ScienceoftheTotalEnvironment624(2018)790–806 ContentslistsavailableatScienceDirect Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv Spatio-temporal variations in climate, primary productivity and efficiency of water and carbon use of the land cover types in Sudan and Ethiopia MuhammadKhalifaa,b,⁎,NadirAhmedElagiba,LarsRibbea,KarlSchneiderb aInstituteforTechnologyandResourcesManagementintheTropicsandSubtropics(ITT),TechnischeHochschuleKöln-CologneUniversityofAppliedSciences,Cologne50679,Germany bDepartmentofGeography,UniversityofCologne,Albertus-Magnus-Platz,D-50923Cologne,Germany H I G H L I G H T S G R A P H I C A L A B S T R A C T • Interactionbetweenclimateandvege- tationecosystemsiscomplexandstill unclear. • Correlatingclimatevariabilityandvege- tationproductivityprovidesusefulin- sights. • Spatio-temporalvariationsinclimate, NPP,WUEandCUEusingremotesens- ingdata • NPP,WUEandCUEvarywidelyamong landcoversindifferentclimatecondi- tions. • Thisvariationshouldbeconsideredin policy-making for water and food security. a r t i c l e i n f o a b s t r a c t Articlehistory: TheimpactofclimatevariabilityontheNetPrimaryProductivity(NPP)ofdifferentlandcovertypesandthere- Received9October2017 actionofNPPtodroughtconditionsarestillunclear,especiallyinSub-SaharanAfrica.Thisresearchutilizespub- Receivedinrevisedform7December2017 lic-domaindatafortheperiod2000through2013toanalyzetheseaspectsforseverallandcovertypesinSudan Accepted7December2017 andEthiopia,asexamplesofdata-scarcecountries.Spatio-temporalvariationinNPP,wateruseefficiency(WUE) Availableonline27December2017 andcarbonuseefficiency(CUE)forseverallandcoverswerecorrelatedwithvariationsinprecipitation,temper- atureanddroughtatdifferenttimescales,i.e.1,3,6and12monthsusingStandardizedPrecipitationEvapotrans- Editor:D.Barcelo pirationIndex(SPEI)datasets.WUEandCUEwereestimatedastheratiosofNPPtoactualevapotranspirationand Keywords: NPPtoGrossPrimaryProductivity(GPP),respectively.ResultsofthisstudyrevealedthatNPP,WUEandCUEof NetPrimaryProductivity thedifferentlandcovertypesinEthiopiahavehighermagnitudesthantheircounterpartsinSudan.Moreover, Remotesensing theyexhibithighersensitivitytodroughtandvariationinprecipitation.Whereassavannahrepresentsthe Climatevariability mostsensitivelandcovertodrought,croplandsandpermanentwetlandsaretheleastsensitiveones.The Wateruseefficiency inter-annualvariationinNPP,WUEandCUEinEthiopiaislikelytobedrivenbyadroughtoftimescaleof Carbonuseefficiency threemonths.NostatisticallysignificantcorrelationwasfoundforSudanbetweentheinter-annualvariations Drought intheseindicatorswithdroughtatanyofthetimescalesconsideredinthestudy.Ourfindingsareusefulfrom theviewpointofbothfoodsecurityforagrowingpopulationandmitigationtoclimatechangeasdiscussedin thepresentstudy. ©2017ElsevierB.V.Allrightsreserved. ⁎ Correspondingauthorat:InstituteforTechnologyandResourcesManagementintheTropicsandSubtropics(ITT),TechnischeHochschuleKöln-CologneUniversityofApplied Sciences,Cologne50679,Germany. E-mailaddress:[email protected](M.Khalifa). https://doi.org/10.1016/j.scitotenv.2017.12.090 0048-9697/©2017ElsevierB.V.Allrightsreserved. M.Khalifaetal./ScienceoftheTotalEnvironment624(2018)790–806 791 1.Introduction theAfricancontinent(Williamsetal.,2007).Monitoringthevariation invegetationproductivity,WUEandCUE,andcorrelatingthisvariation NetPrimaryProductivity(NPP)isdefinedastheamountofatmo- withclimatevariabilityforlargeareasisachallenge,particularlyagainst sphericcarbonthatiscapturedbyplantsandtransformedintobiomass thebackdropofthegivenlimitationsofgrounddata,especiallyinthe (ZhaoandRunning,2010).Thetotalamountcorrespondingtothepho- countriesofSub-SaharanAfrica,forinstance,wheregroundweather tosynthesisprocessiscalledGrossPrimaryProductivity(GPP).Thedif- stationsarefewandsparselydistributed. ferencebetweenNPPandGPPisreferredtoasrespiration(Ardö,2015), WUEandCUEareusefulindicatorsfortheassessmentofthepattern whichistheamountofcarbonpreviouslyassimilatedbytheplantand ofwateruseandcarbonsequestrationbyplants.WUEisdefinedas“The subsequentlyusedformaintenanceofthebiomassorforgrowth.Mon- rateofcarbonuptakeperunitofwaterlost”(Tangetal.,2014).Itcanbe itoringofthevariabilityinprimaryproductioniscriticalduetothefact calculatedbydifferentapproachesaccordingtothepurposeofinvesti- thatNPPprovidesvitalservicesforhumansurvival(Ardö,2015).Re- gation(ItoandInatomi,2012).Here,WUEisdefinedastheamountof ductioninNPPpotentiallyjeopardizesfoodsecurityandmayincrease waterevaporatedforeverygcarbon/m2ofNPPproduced,i.e.NPP/ET, a globalwarmingsinceareductioninNPPmightdecreasetheavailable where ET is theactual evapotranspiration (Kuglitschet al.,2008). a carbonsinks(ZhaoandRunning,2010).Whereasthespatialvariation CUEisdefinedasaratioofNPPtoGPP.Theearliestconceptionofthe of NPP depends on vegetation type, soil, climate conditions and CUE is that it ideally equals 0.5 (DeLucia et al., 2007; Zhang et al., humanactivities,thetemporalvariationofNPPdependsmostlyon 2009).However,CUEshouldnotbeconsideredasaconstantvalue thevariabilityofclimaticfactors(Lietal.,2016).Severalclimatefactors since the driving processes of photosynthesis and respiration are controltheNPP,suchastemperature,precipitationandshortwavesolar nonlinearlygovernedbydifferentenvironmentaldrivers;thus,the