基于MODIS干旱指数与RBFNN方法的江苏冬小麦需水关键期土壤水分遥感监测应用

利用遥感指数反演土壤水分已成为监测干旱的重要手段之一，而单一的遥感干旱指数对于反演土壤水分存在一定局限，本研究从7种不同MODIS遥感干旱监测指数中选取适宜的5种并结合径向基函数神经网络（RBFNN）协同反演江苏省2018年冬小麦需水关键期的土壤相对湿度。结果表明：与单一的遥感干旱指数相比，协同RBFNN的遥感干旱指数反演的模型效果更好，与10 cm和20 cm深度的实测土壤相对湿度的相关系数分别达到0.5161和0.4307，能综合多种通道的遥感信息反映当地土壤水分的变化；同时研究利用RBFNN对2017年5月江苏冬小麦10 cm深度土壤相对湿度进行反演，得到的土壤相对湿度分布图与实测土壤墒情结果较为接近，说明利用RBFNN反演模型有效。研究结果提高了土壤湿度的反演精度，为当地农业干旱的实时监测提供了新思路。

Retrieval of soil moisture at critical period of water demand of winter wheat in Jiangsu Province using MODIS drought index and RBFNN

The use of remote sensing indices to retrieve soil moisture has become one of the most important means for monitoring drought. However, a single remote sensing drought index has certain limitations in retrieving soil moisture. In this study, 5 suitable MODIS remote sensing drought monitoring indices were selected. Combined with Radial Basis Function Neural Network (RBFNN), the soil relative moisture in the critical water demand period of winter wheat in Jiangsu Province in 2018 was retrieved synergistically. The research showed that the soil moisture model retrieved by RBFNN collaborative remote sensing index had better inversion effect than a single remote sensing index. The correlation coefficients with the measured soil relative humidity at different depths of 10 cm and 20 cm reach 0.5161 and 0.4307, respectively, which integrated the remote sensing information of multiple channels to reflect the change of local soil moisture. Meanwhile, the relative moisture distribution map of winter wheat at the depth of 10 cm retrieved by using RBFNN in May 2017 of Jiangsu Province was close to the observed soil moisture results, which indicates that the inversion model was effective. The research results improve the inversion accuracy of soil moisture and provide a certain service reference for local soil moisture inversion.