspecific spatial area.We estimate this indicator using theModel of Agricultural Production and itsImpactontheEnvironment(MAgPIE)(Biewaldetal.,2014;Lotze‐Campenetal.,2008;Schmitzetal.,2012), a spatially explicit agroeconomic land use model. This indicator is combined with spatiallyexplicit hunger projections for the year 2030, in order to develop an Agricultural VulnerabilityIndicator.Weanalyzethelatterundertwodifferentsocioeconomicscenarios,oneofpovertyandoneofprosperity.Thisreportisstructuredasfollows.First,wepresentaliteraturereviewonotherstudiesinvestigatingtheeffectsofclimate‐inducedyieldchangesonpoverty.Wedescribehowourstudyextendspreviousanalyses and the scope and limitations of our approach. Then, we introduce the methodology,indicators and scenarios that we use in this study. Third, we present results and discuss theirimplications,followedbytheconclusionsandpolicyimplicationsinthelastchapter.

1.1 OtherstudiesmodellingclimatechangeimpactsonagriculturalproductionandpovertyOneofthemostimportantimpactchainsthroughwhichclimatechangeaffectstheriskofhungeristhechangeofpotentialcropyields.Changedprecipitationandtemperaturewillaltertheglobalpatternsofpotentialyields,andaltertheallocationofcropproductionaswellastheproductioncosts.Thiswillinturnchangemarketprices,affecthouseholdseconomicsituationandfinallypovertylevelsandtheriskofhunger.Thisimpactchainhasbeenexploredtovariousdegreesintheliterature,whilesomestudiesweremodellingthewholeimpactchainfromclimatetohunger,othersfocusedonlyonpartsofit.AmongthefirststudiesanalyzingthischainhasbeenTobeyetal.(1992),usinganeconomicmodeltoanalyzeyieldshocksinagriculture.Theirresultsshowthatfoodmarketscaneffectivelydampensuchshocksthroughadaptedproductionandconsumption.WhileTobeyetal. (1992)onlysimulatedpartsof the impactchain,Parryetal. (2005)analyzedthewholeimpactchain,feedingclimatesimulationresultsintoprocess‐basedcropmodels,whichinturninformed an economic model that estimated the market outcomes on supply and demand. Theeconomicmodelalsoprovidesasimpleindicatorfortheriskofhunger.Thestudyconcludesthatinascenariowithhighclimateimpactsandlittleadaptivecapacity,Africawillbetheworldregionatthegreatest risk of hunger. Similar results are reported by Parry et al. (2009). By 2050, the semi‐aridregions north and south of the equator in Africa will be especially vulnerable since the projectedincreasedaridity isexpectedtooverlapwith low income levelsof the localpopulation.Additionally,

7

thepoorestpartsofSouthandSouth‐EastAsiaarelikelytobesubstantiallyaffectedbyclimatechange.Hasegawa et al. (2014), use a similar methodology, and argue that population and economicdevelopmenthaveagreater impacton the riskofhunger thannegative climatic conditionsas such.The impact on the risk of hunger varies across regions not only as a result of different climaticconditionsbutalsoduetoadifferentcalorieintakeandagriculturallandavailability.The agriculturalmodel intercomparisonproject (AgMIP) uses 2 climatemodels, 5 cropmodels andnineglobaleconomicmodelstoexploretheimpactchainfromclimatechangetomarketoutcomes.Intheyear2050,theclimateimpactsofaveryhighemissionscenario(RCP8.5)resulted–intheaverageof all simulations ‐ in the reductionof averagepotential yieldsby17%, anarea increaseby11%, areductioninconsumptionby3%andapriceincreaseby20%(Nelsonetal.,2014).Anupdateofthisstudyfocusedonmorelikelymediumandhighemissionscenarios(RCP4.5andRCP6.0),resultinginmuch lower climate impacts, approximately halving the yield and price impacts (Wiebe et al.,accepted).ClimateimpactsmaybeevenlowerifCO2fertilizationeffectswouldbeincluded,whicharestill excluded inmost analysisdue to thehighuncertaintyof the effect. Poverty andundernutritionimpactswerenotanalyzed.Ahmed et al. (2009) extend previous studies by having a more detailed representation of marketimpactsonpoverty,combininganeconomicmodelwithapovertymodule.Thismoduleutilizesmicro‐simulation for representative households at the poverty line in each socioeconomic stratum todetermine changes in poverty headcount based on changes in real income. They quantify thevulnerabilityofthepoortopotentialchangesinclimatevolatility,inthecontextofthefrequencyandmagnitudeofdifferentclimateextremes.Theauthorsuseagainascenariowithstrongclimateimpactsandlowadaptivecapacitytoanalyzeagriculturalproductivitychangesintheperiod2017to2100in16developingcountries.Intheirstudytheeffectsofclimateshocksarevisiblethroughtwochannels:changesinearningsandchangesintherealcostsoflivingatthepovertyline.TheresultsshowthatthehighestsharesofpopulationenteringpovertyasaconsequenceofclimateextremesareinBangladesh,Mexicoandinthesouth‐westofAfrica.Theurbanlaborgroupappearedtobethemostvulnerabletoextremeclimateeventsandtoanyresultingfoodpriceincreases.Agriculturalhouseholdsontheotherhandweremuchlessexposed.Usingasimilarapproach,Herteletal.(2010)usedisaggregateddataonhouseholdeconomicactivitywithinindividualcountriesandembedthisdatawithintheGeneralEquilibriummodeltoexploretheimpactsonpovertyofchangesinagriculturalproductivityduetoclimatechangeintheperiod2000‐2030.Theauthorscomparealowagriculturalproductivityscenarioinaworldofrapidtemperaturechanges,withahighproductivityscenariowithout temperature increase.Here,by2030thepoorestcountriesaretheoneshithardestbytheagriculturalproductivitychanges,sincetheireconomiesaremoredependentonagriculturalproduction.Themodelresultsshowthatthehighestnegativeimpactsof climate change on crops are in Sub‐Saharan Africa, but high losses also occurred in the US andChina.Householdsurveydatafrom15developingcountrieswerethenusedtoestimatetheimpactofagricultural price changes ofwelfare ondifferent groups of poor households. The authors find thatglobalcerealpricesincreaseby32%inthelowproductivityscenarioand16%inthemoreoptimisticproductivityscenario.Inthelowproductivityscenario,povertyincreasedbyasmuchasone‐thirdinthe urban labor social strata in Malawi, Uganda, Zambia and in the non‐agricultural self‐employedstratuminBangladesh.

1.2 ScopeandlimitationsSimilar toParryet al. (2005),Hasegawaet al. (2014),Ahmedet al. (2009),Hertel et al. (2010)andNelson et al. (2013), we use climate projections to determine future crop yield potentials underclimate change, which in turn inform an economic model that estimates the climate impacts onproductioncosts.Whileotherstudies(Ahmedetal.,2009;Hasegawaetal.,2014;Herteletal.,2010,

8

IvanicandMartin,2008)trytotranslateeconomicimpactsintoindicatorsofpovertyorfoodsecurity(e.g.populationexposedtoundernutritionorriskofhunger),ourapproachdoesnotexplicitlymodelthe impact chain that leads from market outcomes to poverty. Instead, the major strength of thisanalysis is to observe the effects of climate change on a much finer scale (0.5°) than other globalmodels,whichusuallyoperateonanationalor (world)‐regionalscale.To investigate the impactsofglobalenvironmentalchangesonpovertyinspecificlocations,differentresearchmethodshavetobeused. This report proposes a new approach.We combinemodel‐based global agricultural indicatorprojectionsthatareavailableatthe0.5°gridcelllevelpopulationandhungerprojectionsatthesameresolutioninordertoidentifyoverlapsbetweentheareaswiththehighestclimatechangeimpactsonagriculturalproductionandtheareasthatareprojectedtobemostpronetotheriskofhunger.Weusethe spatially explicit hunger index instead of poverty indicators, since such indicators such as thefrequentlyusedratioofpopulationlivingbelowthepovertylinearemostlyprovidedonregionallevel.Our study places a special emphasis on modelling on a high spatial resolution, since recent trendanalysis shows that poverty will be increasingly concentrated in particular areas in the future(Shepherdet al., 2013). For example,Amarasingheet al. (2005) show that in Sri Lanka thepooresthouseholds are located in dry areas where small‐size agricultural holdings depend on rainfedproduction andwhere incomediversificationopportunities are scarcedue to a longdistance to thenext urban infrastructure. Furthermore, the urban poor as the net buyers of food are particularlyvulnerabletofoodpricespikesthatoftenfollowclimateinducedproductionshocksanddeclines(Heinetal.,2009;McMichaeletal.,2012;SmitandParnell,2012).Barrettetal. (2006)pointoutthat it isimportanttoidentifysuchless‐favoredareas.Theyneedmoredirectinterventionsforimprovingthecapability of households to build and protect their assets and to improve their access to financialservices. The most appropriate interventions will depend on the local context. The ability ofhouseholds in such locations to copewith shocks anddisasterswill be particularly challenged. Theinadequate capacity of households to recover from such shocks can lead to a cycle of losses andmaladaptivestrategiesincludingdivestmentofproductiveassetssuchaslivestockandsellinglandforfood(UNDP,2007).Ourstudyfocusesonthelong‐termeffectsofclimatechangethroughcropyieldsonagriculture,anddoesnotaccountforclimateimpactsthroughotherimpactchains, likeextremeeventsanddisasters(Shepherdandetal.,2013),unforeseenclimateshifts(FAO,2013),negativeimpactsonhumancapital(Behrman, J.etal.,2004;ClarkeandHill,2013;FosterandRosenzweig,1993;GlickandSahn,1998;Hoddinott,2006)orphysicalcapital(Carteretal.,2007)employedinagriculture.

2 Methods2.1 ModelsThemethodologicalcornerstoneofourstudyisthecoupledmodelsystemofthebiophysicalcropmodelLPJmLandthespatiallyexplicitagroeconomiclandusemodelMAgPIE.

