基于随机森林模型的呼中冻土区土壤碳空间分布特征OA北大核心

Taking the typical frozen soil area of Da Xing'an Ling Huzhong as example,the spatial distibution law of soil carbon storage in the suface layer(0-15,15-30 cm)was studied.Through field investigation,sampling and indoor anlysis,the reflectance of seven bands based on Landsat 8 OLI images was modeled using the Random Forest(RF)algorithm to analyze the spatial distribu-tion of surface soil organic carbon in the permafrost region of the Da Xing'an Ling.The highest correlation variable was selected among 13 remote sensing variables,and six environmental variables of altitude,ARVI,VIGreen,EVI,OSAVI,NDVI were identi-fied as the highest correlation variables and could be used as independent variables.The results showed that with a leaf value of 5 and a tree value of 900,the training and test sets of the prediction model were closest,indicating that the model stability was the best at this time.The fitting accuracy of training set R2 of 0-15,15-30 cm soil layer was 0.23 and 0.43,the fitting accuracy of test set R2 was 0.35 and 0.24.Using this parameter to predict the spatial distribution characteristics of soil carbon stock in the surface layer(0-15,15-30 cm),0-15 cm and 15-30 cm average soil carbon density was 7.40 kg/m2 and 14.80 kg/m2,respectively.The distribution of soil carbon content in 0-15 cm soil layer gradually increased from north to south,and there was no significant difference in soil carbon distribution in 15-30 cm soil layer,but slightly higher in the south than in the north.