radiation(Lietal.,2016).Temperatureandprecipitationhavemorein- ratioofthesefluxesvaries.Photosynthesisandrespirationareprimarily fluenceonNPPinaridandsemi-aridareaswhereassolarradiationisthe governedbytheabsorbedphotosyntheticallyactiveradiation(APAR) main controllingfactorin humid andsemi-humid areas(Liu et al., and temperature, respectively. Several researchers noted that CUE 2015).TemperatureplaysaroleinraisingNPP(ZhaoandRunning, mightvarydependingonclimatefactors(e.g.precipitationandtemper- 2010).Althoughtheperiodfrom2000through2009wasthewarmest ature)andgeographicallocation(Zhangetal.,2009).Plantsareconsid- decadeintherecordssince1880(NOAA,2016),ZhaoandRunning eredcarbonsinks.ExaminingthevariabilityofCUEisthusimportantfor (2010)foundthatglobalNPPhasdeclinedby0.55petagramcarbon climatechangeandCO emissionsstudies(Chenetal.,2013).Also,a 2 (PgC)duringthesamedecade.Theysuggestedthatadryingtrendin betterunderstandingofWUEandCUEmayleadtoabettermanage- thesouthernhemispherewasthemaindriverforthisreduction.There mentofecosystems(Tangetal.,2014;Zhangetal.,2009). hasbeenadebateregardingthesefindingsastowhethertherewasa Variabilityinprimaryproductivityaffectsfoodavailabilityandfood decreaseinNPPorwhetherthisdecreasewasduetoartifactsfrom security.Atthesametime,themagnitudeofprimaryproductionstrong- theappliedmodel(Samantaetal.,2011;ZhaoandRunning,2011).If lyaffectsthecarboncycle(Zhaoetal.,2005).However,lackofcontinu- NPPisaffectedbyclimateassuggestedinthatmodel;then,NPPshould ousgroundobservationhindersthelong-termanalysisofthedynamics havedecreasedduringthisperiod(Medlyn,2011). ofvegetationdevelopment.Luckily,manyremotesensingormodel Ecosystemsdifferintheirresponsestoclimatevariability(Knapp basedpublic-domaindatasourcesnowadaysprovidecontinuousspatial andSmith,2001).Differentplantspeciesresponddifferentlytodrought climateobservationsandotherkeyenvironmentalvariables(e.g.LAI, conditionsbasedontheirphysiologicalandstructuralcharacteristicsin biomass,soilmoisture).Thegeneralavailabilityofthesepublic-domain ordertopreventlossofwater(VanDerMolenetal.,2011).Understand- datasourcesprovidesauniqueopportunityforexaminingtheclimate- inghowlandcovertypesrespondtodroughtandtoclimatevariability plantproductivityrelationship,whichisessentialforareproducible canpromotemoreefficientmanagementoftheselandcoversandcan, analysisoftheimpactofclimatevariationonprimaryproductivity. inturn,assistsignificantlyinsecuringwaterandfoodinthefuture. However,incomparisonwithclimaticvariables,onlyfewpublic-do- Manystudieshavebeenconductedonthismatter,yetmostofthem maindatabasesprovidedataonprimaryproductivity.MODISdata, havefocusedontheclimatedriveroftheNPPvariationonaglobalscale suchastheprimaryproductivity(MOD17)product,areoftenusedto (e.g.Huangetal.,2016;Liuetal.,2015).Recentanalysesshowedthat detectthevariabilityofprimaryproductivityaswellasanalyzingthe semi-aridareasareamaincontrolofglobalNPPvariations(Huanget WUEandCUEoflandcovertypesorentireecosystemsandtheirassoci- al.,2016;Ahlströmetal.,2015).However,itisimportanttoexamine ationwithclimateconditions.Forasummaryofthemostimportant whetherthesamepatternofNPPfoundbyZhaoandRunning(2010) studies,refertoTable1. onaglobalscalealsoappliesonregionalandlocalscales(Chenetal., Tothebestofourknowledge,apartfromthestudyofPengetal. 2013).Asthescalingoftherelevantprocessesistypicallynon-linear,dif- (2017)whostudiedtheimpactofdroughtonNPPonacountryscale ferentpatternsarelikelytoemergeattheregionalorlocalscale.Drought forthewholeglobe,nootherstudywasconductedtoanalyzethere- isexpectedtobemoresevereinthefuture(Aultetal.,2014).Therefore,it sponseofNPP,WUEandCUEofdifferentlandcovertypestodrought isimportanttoinvestigatetheeffectofdroughtonprimaryproductivity andclimate variability in Sub-SaharanAfricabasedupon generally andefficiencyofthelandcovertypesintermsofwaterandcarbonuse. availabledatasets.Inthisstudy,inter-annualvariationsinclimatecon- Analysisoftheseinteractionsatregionalandnationallevelsisuseful ditionsanddroughtondifferenttimescalesfortheperiodfrom2000 andprovidesessentialinformationforlandcovermanagementandcli- through 2013 were correlated with inter-annual variation in NPP, matepolicy-making(Pengetal.,2017;Liuetal.,2015).Recently,acoun- WUEandCUEinvariouslandcovertypesinSudanandEthiopia.The try-scaleanalysisoftherelationshipbetweenNPPanddroughtwas twoselectedcountriesareexamplesofSub-Saharancountrieswithse- publishedbyPengetal.(2017).ThedatausedintheirworkwereModer- veredata-scarcityandhighvulnerabilitytoclimatevariationandfood ateResolutionImagingSpectroradiometer(MODIS)NPPandthedrought insecurity. Understanding the spatio-temporal variability of NPP, StandardizedPrecipitationEvapotranspirationIndex(SPEI).Theyfound WUE,andCUEisessentialtoachieveabetterlandmanagementin thatcountriesshowdifferenttrendsinNPPfortheperiod2000to2014, thesecountriesandtoimprovefoodsecurity. andonly35countriesaccountedforN90%oftheglobalNPP. Onthecontinentalscale,AfricahaswitnessedanincreaseinNPP 2.Materialsanddata duringthesameperiod,i.e.from2000through2009,by0.189PgC. Thiswasmostlyduetoadecreaseinvaporpressuredeficit(Zhaoand 2.1.Studyareaanditsimportance Running,2010).Africanecosystemsproducearound20%ofthetotal globalNPP(Ciaisetal.,2011),andalargefractionoftheinter-annual