Remote Sensing of Evapotranspiration for Operational

University of Nebraska - LincolnDigitalCommons@University of Nebraska - Lincoln

USGS Staff -- Published Research US Geological Survey

2012

Remote Sensing of Evapotranspiration forOperational Drought Monitoring Using Principlesof Water and Energy BalanceGabriel B. SenayU.S. Geological Survey

Stefanie BohmsU.S. Geological Survey

James P. VerdinU.S. Geological Survey

Follow this and additional works at: http://digitalcommons.unl.edu/usgsstaffpub

Part of the Geology Commons, Oceanography and Atmospheric Sciences and MeteorologyCommons, Other Earth Sciences Commons, and the Other Environmental Sciences Commons

This Article is brought to you for free and open access by the US Geological Survey at DigitalCommons@University of Nebraska - Lincoln. It has beenaccepted for inclusion in USGS Staff -- Published Research by an authorized administrator of DigitalCommons@University of Nebraska - Lincoln.

Senay, Gabriel B.; Bohms, Stefanie; and Verdin, James P., "Remote Sensing of Evapotranspiration for Operational Drought MonitoringUsing Principles of Water and Energy Balance" (2012). USGS Staff -- Published Research. 979.http://digitalcommons.unl.edu/usgsstaffpub/979

http://digitalcommons.unl.edu?utm_source=digitalcommons.unl.edu%2Fusgsstaffpub%2F979&utm_medium=PDF&utm_campaign=PDFCoverPages

http://digitalcommons.unl.edu/usgsstaffpub?utm_source=digitalcommons.unl.edu%2Fusgsstaffpub%2F979&utm_medium=PDF&utm_campaign=PDFCoverPages

http://digitalcommons.unl.edu/usgs?utm_source=digitalcommons.unl.edu%2Fusgsstaffpub%2F979&utm_medium=PDF&utm_campaign=PDFCoverPages

http://digitalcommons.unl.edu/usgsstaffpub?utm_source=digitalcommons.unl.edu%2Fusgsstaffpub%2F979&utm_medium=PDF&utm_campaign=PDFCoverPages

http://network.bepress.com/hgg/discipline/156?utm_source=digitalcommons.unl.edu%2Fusgsstaffpub%2F979&utm_medium=PDF&utm_campaign=PDFCoverPages





http://digitalcommons.unl.edu/usgsstaffpub/979?utm_source=digitalcommons.unl.edu%2Fusgsstaffpub%2F979&utm_medium=PDF&utm_campaign=PDFCoverPages

Published in Remote Sensing of Drought: Innovative Monitoring Approaches, edited by Brian D. Wardlow, Martha C. Anderson, & James P. Verdin (CRC Press/Taylor & Francis, 2012).

This chapter is a U.S. government work and is not subject to copyright in the United States.

Authors:

Gabriel B. SenayEarth Resources Observation and Science CenterU.S. Geological SurveySioux Falls, South Dakota

Stefanie Bohms Earth Resources Observation and Science Center U.S. Geological Survey Sioux Falls, South Dakota

James P. VerdinEarth Resources Observation and Science CenterU.S. Geological SurveySioux Falls, South Dakota

123

6 Remote Sensing of Evapotranspiration for Operational Drought Monitoring Using Principles of Water and Energy Balance

Gabriel B. Senay, Stefanie Bohms, and James P. Verdin

CONTENTS

6.1 Introduction..................................................................................................1246.2 MaterialsandMethods.................................................................................124

6.2.1 DataRequirements........................................................................... 1256.2.1.1 PrecipitationData.............................................................. 1256.2.1.2 LandSurfacePhenologyData........................................... 1256.2.1.3 SoilData............................................................................. 1256.2.1.4 ReferenceET...................................................................... 1266.2.1.5 ThermalRemoteSensingData.......................................... 126

6.2.2 ModelDescriptions........................................................................... 1266.2.2.1 WaterBalanceModel:VegET........................................... 1266.2.2.2 EnergyBalanceModel:SSEB........................................... 1296.2.2.3 ComparisonofVegETandSSEB...................................... 132

6.2.3 ETaAnomalies.................................................................................. 1336.3 VegETandSSEBOutputovertheCONUS................................................. 134

6.3.1 CumulativeETa................................................................................. 1346.3.2 ETaAnomalies.................................................................................. 1356.3.3 CaseStudy:SeasonalETTimeSeries.............................................. 137

6.4 ApplicationsofVegETandSSEBfortheFamineEarlyWarning SystemNetwork............................................................................. 138

6.5 Conclusions................................................................................................... 141Acknowledgment................................................................................................... 142References.............................................................................................................. 142

124 Remote Sensing of Drought: Innovative Monitoring Approaches

6.1 INTRODUCTION

Evapotranspiration(ET)isanimportantcomponentofthehydrologicbudgetbecauseitreflectstheexchangeofmassandenergybetweenthesoil–water–vegetationsystemandtheatmosphere.PrevailingweatherconditionsinfluencepotentialorreferenceET through variables such as radiation, temperature, wind, and relativity humid-ity.Inadditiontotheseweathervariables,actualET(ETa)isalsoaffectedbylandcovertypeandcondition,aswellassoilmoisture.ThedependenceofETaonlandcoverandsoilmoisture,anditsdirectrelationshipwithcarbondioxideassimilationinplants,makes it an importantvariable formonitoringdrought, cropyield, andbiomass—acriticalcapabilityfordecisionmakersinterestedinfoodsecurity,grainmarkets,waterallocation,andcarbonsequestration(Bastiaanssenetal.,2005).

Because ET can be difficult to measure accurately, especially at large spatialscales, several different hydrologic modeling techniques have been developed toestimateETausingsatelliteremotesensing.Ingeneral,theETmodelingtechniquescanbegroupedintotwobroadclassesthatincludemodelsbasedonsurfaceenergybalance(e.g.,Bastiaanssenetal.,1998;Suetal.,2005;Allenetal.,2007;Andersonetal.,2007;Senayetal.,2007)andwaterbalance(e.g.,Allenetal.,1998,Senay,2008)principles.Whilewaterbalancemodelsfocusontrackingthepathwaysandmagnitudeofrainfallinthesoil–vegetationsystem,mostremotesensingenergybal-ancemodelsuselandsurfacetemperature(LST)asaprimaryconstraintinpartition-ingradiantenergyavailableatthesurfacebetweenheatandwaterfluxes.

This chapter describes two ET models representing each of these approaches:theVegetationET(VegET)waterbalancemodel(Senay,2008)andtheSimplifiedSurface Energy Balance (SSEB) approach (Senay et al., 2007, 2011a), comparingtheir utility for operational drought monitoring and agrohydrologic applications.BothmodelsusetheconceptofareferenceET(ETo)toestimatethepotentialET(ETp)expectedunderunlimitedwaterconditions,assuminganidealizedreferencecropwithstandardizedbulkandaerodynamicresistancefactorsforvaportransport.Themaindifferencebetweenthetwoapproachesisinthecalculationofacorrectionfactoraccountingforsoilmoistureimpactsonevaporation,estimatingETaasafrac-tionofETo.VegETusesavegetationwaterbudgetingapproachtotracksoilmoisturechanges,whereastheenergybalancemodelusesspatialvariationsinLST.

Bothmodelsweredesignedforglobaloperationalapplicationsandarethereforeintentionallysimplifiedintheirrepresentationofsurfacephenomenaandmodestintheirinputdatarequirements—basedonlyonreadilyavailableremotesensingdata.Thesimplifiedapproachesfacilitatereal-timeimplementationindata-limitedpartsoftheworld,providingtimelyinformationforoperationaldroughtandfoodsecurityanalyseswithminimalmanualinterventionandexpertguidance.