2.1.1 ThebiophysicalcropmodelLPJmLLPJmLsimulatescarbonandwatercyclesaswellasvegetationgrowthdynamicsdependingondailyclimaticconditionsandsoiltexture.NaturalvegetationisrepresentedinLPJmLatthebiomelevelbyninePlantFunctionalTypes(PFTs)(Sitchetal.,2003).Themodelcalculatesclosedbalancesofcarbonfluxes(grossprimaryproduction,auto‐andheterotrophicrespiration)andpools(inleaves,sapwood,heartwood, storageorgans, roots, litter and soil), aswell aswater fluxes (interception, evaporation,transpiration, snowmelt, runoff,discharge) (Gertenet al., 2004;Rostet al., 2008).Photosynthesis is

9

simulated following the Farquhar model approach (Farquhar et al., 1980). Processes of carbonassimilationandwaterconsumptionareparameterizedontheleaflevelandscaledtotheecosystemlevel. Carbon and water dynamics are closely linked so that the effects of changing temperatures,decliningwateravailabilityandrisingCO2concentrationsareaccountedforandtheirneteffectcanbeevaluated (Gerten et al., 2004, 2007). Physiological and structural plant responsesdeterminewaterrequirementsandconsumption.CompetitionbetweenPFTsduetodifferencesintheirperformanceundergivenclimateconditions,canlead to changes in vegetation composition as less adaptedPFTs canbe out‐competed and replaced.Subsequently to alterations in vegetation composition, i.e. in the PFT distribution, changes inproductivity and respective carbon fluxes can also be quantified. This applies to long‐term climatetrendsaswellas interannualclimatevariability, includingtheimpactsofextremeevents.Therefore,the LPJmL model is indeed capable of capturing dynamic responses to, e.g., single or consecutivedrought events. The suitability of the LPJmL framework for vegetation andwater studies has beendemonstratedbyvalidatingsimulatedphenology(Bondeauetal.,2007),riverdischarge(Gertenetal.,2004;Biemansetal.,2009),soilmoisture(Wagneretal.,2003)andevapotranspiration(Sitchetal.,2003;Gertenetal.,2004).

2.1.2 TheagroeconomiclandusemodelMAgPIEMAgPIE (Model ofAgricultural Production and its Impact on theEnvironment) is a global, spatiallyexplicit,economic landusemodelsolving inarecursive‐dynamicmode(Biewaldetal.,2014;Lotze‐Campenetal.,2008;Schmitzetal.,2012),Themodeldistinguishestenworldregionsonthedemandside(Figure1)andusesinputdataof0.5degreeresolutiononthesupplyside.Duetocomputationalconstraints,allmodelinputsonthesupplysideareaggregatedtoclustersfortheoptimizationprocessbased on a k‐means clustering algorithm (Dietrich et al., 2013). With income and populationprojections(seesection2.3.1)asexogenousinputs,demandforagriculturalcommoditiesisprojectedinthe futureandproducedby15foodcrops,5 livestockproducts,1 fibercrop,andbyfodderasanintermediate input. While demand is scenario‐specific and changes over time, it does not react tochangesinsupplyoranyothervariablesateachtimestep.Feedrequirementsforlivestockproductionactivitiesconsistofamixtureofgrazedbiomassfrompasturesaswellasfodderandfoodcrops.Thelivestock‐specific feedenergyrequirementsdependonbiologicalneedsformaintenanceandgrowthbut also temperature effects and the use of extra energy for grazing (Wirsenius, 2000). Themodelsimulates time steps of 10 years and uses in each period the optimal land‐use pattern from thepreviousperiodasinitialcondition.Onthebiophysicalside,themodelislinkedtothegrid‐baseddynamicvegetationmodelLPJmLwhichprovides important biophysical inputs like crop yields under both rainfed and irrigated conditions,relatedirrigationwaterdemandpercrop,andwateravailabilitydependingonclimaticconditionsona0.5degreeresolution.SpatiallyexplicitlandtypesinMAgPIEcomprisecropland,pasture,forest,urbanareas, and other land (Krause et al., 2013). The objective function of MAgPIE minimizes globalagricultural production costs,which involves factor costs for labor, capital, and intermediate inputsderived from the GTAP database (Narayanan andWalmsley, 2008), investments into research anddevelopment(R&D),landexpansioncostsaswellastradeandtransportcosts.R&DinvestmentsallowMAgPIE to increase cropyields in aparticular region.Thisendogenous implementationof technicalchange is based on a surrogate measure for agricultural land use intensity (Dietrich et al., 2014).Expansion of cropland is the alternative to increase the production level. Expansion involves landconversioncostsforeveryunitofconvertedland,whichaccountforthepreparationofnewlandandbasic infrastructure investments (Krause et al., 2013). Land conversion costs arebasedon country‐levelmarginalaccesscostsgeneratedbytheGlobalTimberModel(GTM)(Sohngenetal.,2009).

10

Figure1:ThetenworldregionsinMAgPIE:AFR=Sub‐SaharanAfrica,CPA=CentrallyPlannedAsia(incl.China),EUR=Europe(incl.Turkey),FSU=FormerSovietUnion,LAM=LatinAmerica,MEA=MiddleEastandNorthAfrica,NAM=NorthAmerica,PAO=PacificOECD,PAS=PacificAsia,SAS=SouthAsia(incl.India).

InternationaltradeinMAgPIEisimplementedbyusingflexibleminimumself‐sufficiencyratiosattheregionallevel.Self‐sufficiencyratiosdescribehowmuchoftheregionalagriculturaldemandquantityhastobeproducedwithinaregion.Forinstance,aratioforcerealsof0.8meansthat80%ofcerealsareproduceddomestically,whereas20%areimported.Torepresentthetradesituationof1995,wecalculatedtheself‐sufficiencyratiosforeachregionandcommoditybasedonFAOdata.Therearetwovirtual tradingpoolswhichallocate theglobaldemand to thedifferent supply regions.Thedemandwhichentersthefirstpoolisallocatedaccordingtofixedcriteria.Self‐sufficiencyratiosdeterminehowmuch is produced domestically, and export shares determine the share of each region in globalexports.Theexport shares aregenerated forevery crop for theyear1995andare taken fromFAOdataaswell(Schmitzetal.,2012).However, although the initial self‐sufficiencies for this pool stay constant over time, the final self‐sufficienciesdochangesincedomesticdemandandpopulationchangeovertime.Thedemandwhichentersthesecondpoolisallocatedaccordingtocomparativeadvantagecriteriatothesupplyregions.Thecriteriaarebiophysicalyield,productioncostsandcostsofintensificationthroughinvestmentinyield‐increasingtechnologicalchange.Thisimpliesthatthemodelminimizesglobalproductioncostsandproducesinthosecellswhereitismosteconomicalcomparedtoothercells.Whenanalyzingmodel‐basedchangesofspatialproductionpatternsandproductioncosts,asdoneinthisstudy,thefollowingmodel‐inherentmechanismsshouldbeconsidered:1.) Notwithstanding the negative impacts of climate change on agriculture, a region and scenario‐

specificnumberofpeoplehastobefed.Therearetwomodel‐inherentadaptationmechanismstocompensate for decreases in yields and water availability: 1. More imports, compensatingdecreases in domestic production, 2. Stabilization of regional production through expandingagricultural area, investing in yield‐increasing technical change, or investing in irrigationinfrastructure.

2.) Thepossibilityofusingtradeasameasuretoalleviatetheimpactsofclimatechangeislimitedinascenario with high trade barriers. A globalized world is better able to compensate locallyheterogeneousimpactsonyieldsthroughadjustingtradeflows.

11

3.) Climatechangecanleadtoloweryieldsaswellasalteredwateravailability.Loweravailabilityofwaterreducesirrigatedproductionwithhigheryieldscomparedtothecaseofnoclimatechange,butalsoreducesthepossibilityofexpandingirrigatedcroparea.

2.2 DevelopingtheAgriculturalVulnerabilityIndicatorIn order to estimate the agricultural vulnerability of poor people related to climate change, wecombine the Vulnerability to Hunger Index with Costs of Cfood as a socioeconomic agriculturalindicator. In the followingtwosectionswedescribe1. themethodologybehindthespatiallyexplicitVulnerabilitytoHungerIndexanditsfutureprojectionand,2.themodel‐basedagriculturalindicatorCostsofFood.Sinceitisnotonlyimportanttoknowwherepeoplearevulnerable,butalsohowmany,wedescribeinthelastpartaspatiallyexplicitdatasetonpopulationwhichwecanusetoestimatethenumberofpeopleimpacted.

2.2.1 SpatiallyexplicitVulnerabilitytoHungerIndexFor the Global Hunger Index described in the publication by von Grebmer et al. (2011) from theInternationalFoodPolicyResearchInstitute(IFPRI),threeequallyweightedindicatorsarecombinedinoneindex,namelytheproportionofpeoplewhoareundernourished,theprevalenceofunderweightin children younger than five and themortality rate of children younger than five. All three indexcomponentsareexpressed inpercentagesandweightedequally.HigherGlobalHunger Indexvaluesindicatemorehunger,lowervaluesindicateless.Ofthethreecomponentsmentionedabove,onlychildunderweight and childmortality areprovided on grid cell level for the years1998‐2002and2000,respectively (Center for International Earth Science Information Network ‐ CIESIN ‐ ColumbiaUniversity, 2005a, 2005b). The Food and Agriculture Organization of the United Nations (FAO)providesdataonprevalenceofstuntingamongchildrenunder five foraresolutionof5arc‐minutesfor varying years before 2007 (FAO, 2014), which we use as a replacement for the non‐spatiallyexplicitundernourishmentindicatoroftheGlobalHungerIndex.Basedonthesethreespatialdatasets(childunderweight,childmortalityandstuntingamongchildrenbelowfive),wederive thespatiallyexplicitVulnerabilitytoHungerIndexusedinthisstudyaccordingtotheabove‐mentionedmethodofIFPRI. Due to the lack of the availability of a spatially explicit data set representing the entirepopulation,theVulnerabilitytoHungerIndexpertainsnowonlytochildren.Inordertobecomparabletootherstudies,weusethehungerlevelcategoriesdefinedbyvonGrebmeretal.2011(Table1).WeusethetermVulnerabilitytoHungerbecause itreflectswhetherpeoplepossesssufficientmoneytoafford the necessary food to avoid suffering from hunger. The Vulnerability to Hunger Index isprojected based on economic poverty data, thus representing changes in income. Changes on thesupply side through climate impacts are represented through the Costs of Food, discussed andexplainedbelow.Table1:CategoriesfortheVulnerabilitytoHungerindexdefinedbyvonGrebmeretal.(2011).Theindexranksspatialunitsona100‐pointscale.Zeroisthebestscore(nohunger),and100istheworst.Neitherofthetwoextremesisreachedinpractice.