EastAfricaisoneofthemostchallengingareasformanagingnatural variabilityintheglobalcarboncycleisduetoecologicalprocesseson resourcesduetomanyfactors.Itisaregionhighlyvulnerabletoclimate 792 M.Khalifaetal./ScienceoftheTotalEnvironment624(2018)790–806 Table1 ReviewofsomeresearchusingMOD17dataofNPPandGPP. Author(s) Scale Time Mainobjective(s) Mainfindings period Huanget Global 2000–2013 ToexaminetheimpactofdroughtonNPP NPPishighlycontrolledbydrought,andsemi-aridecosystemsplaythemost al.,2016 importantroleintheinter-annualvariabilityontheglobalscale. Lietal., Global 2000–2014 TostudytheclimatefactorsaffectingNPPvariabilityand NPPiscorrelatedpositivelywithETo,anditrespondsdifferentlyinthe 2016 feedbackofthisvariabilityonactualevapotranspiration northernandsouthernhemispheresaccordingtodominantclimatefactorsin (ETo). eachhemisphere. Ahlström Global Tofigureouttheroleofsemi-aridecosystemsinthe ThetrendandvariabilityofthegloballandCO2sinksarelargelyderivedby etal., trendandvariabilityoflandCO2sinks. variationintemperatureandprecipitationvariationoccurringoversemi-arid 2015 ecosystems. Ardö,2015 Africa 2000–2010 Tocompareprimaryproductiondatafromremote GPPestimationsderivedfromremotesensingdata(i.e.MOD17)arehigher sensinganddynamicvegetationmodels. thanthosederivedfromdynamicvegetationmodels,whileNPPestimations arelower.Whenvalidatedagainstground-baseddata,bothestimationsshow significantpositivecorrelation. Liuetal., China 2000–2011 ToassesstheWUEofecosystemsandtheirresponseto DroughthasanimpactonWUE,andtheresponseofWUEtodroughtvaries 2015 drought amongecosystemtypes,geographiclocationsandclimateconditions. Abdietal., Sahel 2000–2010 ToestimateandanalyzethesupplyanddemandofNPP ThedemandofNPPincreasedonanannualrateof2.2%,butwithanear 2014 region, intheSahelcountries. constantsupply.Themajorincreaseindemandisforfoodrequirement. Africa Tangetal., Global 2000–2013 ToinvestigatetheWUEofdifferentecosystemsandto WUEvariedgreatlyamongecosystemtypesandamongecosystemslocatedin 2014 studytheirvariationandtrends. differentclimatezones.Recentchangesinlandcoverledtodeclineinglobal WUE. Zhanget Lower 2000–2011 Toassesstheeffectofdroughtonvegetation Droughtswithvariedintensitieshavedifferentimpactsonecosystems,which al.,2014 Mekong productivity showvariationinresponsetodrought. Basin Chenetal., Global 1997–2009 ToanalyzetheimpactofdroughtonNPP NPPanddroughtarepositivelycorrelatedinaridregions,whereasboreal 2013 (sub-arctic)areasshownegativecorrelation,andsomeareasshowno correlation. Zhaoand Global 2000–2009 TodetectthetrendofNPPanditsrelationtodrought AglobaldecliningtrendintheaverageNPPisdetectedduringthe Running, investigationperiod.Themaindrivingforceofthisdeclineisdrought. 2010 Zhanget Global 2000–2003 ToinvestigatethepatternofCUE(GPP/NPP)indifferent CUEvariesconsiderablybasedonecosystemtype,geographicallocationand al.,2009 ecosystems,geographicallocationsandclimate climateconditions. conditions. Turneret Global 2000–2004 ToevaluatetheperformanceofMOD17productsacross MOD17productsprovidegooddatatodetectthegeneraltrendofprimary al.,2006 differentbiomesandcompareitto9Eddycovariance productivity,butshowsoverestimationsinlowproductivitylocationsandvice fluxtowersdata. versa. Fig.1.LocationmapofEastAfricashowingtheboundariesofthetwocasestudies(SudanandEthiopia)andthedifferentlandcovertypeslocatedintheregion.Landoverdatainthismap arethatofMCD12Q1product. M.Khalifaetal./ScienceoftheTotalEnvironment624(2018)790–806 793 changeimpacts(Abebe,2014).Moreover,mostofthecountriesinthis fromMODISsatellite.MODISisoneofthesensorsonNASA'sEarthOb- regionareconsideredamongtheleastdevelopedcountries,witha servingSystem(EOS)satellites.Itprovidescontinuousglobalmonitor- highandrapidlyrisingpopulation(UNECA,2016),consequentlyput- ingdataofprimaryproductivitywithaspatialresolutionof1kmand tingmorepressureonthenaturalresourcesinthefuture.Withatotal at a temporal resolution of 8-day, monthly and annual intervals. areaofabout3millionkm2,SudanandEthiopiaareagoodexampleof MOD17version055dataofannualestimatesofNPPandGPPwere landcovertypestypicalforthisregion(Fig.1).Thetwocountriesto- downloaded from the Numerical Terradynamic Simulation Group getherarecharacterizedbyagreatdiversityinlandcovers,includingsa- website.Invalidvalueswereremovedfromtherasterfiles;then,each vannah, permanent wetlands, croplands, shrublands, and forests. rasterfilewasmultipliedbyascalefactorof0.1torestoretheoriginal Accordingly,theyshowsignificantspatialvariationinclimatecondi- NPPandGPPvalues,asinstructedinthemetadatafileofthisdataset. tions,rangingfromhyper-aridinnorthernSudantothehumidcondi- Lastly,using“extractbymask”toolinArcGIS,separaterastertimeseries tionsinsomepartsoftheEthiopianhighlands.Thesefeaturesmake ofNPPandGPPwereproducedforeachofthetwocountries,i.e.Sudan thetwocountriesparticularlysuitableforourresearch.Mostofthe andEthiopia. area in Sudan is bare or sparsely vegetated land (62% of the total WithlimitedavailablegrounddataincomparisonwithMOD17data, area).Theseareasaremainlylocatedinthenorthernhalfofthecountry validationofMODISdataisachallenge(Zhaoetal.,2005).Numerous (Fig.2).InSudan,grasslandisthedominantlandcovertype,covering studieswereconductedtocomparetheMOD17datawithfieldmea- around20%ofthetotalarea.InEthiopia,openshrublandsrepresent surements.Adetailedoverviewoftheperformanceofthisproductisbe- thelargestlandcovertype,coveringapproximately27.6%ofthetotal yondthepurposeofthisresearch.Nevertheless,itisworthmentioning area.Woodysavannah,grasslandsandcroplandsalsorepresentimpor- forinstancethatTurneretal.