6.2 MATERIALS AND METHODS

Hereweprovide abrief introduction to theVegETandSSEBETmodeling algo-rithms.Thetwoapproacheseachhavetheirownmeritsandlimitations,andtheycanbeusedindependentlyorincombination.Thechoiceofmodeldependsontheavail-abilityofdataandtheobjectiveoftheproject.BothmethodsrequireanETodataset,

125Remote Sensing of Evapotranspiration for Operational Drought Monitoring

whichcanbegeneratedusingmeteorologicaldata(netradiation,temperature,windspeed,relativehumidity,andairpressure).Inaddition,theavailabilityofrainfallandlandsurfacephenology (LSP)data iscritical for theVegETwaterbalancemodel,whiletheSSEBenergybalanceapproachrequiresLSTinformationretrievedfromthermal infrared satellitedata.Thesedifferences indata inputsare importantanddefine the applications and constraints that apply to each modeling approach. Forexample,thepresenceofcloudcoveradverselyaffectstheSSEBmodelbecauseLSTcannotberetrievedundercloudyconditionsusingthermalimaging.Incontrast,theVegETmodeldoesnotusethermaldataandislessaffectedbycloudcover,whichcanbeasignificantadvantageduringthegrowingseasoninmanypartsoftheworld.Ontheotherhand,VegETconsidersonlyrain-fedwaterinputstothelandsurfacesys-tem,whereastheLSTinputstoSSEBprovidediagnosticinformationaboutmoistureinputsfromallsources,includingirrigationandshallowwatertables.Anotheradvan-tageoftheSSEBapproachisthatitdoesnotrequireprecipitationdataandthusislesspronetoerrorsassociatedwiththequalityoftheavailableprecipitationdatasets.

6.2.1 Data RequiRements

Tofacilitateglobalapplications,boththeSSEBandVegETmodelingsystemshavebeendesignedtousereadilyavailableglobalremotesensingandweatherdatasets.Inputdatarequirementsbyeachmodel,andrationalethereof,aredescribedinthefollowing.

6.2.1.1 Precipitation DataPrecipitationisakeydriverofthewaterbalanceVegETmodel.Acombinationofcoarse(25kmfor1996–2004)andfiner(5kmfor2005tocurrent)spatialresolutiondailytotalrainfalldatafromtheNationalOceanicandAtmosphericAdministration(NOAA) National Weather Service (NWS) (http://www.srh.noaa.gov/rfcshare/precip_about.php)isbeingusedbasedondataavailability.BothprecipitationdatasetsyieldcomparableseasonalETaestimatesfromVegET(datanotshown).SpatialresolutionofETaoutputfromVegETisnotsignificantlylimitedbytheinputprecipi-tationdatasetbutratherbythescaleoftheLSPdatausedinthemodel.Furthermore,rainfallisrelativelyhomogeneousatthesubwatershedscalewhenaggregatedovermonthlyorlongertimescales.

6.2.1.2 Land Surface Phenology DataAsdescribedinthefollowing,LSPparametersusedinVegETaredefinedusingatimeseriesof1kmNormalizedDifferenceVegetationIndex(NDVI)dataderivedfrom the NOAA Advanced Very High Resolution Radiometer (AVHRR) satelliteimageryfortheperiodof1989–2004(Eidenshink,1990).ThesedatasetshavebeennormalizedovermultipleAVHRRinstruments.

6.2.1.3 Soil DataSoilwaterholdingcapacity,usedinVegET,isderivedfromtheStateSoilGeographicDatabase (STATSGO) (http://www.ncgc.nrcs.usda.gov/products/datasets/statsgo/)fortheUnitedStates,whiledatafromtheFoodandAgricultureOrganization(FAO)DigitalSoilsMapoftheWorldareusedforglobalapplications.


6.2.1.4 Reference ETOverthecontinentalUnitedStates(CONUS),theETodatausedbybothVegETandSSEBareproducedatadailytimestepasdescribedbySenayetal.(2008)usingthestandardizedPenman-Monteithequation(Allenetal.,1998).Global,sixhourlyweatherdatasetsofnetradiation,wind,relativehumidity,andairtemperatureandpressurefromtheGlobalDataAssimilationSystem(GDAS)(Kanamitsu,1989)areusedtogenerateaglobaldailyEToat1°spatialresolution.

6.2.1.5 Thermal Remote Sensing DataTheSSEBenergybalancealgorithmismainlydrivenbyLSTderivedfromthermalbandobservationsacquiredbytheModerateResolutionImagingSpectroradiometer(MODIS). Day-time, 8 day average LST tiles at 1km resolution from the NASATerraplatform(MOD11A2),acquiredfromMarch2000topresent,havebeendown-loadedfromtheLPDAAC(LandProcessesDataActiveArchiveCenter)andrepro-jectedandmosaickedusingtheMODISreprojectiontool.AlthoughinstantaneousLSTdataretrievals(e.g.,fromtheMOD11_L2swathproduct)aretechnicallymoreappropriateforapplicationofSSEBalgorithms,the8daycompositeproductisusedheretoreducecomputationalanddatademandsforoperationalglobalapplications.Furthermore,useofthe8dayproductreducesdatagapscausedbycloudcontamina-tion.RamificationsofusingtheMODIS8dayLSTproductarediscussedfurtherinSection6.2.2.2.

6.2.2 moDel DescRiptions

6.2.2.1 Water Balance Model: VegETTheVegETapproachisbasedonthemostwidelyusedwaterbalancetechniqueforoperationalcropperformancemonitoring:theFoodandAgriculturalOrganization(FAO) algorithm for computing the crop Water Requirement Satisfaction Index(WRSI; FAO, 1986). The WRSI reflects the relative relationship (ratio/percent)betweenwatersupply(fromrainfallandexistingsoilmoisture)anddemand(croptranspirationdemandtomeetitsphysiologicalneeds)usingobserveddatafromthebeginningofthecropseason(planting)untilthecurrentdate.WRSIiscalculatedastheratio(orpercentage)betweentheseasonalETandtheseasonalwaterrequirementofthecrop.TheseasonaltotalwaterrequirementiscalculatedastheETpadjustedbyacropcoefficient(Kc),whichvariesbycroptypeandphenologicalstage.Kcgen-erallyvariesbetween0.3and1.2formostcerealcropsduringthegrowingseason(FAO,1998).

The Famine Early Warning System Network (FEWS NET) demonstrated aregionalimplementationoftheFAOWRSIoveramodelingdomaininsouthernAfrica(VerdinandKlaver,2002).SenayandVerdin(2003)furtherenhancedthegeospatialmodelby introducing theconceptofMaximumAllowableDepletion(MAD)andasoilwaterstressfactorfromirrigationengineeringforbetterestima-tionofETaasafunctionofsoilwatercontent.TheSenayandVerdin(2003)versionofthemodelhasbeenoperationalsince2000,withdailyand10dayoutputsforAfrica,CentralAmerica,andAfghanistanat0.1°(∼10km)resolution.Graphicsofmodeloutputarepostedoperationallyathttp://earlywarning.usgs.gov/fews/.


BuildingontheWRSIconcept,theVegETmodelingstrategy(Senay,2008)wasrecentlydevelopedforestimatingETainnonirrigatedcroplandandgrasslandenvi-ronments as anenhancement to theU.S.GeologicalSurvey (USGS)/FEWSNETcropwaterbalancemodel(SenayandVerdin,2003).VegETblendsconceptsfromirrigationengineeringwitharemotesensingdatastreamtoestimateETaquicklyatlowcomputationalanddatacostsforsitesanywhereintheworld.Figure6.1showsaschematicrepresentationoftheVegETmodelingframework.

Akeyinnovationin theVegETmodel is the inclusionof theLSPparameter,whichdescribestheseasonalprogressionofvegetationgrowthanddevelopment.The LSP allows the VegET model to be location (pixel)-specific, accommodat-inglocalizedvariationsinvegetationgrowthpatterns,ascomparedtotheregion-specific Kc function used in traditional agrohydrologic modeling. LSP can beobservedbyspacebornesensorsandisakeybiophysicalparameterthatlinksthewater and carbon cycles with anthropogenic activities, providing an importantapproachtochangedetectioninterrestrialecosystems(e.g.,Gowardetal.,1985;Reedetal.,1994;Tuckeretal.,2001;deBeursandHenebry,2005).IntegrationofLSPinformationintoaphenology-basedcropcoefficient(Kcp)isdescribedlaterinthissection.