Descriptionhungerlevel Abbreviations usedingraphs Classificationsextremelyalarming EA ≥30.0alarming A 20‐29.9serious S 10‐19.9moderate M 5‐9.9low L ≤4.9

Inasecondstep,weprojecttheVulnerabilitytoHungerIndexfortheyear2000totheyear2030usingscenario‐basednationalpovertyprojectionsprovidedby theWorldBank (RozenbergandHallegate,

12

2015).WhiletheVulnerabilitytoHungerIndexhasaspatialresolutionof0.5°,thepovertyprojectionsare on a country basis. We therefore assume that the subnational pattern of the Vulnerability toHungerIndexstaysconstantovertimeandtherebyneglect that thespatialdistributionofeconomicprosperityandfailuremightchangeovertime.OfthedatasetsprovidedbytheWorldBank,thefollowingarecorrelatedtothenationalaveragesoftheVulnerabilitytoHungerIndexinameaningfulway:povertygapreferringtothe1.25US$povertyline(povgap)withR2=0.32,percentageofpeoplelivingonlessthan1.25US$aday(poor)withR2=0.51,percentageofpeoplelivingonlessthan4$aday(nearpoor)withR2=0.67.Thecombineddatasetsarecorrelatedto theVulnerability toHunger IndexwithR2=0.70.Basedontheserelations,wecreateda linearregressionwithseveralvariableswhichresultedina functiondepictingtherelationbetweenthenationalvalues fortheVulnerabilitytoHunger Index in2000andthedifferentpovertyindicators:VulnerabilitytoHungerIndex=4.237+(‐52.899*povgap+38.285*poor+16.151*nearpoor)

Weusethisfunctionthenagainfortheyear2030totranslatenationalvaluesofthepovertyindicatorsintonationalvaluesoftheVulnerabilitytoHungerIndex.AssumingthatthespatialdistributionoftheVulnerability toHunger Indexhas not changed over time,wedistribute the projected values of theVulnerabilitytoHungerIndexaccordingtotheoriginalspatialdistribution(Figure16intheAppendixshowsthecorrelationbetweenthefittedandtheoriginalvalues).We substitutedmissing data of the poverty indicators by creating a list of countries sorted by thenational averages based on the Vulnerability to Hunger Index. Based on the linear regressiondescribedintheprevioussection,eachaverageVulnerabilitytoHungerIndexvaluewasthenpairedwiththerespectivepovertyvalue(poor,nearpoor,povgap)fortheyear2000.Forcountrieswithoutthesepovertyvalues,weusedaweightedmeanofthetwopovertyindicatorsofthecountriesclosesttothecountrywithoutvalueintheorderedlist.Furthermore,newvaluesfor2000wereprojectedto2030asdescribedabove.Figure2displaystheresultingdistributionofthecompletedVulnerabilitytoHungerIndexforthepresent.

Figure2:VulnerabilitytoHunger Indexbasedonchildundernourishment,childunderweightandchildmortalityrate.TheCIESINdataonchildunderweightandchildmortalityratereflectdata from2004‐2011. The child undernourishment data from the FAO are assembled from varying years. Light greyindicatesindustrialcountries(I),darkgreyreflectsthatnodataareavailablenorcanbededucted(NA).(Hungerlevels:EA=extremelyalarming,A=alarming,S=serious,M=moderate,L=low.)

13

Finally,theVulnerabilitytoHungerIndexwillbecombinedwithCostsofFoodfromtheMAgPIEmodel(see section2.2.2 foranexplanation) inorder tobeable to shownotonlywherepeople relyingonagriculturalproductionaremostaffectedbyclimatechangeimpacts,butalsowherepeopleexposedtohungeraremostaffected.

2.2.2 Theagriculturalindicators:Yields,productionandCostsofFoodInordertoseehowclimatechangecanimpactfarmersinpoorregions,weidentifiedthreerelevantindicatorsforagriculturewhicharesensitivetoclimatechange,namelycropyields,cropproduction,andCostsofFood.WhileyieldsaresimulatedwiththebiophysicalmodelLPJmL,productionandCostsofFoodaregeneratedwiththeagroeconomicmodelMAgPIE.Theyarebasedonanaggregateoffoodcrops which include1: temperate cereals (wheat, barley, rye, mixed grain, oats, triticale), maize,tropical cereals (millet, sorghum,canaryseed, fonio,quinoa), rice, soybean, rapeseed (rapeseedandmustard seed), groundnut, sunflower, oilpalm, pulses (bambara beans, beans dry, broad beans dry,chickpeas,cowpeadry, lentils, lupins,peasdry,pigeonpeas,otherpulses,vetches),potato,cassava,sugarcane,sugarbeetandothers(vegetablesandfruits).Potential crop yields are generated with the biophysical model LPJmL and show how a changingclimate impacts agricultural yields. Yields shown are the area‐weighted average of irrigated andrainfed yields, before economic interventions such as changes in land management or technology.Therefore, these results donot necessarily reveal informationon changes in production levels. Thevaluesoftheindicatorarepresentedingigajoule(GJ)perhectare(ha)foreach0.5°gridcell.Weshowjouleratherthantonneinordertoprovideacomparableunitacrossdifferentcrops.Availabilityofirrigationwaterisinsomepartsoftheworldanimportantpreconditionforagriculturalproduction. Since irrigation depends on biophysical conditions such as vegetation cover, aswell asclimaticconditionssuchasprecipitation,itisalsosensitivetoclimatechange.Thebiophysicalwateravailabilitydiscussedinsection3.1doesnotrepresentirrigationwateravailability.Inordertoaccountfornon‐agriculturalwateruseandminimumenvironmental flows, theadequateamountofwater isdeductedfromthetotalavailablewaterinrivers,lakesandwetlands.TheCostsofFoodaretheaggregatedcostsfortheproductionoffoodandfeedcropsperGJforeach0.5°gridcell.Feedproductioncostsareincludedbecausefoodandfeedcropsare(inthemodelandmostly in reality) identical, the costs for production are therefore the same. Production costs forbioenergy or fiber arenot included. The costs include land conversion costs, transport costs, factorrequirements, investments in irrigation infrastructure, and agricultural R&D investments. RegionalR&Dinvestmentsarespatiallydistributedaccordingtoproductionpatterns.Althoughitisnotpossibletoderivedirectconsumerpricesasaresultoftheinterplaybetweendemandandsupplyatthegridlevel,thechangesinaggregatedproductioncostscanbeinterpretedaschangesinprices,ifweassumethatthereisaconstantmarkup(profitmargin)fortheanalyzedfoodcrops.SinceCostsofFoodaredirectly influenced by climate impacts on yields, they are a good indicator for our study. SinceagriculturalR&DinvestmentsareincludedinCostsofFood,theycannotbeinterpretedcompletelyasprivatemarginalcostsfacedbyafarmer,sinceapartofthecostsarepaidonthenationallevel,e.g.bytaxpayers,oron theglobal levelbyaid foundations.TheCostsofFoodarepresented inUS$/GJ foreach0.5°gridcell2.

2.2.3 SpatiallyexplicitpopulationInordertoestimateclimatechangeimpactsonpoverty, it isdecisivetoknowhowmanypeopleareaffectedbytheseimpacts.Thereforeadatasetofpopulationdensityona0.5°gridfortheyear2005isused(Center for InternationalEarthScience InformationNetwork(CIESIN),Centro Internacionalde1ThecropsinbracketsexplainthecropgroupsusedinMAgPIE.2Togiveanexample,onedrymattertonneofcerealscorrespondstoapproximately17.9GJ.

14

AgriculturaTropical(CIAT),2005).Inordertoprojectthegriddedpopulationintothefuture,weusethe population scenarios developed for the different Shared Socioeconomic Pathways (SSPs)(International Institute forAppliedSystemsAnalysis(IIASA),2013).Duetoa lackofbetterdata,weassumed that the distribution of people in each country stays the same, but that population sizechanges ineachgridcellproportionallytothechangeofpopulation intheentirecountry.ThemapsshowingthepopulationprojectionscanbefoundintheAppendix(Figure17).

2.3 Scenarios2.3.1 SocioeconomicscenariosInordertoanalyzehowpeopleexposedtohungerareaffectedbyclimatechangeindifferentfuturesocioeconomicsettings,weusetwoextremescenariosbasedoneconomicdevelopments.Ontheonehandascenarioofprosperity,representingafutureworldwithhighGDPandlowpopulationgrowth,andontheotherhandascenarioofpovertywhereGDPgrowthislowandpopulationgrowthishigh.WebasekeyindicatorsofthesescenariosontheSSPs,whichhavebeendevelopedbythedifferentkeyclimate research communities in order to be able to explore the long‐term consequences ofanthropogenic climate change and possible responses (Kriegler et al., 2014; Moss et al., 2010;Nakicenovicetal.,2014;O’Neilletal.,2014;vanVuurenetal.,2012).Therelevant indicatorsof thepovertyscenarioarebasedonSSP4,therelevantindicatorsfortheprosperityscenarioarebasedonSSP5. The qualitative indicators defined for each SSP have to be translated into quantitativemodelinput for the agroeconomic model MAgPIE, including trends in environmental awareness,globalizationandtechnologicalchange.Thequantitativedataforthesescenarios(populationandGDPgrowth)havebeenmadepubliclyavailablebytheInternationalInstituteforAppliedSystemsAnalysis(2013).Inthecontextofagroeconomicmodeling,populationscenariostranslateintooveralldemandforcropandlivestockproducts,whiletheregion‐specificdevelopmentofGDPinfluencesdietaryhabitssuchasconsumptionof livestock‐relatedproductsand theamountofcaloriesconsumed,orcalorieswastedper person. Direct demand for food and indirect demand for feed crops depend specifically on theamount of livestock‐related products in the diets. If less meat and other livestock products areconsumedinascenario,vegetalpartsinhumandietsincrease,buttheamountofrequiredfeedcropsdecreases, and vice versa. The exogenously given projections for food demand in the model arederivedfromscenarioinformationonkcalconsumptionpercapitaandpopulationgrowth(Valinetal.,2014).BasedonhistoricalnationaltimeseriesonGDPandfoodandlivestockdemand,theamountofconsumptionpercapitavaries.TheSSPindicator“environment”isnotrelevantforourstudy,astherearenodifferencesbetweenthetwoscenarios.IntheenvironmentallynotverysensitiveSSP4andSSP5,protectedareasincreaseuntil2100by50%compared to2010.The indicator “technology”manifests itself as soilnitrogenuptakeefficiencyandlivestockefficiency(theamountoffeedneededtoproduceacertainamountoflivestockproducts).Theindicator“globalization”isimplementedinourmodelingframeworkthroughdifferentratesoftradeliberalization(Table2).

