简介：Forestsareamongthemostimportantcarbonsinksonearth.However,theircomplexstructureandvastareasprecludeaccurateestimationofforestcarbonstocks.Datasetsfromforestmonitoringusingadvancedsatelliteimageryarenowusedininternationalpolicyagreements.DatasetsenabletrackingofemissionsofCO2intotheatmospherecausedbydeforestationandothertypesofland-usechanges.TheaimofthisstudyistodeterminethecapabilityofSPOT-HRGSatellitedatatoestimateabovegroundcarbonstockinadistrictofDarabkolaresearchandtrainingforest,Iran.Preprocessingtoeliminateorreducegeometricerrorandatmosphericerrorwereperformedontheimages.Usingclustersampling,165sampleplotsweretaken.Of165plots,81wereinnaturalhabitats,and84wereinforestplantations.Followingthecollectionofgrounddata,biomassandcarbonstockswerequantifiedforthesampleplotsonaperhectarebasis.Nonparametricregressionmodelssuchassupportvectorregressionwereusedformodelingpurposeswithdifferentkernelsincludinglinear,sigmoid,polynomial,andradialbasisfunction.Theresultsshowedthatathird-degreepolynomialwasthebestmodelfortheentirestudiedareashavinganrootmeansquareerror,biasandaccuracy,respectively,of38.41,5.31,and62.2;42.77,16.58,and57.3%forthebestpolynomialfornaturalforest;and44.71,2.31,and64.3%forafforestation.Overall,theseresultsindicatethatSPOTHRGsatellitedataandsupportvectormachinesareusefulforestimatingabovegroundcarbonstock.