VegETmonitorssoilwaterlevelsintherootzonethroughadaily(orlongertimestep)waterbalancealgorithmandestimatesETainrain-fedcroplandandgrasslandenvironments.Key inputdata toVegETareprecipitation,ETo, soilwaterholdingcapacity,andLSP.ETa(inunitsofmm/day)iscalculatedastheproductoftheETo(mm/day),asoilmoisturestresscoefficient(Ks),andaphenology-basedwater-usecoefficient(Kcp),asshowninEquation6.1(Senay,2008):

ET K Ka s cp= ∗ ∗ETo (6.1)

PPT ETο

NDVI(LSP) Kcmin

Kcmax Waterbalancemodel

Runo�

KsKcp

ETa = Kcp * Ks * ETo

FIGURE 6.1 SimplifiedconceptualdiagramoftheVegETmodel.Majorinputsareprecipi-tation(PPT),referenceET(ETo),andNDVI.Estimatedparametersareaphenology-basedcropcoefficient(Kcp)andasoil-waterstressfactor(Ks).ModeloutputsareETaandrunoff.


TheKsparameter isdeterminedfromavegetation–soil–waterbalancemodelandhasavaluebetween0(drysoil)and1(moistsoil).Thewaterbalancemodelworkswithadailysoilmoistureaccountingprocedureoverasoilbucketthatisdefinedbythewaterholdingcapacityofthesoilonagrid-cellbasis.TheLSPcoefficient(Kcp)iscomparabletotheKcwidelyusedbyagronomists(Allenetal.,1998)butincludesanLSPdependencederivedfromremotelysensedtimeseriesofNDVI(Senay,2008).Kcprepresentsboththespatialandtemporaldynamicsof thelandscapewater-usepatternsonagrid-cell (orpixel)basis.TheKcpparameter isscaledbetweenpub-lishedcropcoefficientminimum(Kcmin)andmaximum(Kcmax)valuesbasedoncur-rentandclimatologicalNDVIdata:

K

Kc KcNDVI NDVI

NDVI NDVIcpmax min

max mini o= −

−−*( ) (6.2)

whereKcmaxisthemaximum(mature)Kcvalueforaparticularvegetation/croptypeKcministheminimum(earlystage)KcvalueNDVImin and NDVImax are the climatological minimum and maximum NDVI

valuesinayear,respectivelyNDVIiistheclimatologicalNDVIvalueforagivenperiod“i”(averageweekly

maximumvalueinthiscase)NDVIoistheminimumreferenceNDVIvaluethatisassociatedwiththemini-

mumKcvalue

ThecalculationofNDVIodependson theNDVImax specifiedateachpixeland isdeterminedusingoneoftwofollowingcases:

CaseI:IfNDVImax>=0.40,then

NDVI 3o = 0. (6.3)

CaseII:IfNDVImax<0.40,then

NDVI 33 NDVI NDVI NDVIo max min min= ∗ − +0. ( ) (6.4)

Equations6.3and6.4wereformulated tohandlesparselyvegetatedsemiaridandarid regions differently from well-vegetated areas. Even a low maximum NDVIregionwillshowawater-usephenologyifitisrescaleddifferentlyinrelationtoitsownminimumratherthanthe“global”minimumofNDVI=0.3.OtherresearchershaveusedadifferentformulationtoestimateKcvaluesorcomparablecoefficientsfromNDVI(e.g.,Nagleretal.,2005;Groeneveldetal.,2007;Allenetal.,2011)forthesamepurposeofestimatingETa.

AmajorassumptioninthespecificationofKcpisthattherehavebeennomajorclimate or land cover changes over the remote sensing data record to affect thewater-use dynamics of a given individual modeling cell (or pixel) as represented


by theLSP.This limits theutilityofVegET(with thecurrentsetup) formonitor-inghighlymanagedlandscapessuchasirrigatedagricultureandurban/ruralfringeareas.However,withamodificationofthewaterbalancecomponentofthemodel,currentNDVIvaluesarestillcapableofestimatingETfromirrigatedlands,asisdemonstratedbyNagleretal.(2005).

For operational monitoring over the United States, the VegET model is runat 10km spatial resolution (chosen to reduce computational time for a regionalapplication) with operational products updated and posted daily at 7:00 pm(http://earlywarning.usgs.gov/usewem/swi.php). The operational products focus onthegrowingseasonperiod,definedasApril1–October31.

6.2.2.2 Energy Balance Model: SSEBSurfaceenergybalancemethodshavebeensuccessfullyappliedbyseveralresearch-ers (Bastiaanssenetal.,1998;Suetal.,2005;Allenetal.,2007;Andersonetal.,2007)toestimatecropwateruseinirrigatedareasandacrossthegenerallandscape.Theapproachtakeninthesemodelsrequiressolutionoftheenergybalanceequationatthelandsurface(Equation6.5),computingthelatentheatflux(ETaconvertedintounitsofenergy,W/m2,asaresidual):

λE R G Hn= − − (6.5)

whereλEisthelatentheatflux(energyconsumedbyET;W/m2)Rnisthenetradiationatthesurface(W/m2)Gisthegroundheatflux(energystoredinthesoilandvegetation;W/m2)Histhesensibleheatflux(energyusedtoheattheair;W/m2)

MostthermalenergybalancealgorithmsintendedforoperationalETmonitoringhavebeenexplicitlydesignedtominimizesensitivitytoerrorsintheabsolutecalibra-tionandatmosphericcorrectionoftheLSTdata.Allenetal.(2007)describeasur-faceenergybalancemethodthatemploysthehot(dry)andcold(wet)pixelapproachofBastiaanssenetal.(1998)intheSEBAL(SurfaceEnergyBalanceAlgorithmfortheLand)model,constrainingETaestimatesbetweenreasonableboundsasdefinedattheseend-memberpixels.Assuch,thesemethodsdonotrequireabsoluteaccuracyinLSTbutonlyrelativeaccuracyinvariabilityacrossthescene.Fornetradiation,SEBALrequiresmeteorologicaldataon incomingradiation,alongwith theasso-ciatedsurfacealbedoandemissivityrequired tocomputeoutgoingradiation.Thegroundheatfluxisestimatedusingremotesensingestimatesofsurfacetemperature,albedo,andNDVI,whilethesensibleheatfluxisestimatedasafunctionoftempera-turegradientabovethesurface,surfaceroughness,andwindspeed.

AlthoughthefullenergybalanceapproachemployedinSEBALhasbeenshowntogivegood results inmanypartsof theworld (Bastiaanssenet al.,2005),well-trained operators are required to perform the selection of hot/cold end-memberpixels,andinputdatarequirementscanbeprohibitive,especiallyoverlarge,data-sparseregions.Asanalternative,theSSEBapproachwasdevelopedatUSGSEarth


Resources Observation and Science (EROS) Center for operational application(Senayetal.,2007,2011a).

TheSSEBapproachinvolvestwobasicsteps(Figure6.2).ETaiscomputedasaproductofthereferenceETfraction(ETf)andthereferenceET(ETo):

ET ET ETa f o= ∗α (6.6)

whereαisamultiplyingfactorthatisgenerallysetto1.2ifEToisfromthestandardizedclippedgrassreferenceor1.0ifanalfalfa-basedreferenceETisused.Localcalibrationusinglysimeterdataisrecommendedtoaccuratelyestimateαiftheabsolutemagni-tudeofETiscriticalforthestudy.Fordroughtmonitoringpurposes,whereanomalieswithrespectto“average”conditionsaremoreimportant,theconsistencyofthemethodanddataismoreimportantthantheabsoluteaccuracyofET(see Section6.2.3).