15

Table2:TranslationofSSPindicatorsintomodelparametersandtheirimplementationsforthepovertyandtheprosperityscenario.

Modelparameters(SSPindicators)

Povertyscenario (basedonSSP4) Prosperityscenario(basedonSSP5) Sub‐

SaharanAfrica

MiddleEastandNorthAfrica

SouthAsia

Sub‐SaharanAfrica

MiddleEastandNorthAfrica

SouthAsia

Populationinmillionpeoplein2030

1396 511 2054 1240 508 2025Kcalpercapitain2030(basedonGDP)

2531 3245 2621 2707 3394 2763DemandforfoodcropsinPetaJoulein2030(basedonpopulation/GDP)

4462 2460 7013 4191 2100 7088

Shareoflivestockproductsinthedietin2030(basedonGDP)

0.09 0.15 0.14 0.10 0.15 0.14

Tradeliberalization(Globalization)

Startingfrom2010,tradebarriersarerelaxedby10%perdecadefordevelopedregions,butarekeptconstantfordevelopingregions.

Startingfrom2010,tradebarriersarerelaxedby10%perdecadeglobally.

Livestockintensification(Technology)

Slow FastNutrientefficiency(Technology)

High MediumThetwoextremesocioeconomicscenarios leadtodifferent futuredevelopments.Thepreference foreconomic growth leads in the scenario of prosperity to a globalizedworldwith open trade and anincreaseinthestandardoflivinginallpartsoftheworld,exemplifiedbyhighpercapitalivestockandcalorieconsumption.Thescenarioofpovertyisrepresentedthroughaworldofincreasinginequalitywhere especially poor countries suffer under high population aswell as lowGDP growth,with theconsequenceofalowkcalpercapitaconsumption.FordevelopedcountriesthenarrativeissimilartotheprosperityscenariowithliberalizedtradebetweenthemandhighGDPgrowth.Table2showshowthequalitativeSSPindicatorsaretranslatedintomodelparameters.Theoverallregionaldemandfor foodcrops ispartlyverysimilarbetweenthetwoscenarios(e.g. inSouthAsia). This is due to the fact that overall demand is a combinationof per capitademandandpopulation, since in thepovertyscenariopopulation ishighwhileGDP‐relatedpercapitademand islow,andviceversaintheprosperityscenario:theeffectsoffseteachother.LivestocksharesarepartlybasedonGDPprojections.WhileinAfricatheimpactofabettereconomicdevelopmentinaprosperityscenarioisalreadyvisiblein2030,theshareoflivestockconsumptioninSouthAsiaandMiddleEastandNorthAfricaisequalinbothscenariosin2030duetoarelativelyhighGDPin2005andaresultinglower increase. Differences between the scenarios regarding the livestock share in the two regionsbecomemoreapparenttowardsthemiddleofthecentury.

16

Figure3:SpatiallyexplicitVulnerabilitytoHungerIndexin2030forthepovertyandprosperityscenario.Lightgreyindicatesindustrialcountries(I),darkgreyreflectsthatnodataareavailablenorcantheybededucted (NA). (Hunger levels:EA=extremelyalarming,A=alarming,S=serious,M=moderate,L=low.)

In our study, not only themodel implementation is scenario‐specific, but also the projection of theVulnerability to Hunger Index. The Vulnerability to Hunger Index is based on scenario‐dependenteconomic poverty data explained in section 2.2.1. Figure 3 depicts Vulnerability to Hunger Indexprojectionsforthepovertyandprosperityscenariofor2030.CountrieswherenodatawereavailablefortheVulnerabilitytoHungerIndex,likeSouthSudan,areshownindarkgreyonthemap,countrieswhichareindustrializedandthereforeingeneralnotpronetohungeratalargerscaleareshowninlightgrey.

2.3.2 ClimatescenariosInthisstudy,thehigh‐endRepresentativeConcentrationPathway(RCP8.5)witharadiativeforcingof8.5W/m2intheyear2100relativetopre‐industrialvalues(resultingaveragetemperatureincreaseattheendofthecentury3.7°C)andascenariowithnofutureclimatechangearecompared(Mossetal.,2010). The RCP8.5 emission scenario has been implemented in 5 general circulation models andclimate results have been provided by the CMIP5 project available at http://cmip‐pcmdi.llnl.gov/cmip5/ (Tayloret al., 2012).Climate scenarioswere selectedbasedonavailabilityof

17

bias‐correcteddatasetsfromtheISI‐MIPprojectforthefollowinggeneralcirculationmodels:GFDL‐ESM‐2.0, Hadley‐GEM 2, IPSL‐CM5A‐LR, MIROC‐ESM‐CHEM, Nor‐ESM1‐M. Results were supplied toLPJmLasmonthlydatafieldsofmeantemperature,precipitation,cloudinessandnumberofwetdays(Hempeletal.,2013).Yieldandwateravailabilityvaluesusedinthisstudyarethemeanfortheresultsofthe5generalcirculationmodels.Weusethemeanofthe5generalcirculationmodelsaswedonotknowwhichofthemodelsproducesthemostreliableresult.LPJmL simulations of crop yields and water availability used as input in the MAgPIE model aregenerated without CO2 fertilization. The higher atmospheric carbon dioxide concentration, due tohuman induced increases in emissions, will possibly lead to enhanced crop growth. Leaving it outmight therefore lead to overestimating thenegative effectof climate changeon cropyields.But theimplementationofCO2fertilizationissubjecttoahighlevelofuncertainty(Long,2006;Tubielloetal.,2007). The beneficial effect of CO2 fertilization on crop yields require an adjusted management toenable the production of higher quantities, otherwise nitrogen could become limiting (Bloomet al.,2010; Leakey et al., 2009; Parry and Hawkesford, 2010). Changes in the chemical composition ofplantsunderhigheratmosphericCO2concentrationscouldalsohavenegativeeffectsonyields,astheywereshowntoimpedetheplants’defensemechanismsagainstinsectdamage(Dermodyetal.,2008;Zavalaetal.,2008).MüllerandRobertson(2014:46)concludetherefore:“Asbothcropandeconomicmodelsonaglobalscalearenotcapableofaddressingthismuchdetailedfeedback,theassumptiontoignoreCO2fertilizationeffectsisnotunreasonableatthistime”.Inordertoshowtherangeofresultsfromthedifferentclimatemodelsandmethodologies,Figure4showsdifferencesinyieldsbetweenclimatechangeandnoclimatechangeforeachgeneralcirculationmodelwith andwithout CO2 fertilization for all tenMAgPIE regions. These results showmaximumpotentialyieldsfromLPJmLandmightdifferfromtheyieldsusedasMAgPIEinput,sinceLPJmLyieldsare calibrated in MAgPIE in the first time step to take account of imperfect management. Whilespatially explicit projections of yields might be quite different between the 5 general circulationmodels,projectionsofaveragefoodcropyieldswithandwithoutCO2fertilizationdifferbynotmorethan 7% between the different general circulation models. While in 2030 the sign of the yielddifferenceisthesameineachofthe10regions,in2080itdiffersforthesimulationresultswithCO2fertilization for some regions, thus indicating that the uncertainty of themodeling results is clearlyhigher(Figure18intheAppendix).

18

Figure 4: Regional average difference in biophysical yields of food crops for the 5 different generalcirculationmodels(GCM)withandwithoutCO2fertilizationforRCP8.5comparedtoanoclimatechangescenario(climatechange–noclimatechange)fortheyear2030andforthe10MAgPIEregions.(Namesofregions:AFR=Sub‐SaharanAfrica,CPA=CentrallyPlannedAsia,EUR=Europe,FSU=FormerSovietUnion,LAM=LatinAmerica,MEA=MiddleEastandNorthAfrica,NAM=NorthAmerica,PAO=PacificOECD,PAS=PacificAsia,SAS=SouthAsia.)

2.3.3 MarketaccessscenariosMarketaccessisanimportantfactorforpovertysinceitdetermineshoweasilyruralfarmersareableto sell their products. Accessibility of markets is represented in MAgPIE through intra‐regionaltransportcosts.Hightransportcostsrepresentbadmarketaccessandlowtransportcostsrepresentgoodmarketaccess for farmers. In themarketaccessscenarios, thedifferent transportcostsatgridlevel are uniformlymultiplied by a global parameter. Thismeans that in poor and rich regions thechange in transport costs is the same, but since rich regions can be assumed to already have lowtransportcosts,poorregionsaregoingtobenefitover‐proportionally.

19

TransportcostsinMAgPIEarecalculatedbasedonthetravel‐timefromanareaofproductiontothenext citywithmore than50,000 inhabitants and on transport costs per tonne andminute for eachcommodity(Nelson,2008).Inthescenariowithgoodmarketaccess,weimplementthetransportcostsfrom the cellwith the cheapest transport costs globally and therefore the cellwith thebestmarketaccess, everywhere. This is preferred to using no transport costs at all, since a no transport costscenariocanleadtoanimplausiblelandusepatternwherelandexpansiondoesnotfollowthealreadyexistinginfrastructure,butagricultural landappearsrandomlyandremotefromcurrentproduction.Forthebadmarketaccessscenario,weusetransportcostswhicharetwiceashighasinthedefaultcaseandthereforeworsethaninthecurrentsituation.Inourimplementation,transportcostsremainconstantovertime.

