基于sailh模型的參數(shù)化模型及其應(yīng)用

基于sailh模型的參數(shù)化模型及其應(yīng)用

1非織造工藝參數(shù)laincipalitorpoelinglai（leaf區(qū)指數(shù)）是將拉米區(qū)域或從側(cè)面區(qū)域中拉米區(qū)域的特征之一（chen，1992）。拉米是一個(gè)緩慢的區(qū)域，有一個(gè)簡(jiǎn)單的區(qū)域，沒(méi)有時(shí)間，有一個(gè)接一個(gè)的結(jié)構(gòu)。它是一個(gè)不同于微電機(jī)的過(guò)程，微電機(jī)是一個(gè)緩慢的過(guò)程。體積學(xué)、離體和微電機(jī)的特征是可以概括的。它可以被視為微電機(jī)進(jìn)程中的一個(gè)步驟，微電機(jī)可以被視為一個(gè)步驟中的一個(gè)步驟，微電機(jī)可以被視為一個(gè)步驟中的一個(gè)步驟。Atpresent,thequantitativeLAIretrievalmethodsusingremotesensingaremainlyclassifiedastheempiricalstatisticalmodelbasedmethodandthephysicalmodelinversionbasedmethod.TheempiricalstatisticalmodelsusethevegetationindexasthevariableandretrieveLAIbyestablishingthestatisticalrelationshipbetweenVIsandLAI(Fang&Zhang,2003).ThismethoddominatestheLAIretrievalduetoitssimpleexpressionandlessparameters.Otherwisethismethodislackofphysicalmeaning,andthedevelopedrelationshipsbetweentheVIsandLAIaredependentonthespecifictimeandspecificregion.Asaresult,itishardtoexpanditsuseintemporalandspatialdimensions.Onthecontrary,thephysicalmodelsaremorerealisticandapplicable.Alltheparametersofthemodelshavephysicalmeanings.Therearesomeshortcomingsinphysicalmodels,includingtoomuchparameterswhichareintroducedintothemodelsinordertodescribetheearth’ssurfacefeaturesaccuratelyasmuchaspossible,andtheirexpressionarealsoverycomplicated.Althoughthemulti-spectralandmulti-angleremotesensingtechnologyprovidesmoreobservationalinformation,theremotesensingretrievalstillfacestheill-posedproblem,andthephysicalmodelretrievalistooslowtomeetawiderangeofapplicationrequirements.Therefore,itisimportanttofindaparameterizedmodelwithcertainphysicalmeaningandsimpleformatthesametimeforthegenerationoflargeareaLAIproducts.Inthispaper,weproposesaparameterizedmodelbasedontheSAILHmodel,andthenwecomparetheretrievalaccuracy,efficiencyandstabilitybetweentheparameterizedmodelandtheSAILHmodelwithsimulateddataandground-basedmeasurementswhicharetakenintheHeiheriverbasin.2標(biāo)準(zhǔn)4.3.3回亞國(guó)里亞國(guó)里亞國(guó)里亞國(guó)里羅胺政府分類(lèi)性別與同人民群眾sigsifeci能夠sofracienra能,sofracienra能,sofracienra能,sofracienra能,soinratiinra能,sofunsiinra能,soinratiin,etizact.sigCanopyreflectancemodelscanbedividedintothreekindsofmodelinaccordancewithwhetheritisestablishedaccordingtothetheoreticalanalysis,thatistheempiricalmodels,semi-empiricalmodelsandphysicalmodels(Zhao,2007),whilethephysicalmodelscanbedividedintoradiativetransfermodels,geometricopticsmodels,andcomputersimulationmodels.SAILHmodelisonekindoftheradiativetransfermodels.Thebasisoftheradiativetransfermodelsistheradiativetransferequation,whichisanintegral-differentialequation.Intheory,theequationissolvableiftheboundaryconditionsaredetermined.Sofar,thereisnorigorousanalyticalsolutiontotheradiationtransferequationbutonlyavarietyofapproximatesolutions(Xu,2006).ThecommonlyusedsolutionistheK-MequationproposedbyKubelkaandMunk.TheincidentlightisdividedintofourpartsintheSAILmodel(ScatteringbyArbitrarilyInclinedLeaves)(Verhoef,1984)basedontheK-Mequation,thataretheupwardanddownwardradiationfluxdensityaswellastheupwardanddownwardparallelradiationirradiation.Bysolvingnineintermediatevariables(includingtheextinctioncoefficient(ks)ofthedirectradiationfluxdensity,attenuationcoefficient(att),back-scatteringcoefficient(sig),forwardandbackwardscatteringcoefficientofthedirectradiation(sf,sb),conversioncoefficientsfromupward,downwardradiationfluxdensityandupwardparallelradiationtotheobservedradiance(uf,ub,ω),theextinctioncoefficientoftheradiationfluxdensityontheobserveddirection(ko)),thecanopyreflectancecanbegot.SoSAILmodelisafour-streamlineardifferentialequationwithninecoefficients.Itsinputparametersincludethreestructuralparametersandfourspectralparameters,wherethethreestructuralparametersaretheleafareaindexandthetwoparameterswhichareusedtodescribetheleafinclinedangles.Thefourspectralparametersaretheratioofskylight,leafreflectance,leaftransmittanceandsoilreflectance.AstheSAILmodelcannotsimulatethehot-spoteffectwell,Kuuskaddedthehot-spoteffecttotheSAILmodelanddevelopedtheSAILHmodel(Kuusk,1991).Hotspoteffectiscausedbythesinglescatteringoftheilluminatedcanopy.Therefore,Kuuskestablishedtherelevantprobabilitymodelbetweenthedirectionofobservationandthedirectionoftheincidentlighttoconsiderthecanopyhot-spoteffectbasedontheSAILmodel.InSAILHmodel,thecontributionofthesinglescatteringofthewholecanopyisdecomposedintomulti-layersingle-scatteringcontributionandsummedtogetherasthecontributionofthewholecanopy.Inthecalculationofthesinglescatteringcontributionofeachlayer,abi-directionaltransmissiondensityfunctionisused.Thebi-directionaltransmissiondensityfunctioncalculationneedstointroduceanewparameter—thehot-spoteffectfactor.ThenthereareeightinputparametersinSAILHmodel(SeeTable1,themeaningsoftheseparameterscanbefoundintheappendix).3sailhmoligactrafteribution、非織造業(yè)的slain建模Inpracticalapplications,remotesensingisthemosteffectivetechniquewhichisabletogetlarge-scaleandtimeseriesLAIofdifferentsurfacetypes.AlargenumberofstudieshaveshownthatusingremotesensingtechnologycanextractregionalandglobalLAIrapidlyandperiodically,andcanprovidespatialandtemporaldistributionofLAI(Huietal.,2003).AlthoughSAILHmodelisawidelyacceptedmodeltodescribethedirectionalreflectanceofcontinuousvegetationcanopy,itspracticalityislimitedduetoitscomplicatedcalculationandlowretrievalefficiency.Soinordertoimprovetheretrievalefficiencyandkeeptheaccuracy,itisnecessarytomakesomereasonablesimplificationstothemodel.WeproposedaparameterizedmodelbasedonSAILHmodelinthispaper,andevaluatedtheretrievalaccuracy,efficiencyandstabilityoftheparameterizedmodelbasedonthesimulateddataandground-basedmeasureddata.Thisworkaimedatthefollowingtwosteps:(1)SimplifyingthecalculationofthenineintermediatevariablesoftheSAILHmodel.