TheETfvariable is thekeyto theSSEBapproachsince itcaptures the impactofsoilmoistureonETa,whileETodeterminesthepotentialETundernonlimitingwatersupplyconditions.IntherevisedSSEBapproachpresentedinthischapter,ETfiscalculatedfromtheLSTandairtemperaturedatasetsbasedontheassumptionsthatahotpixelexperienceslittleornoET(Bastiaanssenetal.,1998;Allenetal.,2005),andacoldpixelrepresentsmaximumET.AnassumptionismadeinSSEBthatETcanbescaledbetween these twoend-pointvaluesofETinproportion tothedifferencebetweenLSTandairtemperature(Ta)measuredateachpixel.NotethatthemethoddoesnotrelyonabsoluteaccuracyineitherLSTorTa;however,itisrequiredthatthedifferencebetweentheLSTandTaberelativelyaccurateacrossthestudyregion.ThemaindriverfortheETfisthedifferencebetweenLSTandairtemperatureinrelationtothesamedifferencemeasuredatthereferencelocations(hotandcoldpixels).AcrosstheLSTscene,SSEBassumesthatpixelswithlargersurface-to-airtemperaturedifferenceshavehighersensibleheat(lowerET),whilepixelswithsmall(LST−Ta)havelowersensibleheat(highET).TheinclusionofairtemperatureintheETfcalculationintherevisedSSEBapproachisintendedto

LST

Ta

NDVI

ETf

ETa

αETο

Weather data:Rn, Ta, U, RH, P

FIGURE 6.2 SchematicrepresentationoftheSSEBmodelingsetup.Suggestedvalueforαis1.2whenEToisbasedonclippedgrassreferenceET.Rnisnetradiation,Taisairtempera-ture,Uiswindspeed,RHisrelativehumidity,andPisatmosphericpressure.


facilitatecontinentalapplicationoftheSSEBapproach,reducingtheneedtoselectmultipleend-memberLSTpairsacrossthecontinentindifferentclimaticregions,whichisatypicalrequirementforhot-coldpixelapproaches.

ThehotpixelsareselectedusinganNDVIimageasaguidetoidentifytheloca-tions of dry and nonvegetated (or sparsely vegetated) areas that exhibit very lowNDVIvalues.Similarly,thecoldpixelsareselectedfromwell-watered,healthy,andfullyvegetatedareasthathaveveryhighNDVIvalues.TheETfraction(ETf,x) iscalculatedforeachpixel“x”as

ET

dT dTdT dT

f,xh x

h c

= −−

(6.7)

wheredThisthedifferencebetweensurfacetemperature(Ts)andairtemperature(Ta)

atthehotpixeldTcisthedifferencebetweenTsandTaatthecoldpixeldTxisthedifferencebetweenTsandTaatagivenpixel“x”

Themethodissensitivetotheselectionofhotandcoldpixels,andcautionshouldbetakeninselectingthesereferencepoints.Thecoldpixelcanbeawaterbodyorwell-watereddensevegetation,preferablywithanNDVIvaluegreaterthanorequalto0.7.Sincetheenergybalancepartitioningofawaterbodyisdifferentfromalandsurface,awaterbodymaybecolder(iffedfromsnowmelt)orwarmer(iffedfromageothermalsource)thanmostdensevegetation,butthiswillonlybringasystematicbiasthatcanbecorrectedbycheckingagainsttheLSTfromawell-wateredvegeta-tioninthesameareaandseason.ThemainadvantageofawaterbodyisthatitisgenerallyavailablemuchoftheyearexceptthewinterseasonofsomeregionswhenETislow.Thisprovidesanadvantageovertherelativelyshortseasonofdenseveg-etation.However,itisimportanttoremainconsistentinthechoiceofthecoldandhotpixelsduringthedifferentpartsoftheyear(i.e.,ifacoldpixelischosenfromawaterbody,itisadvisabletoselectthesamewaterbodyovertime).Thesameprincipleappliestothehotpixel.Inalargeimagescene,itisadvisabletoselectthehotpixelsfromnonirrigatedperpetuallybareareas,withanNDVIvalue<0.2.

ForthisstudyovertheCONUS,TsisobtainedfromtheMODIS8dayLSTprod-uct,whileTaisassignedfromthemonthlymaximumairtemperature(generallymea-suredataround1.5mabovegroundlevel)fromthePRISM(Parameter–ElevationRegressions on Independent Slopes Model; PRISM, 2011) data set, selecting themonthly interval closest to the8day time period corresponding to theLST dataset.Eight-dayEToiscomputedfromdailyGDASETooutput(Senayetal.,2008).Themodelisrunat8daytimeincrementsovertheperiodofrecord.Inthischapter,onlyseasonalproductsfromAprilthroughOctoberarediscussedbecauseoftheirrelevanceforseason-integrateddroughtmonitoring.Atemporallydynamicsetofhotandcoldpixelsselectedfromrepresentativelocations(coldgenerallyfromthesoutheastUnitedStates[wetterarea]andhotpixels[dryareas]inthewesternHighPlainsoftheUnitedStates)hasbeenusedontheentireCONUSdataset.Itshould


benotedthatalthoughthehotandcoldpixelsareconsistentinspace,theLSTvaluesgenerallyvaryfromseasontoseason,soweprepareauniquesetofhotandcoldpixelsforeachperiodfromthesameregionorlocationthatmeetstherequirements.WhatisuniqueinthisapproachistheuseofasinglesetofhotandcoldpixelstoscaleacrosstheCONUSforeach8dayperiod.

Anumberofsimplificationsregardingrepresentationofland–atmosphereexchangeareimplicitintheSSEBalgorithm,andthesewarrantsomediscussion.First,unlikeSEBALorMETRICandmostthermalETmodels,afullenergybalanceisnotcom-putedwithinSSEB.Rather,ETf isscaleddirectlyininverseproportiontoTs−Ta,whileotherenergybalancecomponentsarenotassessed.Thisscalingneglectstheeffectsofvariablesurfaceroughnessandgroundheatfluxacrossthelandscapeonthesurfaceenergybalance.Also,theuseofthe8dayLSTcompositescanintroduceerrorsintothemethodology,becausevariouspixelsinthescenemaybesampledondifferentdaysunderdifferentatmosphericandsurfacemoistureconditions.Finally,localairtemperaturecanbeverydifferentfromTainterpolatedbetweenstationdata(asinthePRISMdataset),andthiswilladduncertaintytotheETestimates.

ThisSSEBmethodisexperimentalandrequiresfurtherevaluationunderarangeofconditions;however,preliminaryassessmentsareencouraging—particularlyforlong-termseasonalETestimates.Gowdaetal.(2009)evaluatedtheperformanceoftheSSEBusinglysimeterdatainnorthwestTexasandfoundthatitexplained84%ofthelysimeterETvariation,withameanbiasof−0.6mm/day,usingpooleddatasetsfromirrigatedandrain-fedagriculturalsystemswithcornandsorghumfieldsovera2yearstudy(2006–2007).Recently,Senayetal.(2011b)evaluatedtheSSEBETovertheCONUSusinganHUC-8(HydrologicUnitCode)levelwaterbalanceapproach.Theannualdifferencesbetweenprecipitation(P)andrunoff(Q)at1,399HUC-8levelwatershedswerecompared toannualSSEBETestimateswithanr2of0.90andameanbiasof−67mmor−11%of thedifferencebetweenobservedPandQ.TheSSEBETshowsageneralunderestimation in the lowerET region(ET<600mm)comparedtohigherETzones.Moreimportantly,thehighr2(0.90)demonstrates the precision and reliability of the approach in diverse ecosystems,especiallywhenusedasananomalyproduct.

Becausethismethodisintendedforeasyimplementationforlarge-areamonitor-ingbynonexperts,asimplifiedapproachwithminimaldatarequirementsisdesired.Additionally,theETanomaliesusedfordroughtmonitoring(seethefollowing)areless sensitive to errors in the simplifiedmodeling approach than are the absolutemagnitudesofET.Inthiscontext,SSEBcanbeconsideredasanindexdescribingrelativechangesinEToverthesatelliteperiodofrecord.

6.2.2.3 Comparison of VegET and SSEBTable 6.1 summarizes differences between the VegET and SSEB modelingapproachesintermsofinputandoutputdatacharacteristics.OperationalVegEToutput is currently produced over the CONUS on a daily basis, while SSEB-based ETa for the CONUS is updated on an 8 day basis since the summer of2011(http://earlywarning.usgs.gov/usewem/eta_energy.php).HistoricalmonthlySSEB ET outputs are currently available from 2000 to 2009 for the CONUSandarebeingvalidatedusingfluxandwaterbalancemodeloutputs.Inaddition,


initialresultsforAfricaandriverbasinsincentralAsiaareshowingpromisingresults(UNEP,2010;Senayetal.,2007).Applicationsforbothapproachesarepresentedlaterinthischapter.

6.2.3 eta anomalies

Fordroughtmonitoring,indicatorsaretypicallyformulatedintermsofamonthlytoseasonalanomaly,representingdeviationsofcurrentconditionswithrespect to“normal” or “average” historical conditions for that time period. This is becauseanomalies (wetterordrier thanusual) are easier tounderstandandmeasure thanare absolute quantities (e.g., rainfall or ET in mm). Anomaly information is alsomorerelevantfordecisionmakersbecauseitprovidesahistoricalcontextforhowcurrentconditionscomparetoconditionsfrompreviousyears.Inaddition,impactsofmodelassumptions,formulationerrors,andbiasesininputdataarereducedinanomalyproducts.Themainreasonforthisisthatthestatisticalnatureofanomaly

TABLE 6.1Modeling and Data Characteristics of VegET and SSEB

VegET SSEB

Modelingapproach Waterbalance Energybalance

Targetmonitored/output ETa,soilmoisture,runoff ETa

Spatialresolution LimitedbyLSPdata Limitedbythermaldata

MODIS:250m (MODIS/AVHRR:1km)

AVHRR:1km Landsat:∼100m(localapplication)

Spatialextent Global(potentially) Global(potentially)

Frequencyofproduct Daily 8-day,dailyispossible

Delay 1day About2weeksforMODIS

Periodofrecord Limitedbyrainfalldata Limitedbythermaldata

1996–current:NexRad/StationBlend

AVHRR:1989–currentMODIS:2000–current

1979–current:GPCP(GlobalPrecipitationClimatologyProject)

Webaccess VegETmodeloutputisonlineathttp://earlywarning.usgs.gov/usewem/swi.php

ETaanomalyonlineAfrica:http://earlywarning.usgs.gov/fews/africa/index.php

CONUS:http://earlywarning.usgs.gov/usewem/eta_energy.php

Geographicprojection Latitude–longitude Latitude–longitude

GISenvironment Yes Yes

Descriptionofproduct Appropriateforrain-fedagricultureorgrasslandenvironments

Bestappliedtoirrigatedsystems

Challenge/limitationforoperationalimplementation

Nomajorlimitationisanticipated

Cloudcoverandlackofclimaticrecord


calculationcancelsmultiplicativeerrors (e.g.,due tomodel formulationand inputdatabiases)thatappearinboththenumeratoranddenominatorinEquation6.8.

Inthisstudy,seasonalETaanomalieswerecalculatedovertheCONUSforbothVegETandSSEBbasedonthemedianoftheseasonal(April1–October31)totalETfrom2000to2009(dataavailableyears)as

ET ano

ET _yearET _median

aa

a

_ = ∗100 (6.8)

whereETa_anoistheETaanomalyforagivenyearinpercentETa_yearistheseasonalETatotalforagivenyearETa_medianisthemedianseasonalETafrom2000to2009

Althoughanomaliescanbecalculatedatdifferenttimescales,thischapterfocusesonseasonal timescales tohighlight theutilityofanomalyproducts forassessingagriculturaldroughtimpacts,whicharegenerallyfeltataseasonaltimescale.AnexampleofaninternationaloperationalETaanomalyproductforAfricausingtheSSEBmodelispresentedlaterinthechapter.

6.3 VegET AND SSEB OUTPUT OVER THE CONUS

6.3.1 cumulative eta

SeasonalcumulativeETmapsfor2009overtheCONUSgeneratedwiththeVegETandSSEBmodelsarepresented inFigure6.3aandb, respectively.Generally, thetwomapsarecomparableboth inmagnitudeand spatialpatterns in thepredomi-nantlyrain-fedsystemoftheeasternUnitedStates.OutputfrombothmodelsinthisregionexhibitshighseasonalETinexcessof500mm,particularlyinthesoutheast.MorenotabledifferencesbetweenmodelsareobservedinthewesternUnitedStates,forreasonsthatmayvarybylocation.Forexample,themodelspredictsignificantlydifferentfluxesinirrigatedregionssuchastheCentralValleyofCalifornia,wherecropsareexpectedtohavehighETbecauseoftargetedwaterapplications.Theesti-matefromtheVegETwaterbalancemodelislowbecauseirrigationwaterinputsarenotaccountedforinthismodelingapproach.Incontrast,thecontributionofirriga-tion is reflected in theMODISLSTdata input into theSSEBmodel, resulting inhigher,morerepresentativeETvaluesovertheseareas.TherearealsodifferencesintheETresultsforsomeareasofthenorthwestwherevegetationwithhighETmaybebenefitingfromsnowmelt/soilmoisture/groundwaterstorageprocessesduringtheApril–Octobergrowingseason.Thismaybeanotherexampleofimpactsofnonrain-fall-relatedmoistureinputsthatarecapturedbytheSSEBmodelbutnotbytheVegETmodel,whichisdrivenbyrainfallaloneanddoesnotaccountforsnowmeltorrunoff.Significantdifferences in theVegETandSSEBETestimatesoverMinnesotaandWisconsinrequirefurtherinvestigation.Extensivelakes,wetlands,andnear-surfacegroundwatercontributionstotheevaporativefluxmaybecontributingtothehigher


ETfluxespredictedbySSEBinnorthernMinnesota,butingeneral,differencesinthisareamayreflectregionalsurfacepropertiesandland-atmospherecouplingsthatarenotproperlyaccountedforinthesimplifiedenergybalanceapproach.

6.3.2 eta anomalies

SeasonalETanomalymapsforbothVegETandSSEBmodelsarepresentedinFigure6.3canddfor2009,respectively.TheseveredroughtinsouthTexasandthesouth-westUnitedStates, inpartsofArizonaandCalifornia, isclearlydepictedinbothmaps, where below-average conditions (<50% normal ET) prevailed. These areasalsocomparewellwiththedroughtdepictionbytheU.S.DroughtMonitor(USDM)ofmoderatetoseveredroughtformuchofthegrowingseason(datanotshownbutavailable at http://drought.unl.edu/dm/archive.html). In contrast, above-averagemoistureconditionsareindicatedbybothmodelsformuchoftheHighPlainsregionspanningpartsofNorthDakotatowesternTexas.Similarabove-averagemoisture

0 500km

1000 N

0 500km

1000 N

VegETET (mm)

0–5050–100100–200200–300300–400400–500500–600600–700700–800800–900900–13001300–1300No data

(a)

(b)

VegETanomaly (%)

0–5050–7070–9090–110110–130130–150>150No data

FIGURE 6.3 (See color insert.)Seasonal(April–October)totalETa(mm)andETanoma-lies(%)forCONUSin2009:(a)seasonalVegETETa,(b)seasonalSSEBETa,(c)seasonalVegETETaanomaly,and(d)seasonalSSEBETaanomaly.