2.3.4 AgriculturaltechnologicalprogressscenariosInordertorepresentanoptimisticandpessimistictechnologicalprogressscenario,wevarytheyieldelasticitywithrespect to investments intoagricultural technical change.Thecurrently implementedyieldelasticityis0.27(Dietrichetal.,2014).Forrepresentingfastandslowtechnologicalprogress,wechoose two alternative scenarios in which we set the elasticity to 0.32 (cheap technical changecorrespondingtofasttechnologicalprogress)and0.22(expensivetechnicalchangecorrespondingtoslowtechnologicalprogress).ThesevaluesweretakenfromthepaperbySchmitzetal.(2012),wherethe authors tested the range for possible yield elasticities based on empirical data. A high yieldelasticity impliesthat investmentsarehighlyprofitableand leadtoa fast increaseinyieldsandviceversa fora lowyieldelasticity.Theyieldelasticity isaglobalparameter,but resulting technologicalprogress differs across world regions. For regions with currently low land‐use intensity, yield‐increasingtechnologicalchangecanbeachievedatlowercosts(Dietrichetal.,2014).

2.3.5 ReferencescenariosWeshowourresultsalwaysrelativetoreferencescenarioswithoutclimatechange,butwiththesamesocioeconomic setting, in order to estimate the impact of climate change – ceteris paribus.Marketaccessandtechnicalprogressscenariosarebasedonthepovertyscenarioinordertoseetheeffectsofchangingparametersforthescenariomorepronetopoverty(Table3).Table3:Overviewofscenariosettingsusedinthisstudy.

Climatescenarios Socioeconomicscenarios Marketaccess Technologicalprogressnoclimatechange poverty medium medium

RCP8.5 poverty medium mediumnoclimatechange prosperity medium medium

RCP8.5 prosperity medium mediumnoclimatechange poverty good medium

RCP8.5 poverty good mediumnoclimatechange poverty bad medium

RCP8.5 poverty bad mediumnoclimatechange poverty medium fast

RCP8.5 poverty medium fastnoclimatechange poverty medium slow

RCP8.5 poverty medium slow

20

3 ResultsanddiscussionInsections3.1and3.2,wedescribeanddiscusstheresultsfortheagriculturalindicatorsandcombinetheminsection3.3withtheVulnerabilitytoHungerIndex,resultingintheAgriculturalVulnerabilityIndicator.Sincehunger is relevantonly insomepartsof theworld,weshowresults for theregionscurrentlymostaffected:Sub‐SaharanAfrica,MiddleEastandNorthAfricaandSouthAsia.WeusetheworldregionsasdefinedbytheWorldBankGroup,butbasedonthealmostidenticalMAgPIEregions.Sub‐SaharanAfrica (SSA)corresponds to theMAgPIEregionSub‐SaharanAfrica (AFR),but includesWestern Sahara. Middle East and North Africa (MNA) correspond to the MAgPIE region MiddleEast/NorthAfrica(MEA)excludingWesternSahara.SouthAsia(SAS)withoutMyanmarcorrespondstotheMAgPIEregionSouthAsia(SAS).The agricultural indicators discussed in the sections are crop yields, crop production, and Costs ofFood. Changes in yields are generated with the biophysical model LPJmL and are thereforeindependent of socioeconomic assumptions,while production and Costs of Food are shown for thepovertyandprosperityscenario.

3.1 TheimpactofclimatechangeonbiophysicalindicatorsInall threeregionsunderconsideration(MiddleEastandNorthAfrica,SouthAsia,andSub‐SaharanAfrica),theaverageyieldoffoodcropsdecreaseswithclimatechangecomparedtonoclimatechange.Thebiophysicalprojectionsshowthatby2030averageyieldsoffoodcropswilldecreaseintheMiddleEastandNorthAfricabymorethan7%,inSouthAsiabymorethan5%,andinSub‐SaharanAfricabymorethan4%(Figure8,leftpanel).Not only at an average regional level but also locally yields predominantly decrease in the climatechange3scenario.ThecountrieswhicharemostaffectedareSudan,OmanandpartsofSaudiArabia,Somalia,EthiopiaandKenya.Significantyielddecreasesofmorethan7%canalsobefoundwithintheSahelzone.Only inBotswana,NamibiaandpartsofSouthAfricaandEthiopiadoaverage foodcropyieldsincrease(Figure5,upperpanel).

3Whenanalyzingyieldresults,oneshouldkeepinmindthattheseyieldsreflectcurrentlevelsofmanagementintensityandcurrentgeographicaldistributionofmanagedland.Thismeansthatcell‐specificchangesinyieldsdonotnecessarilycorrespondtocell‐specificchangesinlevelsofagriculturalproduction.

21

Figure5:Cellularaveragedifferenceinyieldsoffoodcropsandbiophysicalwateravailabilityin2030forRCP8.5comparedtoanoclimatechangescenario(climatechange‐noclimatechange).Positivevalues(green)indicatethatyields/wateravailabilityincreasewithclimatechange,whilenegativevalues(red)showthatyields/wateravailabilitydecrease.

Biophysicalwateravailability,whichisapreconditionforirrigatedproduction,decreasesmostinthewesternandcentralpartoftropicalAfrica,southoftheSahel,aswellasinthenorthofIndia.InEgypt,Ethiopia,Kenya,Tanzania,Somalia,aswellas inPakistan,wateravailability isprojected to increasewithclimatechange.Regionswhicharealreadywaterscarcearenotimpacted(Figure5,lowerpanel).

22

3.2 TheimpactofclimatechangeonagroeconomicindicatorsWe use MAgPIE here as a tool to explore future agricultural production and Costs of Food underdifferentclimateandsocioeconomicconditions.Duetotheuncertaintyininputdatasuchasyieldsandwateravailabilityaswellasmodellimitations(Section4.3),quantitativeresultsshouldbeconsideredwithcare.

3.2.1 Climate‐inducedchangesinagriculturalproductionInSub‐SaharanAfrica,averageregionalproductiondecreaseswithclimatechangeinthepovertyandprosperityscenarioby1%and2%,respectively.ProductiondecreasesintheSahelandsouthofitinboth scenarios. The increase in yields in Namibia leads to an increase in production in bothsocioeconomicscenarios,butproductionincreasesalsoinTanzaniawhereyieldsgodown.InMalawi,Zimbabwe and parts of the Democratic Republic of the Congo and Zambia as well as in Egypt,productionincreasesinthepovertyscenario,butdecreasesintheprosperityscenarios.Inmostpartsof Central Africa (Congo), there is no change in production because large parts are covered withtropical forests.Althoughyields increaseinBotswana,thelackofchangeinproductionisduetothelackofcropproductionwithandwithoutclimatechange.Total production in South Asia decreases only by 1% in the poverty scenario, but by 5% in theprosperity scenario. Although yields decrease everywhere in South Asia, production increases inPakistanandpartsofIndiainbothsocioeconomicscenarios,buttheproductionincreaseisstrongerinthepovertyscenario.ProductiondecreasesintherestofIndia.Overall production in Middle East and North Africa increase by 1% in the poverty scenario withclimatechangeandby6%intheprosperityscenario,buttheproducingcountriesdiffer.Agriculturalproduction increases inMorocco in thepovertyscenarioand there isamixedpatternofdecreasingandincreasinglocalproductionintheprosperityscenario.InLibyathereisnosignificantdecreaseinyieldsandproduction increases inbothscenarios. InEgyptyieldsdecreasewithclimatechangeandproductiondecreasesintheprosperityscenarioandincreasesinthepovertyscenario.InSaudiArabiayieldsdecreaseduetoclimatechange,butproductionincreasesintheentirecountryintheprosperityscenarioandinpartsofthecountryinthepovertyscenario(Figure6andFigure8).

23

Figure6:Cellularaveragedifferenceinproductionoffoodcropsforeach0.5°gridcellin2030forRCP8.5compared toanoclimatechange scenario (CC‐noCC)and twoSharedSocioeconomicPathways (SSP4,SSP5).Positive values (green) indicate thatproduction increaseswithCC,whilenegative values (red)showthatproductiondecreases.

3.2.2 AdaptationstrategiesaredifferentforthefocusregionsModel‐basedreactionstoclimate‐induceddecreasesinproductivityaredifferentforeachofthethreeconsidered world regions as well as for the different socioeconomic scenarios. In the prosperityscenario, the regions Sub‐Saharan Africa and South Asia may adapt by changing trade patterns(increasing imports and decreasing exports). Due to high trade barriers in the poverty scenario(Figure 23), expanding agricultural area is the prevailing strategy. In the poverty scenario, Sub‐Saharan Africa and South Asia use 9% and 7%more agricultural land compared to the no climatechangescenario,whileintheprosperityscenarioagriculturalareaincreasesonlyby5%inSouthAsiaand even decreases by 8% in Sub‐Saharan Africa (Figure 22 in the Appendix). The seeminglycontradictionbetweenadecreaseinproductioninthepovertyscenarioandanincreaseinproductionareaisaresultofthecrop‐specifictradebarriers.Evenif,asaresultofclimatechange,cropyieldsarevery low in this region, demand has to be fulfilled and production has to be adjusted. Adjusted

