基于GEE的洱海流域土地利用/覆被分类算法对比研究OACSTPCD

In complex highland and mountainous areas,rapid and accurate long-term automatic land cover classification can serve as a foundation for land planning and resource utilization.In this paper,four spatial datasets,namely Landsat image surface reflectance,vegetation index,water body index,and DEM,were se-lected as the basis and supporting data for land cover classification by using the GEE platform.Three clas-sification algorithms,CART,RF,and SVM,were applied to automatically extract and compare the accuracy of land cover type information in the Erhai watershed.The results showed that 1)the classification accura-cy of RF was the highest,and the overall accuracy of SVM was the lowest among the three classification al-gorithms.RF was a optimal classification algorithm for LULC in the Erhai basin.2)The use of supporting data sets further improved the accuracy of the interpretation.The selection of sample points was the key in-fluence.3)Erhai_RF was capable of achieving higher accuracy.Meanwhile,the detailed features were more prominent.It will be more accurate in terms of local actual classification.This study can provide methodo-logical and technical support for intelligent and rapid extraction of long-time series land cover data products and screening of optimal classification algorithms in the Erhai watershed.