(continued)

0 500km

1000 N

0 500km

1000 N

VegETET (mm)

0–5050–100100–200200–300300–400400–500500–600600–700700–800800–900900–13001300–1300No data

(a)

(b)

VegETanomaly (%)

0–5050–10070–9090–110110–130130–150>150No data

SSEB ET (mm)0–5050–100100–200200–300300–400400–500500–600600–700700–800800–900900–13001300–1300No data

(c)

(d)

0 500km

1000 N

SSEB ET(%)

50–7070–80

0–50

90–110110–130130–150>150No data

0 500km

1000 N

FIGURE 6.3 Seasonal (April–October) total ETa (mm) and ET anomalies (%) for CONUS in 2009: (a) seasonal VegET ETa, (b) seasonal SSEB ETa, (c) seasonal VegET ETa anomaly, and (d) seasonal SSEB ETa anomaly.


conditionswerealsodetectedovermuchofthenorthernpartofthesemiaridwest-ernUnitedStatesinbothmodelresults.BothoftheseresultsareconsistentwiththeUSDM,whichassignedmostoftheseareasanondroughtdesignationovertheyear.

Althoughthereisageneralagreementbetweenthetwomaps,someregionsexhibitsignificantdiscrepancies,includingtheareasinMinnesotaandWisconsinthatwerehighlightedintheprevioussection.WhileSSEBindicatesnormalconditions,VegETsuggestsETisbelowaveragefromtheviewpointofrainfalldistributionduringthegrowingseason.ThisisinagreementwiththeUSDM,whichclassifiedtheregionasexperiencinghydrologicaldrought formuchof the2009growingseason.Thissuggeststhatthewaterbalanceandenergybalanceapproachesmayberespondingdifferently tovaryinghydrologicprocesses that affect the timing,magnitudeandseverityofagriculturalandhydrologicaldroughts.

Theobvioustexturaldifferenceinthespatialpatternsrepresentedinthetwomapsresultsfromdifferencesinspatialresolution.TheSSEBismodeledat1kmwhiletheVegETisproducedat10km,butthisshouldnotaffectresultsandconclusionsmadeataregionalscale.Theseresultsillustratethepotentialforbothapproachestogener-atevaluableETinformationforoperationaldroughtmonitoring,butmoreinvestiga-tionisrequiredtobetterunderstandtheETestimationdifferencesbetweenthetwomodelingapproachesanddeterminehowtheycanbestbeusedascomplementarydatasources.

SSEB ET (mm)0–5050–100100–200200–300300–400400–500500–600600–700700–800800–900900–13001300–1300No data

(c)

(d)

0 500

km

1000 N

SSEB ET(%)

50–7070–80

0–50

90–110110–130130–150>150No data

0 500km

1000 N

FIGURE 6.3 (continued)


6.3.3 case stuDy: seasonal et time seRies

Todemonstrate the relativeutilityof theETandanomalyproductsgeneratedby theSSEBandVegETmodels,acounty-basedanalysiswasconductedfortwoselectedcoun-tieswithcontrastingconditionsin2009.Onecountywaslocatedinadrought-affectedpartofsouthTexas(DuvalCounty)andanotherfromcentralNebraska(CusterCounty),whichhadabove-averagerainfalloverthegrowingseasonthatyear.

Figure6.4aandbshowmonthlyETtotalsandanomalies,respectively,fromthetwomodelsforthetwocounties.AcloserlookatFigure6.4ashowsthatbothmodels,asexpected,predicthigherETforCusterCountythanfordrought-strickenDuval

160

Monthly ET (mm)Custer County, NE and Duval County, Texas

ET anomaly (%)Custer County, NE and Duval County, Texas

140

120

Mon

thly

actu

al E

T (m

m)

Mon

thly

ET

anom

aly (

%)

100

80

60

Apr(a)

(b)

May Jun JulMonths, 2009

Months, 2009

Aug Sep Oct

40

20

0

200

150

100

50

0

VegET NE

VegET TXSSEB TX

SSEB NE

VegET NE

VegET TXSSEB TX

SSEB NE

Apr May Jun Jul Aug Sep Oct

FIGURE 6.4 Monthlycounty-averageET(mm)andthecorrespondinganomalies(%)fortwoselectedcountiesinthecentralUnitedStatesin2009usingVegETandSSEBmodels:(a)monthlytotalsforCusterCounty,Nebraska,andDuvalCounty,Texas,and(b)monthlyanomaliesforCusterandDuval.


County.BetteragreementbetweenmodeloutputswasobtainedinCusterCounty,withaseasonalmonthlyaverageof74mmfrombothmodels.Incontrast,themodelsgavesignificantlydifferentresultsforDuvalCounty,withVegETandSSEBestimat-ingseasonalmonthlyaveragesof30and63mm,respectively.

ThisillustratespotentialdifferencesbetweentheVegETandSSEBapproachestoETestimationanddroughtmonitoring.Accordingtoamapofirrigatedagriculturallandareafor theUnitedStates(Brownetal.,2009),agriculture inCusterCountyis generally under a predominantly rain-fed system, and rainfall moisture inputsare well represented in both modeling approaches. In comparison, Duval Countyappearstocontainahigherfractionofirrigatedlandareaintheirrigatedagricul-turallandmap.TheincreasedETduetoappliedirrigationwaterinDuvalCountywouldbecapturedbyLST-basedSSEBbutisnotaccountedforintheVegETmodel.ThelargestdifferencebetweentheVegETandSSEBETcurvesforDuvalCountyoccurinJuneandJuly,whichisgenerallyatimeofpeakirrigationformostcrops.BySeptember,whenirrigationisnotasreadilyusedandmostoftheETismetbyrainfall,theVegETandSSEBETresultswereinbetteragreement(within12%).ThisresultsuggeststhatacomparisonofSSEBandVegETmapsmaybeavaluabletoolforidentifyingirrigatedagriculturalareas.Furthermore,inprinciple,thedifferencebetweenthetwoapproachesmaybeusedtoestimatetheamountofirrigationthatisapplied(i.e.,theSSEBETwouldprovidethetotalETirrespectiveofthewatersourcewhiletheVegETETacanaccountfortheamountofETsuppliedbyrainfall).

AlthoughmonthlyETtotalsinabsolutetermsareimportantforagrohydrologicanalysis,wecannotinferfromtheplotsinFigure6.4awhetherthecountiesareinadroughtorhowseverethemoisturedeficitsmightbe.ThemonthlyETanomaliesforbothcountiespresentedinFigure6.4bareamorevaluabletoolforthisapplication.TheanomalieswerecalculatedbycomparingthemonthlyETin2009tothehistori-cal10yearmedianmonthlyETvalues(2000–2009)forthesamemonth.ThisplotshowsthatDuvalCountyhadbelow-averageETduring2009,reflectingtheobserveddrought,whiletheETforCusterCountywasaboveaverageformostoftheseasonbecause of more favorable weather conditions. Furthermore, the anomalies frombothmodelsareinbetteragreementthanarethemonthlyETtotals,whichfurtherillustratesthevalueofusingETanomalyinformationindroughtdetection.DespitethelargediscrepanciesobservedbetweentotalmonthlyETfromVegETandSSEB(30mmvs.63mm)forDuvalCounty,theseasonalmonthlyanomaliesare62%and65%forVegETandSSEB,respectively(Figure6.4b).

6.4 APPLICATIONS OF VegET AND SSEB FOR THE FAMINE EARLY WARNING SYSTEM NETWORK

Thelivelihoodofmostruralpopulationsinsub-SaharanAfricaisbasedontraditionalrain-fedagriculturethatisdependentonseasonalrainfall.KnowledgeofcropwaterusageandsoilmoisturestatusthatcanbeobtainedthroughremotelysensedETprod-uctsprovidesvaluableinformationformanagingwaterresourcesandanticipatingcropfailure (Tadesse et al., 2008). The FEWS NET (http://earlywarning.usgs.gov/fews/)


has developed various tools that use readily available satellite-derived and model-assimilatedweatherdatasetstomonitorhealthandproductivityofrain-fedagricul-turalareas.Basedoneaseofimplementationandminimalinputdatarequirements,theVegETandSSEBmodelsarebeingintegratedasoperationalmonitoringtoolswithintheFEWSNETsystem.