24

productionoflowyieldingcropswillthereforeleadtoanover‐proportionalexpansionofagriculturalarea.Notonlychangesincropyieldsbutalsothereductionofavailableirrigationwateraltersregionalproduction patterns. In Sub‐Saharan Africa, the decrease of available irrigation water leads to adecreaseofirrigatedproductioninthepovertyandprosperityscenarioby25%and29%respectively,while in SouthAsia irrigated production decreases by 3% for the poverty scenario and 5% for theprosperity scenario (Figure 21 in the Appendix). Both regions, but especially Sub‐Saharan Africa,where climate change has a partially positive impact, react to climate change in the model byconcentratingproductioninareaswhereyieldsarepositivelyormarginallyaffected,thusrelievingthepressureonareaswhereproductivityhasdecreased.In order to adapt to climate change, Middle East and North Africa cannot resort to expandingagriculturalarea,since the landpotentiallyavailable foragriculture isalreadyat its limit.Norcan itadaptbyincreasingimports,becausetradeconstraintsinthemodelarealreadybindingevenwithoutclimate change4. There are several crops5for which the domestic self‐sufficiency constraints arealreadybindinginthescenariowithoutclimatechange.Sustainingproductionofthesecropsdespitedecreasing yield levels in the model can only be achieved through massive investments in yieldincreasing technology. The decrease of yields therefore has to be completely compensated byimprovedmanagementandtechnologicalprogress.Requiredinvestmentsintechnicalchangeleadtoayieldincreaseof26%comparedtothenoclimatechangescenariointhepovertyscenarioandof20%in theprosperity scenario. Inourmodel implementation, technical change investments increase theyieldofall cropsproportionally.Thecropgroupwhich triggers themassive investment in technicalchangeistemperatecereals,whichisimpactedbystrongyielddecreases(9%comparedtoanaverageof7%)whileat thesametimerepresenting themost importantcrop in theregion(share inoverallproduction:35%inthepovertyscenarioand40%intheprosperityscenario),thusputtingenormouspressureonagriculturalsystems.The crop group profitingmost from the spillover effect of technical change investments is tropicalcereals. While biophysical yield is little affected by climate change (1%), the resulting yield aftertechnicalchangeismuchhigherinthescenariowithclimatechangecomparedtothescenariowithout(namely30%inthepovertyscenarioand40%intheprosperityscenario)6.Theresultingcomparativeadvantage of a high regional tropical cereals yield leads to an increase of production utilizedexclusivelyforexport.Sinceself‐sufficiencyconstraintsintheprosperityscenarioareonlybindingfortemperatecereals,agriculturalproductionisconcentratedevenmoreontropicalcereals,whileinthepovertyscenarioresourceshavetobeconcentratedintodomesticproduction.Moreover,productionshiftsfromirrigated(highyields)torainfed(lowyields)duetothedecreaseofwater availability in the Middle East and North Africa. This is visible in the decrease of irrigatedproductionof9%inthepovertyscenarioand6%intheprosperityscenario,aswellasintheincreaseof rain‐fed production of about 30% and 35% respectively in the poverty and prosperity scenario,comparedtothenoclimatechangescenario.

4SeeSection2.1.2foranexplanationofthemodelimplementationoftrade.5Thecropswithbindingself‐sufficiencyconstraintsinMiddleEastandNorthAfricainclude:temperatecereals,rice,maize,groundnut,oilpalm,sugarandcassava.6Thehigherrelativeyieldincreaseintheprosperityscenarioresultsfromthefactthatinthescenariowithoutclimate change less technical change investments were necessary resulting in a lower reference yield, theabsolute average yield is for each of the two climate scenarios higher in the poverty scenario than in theprosperityscenario.

25

3.2.3 Climate‐inducedchangesinCostsofFoodThe change in Costs of Food is spatially very diverse and varies also between the differentsocioeconomic scenarios. In thepoverty scenario,more countries suffer fromclimate‐inducedpriceincreasesthanintheprosperityscenario(Figure7).TheaverageregionalcostincreasesarehighestintheMiddleEastandNorthAfricainthepovertyscenario.Here,averageCostsofFoodincreaseby35%underclimatechangecomparedto17%intheprosperityscenario(Figure8).

Figure7:CellularaveragedifferenceinCostsofFoodin2030forRCP8.5andtwosocioeconomicscenarioscompared toanoclimatechangescenario(climatechange–noclimatechange).Positivevalues(red)indicatethatCostsofFoodincrease,whilenegativevalues(blue)showthatCostsofFooddecrease.

IncreasesinCostsofFoodofmorethan10US$/GJinbothsocioeconomicscenariosaffectMorocco,theverynorthernpartofAlgeria,Tunisia,Libya,Egypt,SaudiArabia,YemenandOman.InEgypt,CostsofFoodincreaseinthepovertyscenarioanddecreaseintheprosperityscenario.AverageCostsofFooddecreaseinSub‐SaharanAfricainbothscenarios,inthepovertyscenarioby2%andintheprosperityscenario by 17%. From a national perspective, Costs of Food decrease under climate change inMauritania,Botswana,Somalia,TanzaniaandZimbabweinbothsocioeconomicscenarios,butincrease

26

inthepovertyscenarioanddecreaseintheprosperityscenarioinSudan,partlyinthecountriessouthoftheSahel,Angola,Zambia,thesouthoftheDemocraticRepublicofCongoandMozambique.InSouthAsia,averageCostsofFoodincreaseinbothsocioeconomicscenarios,by12%inthepovertyscenarioand18%intheprosperityscenario.CostsofFoodincreaseinmostofIndiaaswellasinAfghanistanandpatternsarequitesimilarbetweenthepovertyandprosperityscenario(Figure7).

3.2.4 ClimatechangedoesnotnecessarilyleadtoincreasingCostsofFoodOurmodelresultsshowthatCostsofFooddecreaseunderclimatechange inSub‐SaharanAfricaonaverageby2%inthepovertyscenarioandby17%intheprosperityscenario.ThedecreaseofCostsofFoodinmanycountriesinSub‐SaharanAfricaoriginatesontheonehandfromachangeintradeflowsin the model, which leads to a decrease in overall production, and thus to a concentration ofproduction in the most profitable areas. On the other hand, Costs of Food decrease in the modelbecauseclimatechangehasapartiallypositiveimpactonyields.Ifproductionisshiftedtothesecrops(cassava,maize,sugarcaneandsunflower)aswellasmorefavorableareas,CostsofFooddecreases.TheincreaseofCostsofFoodinsomecountriesaswellastheloweraveragedecreaseofCostsofFoodinthepovertyscenariocomparedtotheprosperityscenarioresultsfromlessimports,whichleadstolargeragriculturalareasbeingcultivatedwithlowyieldingcropsandlesspossibilitytogrowthemostprofitablecrops.Due to higherpopulation growth andmore tradebarriers, average absoluteCosts of Food in SouthAsiaarealreadywithoutclimatechangehigherinthepovertyscenariothanintheprosperityscenario(24US$/GJinthepovertyscenario,21US$/GJintheprosperityscenario),resultinginahigherrelativeimpactofclimatechangeontheprosperityscenario(Figure8).Whileproductionof food cropsdecreasesunder climate change in themodel by1% in thepovertyscenario and 5% in the prosperity scenario due to an increase in imports and decrease of exports,CostsofFoodincreasebyasmuchas12%inthepovertyscenariocomparedto17%intheprosperityscenario.ThehighincreaseinCostsofFoodunderclimatechangeintheprosperityscenariodespitedecreasingproductioncanbeexplainedbythehighdemandfor livestockproducts intheprosperityscenario indevelopedregionssuchasNorthAmerica,Europe,Australiaand Japan.WhileexportsoffoodcropsdecreaseinSouthAsiaintheprosperityscenario,feedcropproductionofmainlysugarandsoybean, necessary for the export production of ruminant products, increase by 8% and 23%respectively, compensating thedecreasing cropyields inotherworld‐regions. In comparison, in thepoverty scenario exports of ruminant products are lower under climate change, since the negativeimpact of climate change cannot be compensated by trade, and food crop production replaces feedproduction.Despiteinvestmentsincost‐increasingtechnicalchange,climate‐inducedyielddecreasesarenotcompletelycompensated.Comparedtotherespectivescenariowithoutclimatechange(whereyieldsarealsoenhancedbytechnicalchange),theyare7.5%(povertyscenario)and9.5%(prosperityscenario)lower.Theloweryieldsleadinbothscenariostohigherfactorcostspertonne,resultinginhigherCostsofFood.

27

Figure8:Regionalaveragedifferenceinbiophysicalyieldsoffoodcrops(leftpanel),productionoffoodcrops and Costs of Food (COF) (right panel) for RCP8.5 compared to a no climate change scenario(climatechange ‐noclimatechange) fortwosocioeconomicscenarios in2030.(Regionnames:MNA=MiddleEastandNorthAfrica,SAS=SouthAsia,SSA=Sub‐SaharanAfrica.)

Accordingtooursimulations,MiddleEastandNorthAfricacompensatestheclimate‐induceddecreasein yields by investing in yield‐increasing technical change. Yield increase resulting from technicalchangeinvestmentsishigherinthepovertyscenariothanintheprosperityscenario,duetothehighself‐sufficiencyconstraintsofalmost100%andahighoverallfooddemandduetoalargerpopulation.In thepovertyscenario the increase inCostsofFood isevenhighersince technicalchangebecomesmore expensive when the yield level is already high, meaning that the additional necessary yieldincrease in the poverty scenario is over‐proportionally expensive. This leads to overall technicalchangeinvestmentswhichare103%higherinthepovertyscenarioand84%higherintheprosperityscenario compared to theno climate change scenario (Figure 24 in theAppendix). Thedecrease ofproductioninEgyptandtheresultingdecreaseinCostsofFoodintheprosperityscenarioisduetothefactthattradeislessrestrictedintheprosperityscenario,andproductioncanthereforebeshiftedtomoreproductiveareas.

3.3 AssessingagriculturalvulnerabilityunderclimatechangeInordertobetterunderstandtheimpactsofclimatechangeonpoverty,wecombinetheagriculturalindicatorCostsofFoodwiththeVulnerabilitytoHungerIndexprojectionsforthetwosocioeconomicscenarios (Figure 3). We use the hunger level categories defined by IFPRI (Table 1) and impactcategories for Costs of Food as defined in Table 4 in the following table and maps. Since 1GJcorrespondsto239,000kcal,aclimate‐inducedincreaseofCostsofFoodof0.1US$/GJ(definedasthelowest threshold for being impacted) correlates to an increase of 0.001US$ per 2390 kcal. SinceaccordingtoSmil(2001),2200kcalperdaysatisfythemetabolicenergyrequirementsofanaverageperson,whileanadditional400kcalareneededanddefinedasunavoidablewaste,the2390kcalcanbeinterpretedastherequirednumberofcaloriesperdayforapersontostayhealthy.Therefore,wecanapproximatelysaythatpeoplewillpayadditional0.001US$ormoreperdayfortheirfoodwhenimpactedbyclimatechangein2030andincludedinthelowestimpactcategory.Althoughthisseemsverylow,itmightbeanon‐negligibleamountofmoneyforapersonlivingonlessthan1.25US$perday.

28

Table4:DefiningthecategoriesforchangesinCostsofFoodduetoclimatechangein2030usedintheAgriculturalVulnerability Indicator.The tworight‐handcolumns show the samevalues,butconvertedintodifferentunits.

