论文部分内容阅读
利用遥感技术反演土壤水分对于我国西北地区农业干旱问题研究具有重要意义。该文以新疆焉耆盆地为研究区域,分别利用微波遥感数据(Sentinel-1ASAR)和光学遥感数据(Landsat8)计算土壤后向散射系数(σ0soil)和改进型温度植被干旱指数(MTVDI),并将σ0soil和MTVDI参数作用于支持向量机(SVM)回归算法,探讨了不同参数条件下SVM模型在土壤水分反演中的适应性。实验结果表明,相比只用单因子(σ0soil或MTVDI)作为模型参数,以σ0soil和MTVDI两者共同作为SVM模型输入参数时,土壤水分监测精度显著提高,其建模集决定系数R2=0.81,均方根误差RMSE=3.16%;验证集R2=0.89,RMSE=3.15%。最后,利用最优模型对研究区土壤水分进行了反演,并对不同土地类型含水量进行了评价,可为光学遥感与微波遥感协同反演土壤水分提供参考。
Retrieving soil moisture using remote sensing technology is of great significance to the study of agricultural drought in the northwest of China. In this paper, the soil backscattering coefficient (σ0soil) and modified temperature vegetation drought index (MTVDI) were calculated by using the remote sensing data (Sentinel-1ASAR) and the optical remote sensing data (Landsat8) in the Yanqi Basin, Xinjiang. And MTVDI parameters on the support vector machine (SVM) regression algorithm to explore the adaptability of SVM model in soil moisture inversion under different parameters. The experimental results show that compared with the single factor (σ0soil or MTVDI) as the model parameter and both σ0soil and MTVDI as the input parameters of the SVM model, the soil moisture monitoring accuracy is significantly improved. The determination coefficient of the modeling set R2 = 0.81, Root mean square error RMSE = 3.16%; validation set R2 = 0.89, RMSE = 3.15%. Finally, the optimal model was used to invert the soil moisture in the study area and evaluate the water content of different land types. This study may provide a reference for the synergistic retrieval of soil moisture by optical remote sensing and microwave remote sensing.