Asnotedearlier, theVegETmodelhasitsoriginsintheoriginalFAOWRSI(FAO, 1986). The operational FEWS crop water balance model uses the sameprinciplesofFAOmethodinthecalculationsoftheWRSIvaluesbasedonregion-specificcropcalendarsbutparameterizes thecalculationofETasafunctionofsoilmoisture.VegETfurther improves theFEWScropwaterbalancemodelbyintroducing a location-specific crop water-use coefficient that is derived fromremotelysenseddata.

AlthoughtheAfricaoperationalcropwaterbalancemodelisstillrunningwithaprescribedcropcalendar,aplanisunderwaytointegratetheVegETparameteriza-tion,withamoreobjectivevegetationcalendarbasedonremotelysenseddatathatisspecifictoalocationinsteadofaregion.Inlightofthis,initialworkwasdonetoapplytheVegETmodeltoestimatetheNileBasinwaterbalancedynamics(Senayetal.,2009),highlightingthepotentialoftheapproachnotonlyindroughtmonitor-ingbutalsoforhydrologicstudies.Figure6.5aandbcompareVegETETaestimateswithmeanannualprecipitationover thebasin,derived fromsatellite-based rain-fall estimate (Xie and Arkin, 1997). As expected, high and low rainfall regionsinFigure6.5ashowcorrespondinghighand lowET, respectively, inFigure6.5bas the result of rainfall and vegetation cover. Note that VegET does not capturetheeffectsofintenseirrigationthatoccursalongtheNileRiver,particularlyattheDeltainEgyptwheretheNileRiveremptiesintotheMediterraneanSea(extremenorth).ThediagnosticLSTinputstoSSEBhandlethisbetter throughthe“total”ETestimationapproachinsteadoftherainfall-drivenwaterbalancemodels(datanotshown).

WiththeFEWSNETprincipleofrelianceonaconvergenceofevidence,USGS/FEWSNETjustlaunchedanoperationalimplementationoftheSSEBETmodelingapproachtocomplementtheexistingwaterbalanceproductsusingaMODISdatastream for the entire African continent. An operational ET anomaly product hasbeenproducedandstagedattheFEWSNETwebsite(http://earlywarning.usgs.gov/fews/africa/index.php)sinceJune2011.ThenewSSEBproductsconsistofmonthlyand cumulative ET anomalies at 1km resolution. A sample product is shown inFigure6.6,highlightingtheseveredrought(upto<50%ofnormal)ineastAfricaasaseasonalanomalybetweenJanuary1andJuly3,2011(mostrecentavailabledata).The product shows an above-average ET (>110% of normal) in parts of southernAfricaandnormalconditions(rangingbetween90%and100%ofnormal)inmuchofAfrica,includingtheirrigatedbasinoftheNileRiverDelta.Irrigatedareastendtoshownormalconditionsfromyeartoyearsinceirrigationapplicationisnotaffectedbytheyear-to-yearvariabilityofrainfallaslongasthewaterissourcedfromlargereservoirs,asisthecasefortheNileRiverDelta,whichisregulatedbytheAswanHighDam/LakeNasser.


Nile BasinCountriesLakesRivers

50–200200–300300–500500–700700–900900–11001100–13001300–15001500–18001800–2500

0 400

(a)

(b)

800 1200 1600 km

<= 50

>= 2500

Annual rainfall (mm)


50–100100–200200–300300–400400–500500–600600–700700–800800–900900–1300

0 400 800 1200 1600 km

<= 50

>= 1300

Annual ETa (mm)

FIGURE 6.5 (See color insert.) Spatial distribution of satellite-derived annual rainfallinnortheasternAfrica (medianof2001–2007) (a)andannualETa from theVegETmodel(medianfromthesameperiodastherainfall)(b).


50–200200–300300–500500–700700–900900–11001100–13001300–15001500–18001800–2500

0 400

(a)800 1200 1600 km

<= 50

>= 2500

Annual rainfall (mm)

(b)


50–100100–200200–300300–400400–500500–600600–700700–800800–900900–1300

0 400 800 1200 1600 km

<= 50

>= 1300

Annual ETa (mm)

FIGURE 6.5 Spatial distribution of satellite-derived annual rainfall in northeastern Africa (median of 2001–2007) (a) and annual ETa from the VegET model (median from the same period as the rainfall) (b).

Legend

Cumulative evapotranspiration (ETa) anomalyJanuary 01–July 03, 2011

RiversLakesCountry boundaries

50–7070–9090–110110–130130–150

ETa anomaly (%)< 50

> 150

0 250 500 750 Nkm

FIGURE 6.6 Africa-wide seasonal anomaly of ETa from the SSEB model output for 2011 as of July 3, 2011 (January 1–July 3). SSEB ET anomaly is operationally processed and posted on a FEWS NET website regularly on an 8 day time step.


6.5 CONCLUSIONS

ThemainobjectiveofthischapterwastodemonstratetheuseofremotelysenseddatainsimplifiedenergyandwaterbalancemodelingapproachestoestimatingETfor drought monitoring and agrohydrologic applications. Both VegET and SSEBmodelswereabletocapturethegeneralspatialpatternsofseasonalETovermuchoftheCONUS.However,notabledifferenceswereobservedbetweentheirseasonalETtotals,particularlyoverlocationswherewatersourcesotherthanrainfall(e.g.,irriga-tion,snowmeltrunoff,andsubsurfaceirrigationfromhighwatertables)areavailabletovegetation.Theanomalymapsprovedtobemoreusefulindetectingyear-to-yearchangesthanseasonalETtotals,whicharepronetoerrorsassociatedwithdataandmodelassumptionsandsimplifications.

Althoughbothapproachescanprovidecomparableresultsfordroughtmonitor-ingusingtheanomalyproducts,eachmayhaveuniqueadvantagesinsomespecificapplicationsandlocations.Forexample,thewaterbalanceapproach(rainfallbased)

Cumulative evapotranspiration (ETa) anomalyJanuary 01–July 03, 2011

RiversLakes

Legend

Country boundaries

50–7070–909–110110–130130–150

ETa anomaly (%)< 50

>150

0 250 500 750 Nkm

FIGURE 6.6 (See color insert.) Africa-wide seasonal anomaly of ETa from the SSEBmodeloutputfor2011asofJuly3,2011(January1–July3).SSEBETanomalyisoperation-allyprocessedandpostedonaFEWSNETwebsiteregularlyonan8daytimestep.


provides more information on temporal soil moisture variability and runoff as aby-productoftheVegETmodel,whichisusefulforotherhydrologicapplications.Incomparison,theSSEBETmodelingismoreusefulforquantifyingETfromnon-rain-fed systems such as irrigation and groundwater-fed vegetation systems sinceSSEBETestimatesETregardlessof thewatersource.However,someof thedif-ferencesexhibitedbetweenVegETandSSEBrequirefurtherinvestigationtofullyunderstandtheprocessesinvolvedanddeterminesynergisticapplications.

Thisstudyhighlightsthatsimplifiedmodelingtechniquesandparameterizationthatusereadilyavailableglobalsatellitedataandmodel-assimilatedweatherdatasetscanbe implementedeffectively inanoperationalsetupfor timelyassessmentofdroughthazardsandmonitoringagrohydrologicconditionsindata-poorpartsoftheworld.Recently,FEWSNEThasimplementedanoperationalsetupoftheSSEBoverAfricaforagriculturalmonitoringinAfricausingtheMODISdatastreamaspartoftheconvergenceofevidenceprincipleadvocatedbyFEWSNET.BecauseoftheglobalnatureoftheinputdatasetstoboththeSSEBandVegETmodels,thereareopportunitiestoexpandtheseproductsindifferentpartsoftheworld.Fielddataarerequiredtovalidateandcalibratethesemodelsbeforeusingtheproductsinanabsolutesenseforwaterbalanceapplications.However,themodelscanproducereli-ableanomalyproductsthatcanbeusedfordroughtdetection.

