基于多源遥感协同反演的区域性土壤含水率动态研究

Dynamic Study of Regional Soil Moisture Content Based on Multi-source Remote Sensing Co-inversion

为进一步推动多源遥感技术在农业生产与管理中的应用,以内蒙古河套灌区五原县3500 hm 2研究区为实验对象,对实验区进行实地取样化验,获取并处理了多种遥感影像,综合研究光学影像地表反射率及雷达影像后向散射系数对土壤含水率的响应特征.对遥感影像数据进行各种数学变换,求出其与土壤含水率的相关性强弱;利用SPSS与MATLAB对数据进行回归分析,采用多元线性回归、最小二乘回归、神经网络回归等不同的回归方法进行分析,得出相对应的回归模型方程,建立了多数据源含水率反演模型.ALOS,北京二号和Landsat影像协同反演土壤含水率神经网络的拟合值达到0.892,R 2值达到0.796,模型的稳定性和预测精度较好,可知融合多源遥感数据的神经网络模型可快速精准监测土壤含水率的动态变化规律,为灌区农业生产及土壤盐碱化防治提供基础信息指导.

To further promote the application of the multi-source remote-sensing technology on soil moisture monitoring in cropland soils,a study area with 3500 hectares located at Wuyuan County was used to measure soil moisture and correct multiple remote-sensing images at Hetao Irrigation District,Inner Mongolia,China.The responsive characteristics of soil moisture to the optical image-based surface reflectance and the radar image-based backscatter coefficient were determined.The correlations between remote-sensing image data and soil moisture content were obtained based on the mathematical transformations of remote-sensing image data.Multi-linear regression,least square regression,neural network regression and other regression methods were used to analyze these data with Statistic Package for Social Science(SPSS)and MATLAB.Regression models related to soil moisture content and remote-sensing image data were established for obtaining favorable multi-source image-based soil moisture content inversion models.The neural network regression model of co-retrieval of soil moisture content with ALOS Image,Beijing No.2 Image and Landsat Image gave a good fitting value(0.892)and a coefficient of determination(R 2)of 0.796.The above results demonstrated that the neural network model integrating multi-source remote-sensing data can quickly and accurately predict the dynamic changes of soil moisture and,thus,provide basic information for agricultural production and management of soil salinization and alkalization.