Impactcategory

Description ChangesinCostsofFoodinUS$/GJduetoclimatechange

ChangesinCostsofFoodinUS$perday(US$/2390kcal)duetoclimatechange

EI Extremelyhighimpact >10 >0.1

HI Highimpact 5‐10 0.05‐0.1SI Strongimpact 0.1‐5 0.001‐0.05NNI Nonegative

impact

29

Figure9:AgriculturalVulnerability Indicator,basedon thedifferences inCostsofFood (COF)betweenRCP8.5andanoclimatechangescenarioandcombinedwiththeprojectedVulnerabilitytoHungerIndexfor theyear2030 for the twosocioeconomicscenarios(map).Percentageshareofregionalpopulationaffectedbyclimate‐inducedincreasesinCostsofFood,exposuretohunger,andboth(barplot).(Hunger levels:EA=extremelyalarming,A=alarming,S= serious,LM=moderateand low,Climatechange impactcategories:EI=extremelyhigh impact,HI=high impact,SI=strong impact,NNI=nonegative impact; Region names:MNA =Middle East andNorth Africa, SAS = South Asia, SSA = Sub‐SaharanAfrica.)

3.3.1 Regionalandnationalresults:AgriculturalVulnerabilityIndicatorWhile Middle East and North Africa suffers, according to modelling results, most under theconsequences of climate change in 2030 by experiencing extremely high Costs of Food, a positiveeconomic development leads to an improved situation concerning hunger. People are even lessnegativelyimpactedintheprosperityscenario,duetoalowerincreaseofCostsofFood.Somepartsoftheregionsufferinthepovertyscenariounderahighincreaseinclimate‐inducedCostsofFood,suchasMoroccoandNorthernAlgeria,whiletheirVulnerabilitytoHungerisonlylowormoderate.InLibyaandEgypt,aseriousimpactonCostsofFoodiscombinedwithalowandmoderatevulnerabilitylevel.ThesituationimprovesformanycountriesinMiddleEastandNorthAfricaintheprosperityscenario

30

compared to thepovertyscenario, suchasLibyaandAlgeriawhere thenegative impactonCostsofFoodisalleviated.InSub‐SaharanAfrica,climatechangehasapartlypositiveimpactonagriculture,andduetoincreasedtrade production can be concentrated in themost productive areas, leading to even lower averageCostsofFood.Butas theeconomicsituation,according to theprojections,will improveonlyslowly,manypeoplewillstillsufferfromhunger.Sinceboth,theVulnerabilitytoHungerIndex,aswellastheclimate‐induced change in Costs of Food aremore favorable in the prosperity scenario than in thepovertyscenario, theoverallsituationregardinghunger is improved in theprosperityscenario.Themost severeconsequencesof climatechange inSub‐SaharanAfricacanbeobserved for thepovertyscenario in Sudan, Mozambique, parts of Zambia, Democratic Republic of Congo, Angola andMadagascar,wherepeoplewithextremelyalarmingVulnerability toHungerareextremely impactedby the climate‐induced changes in Costs of Food. The increase in Costs of Food is serious in thecountriesaroundtheSahelwhereVulnerabilitytoHungerisalarming.NoimpactofclimatechangeonCosts of Food aswell as lowVulnerability toHunger canbe seen in SouthAfrica. In theprosperityscenario,thecountriesinSub‐SaharanAfricaaremostlybetteroffthaninthepovertyscenario,partlyduetotheobservationthatclimate‐inducedimpactsonCostsofFooddisappearinalmosttheentireregion,exceptinsmallpartsofZambia,Angola,KeniaandEthiopia,butalsoduetotheloweroverallVulnerability to Hunger, which only remains alarming in Angola and the Democratic Republic ofCongo.In South Asia, Costs of Food will increase by 2030, but the negative effects cannot in all parts bealleviated by a positive economic development. Parts of the region will suffer from seriousVulnerabilitytoHungeraswellashighimpactsonCostsofFood.Inthisregion,differencesbetweenthe poverty and prosperity scenario are smallest. Vulnerability to Hunger in India and Pakistanbecomeslowerintheprosperityscenariocomparedtothepovertyscenario,buttheimpactonCostsofFood,whichpartiallystrongandhigh,remainsalmosteverywhereunchanged.

3.3.2 Regionalandnationalresults:NumberofpeopleaffectedTable 5 shows the number of people affected by climate‐induced changes in Costs of Food. Here,spatially explicit population projections for the poverty and prosperity scenario (Section 2.2.2) arecombinedwiththeresultsfromtheAgriculturalVulnerabilityIndicator.ThetableshowsthenumberofpeoplewhofallinoneofthethreemostnegativecategoriesforimpactsofclimatechangeonCostsofFoodandhungerlevelandomitspeoplewhichareeithernotnegativelyimpactedbyclimatechangeorfall intothelowesthungerlevelcategory. Inthethreeregionscombined,2320Mpeopleseriouslyvulnerable tohungerarealsostronglyaffectedbyclimate‐induced increasesofCostsofFood in thepovertyscenario.Bycontrast,intheprosperityscenariothisnumberisreducedto1448Mpeopleforseveral reasons. There are less people in total, average incomes are higher, a smaller share of thepopulationisvulnerabletohunger,andnegativeclimateimpactscanbebettercompensatedbytrade.

31

Table5:Numberofpeople(inmillion)affectedbyclimate‐inducedincreasesinCostsofFood(differencebetweenRCP8.5andano climate change scenario) sortedbydifferent impact categories (rows)andExposuretohungerfordifferentVulnerabilitytoHungerIndexcategories(columns)in2030,regionallyforapovertyandaprosperityscenario.

Poverty ProsperityVulnerabilitytoHunger Serious Alarming

Extremelyalarming Sum Serious Alarming

Extremelyalarming Sum

Climate‐inducedchangesinCostsofFood

Sub‐SaharanAfrica

Strongimpact 205 224 28 457 164 8 3 175Highimpact 9 14 0 23 26 16 0 42Extremelyhighimpact 13 81 16 110 33 2 0 35Sum 227 319 44 590 223 26 3 252

MiddleEastandNorthAfrica

Strongimpact 0 0 0 0 0 0 0 0Highimpact 0 0 0 0 1 10 0 11Extremelyhighimpact 21 42 0 63 4 22 0 26Sum 21 42 0 63 5 32 0 37

SouthAsia

Strongimpact 767 228 3 998 610 3 0 613Highimpact 573 1 0 574 504 0 0 504Extremelyhighimpact 94 1 0 95 42 0 0 42Sum 1434 230 3 1667 1156 3 0 1159InMiddleEastandNorthAfricainthepovertyscenario,63Mpeopleor12%oftheregionalpopulationareexposedtoatleastserioushungerandareextremelyhighimpactedbyincreasesinCostsofFood.Intheprosperityscenario,lifeimprovesforpeoplebybeinglessimpactedbyclimatechange,where26Mpeopleareextremelyhighimpactedand11Mpeoplearehighlyimpacted,aswellas forpeoplevulnerabletohunger,since49Mpeoplelessarevulnerabletoatleastserioushungerintheprosperityscenariocomparedtothepovertyscenario.Consequently,theimpactofclimatechangeonpeopleatriskofsufferingunderatleastserioushungerisreducedto37Mpeopleor8%ofthepopulationintheprosperityscenario.Thisis incontrasttothehighnumberof443MpeopleimpactedbyincreasesinCosts of Food in the poverty scenario and 371M in the prosperity scenario. The relatively smallconcurrenceofpeoplevulnerabletohungerandbeingimpactedbyclimatechangeisduetopositiveeconomicprojectionsforthisregion.A large share of the population will be at risk of suffering at least serious hunger in Sub‐SaharanAfrica:90%ofthepopulation(1267Mpeople)inthepovertyscenarioand,duetopositiveeconomicdevelopment,60%(903Mpeople)intheprosperityscenario.Onlyapartofthesepeoplewillalsobeaffectedbynegativeclimatechange,namely590Mpeopleinthepovertyscenarioand252Mpeopleinthe prosperity scenario. This is due to the relatively good ability of the region to adapt to climatechange,mainly by shifting production tomarginally or positively affected areas. Unfortunately, thiswill only work for the short‐time horizon until 2030. In the long run (until 2080), most of thepositivelyaffectedareasin2030willalsosufferfromdecreasingyields(Figure31intheAppendix).

32

InthepovertyscenarioinSouthAsia,81%ofthepopulationliveinregionswhereCostsofFoodareatleast strongly impacted by climate change andVulnerability toHunger is serious orworse (1667Mpeople).Thisnumberofpeopleisreducedto1159Mpeopleor59%ofthepopulationintheprosperityscenario, meaning that life improves for 508M people through better economic development andliberalizedtrade.Thedecisiveimprovementisrelatedtoareductionofpeopleexposedtohungerfrom1990Minthepovertyscenarioto1225Mpeopleintheprosperityscenario,whilepeopleaffectedbyatleaststrongincreasesinCostsofFoodareonlyreducedfrom1627Mpeopleinthepovertyscenarioto1623Mintheprosperityscenario.Infact,mostpeopleinthepovertyscenarioareinthelowestimpactcategory, leavingasmallernumberofpeoplehighly impacted,while intheprosperityscenario23%morepeoplearehighlyimpactedthaninthepovertyscenario. Intheglobalizedprosperityscenario,SouthAsiaproduceslivestockandfeedcropsforexportintodevelopedregionsthusputtingpressureondomesticCostsofFood.If we confine our focus and look only at the two most extreme categories for the AgriculturalVulnerability Index we can identify the people whose daily Costs of Food increase by more than0.05US$ andwhich suffer under an at least alarming Vulnerability toHunger. ForMiddle East andNorthAfricatheseare8%ofthepopulationinthepovertyscenarioand7%intheprosperityscenario.InSub‐SaharanAfricathedifferencebetweensocioeconomicscenariosisremarkable,with8%ofthepopulation belonging into the two extreme categories in the poverty scenario and 1.5% in theprosperityscenario.InSouthAsiaonly0.1%ofthepopulationisconcernedinthepovertyscenario.

