基于无人机-卫星遥感升尺度的土壤水分监测模型研究

土壤水分是研究土壤-植物-大气循环系统中能量与物质交换的关键，通过尺度转换方法将无人机遥感数据上推以修正卫星数据，可有效改善卫星遥感反演模型精度。本文以河套灌区为研究对象，分别采用重采样和TsHARP升尺度法，引入多元线性回归(MLR)、BP神经网络(BPNN)和支持向量机(SVM)算法构建不同土壤深度下无人机-卫星升尺度土壤含水率反演模型。研究结果表明：重采样升尺度法在不同土壤深度下模型整体精度由高到低依次为SVM、MLR、BPNN,其中在土壤深度0～60 cm下采用SVM模型最优，R²达到0.571,RMSE为0.022%;TsHARP升尺度法在不同土壤深度下模型整体精度由高到低依次为BPNN、SVM、MLR,其中在土壤深度0～60 cm下采用BPNN模型最优，R²达到0.829,RMSE为0.015%。与升尺度修正前对应土壤深度模型对比，两种升尺度方法均能明显提高卫星遥感对土壤含水率的反演精度，但TsHARP升尺度法整体优于重采样法；重采样法的R²由0.413提升至0.571,RMSE由0.026%降至0.022%...

Soil Moisture Monitoring Model Based on UAV-Satellite Remote Sensing Scale-up

Soil moisture is the key to the study of energy and material exchange in the soil-plant-atmosphere circulatory system. Using the scale conversion method to push up the remote sensing data from the UAV to correct the satellite data can effectively improve the accuracy of the satellite remote sensing inversion model. Taking Hetao Irrigation Area as the research object, the resampling and TsHARP scale-up method were adopted respectively, and algorithms such as multiple linear regression (MLR), BP neural network (BPNN) and support vector machine (SVM) were introduced to construct UAV-satellite scale-up soil moisture content inversion model under different soil depths. The research results showed that the overall accuracy of the resampling scale-up method was SVM, MLR and BPNN from high to low in different soil depths, among which the SVM model was the best when the soil depth was 0~60cm, R2 was 0.571, and RMSE was 0.022%. The overall accuracy of the model of TsHARP scale-up method under different soil depths was BPNN, SVM and MLR from high to low, among which the BPNN model was the best under 0~60cm soil depth, R2 was 0.829, and RMSE was 0.015%. Compared with the corresponding soil depth model before scaling up, both scale-up methods can significantly improve the retrieval accuracy of soil moisture content from satellite remote sensing, but TsHARP scale-up method was better than resampling method as a whole; R2 of resampling method was increased from 0.413 to 0.571, and RMSE was decreased from 0.026% to 0.022% (a decrease of 15.4%); R2 of TsHARP scale-up method was increased from 0.428 to 0.829, and RMSE was decreased from 0.025% to 0.015% (a decrease of 40.0%). The research result can provide theoretical and technical support for high precision monitoring of soil moisture in large-scale irrigation areas.