(2006)usedEddyfluxtowerstovalidate tantlandcoversinEthiopiaintermsofarea(Fig.2). MOD17 dataandfound that MOD17datasetsshow no overall bias whencomparedwithtowersdata.Theyfound,however,thatMOD17 2.2.Dataandmethods datatendtooverestimateprimaryproductivityinthelowproductivity areasandunderestimateitinhighproductivityareas.Zhaoetal.(2005) Usingdataacquiredfrompublic-domainsourcesofferasolutionfor mademanyenhancementsinthemaininputsofthisdatasetandre- suchadata-scarceregion.Alongwithlandcoverdata,timeseriesofthe ported on correlation analyses between MOD17 and ground-based NPP,GPP,NormalizedDifferenceVegetationIndex(NDVI),precipita- data. Ardö (2015) compared MOD17 NPP with Aboveground NPP tion,evapotranspiration,temperatureandStandardizedPrecipitation (ANPP) data collected from ground measurements in 35 sites in EvapotranspirationIndex(SPEI),wereusedinthisanalysis(Table2). Sudan.Whiletheyfoundastrongcorrelationbetweenthemulti-year Allofthesedatawerederivedfromthepublic-domainsourcesfora averageMOD17NPPandtheANPP(r=0.80,RMSE=135g),they timeframeextendingfrom2000through2013.Mostoftheremote alsoreportedasystematicover-estimationofMOD17NPP,whichis sensingdatasetsusedinthisstudyarerecentproductsandavailable attributed to the fact that the ANPP only considers aboveground onlyfortheyearsafter2000,forinstance,theNPP,GPPandET from biomass. a MODIS.Theselectionoftheperiod(2000−2013)wasmainlycon- trolledbytheavailabilityofthedatafromdifferentsourcesforthe 2.2.2.NormalizedDifferenceVegetationIndex(NDVI) sametime period.Ontheotherhand,manystudiesdealtwiththe DataontheNDVIwereobtainedfromthewebsiteoftheFamine samesubjectswereconductedforsomehowthesimilarperiod.We EarlyWarningSystemsNetworks(FEWSNET).Thisdatasetwasdevel- choosetheperiodtobeconsistentwiththeseresearchesinordertofa- opedbytheU.S.GeologicalSurvey(USGS)EarthResourcesObservation cilitatethecomparisonofthefindings.Moreover,asmentionedearlier, andScience(EROS)Center.Thedatausedhereinwere10-daycompos- theperiodbetween2000–2009wasthehottestontheglobalrecord; itedatawithaspatialresolutionof250m.TherawNDVIimagespro- therefore,theseyearsareveryinterestingforstudyingtheNPPpatterns. cessinginvolved:(i)eliminatingtheinvalidvalues,(ii)convertingthe Inthisstudy,thedataprocessingwascarriedoutusingArcGIS10.3 digitalnumbers(DN)providedintherasterfilestoNDVIvalues,using software. theformulaNDVI=(DN−100)/100,asperinstructionoftheproduct documentation;(iii)using“extractbymask”tooltocreateseparate 2.2.1.Primaryproductivity NDVIdataforeachcountryfromtheoriginaltilesofEastAfrica;(iv)ag- Primary productivity data were obtained from MOD17 product gregating10-daycompositedataintomonthlytimesteps,usingmaxi- (Zhaoetal.,2005),whichprovidesNPPandGPPdata(ingcarbonm−2) mumvaluecomposite(MVC)method(Holben,1986),whichselects 1.4 Abbreviations 1.2 B Barren or sparsely vegetated MF Mixed forests CS Closed shrublands OS Open shrublands 2m) 1 C/NV Cropland/Natural vegetation mosaic PW Permanent wetlands k C Croplands S Savannas n 0.8 DBF Deciduous broadleaf forests W Water o milli EGBF GEvraesrgsrlaenedns broadleaf forests UWS UWroboadny savannas 0.6 ( a re 0.4 Ethiopia A Sudan 0.2 0 B CS C/NV C DBF EBF G MF OS PW S U W WS Fig.2.TotalareaofeachlandcoverinSudanandEthiopiaasappearintheMDC12Q1product. 794 M.Khalifaetal./ScienceoftheTotalEnvironment624(2018)790–806 themaximumvalueforeachpixelfromthethree10-daycompositeim- % d 28 dny an ages.IntegratedNDVI(INDVI)forthecultivationseasonwasusedasa Performance fiDatadifferfromeldmeasurementsby CHIRPScorrelationswithgriddedgrounNprecipitationdataiswithR0.75inmaareasoftheworldfiCorrelationcoefcientbetweenMOD16towerdatais0.86(Muetal.,2011) w22cwtdp1hi.09ar2phiiot19ts.iia3Ihxt55cnr.ayh))iaePntt..fishrrItsooteeeoeprncpawaimcarrptwauceciioshtcsarianreultcteihntmarninhoetlccsStnulslyutolustaalhtladattuuenteietdoedddimdynob,anaaadwniigsonoangemafturgaues0aar,ugos.sl0Cemsawd5dHt(amii°FnIotC.RaiganPle.PitrlmsiTSodooehfacnae(eettsFtehonsouasfe,nHnlite.n,oahkixng1zreteia9rotgNrtah9fidcDn5eapsta;lV-rr.Gbl,PeIeyr1g2rfcooi-ii00nopmur-1cnipJdt5eauaa(s)InatEnykienvlfoacrtde–noagorRoiGOmsdbleioca,opditw2tnnooa0PabsG21rfrieoetI.d4r0Ser-,,, oads/Global/CHIRPS%202.0 oads/East%20Africa/eMODIS%20NDVI ata.UDel_AirT_Precip.html#detail dis/modis_products_table/mcd12q1 wCbpmfi2(nrU.yHro2aoeaDm.Istn4AcRhegui.iPpeltoctTs)oSeiloeeUtirwmmmddapintnratppariioogtoseeteed.enlruprdiauBaarssdtcbtoSlteuueataitddrltnatriieuaisettidoycneatinepsnootsrgtGhgf,obirmiiteidesnrhdoeueciiudlnsmlocsesuvtgdeedodeiraddtcisiptneeeatafrivgslosgeseSeroescalutftitneopairp(apsovtiFrriitneenueolady.ycnngtUiiksw(opbeDUniyiesatneSattrdtliGtnShamaipuiSoletn.ra),danogU2tsatvfieo0nnoomit1drni,pavi5eSwssrne)susodr.whivdgsoiEiiaricndttitnhyhdsehdaiiaohiosnennfipdddgd-DihesaEmdeiv-,ttlqrehruoaoeulinwoousdatpupgalhaielihrttltadyyeast. wnl wnl d/d mo airtemperaturedatawithaspatialresolutionof0.5°forthewhole Downloadsite http://www.ntsg.umt.edu http://www.ntsg.umt.edu http://earlywarning.usgs.gov/fews/datado