3.4 Adaptationoptions:TechnologicalprogressandimprovedmarketaccessNegative climate impacts on agriculture and Costs of Food may be counterbalanced by potentialadaptationmeasures.Weinvestigateheretwodifferentoptions,namelyinvestmentsintoagriculturalR&D,resultinge.g.inbetteradaptedcropvarietiesorimprovedmanagementsystems,andimprovedmarketaccessfacilitatedbyestablishinginfrastructure,enablingfarmerstobetterreachmarkets.Weshowhereonlyresultsthepovertyscenario,inordertotesttheadaptationpotentialoftheproposedmeasures.Theseadaptationoptionsaredifferent fromtheonesdiscussed insection3.2.2 insofarasthey are not model‐inherent mechanisms necessary to generate sufficient food production. Theyratherserveasasensitivityanalysisforhowaspecificmeasurecouldbufferclimatechangeimpacts.While technological progress as an adaptation option is already implemented in the model, theeffectivenessofinfrastructureimprovementisanalyzedhereforthefirsttime.3.4.1 Results:TechnologicalprogressIntheslowtechnologicalprogressscenario,foodcropproductiondecreasesinthethreeregionsunderconsiderationandCostsofFoodincreasewithclimatechangeinallcases. InSub‐SaharanAfrica,therelative increase inCosts of Food in the slow technological progress scenario is highestwith250%(40US$/GJ)comparedtothenoclimatechangescenario,MiddleEastandNorthAfricasuffersmostinabsolute terms with Costs of Food increasing by 70US$/GJ (80% in relative terms). The fasttechnologicalprogressscenario,ontheotherhand,increasestheabilitytoadapttoclimatechange.InMiddleEastandNorthAfricaCostsofFoodincreaseonlyby12%comparedto37%inthedefaultcaseInSub‐SaharanAfrica,CostsofFoodevendecreasesinthefasttechnologicalprogressscenario(Figure10).

33

Figure10:RegionalpercentagechangesinfoodcropproductionandCostsofFoodforRCP8.5comparedtoano climate change scenario (climate change ‐no climate change) for the year2030, thepovertyscenarioand two technologicalprogress scenarios.Reference (ref) refers to thepoverty scenariowithmedium technological progress settings. (Region names:MNA =Middle East andNorthAfrica, SAS =SouthAsia,SSA=Sub‐SaharanAfrica.)

3.4.2 FasttechnicalprogresscanalleviatenegativeclimatechangeimpactsThedecreaseofproductionwithclimatechange inalmostallscenariosresults fromyielddecreaseswhicharecompensatedbyincreasesinimportsanddecreasesinexports7.TheenhanceddifficultyofadaptingtoclimatechangewhenR&Disexpensiveisreflectedbythestrongerdecreaseofproductionintheslowtechnologicalprogressscenariocomparedtothereferenceandfasttechnologicalprogressscenario in all regions.While slow technologicalprogress is veryproblematic forour focus regions,fasttechnologicalprogressdoeshavelittleimpactonproduction(SubSaharanAfricaandSouthAsia)oreven reducesproductioncompared to the referencecase (MiddleEastAsia andNorthAfrica). Inthese scenarios it is cost minimizing in the model to rather increase production in regions whereadditionalyieldincreaseisrelativelycheaperandmoreefficient.Although decreasing production in the slow technical progress scenario lowers overall productioncosts,CostsofFoodincreasesinceclimate‐inducedloweryieldshavetobecompensatedbycostlylandexpansion,higherfactorinputperton,andnecessaryinvestmentsinexpensivetechnicalchange.ThehighincreaseinCostsofFoodintheslowtechnologicalprogressscenariowithhightechnicalchangecostsinMiddleEastandNorthAfricacanbeeasilyunderstoodwhenreconsideringthatMiddleEastandNorthAfricaistheregionwhereinthemodelneitheragriculturalexpansion,norchangesintradeareanadaptationoptionandthattheregionisthereforerelyingmostlyonyield‐increasingtechnicalchange to sustain production.While overall production costs increasewith climate change in bothscenariosinMiddleEastandNorthAfricaaswellasSouthAsia,resultinginincreasingCostsofFood,theydecreaseinSub‐SaharanAfrica(Figure24intheAppendix).Thedecreaseinproductionandtheresultingdecreaseofoverallproductioncostsof1.8%inSub‐SaharanAfricaaresufficienttoleadtoadecreaseofCostsofFood in the fast technicalprogressscenario.Although in theslowtechnologicalprogressscenario,overallcostsdecreaseby2.6%andproductionby6.4%,CostsofFoodincreasewith7AnextensivediscussionontheincreaseofproductioninthereferencescenarioinMiddleEastandNorthAfricacanbefoundinsection3.2.2.

34

climatechange.Thiscanbeattributedtothelowinvestmentinexpensivetechnicalchange,leadingtoalowyieldrelativetothefasttechnologicalprogressscenario.Inordertocompensateyielddecreasesduetoclimatechangeandfulfill fooddemand,agriculturalareahastobeexpandedby4Mha,whichincreases production costs and consequentially Costs of Food. The decrease in production iscompensated by imports (Figure 23 in the Appendix). Maps with cellular explicit changes inproduction aswell as Costs of Food can be found in the Appendix (Figure 25 andFigure 26 in theAppendix).

3.4.3 Results:MarketaccessChanges in market access for farmers through manipulating transportation costs impact the wayeconomic regions react to climate change. Production decreases or stays constant under climatechangeinthebadmarketaccessscenarioandincreasesoronlyslightlydecreasesinthegoodmarketaccessscenario.InMiddleEastandNorthAfrica,andSouthAsia,CostsofFoodincreasewithclimatechangeinallscenarios,butincreasemoreinthegoodmarketaccessscenariothaninthebad.InSub‐SaharanAfrica,CostsofFooddecreasewithclimatechange in thegoodmarketaccessscenarioandincreaseinthebadmarketaccessscenario.

3.4.4 BettermarketaccesshelpsproducersbutmighthurtlocalconsumersTransportationcostsdecreaseinthegoodmarketaccessscenarioandthuscanmakefoodproductionprofitable even when the next market is relatively far away. Developing countries profit over‐proportionallyfromlowtransportcostsduetoabadlyfunctioninginfrastructureinthereferencecase.Hence,theresultingcomparativeadvantageaswellasthenegativeimpactofclimatechangeonglobalcropyieldsprovides the incentive to invest inR&Dtoboostyields.Thiseffect leads forMiddleEastand North Africa and for South Asia to an increase in Costs of Food, since in order to increaseproduction,overalltechnicalchangeinvestmentsaswellasoverallcostsforfactorrequirements,landconversionandtransporthavetoincrease.Whileproducerswillprofitfrombettermarketaccess,thesituation for consumers may worsen. In Middle East and North Africa, agricultural area actuallydecreases in thegoodmarketaccessscenariosalthoughproduction increases.This isbecausesomeareascannotbeusedatallforagriculturalproductionunderclimatechange,soareahastodecrease,but it decreases less compared to the bad market access scenario, since some new and moreproductiveregionscannowbeutilized.InSub‐SaharanAfrica,productionandCostsofFoodalreadydecreaseinthedefaultcaseunderclimatechange, as explained above, but with lower transport costs production becomes more profitable,meaningthatproductiondecreasesbylessthan0.5%comparedto1%inthedefaultcase.Thestrongdecrease inCostsofFoodcanbeexplainedbythe fact thatSub‐SaharanAfrica isaregionwithhightransportcostsinthereferencecase.Improvingmarketaccessthereforeprovidestheopportunitytoshiftproductionprofitablytomoreproductiveareas.Figure28intheAppendixshowsthatproductionshiftsfromBotswanaandAngolatoNamibia,whichisoneofthecountriesprofitingmostfromclimatechangein2030intermsofyieldincreases(Figure5).

35

Figure11:RegionalpercentagechangesinfoodcropproductionandCostsofFoodforRCP8.5comparedtoanoclimatechangescenario(climatechange‐noclimatechange)fortheyear2030,forthepovertyscenarioand twomarketaccess scenarios.Reference (ref) refers to thepoverty scenariowithmediummarketaccess.(Regionnames:MNA=MiddleEastandNorthAfrica,SAS=SouthAsia,SSA=Sub‐SaharanAfrica.)

Whiledevelopingregionsprofitover‐proportionally fromthegoodmarketaccessscenario, theyarealsoover‐proportionallynegativelyimpactedbythebadmarketaccessscenario.Inthemodel,MiddleEastandNorthAfricadoesnothavethepossibilitytodecreaseproductionunderclimatechangeduetothelackofanyotheradaptationoption.Therefore,itinvestsintechnicalchangetoincreaseyields,thusleadingto increasesinCostsofFood.SouthAsiaandSub‐SaharanAfricahavethepossibilitytoadapttoclimatechangebyincreasingimportsanddecreasingexports,butsincetradeisnotsufficientto compensate yield decreases, production partly has to be sustained through either investing intechnicalchangeorbyexpandingagriculturalarea,thusincreasingCostsofFood.Mapswithcellularexplicit changes in production as well as Costs of Food can be found in the Appendix (Figure 28andFigure29).

3.5 TheglobalperspectiveAlthough ‐ from a poverty perspective ‐ the three decisive regions have been covered, due to tightlinkagesbetweenworld regions8, the effectsof climate changeonyields andCostsof Food inotherregionsaswellastheeffectsofchangesintradearedecisive.

3.5.1 GlobalimpactofclimatechangeonCostsofFoodWhile all regions, except for North America, are negatively impacted by climate change throughdecreases of average crop yields, the level differs. In Pacific OECD and Pacific Asia, average yieldsdecreasebymorethan8%,whileinEurope,theFormerSovietUnionandLatinAmericadecreasesarelessthan3%.IntermsofCostsofFood,theimpactisverydiversebetweenregionsandsocioeconomicscenarios.TheFormerSovietUnionandLatinAmericaarerelativelylittleimpacted;inEurope,NorthAmerica, Pacific OECD and Pacific Asia in the poverty scenario, Costs of Food even decrease under8ThesevenglobalregionsbesidethefocusregionsarenowlabelledwithMAgPIEregionnames,sinceWorldBankregionsonaglobalscalearenotcomparable.AdefinitionoftheMAgPIEregionscanbefoundinFigure1.

36

climate change.Centrally‐plannedAsia is evenworse impactedby increase