Downscaling SMAP Soil Moisture Data Using MODIS Data

Abstract

Soil moisture level is an important index in studying environmental changes. High resolution soil moisture data is in high demand for agricultural and weather forecasting purpose. Current daily large-scale soil moisture projects fail to provide sufficient resolution for medium or small region research. To acquire high-resolution soil moisture data, different kinds of methods are put into practice, including multivariate statistical regression, weight aggregation and so on. In this research, SMAP (Soil Moisture Active Passive) level 3 data with 36-km resolution are successfully downscaled by MODIS (Moderate Resolution Imaging Spectroradiometer) 1-km LST (Land Surface Temperature) product, NDVI (Difference Vegetation Index) product, SRTM (Shuttle Radar Topography Mission) DEM (Digital Elevation Model), and TWI (Topographic Wetness Index). Three regression models are built based on these supplemental indexes correlated with the SMAP retrieval. All downscaled results are validated with SMAPVEX15 field data. The research aims to establish and validate the multivariate regression method for downscaling low-resolution remote sensing image (such as SMAP) with local field observations. Based on the validation results, the research suggests the regression models have a decent fit. The downscaled soil moisture data indicating the method is applicable to small region research.