资源描述：
《《agenericframeworkforspatialpredictionofsoilvariablesbasedonregression-kriging.》》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。
1、Geoderma120(2004)7593www.elsevier.com/locate/geodermaAgenericframeworkforspatialpredictionofsoilvariables$basedonregression-krigingTomislavHengla,*,GerardB.M.Heuvelinkb,AlfredSteinaaInternationalInstituteforGeo-informationScienceandEarthObservation(IT
2、C),P.O.Box6,7500AAEnschede,TheNetherlandsbLaboratoryofSoilScienceandGeology,WageningenUniversity,P.O.Box37,6700AA,Wageningen,TheNetherlandsReceived27February2003;receivedinrevisedform11July2003;accepted19August2003AbstractAmethodologicalframeworkforsp
3、atialpredictionbasedonregression-krigingisdescribedandcomparedwithordinarykrigingandplainregression.Thedataarefirsttransformedusinglogittransformationfortargetvariablesandfactoranalysisforcontinuouspredictors(auxiliarymaps).Thetargetvariablesarethenfi
4、ttedusingstep-wiseregressionandresidualsinterpolatedusingkriging.Agenericvisualisationmethodisusedtosimultaneouslydisplaypredictionsandassociateduncertainty.Theframeworkwastestedusing135profileobservationsfromthenationalsurveyinCroatia,dividedintointe
5、rpolation(100)andvalidationsets(35).Threetargetvariables:organicmatter,pHintopsoilandtopsoilthicknesswerepredictedfromsixreliefparametersandninesoilmappingunits.Predictionefficiencywasevaluatedusingthemeanerrorandrootmeansquareerror(RMSE)ofpredictiona
6、tvalidationpoints.Theresultsshowthattheproposedframeworkimprovesefficiencyofpredictions.Moreover,itensurednormalityofresidualsandenforcedpredictionvaluestobewithinthephysicalrangeofavariable.Fororganicmatter,itachievedlowerrelativeRMSEthanordinarykrig
7、ing(53.3%versus66.5%).Fortopsoilthickness,itachievedalowerrelativeRMSE(66.5%versus83.3%)andalowerbiasthanordinarykriging(0.15versus0.69cm).ThepredictionofpHintopsoilwasdifficultwithallthreemethods.Thisframeworkcanadoptbothcontinuousandcategoricalsoilv
8、ariablesinasemi-automatedorautomatedmanner.ItopensapossibilitytodevelopabundlealgorithmthatcanbeimplementedinaGIStointerpolatesoilprofiledatafromexistingdatasets.D2003ElsevierB.V.Allrightsreserved.Keywords:Spatialprediction;Logittransformation