(2)Simplifyingthecalculationthecontributionofthesinglescatteringoftheilluminatedcanopy.3.1規(guī)訓(xùn)4.4—Simplificationofthecalculationofthenineinter-mediatevariablesoftheSAILHModelThemainreasonofthelowretrievalefficiencyofSAILHmodelisthatwhencalculatingthecanopyreflectance,theleafinclinedanglesofcontinuouscanopyaredividedinto13anglesandthenineintermediatevariablesareexpressedasthefunctionsoftheseleafinclinedangles,thentheresultiscalculatedbysummingthese13termsiteratively.IfwecansimplifythecalculationsofthesenineintermediatevariablesandkeeptherestcalculationsinSAILHmodel,wemayimprovetheefficiencygreatly.ThroughanalyzingandcomparingthecalculationofthenineintermediatevariablesinSAILHmodel,wehavethesimplifiedcalculationsshowninTable2,where,LAIistheleafareaindex,Inthecalculationofnineintermediatevariables,themoscomplexoneisω,whichindicatesthatiftheilluminationdirectionandtheobserveddirectionintersectatthesamesideoftheleaf,theleafreflectanceshouldbeusedtorepresentthereflectionoftheleaftothelight,otherwisetheleaftransmittanceshouldbeusedtorepresenttheweakenoftheleaftothelight.Inthissense,ωrepresentsasimilarconceptwiththescatteringphasefunction,butthedifferenceisthatitcontainsLAI,sointhecalculationlater,weusethescatteringphasefunctiontoapproximatethecalculationofThetwo-parameterellipticdistributionisusedtodescribetheleafinclinedangledistributioninSAILHmodel.Inordertomakeabetterapproximationtothescatteringphasefunctionofthisdistribution,wechoosethescatteringphasefunctionofsphericaldistributiontoreplaceit.Wecangetabetterapproxi-mationusingthisscatteringphasefunctioninsteadof3.2出適數(shù)字號(hào)和出口品種Supposedthatthecontinuousvegetationcanopyiscomposedbyaseriesofhorizontalanduniformlayers,thenthecontributionofsinglescatteringoftheilluminatedcanopycPwhereΓ(ΩInEq.(3),PwherePs_Listhehotspotfactorparameter,thecalculationofΔis:TotestthecorrectnessoftheEq.(4),wecomparedtheresultcalculatedbyitwiththeresultoftheSAILHmodel(Fig.2).PThen,thecontributionofthesinglescatteringoftheilluminatedcanopycanbecalculatedasfollows:Sofar,thecalculationsofthenineintermediatevariablesandthecontributionofsinglescatteringoftheilluminatedcanopyhavebeensimplified,andtheothercalculationskeepthesameasSAILHmodel.WedesigneagroupofexperimentstocompareandevaluatetheforwardBRDFsimulatedabilityoftheparameterizedmodelWesetthesunzenithangleat35°andthesolarazimuthangleat0°,theobservationzenithanglesarechangesfrom0°to85°withanincrementof5°,andtheobservationazimuthanglesarefrom0