Zheng_2012_Remote-Sensing-of-Environment

Remotesensingofcropresiduecoverusingmulti-temporalLandsatimageryBaojuanZhenga,⁎,JamesB.Campbella,KirstenM.deBeursa,baDepartmentofGeography,VirginiaTech,115MajorWilliamsHall,Blacksburg,VA24061,USAbDepartmentofGeographyandEnvironmentalSustainability,TheUniversityofOklahoma,100E.BoydStreet,Norman,OK,USAabstractarticleinfoArticlehistory:Received5August2011Receivedinrevisedform26September2011Accepted29September2011Availableonline4November2011Keywords:LandsatRemotesensingCropresidueTillageNormalizeddifferencetillageindexMulti-temporalTillagepractices,whichhavedirectimpactsonsoilandwaterquality,havechangeddramaticallyduringthepastseveraldecades.Tillageinformationisoneoftheimportantinputsforenvironmentalmodeling,buttheavailabilityofthisinformationisstilllimitedspatiallyandtemporally.Previousstudieshaveencountereddif-ficultiesindefiningreliablecorrelationsbetweencropresiduecover(CRC)andLandsat-basedtillageindicesbecausetheyneglectedthesignificanceofthetimingoftillageimplementation.Thisstudyexploresrelation-shipsbetweentemporalchangesofagriculturalsurfacesandthenormalizeddifferencetillageindex(NDTI)inCentralIndiana.WefoundthatminimumNDTI(minNDTI)valuesextractedfrommulti-temporalNDTIprofilesreliablyindicatethesurfacestatuswhentillageorplantingoccurred.Simplelinearregressionrevealsacoefficientofdetermination(R2)of0.89betweenCRCandminNDTIforcalibration.Inaddition,apercent-agechange(PC)methodwastestedforclassifyingCRCintothreecategories(CRCb30%;30%bCRCb70%;CRC70%).BoththeminNDTIandPCmethodsresultedinoverallclassificationaccuraciesof90%,produ-cer'saccuraciesof83–100%,anduser'saccuraciesof75–100%.OurresultsindicatedthatbothLandsatTMandETM+imageryarecapableofmappingCRC,however,multi-temporalLandsatimageryisrequired.Toestablishacapabilityforcropresiduemapping,designersoffutureremotesensingplatformsshouldconsiderincreasingtemporalresolution.©2011ElsevierInc.Allrightsreserved.1.IntroductionAgriculturalbestmanagementpractices,suchasconservationtill-ageandcovercrops,havebeenadoptedwidelyinrecentyears.Thebenefitsofconservationtillagearesubstantial,includingimprove-mentofsoilandwaterquality,reductionofsoilerosion,andmaximi-zationofagriculturalwateruseefficiency(USDA-NRCS,2001).Reliableandsystematicsite-specificconservationtillagedatadonotcurrentlyexist,butwouldformanimportantresourcesupportingtheevaluationoftheeffectivenessofthesepractices.Non-conservationtillage(intensive/conventionalandreducedtill-age)leaveslessthan30%cropresiduecover(CRC),whileconservationtillageleavesmorethan30%CRC(CTIC,2010).CurrentCRCdataarenotsurveyedsystematicallyandvaryfromonelocationtoanother.TheUSDANaturalResourcesConservationService(NRCS)collectsCRCdatavisuallyusingaline-transectmethod(Morrisonetal.,1993).TheConservationTechnologyInformationCenter(CTIC)providesassess-mentsofconservationtillagepractices,butcollectsdatausingannualroadsidesurveysofcropresiduelevels,whichissubjective.Itstillagedataareavailableatcounty,state,andregionallevels.Thecounty-leveldatawererecentlyaggregatedto8-digitHydrologicUnit(HU)wa-tersheds(Baker,2011).TheNationalAgriculturalStatisticsService(NASS)datareliesonsurveyrespondentsandisonlyavailableatstateandcountylevel.Theseinventorydataareeithertoocoarse(i.e.,theycannotprovidefieldleveldetail,norreportwithin-fieldspatialvariabil-ity),orareinconsistent,addingmoreuncertaintiesintheenvironmen-talmodelingprocess.Thespatialandtemporalgapsintheseinventorydatarestrictourabilitytosimulatetheimpactofcropmanagementonwaterqualityorcarbonsequestrationatbroadscales(Jareckietal.,2005;Saseendranetal.,2007).Thus,thereisastrongneedtodevelopmethodstomonitoragriculturalpracticesoverlargeareas,overtime,usingconsistentmethods.Multispectralremotesensingoffersanop-portunitytosystematicallyobtaininformationdescribingcropresiduesefficientlyandobjectivelyoverbroadareas.EarlyattemptstouseremotesensingtechniquesformappingCRCcanbetracedbackto1975(Gausmanetal.,1975).Sincethen,thepo-tentialofremotesensingofcropresiduehasbeeninvestigatedbothinthelaboratoryandinthefield(Biard&Baret,1997;Daughtry,2001;Daughtryetal.,1995;Sullivanetal.,2007,2006).Remotesens-ingtillageindices,suchasthecropresidueindexmultiband(CRIM)(Biard&Baret,1997),thecelluloseabsorptionindex(CAI)(Daughtry,2001),andcropresiduecoverindex(Sullivanetal.,2006)aredesignedinthelaboratorytoamplifythedifferencesinthespectralsignalsbetweencropresiduesandsoils(Table1).Mosttillageindicesarebasedonthecelluloseandligninabsorptionfeaturesnear2100nm.Researchershaveappliedthesetillageindices(Table1)toairborne(Daughtryetal.,2005)andsatelliteremotesensingimagery(Daughtryetal.,2006;Gowdaetal.,2003;Serbinetal.,2009a;Sullivanetal.,2008;Thomaetal.,2004;vanDeventeretal.,1997).RemoteSensingofEnvironment117(2012)177–183⁎Correspondingauthor.Tel.:[phone].E-mailaddress:[e-mail](B.Zheng).0034-4257/$–seefrontmatter©2011ElsevierInc.Allrightsreserved.doi:10.1016/j.rse.2011.09.016ContentslistsavailableatSciVerseScienceDirectRemoteSensingofEnvironmentjournalhomepage: