Modeling Of Soil Water Content Inversion Based On Full Polarization Radar Data

At present,the problem of land desertification in China is becoming more and more serious,especially in Northwest China.Land degradation makes the local economic development and ecological environment severely tested.As an important indicator to evaluate and monitor the phenomenon of land desertification,soil moisture data is one of the urgent problems to be solved.In this paper,the bird island area and Gangcha county area around Qinghai Lake Basin in Northwest China are taken as the research areas,and the radarsat-2 full polarization radar data and the measured soil water content data in the same period are used as the data sources/ α The54 polarization parameters obtained from polarization decomposition were used as independent variables to participate in PLS and GA-PLS modeling.The sensitive information of soil water content in polarization parameters was identified and extracted,so as to establish the regression model of soil water content in the basin around Qinghai Lake.The research results are as follows1.In the modeling research of PLS algorithm and GA-PLS algorithm,the two models can effectively deal with the samples with more independent variables and identify the information sensitive to the changes of dependent variables.Combined with the prediction results of the two models,the analysis shows that the two modeling methods can effectively retrieve the soil water content of the basin around Qinghai Lake.PLS algorithm and GA-PLS algorithm are feasible in the field of soil water content inversion.2.Through the comparative analysis of the prediction results of the two models,it shows that the PLS model has higher inversion accuracy and better stability in the study area.The root mean square error of the prediction model is 0.06563 and the fitting accuracy is 0.639;The root mean square error of the prediction model in Gangcha county is 0.07599,and the fitting accuracy is 0.608.3.According to the results of the prediction model,this paper uses PLS algorithm to get the inversion results of soil water content in bird island area and Gangcha county area.The soil water content of the two areas is mainly in the range of 0-10%,and the total soil water content of Niaodao area is higher than that of Gangcha county.Combined with the geographical environment analysis of the two areas,the main reason may be that the precipitation of bird island area is larger than that of Gangcha county area when the image is obtained,and the terrain of bird island area is lower.4.After the end of the modeling,through the analysis of the results,this paper considers that the factors that affect the modeling results of PLS and GA-PLS are soil basic environment,sample size,the selection of calibration set and validation set,and the selection of independent variables.In general,the two models have good application prospects in the inversion of soil water content.