°to355°withthesameintervalchangesof5°,sothereareatotalof1296observationcombinationsintheupwardhemisphere.ThevaluesoftheinputstructuralparametersareshowninTable3.Wedesigntheparametersforthefollowingthreeconsiderations:(1)GroupⅠ:theellipticaleccentricityissetat0.001,inwhichcasetheleafinclinedangledistributionneartoasphericalleafangledistribution,theforwardBRDFsimulationabilityoftheparameterizedmodelisevaluatedwhenLAI=2andLAI=4,respectively;(2)GroupⅡ:theellipticaleccentricityissetat0.9,andtheaverageleafinclinedangleissetatdifferentvalues,whenθTheRMSEsandthecorrelationcoefficientsoftheforwardsimulatedBRDFoftheentireupwardhemispherebetweentheparameterizedmodelandSAILHmodelareusedastheevaluatecriterion.TheresultsareshowninTable4.Throughanalysiswecanseethat,intheentireupwardhemisphere,theRMSEsoftheforwardsimulatedBRDFbetweentheparameterizedmodelandSAILHmodelcanbemaintainedbelow0.01intheredband,andbelow0.04inthenear-infraredband;whetherinredornear-infraredband,thecorrelationcoefficientsoftheparameterizedmodelandSAILHmodelhasreachedmorethan96%.4通過(guò)mehateritydmortgage,grounge/sailhmortrace-acculacemortracemortrace-solega-acculacemorymeasulace.ThereisahighpositivecorrelationbetweentheBRDFsimulationsbetweentheparameterizedmodelandSAILHmodel,sotheSAILHmodelcanberepresentedbytheparameterizedmodeltosomedegree.Thenweevaluatedtheretrievaaccuracy,efficiencyandstabilityofthesetwomodelsusingsimulateddataandground-basedmeasureddata.4.1prior-to保護(hù)的alga:si農(nóng)村天監(jiān),非價(jià)值的自然價(jià)值Thegenerationofsimulateddataisasfollows:supposingagrouporseveralgroupsofinputparameters,givingtheappropriatepriorknowledge,takingtheSAILHforwardsimulatedvaluesastheobservedtruevalues,andthenretrievingparametersusingtheparameterizedmodelwithPowelloptimizationalgorithm.Forthepurposeofcomparison,wealsoretrievedtheLAIvaluesusedtheSAILHmodelwiththesamepriorknowledgeatthesametime.Weadded10%Gaussiannoisetotheobservedvaluesduringretrieval.4.1.1retefteriecevi接法laiWeusetheabsoluteerrorsbetweentheretrievedLAIvaluesandtheinputLAIvaluesinforwardsimulationastheevaluationcriterionoftheretrievalaccuracy.Fig.3showstherelationshipbetweentheinputLAIvaluesandtheretrievedLAIvalues,inwhichtheXaxisisfortheinputLAIvalues,andtheYaxisfortheretrievedLAIvalues.Fig.4indicatestherelationshipoftheretrievalabsoluteerrors’changeswiththeinputLAIchanges.FromFig.3andFig.4wecanconcludethat:thereisnogreatdegreelossoftheretrievalaccuracyfortheparameterizedmodelcomparedwithSAILHmodel;fromtheabsoluteerrorterms,themaximumretrievalerrorsis0.180and0.122inredandnear-infraredbandsrespectivelyforSAILHmodel,whilethemaximumretrievalerrorswas0.179and0.197inredandnear-infraredbandsrespectivelyfortheparameterizedmodel.4.1.2價(jià)格retexThemethodweusetoevaluatetheretrievalefficiencyoftheparameterizedmodelisasfollows:using6observationsofdifferentanglesineachretrievals,settingdifferentretrievaltimesof50times,100times,200timesand500timesrespectively,andthencomputingtherunningtime.Eachtestisrepeatedforfivetimes,andthentaketheaveragedrunningtimeastheevaluationcriterionoftheretrievalefficiency.TheresultsareshowninFig.5.WhatwecanseefromFig.5isthat,inthetermsofretrievalefficiency,theretrievalefficiencyissignificantlyimprovedby10timesoftheparameterizedmodelcomparedwithSAILHmodel.4.1.3relacketin-sitching,sig/sailityForthestabilityevaluationofthemodel,weusethefollowingmethods:intheretrieval,wecomparedtheretrievalaccuracyofLAIbyincreasingthenoiseinthesimulateddata(choosetherelativeerrorastheevaluationcriterion).TheresultsareshowninFig.6.WecanseefromFig.6that,theretrievalerrorisbiggerasthenoiseincreases,andthestabilityoftheparameterizedmodelissuperiortoSAILHmodelwhetherintheredbandorthenear-infraredband.4.2index,measunen,lai,lai,lai,lai,lai,lai,lai,lai,lai,lai,lai,lai,lai,lai,lai,lai,lai,lai,lai,lai:3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.4.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.Theparameterizedmodelisevaluatedbyusingtheground-basedmeasuredcanopyreflectancedataandLAIdatainYingKefromthesatellite-aircraft-groundsynchronousexperimentovertheHeiheriverbasinin2008.ThereweretwomeasurementsetsinJune22andJuly1respectivelyandonlyonemeasurementsetinJuly9intheexperimentalarea.Thus,thereweretotal5groupsofmeasurementsets.WeretrievedtheeightinputparametersmeanwhileusingthesamePowelloptimizationalgorithm.Amongtheseinputparameters,theexpectationsanduncertaintiesofLAI,leafreflectance,leaftransmittanceandsoilreflectancewereobtainedbygroundmeasurements;theexpectationsoftheaverageleafinclinedangleandtheellipticaleccentricitywereretrievedfrommeasuredLADdata,andtheuncertaintiesofthesetwoparametersweresetat0.01;theexpectationsanduncertaintiesofthehot-spoteffectfactorparameterandtheratioofskylightweresetat0.1and0.01,respectively.ThesettingoftheexpectationsanduncertaintiesoftheinputparametersarelistedindetailinTable5.4.2.1respicarespifie-acpo2004和retrie-lai,wspincipalpo底回事件4.InthegroundmeasurementsexceptthattakenonJuly9inHeihefieldcampaign,therewerecanopyreflectancedataoffourplanesineachmeasurementincludingthesolarprincipalplane,theverticalprincipalplane,theparallelrowplaneandthecrossrowplanerespectively(measurementswereonlyavailableatthesolarprincipalplaneandverticalprincipalplaneonJuly9).Inourretrievaltests,firstweusethemeasureddataofeachplanetoretrieverespectively,andthenusethedataofallfourplanestogethertoretrieveLAI.Atlast,wegettheresultsasfollows:comparedwiththegroundmeasuredLAIdata,themaximumretrievalabsoluteerroris1.344ofSAILHmodel,andtheminimumabsoluteerroris0.002;themaximumretrievalabsoluteerroris0.807oftheparameterizedmodel,andtheminimumretrievalabsoluteerroris0.056.Thus,theretrievalerroroftheparameterizedmodelisinanacceptablerange.4.2.2e-terizedmortgageWealsoevaluatetheretrievalefficiencyoftheparame-terizedmodelbyusingthemeasureddata,andgetthesameconclusionthattheretrievalefficiencyisgreatlyimprovedfor8—10timesbytheparameterizedmodel.5sailhmot簡(jiǎn)介Inthispaper,weproposesaparameterizedmodelbasedonSAILHmodel.Inthenewmodel,wesimplifythecalculationofthenineintermediatevariablesinSAILHmodel,andalsosimplifythecalculationofthesinglescatteringcontributionoftheilluminatedcanopyatthesametime,andthenkeeptherestcalculationsthesamewiththeSAILHmodel.Weevaluatestheretrievalaccuracy,efficiencyandstabilityoftheparameterizedmodelwithsimulatedandmeasureddatarespectively,andthenwegetsomeconclusionasfollows.(1)Intheentireupwardhemisphere,thereisastrongcorrelationbetweenthesimulatedBRDFbytheparameterizedmodelandtheSAILHmodel,thecorrelationcoefficientsreachedmorethan96%inbothredandnear-infraredband,andtheRMSEintheredbandcanbemaintainedbelow0.01,whileinthenear-infraredbandcanbemaintainedbelow0.04,whichindicatesthattheparameterizedmodelcanbeusedasthesubstituteofSAILHmodel.(2)Evaluationsbasedonsimulateddataandmeasureddatashowthat:fromthetermsofretrievalabsoluteerror,theretrievalaccuracyoftheparameterizedmodelisequaltoSAILHmodel;buttheretrievalefficiencyisimprovedfor8~10times,meanwhile,thestabilityoftheparameterizedmodelisbetterthanSAILHmodel.However,whenweevaluatetheretrievalaccuracyandefficiencyofthesemodelsbyusingthemeasureddata,theexpectationsanduncertaintiesofsomeinputparametersareassumedbecauseofnoexperimentaldata,whichmightaffecttheretrievalresults.(3)AlthoughtheparameterizedmodelcanmakeverygoodapproximationtoSAILHmodelinsimulatingcanopy’sBRDF,andcangreatlyimprovetheefficiencyofLAIretrievalwhilemaintainingtheretrievalaccuracy,thismodelisbasedontheassumptionthattheleafinclinedangledistributionissphericaldistribution,andonlycanbeappliedtocontinuousvegetationcanopyconditions.Therefore,itsapplicabilityislimitedtosomeextent.論上講,如果確定了邊界條件,那么這個(gè)方程就是可解的。但是到目前為止,還沒(méi)有求得輻射傳輸方程的嚴(yán)格的解析解,只有各種各樣的近似解法(徐希孺,2006)。目前應(yīng)用比較廣泛的解法是Kubelka和Munk提出的K-M方程。SAIL(scatteringbyarbitrarilyinclinedleaves)(Verhoef,1984)模型是在K-M方程的基礎(chǔ)上將入射源分為向上和向下傳輸?shù)妮椛渫棵芏?以及向上和向下傳輸?shù)钠叫休椛漭椪斩?個(gè)部分,通過(guò)求解9個(gè)系數(shù)(包括直射輻射通量密度的削弱系數(shù)(ks),消光系數(shù)(att),背向散射系數(shù)(sig),直射輻射的前向、后向散射系數(shù)(sf,sb)、向上、向下傳輸?shù)妮椛渫棵芏?、向上傳輸?shù)钠叫休椛涞妮椪斩认蛴^測(cè)方向上傳輸?shù)妮椛淞炼鹊霓D(zhuǎn)換系數(shù)(uf,ub,ω)、觀測(cè)方向上的輻射通量密度的削弱系數(shù)(ko)進(jìn)而求得冠層的反射率,因此SAIL模型是一個(gè)4流9參數(shù)的線性微分方程組。它的輸入?yún)?shù)包括3個(gè)結(jié)構(gòu)參數(shù)和4個(gè)光譜參數(shù)。