The relevant feature parameters extracted from the domestic ZY1-02D multispectral satellite were used to characterize the soil salinity over the coastal area of Qinzhou Bay with the support of Ada Boot,LightGBM,XGBoost,RFR,and CatBoost machine leaming algorithms.The performance of each model was evaluated with the coefficient of determination(R2)and root mean square error(RMSE).The results show that the soil total salt content in the research area was measured to range from 0.740 to 10.352g/kg with an average of 1.739g/kg.Model simulation results demonstrate that CatBoost had the best predictive performance over AdaBoost,LightGBM,XGBoost,and RFR,and combined CatBoost with the highest accuracy(R2=0.8317,RMSE=0.396g/kg);and of all variables in a group,the mean of texture features was most sensitive to soil salinity and made the highest contribution;The soil salt content was simulated to range from 0to 8.784g/kg,with an average of 2.478g/kg,in which mild salinity mainly occurred in the western part of the study area and scattered in the eastem part.The combination of domestic resource satellite remote sensing data and CatBoost model has shown good performance in retrieving soil salinity in the coastal area of Qinzhou Bay,providing a new approach to characterizing coastal soil salinity at a large-scale.
关键词
土壤盐分/遥感定量反演/机器学习/资源一号/钦州湾
Key words
soil salinity/remote sensing quantitative inversion/machine learning/ZY-limagery/Qinzhou bay
维普
万方数据