http://www.ntsg.umt.edu/project/mod16 http://earlywarning.usgs.gov/fews/datado http://www.esrl.noaa.gov/psd/data/gridde http://www.sac.csic.es/spei https://lpdaac.usgs.gov/dataset_discovery/ gfwipo2pujnerlv.riso2coe.roeemt.aaAb5Odcr.ttes.e.Iihntt1oAshMw(els9nycpeWretOi0rtfus(sdoe1hDittEalwavul1lTattmtidohe6dniapoy)veAoed2un.ast2tp0Fbpstitgm1.eolioiaimTsrl4rtcnoevhr-.ateabdideDphonmaoeuselMaslamerpotsespiaarinotavdriutedotaivsaarnhsutfeptpailaruoUsoyonrssonratDsideEuraoset(aeT,trnlsEanlecao2,T(terpsfm0deaVsspr0)tar4ooihpwt1ran.daira0)gsoestut.1gheirvscT)oretltithyesdnouugipefodtsdasertfydetahfodm,eeastdiadMrsatptoueaadeutbdOcosoraseseaDntbiedaattny1runasgcv6fecrnaooaetAtiuvturnNw2iadeaopAelaralnpuaSsatsesrtacAvvtouhdeah/wadsErtaepecuaeOtgodogapcrrSteuetheorfspnarrweua.oinrtrlnosroamesddoys--- Spatialresolution 1km 1km 5km 1km 0.25km 50km 50km 0.5km MdalagOyo,DrmiItShosnmatthedllyelivtaeenl.doTphaenedndubaaytlaMtceoummepteoarianl.l(ar2e10so0kl7mu)tisaopnnad.tTiiahmlisrpeprsoroovldeuduticobtnyuasMensudaeanttEa8Tl-. (2011) to calculate ET based on the Penman-Monteith equation Temporalresolution Annual Monthly 10-daycomposite Monthly 10-daycomposite Monthly Monthly – (uMseoMdnhOteeDirt1eh6i,n1p.9rTo6hv5iis)d.persoadsupcaetcoiavleprrcoodmuecstftohrethperoNbilleembaesnincowuhnitcehreisdthbeyotnhee Reference Zhaoetal.(2005) Zhaoetal.(2005) Funketal.(2015) Muetal.(2007);Muetal.(2011) – WillmottandMatsuura,2001 – trtcviheoamgelnuuesmelisadsaereajroMrnieerdsOderwDceeos1xets6croyelArsui2tcndergidemnaatgshtteaeiidnsnoevftrtohaisgrleoiiSdnfruaenvdlgoaEailtonTuncaeao.vsnnPadbrsluoiyEdcetereshersfiimsoonirpngomigvadi.toenhfsgeeMrtDthOsNe,D.wpS1ei6hxpAieca2lhrsariotasefwortnahdseeatseotearf inthisstudy. Source MOD17 MOD17 CHIRPS MOD16 e-MODISNDVI,FamineEarlyWarningSystemsNetwork(FEWSNET)UniversityofDelaware(UDel) StandardizedPrecipitationEvapotranspirationIndex(SPEI)MCD12Q1 wSSdcEemoootTaoeenumtaraonTttai.erhdhtznFalaireAosasdtertrnfips,rroeideroiTndncoarasadedrsnotccy,auafghoMnncctesetdchoyrOtewsevsM,DaatafRPleoOr1s.aemiu6(Dvamnn2ans1mu0dolc6ibn1edeaielEdn5anoatT)ect-wneeoarMf.dtdenoeTsoauseulhtlinanns.isemrti(tdnieget2rwagiena0tatchg1oneagyr4spoamoe)bloeEuydeuordsTndttsdiwuadasyeergdbsdelecrhyeaaEoeosntnTev2acwmadaMi–bnrde7eflyiOtdaaniustmlDnttalexhoc1vmbfeamt6oeeotrrtflEewpdawstTueluaeeexrrvearreean8tstmnmndooednawMeepgaraaaraecOhyssorhrusuDausae.rrmnn1reeIisnndd6---. d e mentsinnorthandnorthwestChina.Validationstudiesofthisproduct Table2fiSpecicationofthedataus Dataset NetPrimaryProductivity(NPP)GrossPrimaryProductivity(GPP)Precipitation(P) ActualEvapotranspi-ration)(ETaNormalizedDifferenceVegetationIndex(NDVI)Temperature(T) Droughtindex Landcover osfiSmo(eMuvveoreedlOvdrrraee-eDEnsdsa1uc.ts6aisTdmtelAhieAff2aefufwe)tyrleriaaceofm(taonpeuuaceornnrernbedasdgaetfbeiltroaohweetnnhsta.wctateAie,lrmeloposeZanavrnattoeyeaedtre)lhlurdtileehtcoeegtewg-tsibor,Gaoa(nlye.hsuae(zeinlt2igdr,ds0haMcE1)maIT6rOlEa)reeTDie,hagas1Msaatvu6vtimOaieroaleDuvanlmgtae1iSolso6eic.rdnnAhiDatt2mehtseemsimeadsptenhiMottdioneenOdneMtDcsdhoeOIueeSnfdDstotEir1hnbtTa6oge-al M.Khalifaetal./ScienceoftheTotalEnvironment624(2018)790–806 795 product,thisproductprovidesessentialknowledgeonthewatercycle respectively(SuliemanandElagib,2012;Elagib,2013).Forthetwo anditsinteractionwithenvironmentalchanges(Muetal.,2007).Itpro- years,10-daycompositeNDVIdataweresummedforthefivecultiva- videseasilyaccessibleinformationforareaswithlimitedsurfacedata tionmonths(June–October),andtheseasonalINDVIwasusedasa liketheregionofSub-SaharanAfrica. proxyofvegetationproductivityasexplainedbefore.Annualaverage WUEandCUEwerecalculatedforeachlandcovertypeastheratiosof 2.2.6.Droughtindexdata NPPtoET andNPPtoGPP,respectively.SAIsofWUEandCUEwere a Becausedroughtisaslowphenomenonthatdevelopsoveralong alsocalculatedforthepurposeofinvestigatingthelinkagebetween timewithoutprecipitation(Gillette,1950;WilhiteandGlantz,1985), themandSAIsoftheclimateelements. droughtindicesthattakedifferenttimescalesintoaccountareveryuse- fulfordroughtassessment.Incorporatingdifferenttimescalesintheas- 3.Results&discussion sessmentofdroughtimpactiswidelyused(Potopováetal.,2015).In the current investigation, we used the Standardized Precipitation 3.1.Climateconditionsduring2000–2013 EvapotranspirationIndex(SPEI)-developedbyVicente-Serranoetal. (2010)-withaspatialresolutionof0.5°andamonthlytimestep.SPEI Ingeneral,precipitationinSudanismuchlessthaninEthiopia.Tem- calculationrequiresprecipitationandtemperaturedatatoaccountfor porally,precipitationshowssomevariationfromyeartoyearinboth thedifferencebetweenprecipitationandpotentialevapotranspiration countries(Fig.4a).TheyearwiththelowestprecipitationinSudan (PET),i.e.asimplewaterbalance.SPEIisamulti-scalarindexthatallows was2004,recording180.3mm.Fortheperiod2000–2013,thearealav- comparisonofdroughtseverityoverdifferenttimescalesandacross eragetotalannualprecipitationis227.9mmandthecoefficientofvar- space.FortheSPEIscalerangesrefertoTable3.AsSPEIisastandardized iation(CV)is0.13forSudan.AsforEthiopiathecorrespondingvalues variable,itcanbecomparedwithotherSPEIvaluesovertimeandspace. are808.8mmand0.08(Fig.4a).Theresultsonaverageprecipitation Themostwidelyusedtimestepsare1,3,6,12and24months,denoted obtainedfromCHRIPSarecomparablewiththe1970–2000averages bySPEI01,SPEI03,SPEI06,SPEI12andSPEI24,respectively(Chenetal., (i.e.225.5mmand799.1mmforSudanandEthiopia,respectively,ases- 2013).Inthecurrentinvestigation,weusedSPEI01,SPEI03,SPEI06 timatedusingWorldClim(FickandHijmans,2017)).Mostlyvegetated andSPEI12. landsinSudanarefoundonaneast-westbeltlocatedinthesouthern partofthecountrywhereascharacterizedbyprecipitationamountsbe- 2.2.7.Landcover tween250andrarelyabove1000mm.Largeareasofthenorthernpart Inthecurrentresearch,MCD12Q1landcoverdataset(Friedletal., ofSudanreceiveb250mmperyear.AsforEthiopia,theannualprecip- 2010)providedbytheLandProcessesDistributedActiveArchiveCenter itationisonaverageN2000mm,withthehighestprecipitationoccur- (LPDAAC)wasused.Thisproductprovidesannuallandcoverdatafor ringinthewesternpartwhereasthelowestprecipitationisrecorded thewholeglobewithaspatialresolutionof1kmfortheperiodspan- inthenortheasternandsoutheasternportionsofthecountry. ning2001to2012.Thelatestlandcoverdata(version051)wasused Inthestudyarea,itisessentialtounderstandthetimingofprecipi- herein, with land cover classification scheme of the International tationasacrucialfactorinthevegetationdevelopment.SudanandEthi- Geosphere-BiosphereProgramme(IGBP).Thegloballandcoverlayer opia are relying mostly on rainfed agriculture for domestic food wasprocessedtogenerateseparatelayersoflandcoverclassesfor production.Therainyseasonandgrowingseasonarealmostidentical, eachcountry. andtheyextendfromJunetoOctober.InSudan,eventhetimingof themaingrowingseasonforirrigatedagriculturesynchronizesthe 2.2.8.Correlationofthevariablesandcalculationofwaterandcarbonuses rainyseason,especiallyinthelargeirrigatedschemes(e.g.Geziraand efficiency Rahad)becausethewatersupplyintheseschemesishighlydependent TheNPPandclimateelementswerefurtherstandardizedaccording ontheRiverNileflowwhichisinturnhighlyvariableduetotheseason- toKraus(1977)toanalyzetheinter-annualvariabilityoftheStandard- alvariabilityinprecipitation. izedAnomalyIndices(SAIs).Beforestandardization,thedataforallvar- Ontheaverage,mostofSudanshowsannualET ofb500mm.The a iablesweretestedfornormalityusingtheShapiro-Wilktest(Shapiro southernpartdisplaysET upto1500mm.Someoftheirrigatedagricul- a andWilk,1965;GhasemiandZahediasl,2012).Employinganonline turalschemes(e.g.Gezira)incentralSudanshowanaverageET be- a calculator(Dittami,2009),thedatawerefoundtobenormallydistribut- tween500mmand1000mm.Thecountry'sannualaverageofET is a ed. For each variable, the SAI was calculated as: {value − average 204.4inSudanand530.2mminEthiopia(Fig.4b).Spatialvariationin (2000–2013)}/standarddeviation(2000–2013).Theannualaverage ET takesthesamepatternasthatofprecipitation.ThehighestET a a ofeachclimatevariableforeachlandcoverclasswascalculatedusing valuesareforwaterbodies,e.g.LakeTanainEthiopia(Fig.4b). the functions of the GIS environment. Fig. 3 shows the procedure The14-yearannualaveragetemperatureis28.1°CinSudanand23.1 followedinthisstudytocorrelatethevariationintheannualNPPand °CinEthiopia.Duringthestudyperiod,thehighestaverageannualtem- climateindices.Thisprocedurewasusedforalllandcoverclassesand peraturedetectedforeachcountrywas29.0°CinSudanin2010and for each year using thenon-parametric Spearman's coefficients (ρ, 23.5°CinEthiopiain2009.Regionally,thecenterofSudanisthearea rho),withtheaidofXLSTATV.19.4software(Addinsoft,2017).Crop- characterizedbythehighesttemperature(Fig.4c). landsandgrasslandswereconsideredintheanalysisoftheimpactof Onamonthlytimescale,themostseveredroughtsintheregiondur- climatevariabilityonfoodproductiononamonthlytimestepfortwo ingthestudyperiodoccurredin2004,2005and2009(Fig.S1).During selectedyears(2007and2009)representingawetandadryyear, theseyears,Sudanwasaffectedbymoderatetoseveredroughtwhile moderate drought conditionsprevailed during only few monthsin these years in Ethiopia. Based on the annual data (Fig. 5a and b), Table3 CategoriesoftheSPEIscale. Sudanwasaffectedbyamoderatedroughtin2009whereastherest oftheyearswerenormal.Ethiopiaexperiencednormalmoisturecondi- Class SPEIvalue tionsduringthesameperiod. Extremelywet ≥2.00 Severelywet 1.5to1.99 3.2.NetPrimaryProductivityduring2000–2013 Moderatelywet 1.00to1.49 Normal 0.99to−0.99 Moderatelydrought −1.00to−1.49 Spatially,thehighestNPPvaluesinSudanarefoundalongthesouth- Severedrought −1.50to−1.99 easternandsouthwesternbordersofthecountrywheremostofthesa- Extremedrought ≤−2.00 vannaharelocated(Fig.6a).Themajorityofthenorthernpartsofthe 796 M.Khalifaetal./ScienceoftheTotalEnvironment624(2018)790–806 Fig.3.Flowchartofthemethodologicalprocedurefollowedinthisstudytocorrelatetheinter-annualvariabilityinNPP,CUEandWUEandtheirresponsetoclimatevariabilityanddrought conditions. countryarebarrenareaswithtoolowNPPratescomparedtothevege- thereasonbehindthisconsiderabledropinNPP.Incomparisonwith tatedareaswithinthecountry.InEthiopia,thehighestNPPvalueschar- Sudan,somelandcoversinEthiopiadisplayedpositiveNPPanomalies acterizethemiddlepartofthecountrythatiscoveredmostlybywoody duringthefirstyearsofthestudyperiod,mainlyduetothehighprecip- savannahandevergreenbroadleafforests(Fig.6b).Theareaswiththe itation.Theimpactofdroughtof2002insomelandcovertypes(e.g. lowestNPPratesarelocatedinthenortheasternpartofthecountry