ACKNOWLEDGMENT

Anyuseoftrade,firm,orproductnamesisfordescriptivepurposesonlyanddoesnotimplyendorsementbytheU.S.Government.

REFERENCES

Allen,R.G.,L.S.Pereira,T.A.Howell,andM.E.Jensen.2011.Evapotranspirationinformationreporting:1.Factorsgoverningmeasurementaccuracy.Agricultural Water Management98:899–920.

Allen,R.G.,L.Pereira,D.Raes,andM.Smith.1998.Crop Evapotranspiration.FoodandAgricultureOrganization(FAO)Publication56.Rome,Italy:FAOoftheUnitedNations.

Allen,R.G.,M.Tasumi,A.Morse,andR.Trezza.2005.ALandsat-basedenergybalanceandevapotranspirationmodelinWesternUSwaterrightsregulationandplanning.Irrigation and Drainage Systems19:251–268.

Allen,R.G.,M.Tasumi,A.Morse,R.Trezza,W.Kramber,I.Lorite,andC.W.Robison.2007.Satellite-basedenergybalanceformappingevapotranspirationwithinternalizedcalibration(METRIC)–Applications.Journal of Irrigation and Drainage Engineering133:395–406.

Anderson,M.C.,J.M.Norman,J.R.Mecikalski,J.A.Otkin,andW.P.Kustas.2007.Aclima-tologicalstudyofevapotranspirationandmoisturestressacrossthecontinentalUnitedStatesbasedonthermalremotesensing:1.Modelformulation.Journal of Geophysical Research112:D10117,doi:10.1029/2006JD007506.

Bastiaanssen,W.G.M.,M.Menenti,R.A.Feddes,andA.A.M.Holtslag.1998.Aremotesens-ing surface energy balance algorithm for land (SEBAL): 1. Formulation. Journal of Hydrology212(213):213–229.

Bastiaanssen,W.G.M.,E.J.M.Noordman,H.Pelgrum,G.Davids,B.P.Thoreson,andR.G.Allen.2005.SEBALmodelwithremotelysenseddatatoimprovewater-resourcesmanagementunderactualfieldconditions.Journal of Irrigation and Drainage Engineering131:85–93.


Brown,J.F.,S.Maxwell,andM.S.Pervez.2009.MappingirrigatedlandsacrosstheUnitedStatesusingMODISsatelliteimagery.InRemote Sensing of Global Croplands for Food Security. eds.P.S.Thenkabail, J.G.Lyon,H.Turral,andC.M.Biradar,pp.177–198.BocaRaton,Florida:Taylor&Francis.

deBeurs,K.M. andG.M.Henebry. 2005.A statistical framework for the analysis of longimagetimeseries.International Journal of Remote Sensing26(8):1551–1573.

Eidenshink, J.C. 1990. The 1990 conterminous US AVHRR data set. Photogrammetric Engineering and Remote Sensing58:809–913.

FAO.1986.Yield Response Function.FAOIrrigationandDrainagePaper3.Rome:FoodandAgriculturalOrganizationoftheUnitedNations.

FAO.1998.Crop Evapotranspiration: Guideline for Comparing Crop Water Requirements.FAOIrrigationandDrainagePaper56.Rome:FoodandAgriculturalOrganizationoftheUnitedNations.

Goward, S.N., C.J. Tucker, and D.G. Dye. 1985. North American vegetation patternsobservedwiththeNOAA-7advancedveryhighresolutionradiometer.Plant Ecology64(1):3–14.

Gowda,P.H.,G.B.Senay,T.A.Howell,andT.H.Marek.2009.Lysimetricevaluationofsim-plifiedsurfaceenergybalanceapproachintheTexasHighPlains.Applied Engineering in Agriculture25:665–669.

Groeneveld,D.P.,W.Baugh,J.Sanderson,andD.Cooper.2007.Annualgroundwaterevapo-transpirationmappedfromsinglesatellitescenes.Journal of Hydrology344:146–156.

Kanamitsu,M.1989.DescriptionoftheNMCglobaldataassimilationandforecastsystem.Weather and Forecasting4:335–342.

Nagler,P.,R.Scott,C.Westenburg,J.Cleverly,E.Glenn,andA.Huete.2005.Evapotrans-pirationonwesternU.S. riversestimatedusing theEnhancedVegetationIndexfromMODISanddatafromeddycovarianceandBowenratiofluxtowers.Remote Sensing of Environment97:337–351.

PRISM. 2011. PRISM Products Matrix. PRISM Climate Group, Oregon StateUniversity, Corvallis, Oregon. http://www.prism.oregonstate.edu/products/matrix.phtml?vartype=tmax&view=maps(accessedonDecember12,2011).

Reed,B.C.,J.F.Brown,D.VanderZee,T.R.Loveland,J.W.Merchant,andD.O.Ohlen.1994.Measuringphenologicalvariabilityfromsatelliteimagery.Journal of Vegetation Science5(5):703–714.

Senay,G.B.2008.Modelinglandscapeevapotranspirationbyintegratinglandsurfacephenol-ogyandawaterbalancealgorithm.Algorithms1:52–68.

Senay,G.B.,K.O.Asante,andG.A.Artan.2009.WaterbalancedynamicsintheNileBasin.Hydrological Processes23:3675–3681.

Senay,G.B.,M.E.Budde,andJ.P.Verdin.2011a.EnhancingtheSimplifiedSurfaceEnergyBalance(SSEB)approachforestimatinglandscapeET:ValidationwiththeMETRICmodel.Agricultural Water Management98:606–618.

Senay,G.B.,M.Budde,J.P.Verdin,andA.M.Melesse.2007.Acoupledremotesensingandsimplifiedsurfaceenergybalanceapproachtoestimateactualevapotranspirationfromirrigatedfields.Sensors7:979–1000.

Senay, G.B., S. Leake, P.L. Nagler, G.Artan, J. Dickinson, J.T. Cordova, and E.P. Glenn.2011b. Estimating basin scale evapotranspiration (ET) by water balance and remotesensingmethods.Hydrological Processes(inpress).DoI:10.1002/hyp.8379.

Senay, G.B. and J.Verdin. 2003. Characterization of yield reduction in Ethiopia usinga GIS-based crop water balance model. Canadian Journal of Remote Sensing29(6):687–692.

Senay, G.B., J.P. Verdin, R. Lietzow, and A.M. Melesse. 2008. Global daily referenceevapotranspiration modeling and evaluation. Journal of American Water Resources Association44:969–979.


Su,H.,M.F.McCabe,E.F.Wood,Z.Su,andJ.Prueger.2005.ModelingevapotranspirationduringSMACEX:Comparingtwoapproachesforlocalandregionalscaleprediction.Journal of Hydrometeorology6(6):910–922.

Tadesse,T.,M.Haile,G.Senay,B.D.Wardlow,andC.L.Knutson.2008.Theneedforintegra-tionofdroughtmonitoringtoolsforproactivefoodsecuritymanagementinsub-SaharanAfrica.Natural Resources Forum32(4):265–279.

Tucker,C.J.,D.A.Slayback, J.E.Pinzon,S.O.Los,R.B.Myneni, andM.G.Taylor. 2001.Higher northern latitude normalized difference vegetation index and growing seasontrendsfrom1982to1999.International Journal of Biometeorology45:184–190.

UNEP.2010.Africa Water Atlas.DivisionofEarlyWarningandAssessment(DEWA).UnitedNationsEnvironmentalProgramme(UNEP).Nairobi,Kenya.

Verdin,J.andR.Klaver.2002.Gridcellbasedcropwateraccountingfor thefamineearlywarningsystem.Hydrological Processes16:1617–1630.

Xie,P.andP.A.Arkin.1997.A17-yearmonthlyanalysisbasedongaugeobservations,satel-liteestimates,andnumericalmodeloutputs.Bulletin of the American Meteorological Society78(11):2539–2558.

Documents

Remote Sensing of Evapotranspiration for Operational