Soil Moisture Estimation by Radar Data Based on Multiple Regression

Abstract

The task of soil moisture assessing by remote (satellite) methods remains relevant up to the present time. Regression models based on the correlation of radar data with ground measurements of soil moisture are constructed to assess soil moisture in the work. Ground-based measurements were taken at two stations in Germany (Falkenberg and Gevenich), which are part of the ISMN (International Soil Moisture Network) worldwide soil moisture data network. Sentinel-1 satellite data is used as radar data. Multiple regressions with a determination coefficient up to 0.91 are constructed. It is proposed to use not only radar, but also meteorological data in regressions, which allows increasing the coefficient of determination and reducing the standard error of regression. For the possible spread of regressions obtained for one territory to another territory, two criteria are selected: the proximity of the values of the Selyaninov hydrothermal coefficient and the similarity of the soil texture. According to these conditions, two stations in the Ryazan region and in Kalmykia were chosen, whose archival information on soil moisture is contained in the ISMN database up to 1998. Each of the selected stations satisfies only one of the criteria.