closedshrublands,croplandsmixedforestandsavannah)wasnotable which is mostly barren, sparsely vegetated or covered by open asNPPintheselandcoversshowednegativeanomalies. shrublands.TheannualaverageNPPis87.24and501.92gcarbonfor Asmentionedearlier,manydriversregulatetheinter-annualvari- SudanandEthiopia,respectively.Allof thelandcoversin Ethiopia abilityofNPP.Inaridandsemi-aridareas,suchasthestudyarea,pat- showhigherNPPthantheircorrespondentsinSudan.Thiscouldbeat- terns of precipitation and temperature are likely to be the most tributedtothehigherprecipitationandlargervegetatedlandsinEthio- importantclimaticfactors,buttheinteractionsoftheseclimaticfactors piacomparedtoSudan.Onaverage,evergreenbroadleafforestsand withvegetationactivitiesarecomplex(Lietal.,2016).Fromtheanaly- woodysavannahinEthiopiashowthehighestannualNPP,withNPP sis,itcanbenotedforEthiopiathatthevariabilityinNPPisinfluenced valuesof1279.5and845.3gCarbon/m2,respectively. directlybythevariabilityinprecipitation(Fig.7).Thisismanifestby the in-phase response of NPP to changes in precipitation and by 3.3. Variation of primary productivity and correlation with climate Spearman'sρ(Fig.S2).AmongalllandcovertypesinEthiopia,grass- variability landsrevealedthestrongestcorrelationρvalue(ρ=0.82,p=0.05). Thecorrelationwithclosedshrublands,croplands,deciduousbroadleaf Throughouttheperiodfrom2000to2013,theNPPdisplayedhigh forests,grasslandsandpermanentwetlandsweresignificantwhereas inter-annualvariabilityinbothcountries.InSudan,mostoftheland thecorrelationforevergreenbroadleafforests,mixedforests,open covertypesshowednegativeanomaliesduringtheyears2000–2008 shrublands,savannahandwoodysavannahwerestatisticallyinsignifi- (Fig.7a).Theyear2007wasanexception,probablybecauseitwasa cant.InSudan,thecorrelationwasstatisticallyinsignificantatp= yearwitharelativelyhighprecipitation.Thelastfiveyears(2009– 0.05(Fig.S3).Zhangetal.(2014)listedmanybioticandabioticfactors 2013)witnessedanincreaseinNPPofalllandcovertypes.NPPforthe (e.g.soilproperties,nutrient,availabilityandtemperature)tobere- year2002 wasassociatedwiththelargestdrop forall landcovers. sponsibleforthelackofimmediateresponseofNPPtothecurrent- Droughteffectintwosuccessiveyears(2001and2002)islikelytobe yearprecipitation.Spatially,thehighestcorrelationbetweenNPPand M.Khalifaetal./ScienceoftheTotalEnvironment624(2018)790–806 797 Fig.4.TemporalandspatialvariationinclimateelementsinSudanandEthiopia:(a)precipitation,(b)actualevapotranspirationand(c)temperature. precipitationisdetectedintheeasternandsouthernEthiopia(Fig.8). et al. (2017) for many countries (e.g. Indonesia, Philippines and TheSpearman'sρintheseareasisN0.6.Theseareasaredominatedby Malaysia). By correlating SPEI at different time scales (1, 3, 6, and croplands,grasslandsandsavannah.Onlysmallspotsinthecentral 12 months) with NPP anomalies, we found the highest correlation (GedarefandBlueNilestates)andwesternpartsofSudanexhibitedρ withtheannualNPPanomaliesforSPEI03inEthiopia.Thus,drought valuesN0.6.Theformerischaracterizedbyextensivecroplands(both eventsonatimescaleof3monthslargelycontrolNPPinthiscountry. irrigatedandrainfedagriculture).However,thecorrelationbetween Thishighdependencecanbedetectedspatially.LargeareasinEthiopia NPPandprecipitationintheseareaswerestatisticallyinsignificant.Sta- showedstatisticallysignificantcorrelationbetweenNPPandSPEI03 tisticalanalysisshowednosignificantcorrelationbetweentheinter-an- comparedtodroughtindicesatothertimesteps(Fig.8).Savannah nualvariationinmeanannualtemperatureandNPPforalllandcovers alsoshowedaverystrongpositivecorrelationbetweenNPPanomalies inbothcountries.Severallandcovertypesshowedinsignificantcorrela- andSPEIatatimestepofthreemonths(ρ=0.93,p=0.05).Open tionbetweentheinter-annualvariationintemperatureandNPP(Fig. shrublandsinEthiopiaseemtobemoresensitivetodroughtastheir 8).Lackofcorrelationbetweeninter-annualvariabilitiesintemperature NPPshowedstrongpositivecorrelationwithadroughtofonemonth andNPPinaridandsemi-aridregionsweredetectedinsomepartsof (SPEI01)(ρ=0.84,p=0.05).TheNPPforcroplandsandpermanent theworld,asreportedbyLiangetal.(2015)forChina.Thesefindings wetlandsdisplayedtheweakestrelationshipwithSPEI03(ρ=0.66 maysuggestthatthelandcovertypesinEthiopiaaremoresensitive and0.63respectively,p=0.05).Thisrelativelylowercorrelationsug- tovariationinprecipitationthanthosecharacterizingSudan.However, geststhatthesetwolandcoversarelesssensitivetodroughtthanthe sincethisanalysiswasconductedononlyanannualbasis,furtheranal- other land covers due to agricultural management (e.g. irrigation) ysis on the varying seasonal effects on vegetation development is and/orsufficientwatersupplyfromtributariesorgroundwater.Ever- deemedimperativetodrawamoresolidconclusion. greenbroadleafforestsseemtobequiteresistanttodroughtsince theirinter-annualNPPanomaliesshowedamoderatecorrelationwith 3.4.Droughtimpactonprimaryproductivity SPEI(ρ=0.54withSPEI01,p=0.05).Deeperrooting(Songetal., 2017) and access to larger water stores in soils and groundwater DroughtaffectstheannualNPPinthelandcoversofSudanandEthi- couldbeareason.AllofthelandcovertypesinSudanshowednostatis- opiadifferently.WhilealllandcovertypesinEthiopiashowapositive ticallysignificantcorrelationbetweentheirannualNPPanomaliesand