3個(gè)結(jié)構(gòu)參數(shù)分別是葉面積指數(shù)和描述葉傾角分布的2個(gè)參數(shù);4個(gè)光譜參數(shù)分別是天空光比例,葉片反射率,葉片透過(guò)率和土壤反射率。由于SAIL模型不能很好的對(duì)熱點(diǎn)效應(yīng)進(jìn)行模擬,因此Kuusk(1991)在SAIL模型的基礎(chǔ)上加入熱點(diǎn)效應(yīng)從而發(fā)展了SAILH模型。熱點(diǎn)效應(yīng)是由葉簇對(duì)光的一次散射造成的,因此Kuusk在SAIL模型的基礎(chǔ)上通過(guò)建立光線方向與觀測(cè)方向間間隙率的相關(guān)概率模型考慮冠層的熱點(diǎn)效應(yīng)。在SAILH模型中整個(gè)冠層的單次散射被分解為多層冠層的單次散射貢獻(xiàn)之和;在計(jì)算每層冠層的單次散射的貢獻(xiàn)時(shí)采用了雙向透過(guò)率密度函數(shù)計(jì)算其單次散射的貢獻(xiàn)。該雙向透過(guò)率密度函數(shù)的計(jì)算需要引入一個(gè)新的參數(shù)——熱點(diǎn)效應(yīng)因子,這樣SAILH模型的輸入?yún)?shù)共8個(gè)(表1)。3sailh模型的建立在實(shí)際應(yīng)用中,需要不同地表類(lèi)型的大范圍、長(zhǎng)時(shí)間的LAI,遙感技術(shù)是實(shí)現(xiàn)這一目標(biāo)的途徑。大量的研究表明,利用遙感技術(shù)可以快速、大范圍、周期性地提取區(qū)域乃至全球的LAI,并能夠提供LAI空間和時(shí)間的分布狀況(惠鳳鳴等,2003)。SAILH模型雖然是被廣泛接受的描述均勻植被冠層方向性反射率的模型,但是其計(jì)算過(guò)程復(fù)雜,反演效率不高,實(shí)用性不強(qiáng)。因此以在保證模型反演精度的基礎(chǔ)上提高模型的反演效率為目的,對(duì)模型進(jìn)行合理的簡(jiǎn)化是非常必要的。本文從這個(gè)目的出發(fā),提出了一種基于SAILH模型的參數(shù)化模型,分別基于模擬數(shù)據(jù)和實(shí)測(cè)數(shù)據(jù)對(duì)該參數(shù)化模型的反演精度、反演效率和程序穩(wěn)定性進(jìn)行了評(píng)價(jià)。3.1簡(jiǎn)化計(jì)算過(guò)程SAILH模型的反演效率很低,主要原因是SAILH模型在計(jì)算冠層反射率時(shí)將連續(xù)植被冠層的葉傾角離散為13個(gè)角度,將這9個(gè)中間變量分別表示為葉傾角的函數(shù),通過(guò)對(duì)這13個(gè)角度進(jìn)行迭代求和計(jì)算得到。如果能夠?qū)⑦@9個(gè)中間變量的計(jì)算過(guò)程簡(jiǎn)化,其余的計(jì)算過(guò)程仍然采用SAILH模型中的計(jì)算過(guò)程,那么SAILH模型的計(jì)算效率會(huì)得到提高。通過(guò)比較SAILH模型中9個(gè)中間變量的計(jì)算過(guò)程,得到的簡(jiǎn)化計(jì)算過(guò)程如表2。在表2中,LAI為葉面積指數(shù),在9個(gè)中間變量中,計(jì)算過(guò)程最復(fù)雜的就是ω,其表示的含義是如果光照方向和觀測(cè)方向相交于葉片的同側(cè),則用葉片的反射率表示葉片對(duì)光的反射作用如果光照方向和觀測(cè)方向相交于葉片的異側(cè),則用葉片的透過(guò)率表示葉片對(duì)光的削弱作用。從這個(gè)意義上講,ω表示的是與散射相函數(shù)類(lèi)似的一個(gè)概念,不同的是它包含了LAI,因此在后面的計(jì)算中,用散射相函數(shù)近似計(jì)算SAILH中的所使用的SAILH模型中的葉傾角分布函數(shù)是用雙參數(shù)橢圓分布描述的,為了能夠?qū)υ摲植嫉娜~傾角的散射相函數(shù)進(jìn)行比較好的近似,選擇用球形分布的散射相函數(shù)代替3.2參數(shù)化模型的模擬能力評(píng)價(jià)假設(shè)連續(xù)植被冠層由一系列的水平的均勻?qū)盈B加而成,各層之間相互獨(dú)立,那么整個(gè)冠層對(duì)直射光的單次散射的貢獻(xiàn)式中,Γ(Ω式中,P式中,P其中:s_L為熱點(diǎn)效應(yīng)因子,Δ的計(jì)算公式如下:為了檢驗(yàn)式(4)的計(jì)算結(jié)果是否正確,將其與SAILH模型的計(jì)算結(jié)果比較,如圖2。P得到了P至此,9個(gè)中間變量及光照冠層單次散射貢獻(xiàn)的計(jì)算過(guò)程已得到簡(jiǎn)化,其余的計(jì)算過(guò)程仍然使用SAILH的計(jì)算過(guò)程。設(shè)計(jì)了一組簡(jiǎn)單的實(shí)驗(yàn)參數(shù)化模型正向模擬BRDF的能力進(jìn)行評(píng)價(jià)和比較。假設(shè)太陽(yáng)天頂角為35°;太陽(yáng)方位角為0°;觀測(cè)天頂角從0°—85°,以5°的角度間隔變化;觀測(cè)方位角從0°—355°,同樣以5°的角度間隔變化;這樣整個(gè)上半球空間總共有1296個(gè)觀測(cè)角度組合。結(jié)構(gòu)參數(shù)的輸入值如表3。參數(shù)的設(shè)計(jì)是出于以下3個(gè)方面的考慮:(1)第1組:將橢圓離心率ε設(shè)定為0.001,這種情況下葉傾角的分布接近于球形分布,分別比較當(dāng)LAI=2和LAI=4時(shí)參數(shù)化模型正向模擬BRDF的能力;(2)第2組:將橢圓離心率ε設(shè)定為0.9,平均葉傾角θ(3)第3組:其他參數(shù)設(shè)置同(2),比較當(dāng)LAI=4時(shí)參數(shù)化模型的正向模擬能力。對(duì)于光譜參數(shù),將紅光波段的天空光比例、葉片的反射率、葉片透過(guò)率和土壤反射率分別固定為0.1,0.1,0.12和0.1,近紅外波段的值則分別固定為0.1,0.45,0.5和0.2。對(duì)參數(shù)化模型的模擬能力進(jìn)行評(píng)價(jià)的指標(biāo)為參數(shù)化模型與SAILH模型正向模擬的整個(gè)上