significantcorrelation(p=0.05)betweenNPPanddroughtseverity, SPEIatanyofthetimescales. nostatisticallysignificantcorrelationwasdetectedforSudan.Lackof ThetotalNPPinthedryyear2009increasedby20.1%inSudanand correlationbetweenannualNPPandtheSPEIwasalsofoundbyPeng decreasedin Ethiopia by11.4%from the14-yearcountry'saverage 798 M.Khalifaetal./ScienceoftheTotalEnvironment624(2018)790–806 Fig.5.TimeseriesofSPEIforinthestudyareawithdifferenttimesteps1,3,6and12months,(a)Sudanand(b)Ethiopia. (2000–2013).TheincreaseofNPPindryyearsisreportedformany yield(Bussmannetal.2016;Elagib,2015;Elagib,2013)andNPP.In areasaroundtheworld,forexampleinNortheastChina(Liuetal., NortheastChina,autumndroughtwasfoundtoleadtolargerreduction 2015;Sunetal.,2016;Peietal.,2013).Generally,therearethreepoten- inNPPwhereasspringdroughthadinsignificantimpact(Liuetal., tialcausesofthisphenomenonasreportedbyYangetal.(2016),Liuet 2015).Frolking(1997)foundthatlatesummerdroughtincreasedNPP al.(2015)andPeietal.(2013).Thesecausesare(i)theassociationofin- by about 20% due to reduced respiration. The present analysis of creaseintemperaturewithdroughtconditions,(ii)thememoryeffect droughtcharacteristicsinSudanshowsastrongerintensitybutlesser ofthepreviousyeardroughtonthecurrentyearNPPand(iii)thechar- spatialextentofdroughtin2009comparedtoastrongerdroughtin acteristicsofdrought.Inthecaseofthedryyear2009inSudan,allthese otherdryyears(e.g.2002)duringthebeginningoftheseason(e.g. factorsseemtohaveplayedaroleinincreasingtheNPP.Itcanbeex- July).ThesedroughtcharacteristicsresultedinarelativelyhigherNPP plainedpartlybythenotableincreaseintheannualtemperaturein inmanylandcovers(e.g.croplandsandshrublands)inJulyin2009 thisyear(Fig.7),whichwasthehottestontherecordasreportedby whichislessby33%fromtheNPPofthesamemonthinthewetyear, SuliemanandElagib(2012)foreasternSudan.Temperatureplaysa 2007. keyroleintheplantrespirationprocess,thuscontributingtoincreasing NPP (Zhao and Running, 2010). In particular, mild drought when 3.5.Intra-annualvariabilityofprimaryproductivityanddrought coupledwithhightemperature,itmightoffsetthedecreaseofNPPin- ducedbywaterdeficit,i.e.leadingtoanincreaseintheNPP(Sunetal. Results of monthly GPP and NDVI for croplands and grasslands 2016;Liuetal.2015).Incontrast,temperatureduringthedryyearof showedthesametemporalpatternsin2007and2009.GPPandNDVI 2004wasbelowthemulti-yearaveragewhichmighthavecontributed starttoincreaseremarkablyatthebeginningoftherainyseason(i.e. tothedecreaseintheannualNPP.Additionally,thememoryeffectof June)asshowninFig.9.TheaverageNDVIshowedlowervaluesduring droughtalsoimpactstheNPP(Yangetal.,2016),likelydecreasingthe thedroughtyear2009inbothcountriesandforbothtypesofland NPPinthedryyearof2002duetocumulativedroughtoftheyears covers.AstrongcorrelationwasfoundbetweenthemonthlyNDVI 2001and2002.Contrarytothis,therelativelymilderdroughtcondition andmonthlyGPPforthetwolandcoversinbothcountries(Fig.S4). in2008probablyweakenedthedroughtimpactof2009onNPP.The Thedroughtconditionof2009reducedboththemonthlyNDVIand timingofdroughtduringtheyear,especiallythebeginningofthegrow- GPPduringtherainymonths.ThereductionintheNDVIwasmoreno- ingseason,isalsoanimportantfactordeterminingthelevelsofcrop ticeableinSudancomparedtoEthiopia.Accordingly,thedeclinein M.Khalifaetal./ScienceoftheTotalEnvironment624(2018)790–806 799 Fig.6.(a)SpatialvariationoftheaverageNPPinSudanandEthiopiaasmodeledbyMOD17.(b)Multiyearannualaverage(2000–2013)ofNPPforlandcovertypesinSudanandEthiopia. iNDVIforcroplandsandgrasslandsin2009comparedto2007were16.9 H O)asreportedbyLiuetal.(2015),andforthesouthernUnitedStates 2 and14.9%inSudanand7.1and16.1%inEthiopia,respectively.Statisti- (0.71gCkg−1H O)asindicatedbyTianetal.(2010),buthighervalues 2 calcorrelationbetweentheintra-annualvariationsinGPPandintra-an- fortheglobalaverage (0.92 gCkg−1H O),asreportedbyItoand 2 nualvariationsinprecipitation,temperatureandSPEIrevealedweak Inatomi(2012). correlationforcroplandsandshrublandsinbothcountries.Thehighest Variationsarenoticeableforthesamelandcovertypesunderdiffer- correlation between GPP and SPEI03 was detected for Sudan with entclimateconditionsinSudanandEthiopia.Evidenceoflargevaria- Spearman's ρ of 0.58 and 0.59 for croplands and shrublands, tionsinWUEwerealsopresentedbyotherresearchersfordifferent respectively. landcovertypesduetodifferencesincarbonuptakeandwatercon- sumption(Liuetal.,2015).Thisdissimilarityismainlyduetophysiolog- 3.6.Wateruseefficiency(WUE) icaldifferencesandclimateconditions(Tangetal.,2014).Generally,all ofthelandcovertypesinEthiopiashowhigherWUEthantheircounter- ThemagnitudeofWUEvariesdependingonthemagnitudesofNPP partsinSudan(Fig.10).Forinstance,thesavannahshowsanaverage andevapotranspiration.ThenationalmultiyearaverageWUEforSudan WUEof0.31gCkg−1H OinSudanbut0.75gCkg−1H OinEthiopia. 2 2 is lower (0.24 g C kg−1 H O) as compared to that for Ethiopia InEthiopia,theevergreenforestshavehigherWUEthanthedeciduous 2 (0.74gCkg−1H O)duetolowerNPPandhigherevapotranspiration forests.Thesefiguresareinagreementwithresultsreportedforforests 2 fortheformer.LiteratureshowscomparablevaluesfortheEthiopianav- insimilarlatitudes(Tangetal.,2014).Amongallthelandcovertypesin erageWUEandthenationalaverageWUEsforChina(0.79gCkg−1 theregion,evergreenbroadleafforestsandwoodysavannahexhibitthe