Improving Spatial Soil Moisture Representation Through Integration of AMSR-E and MODIS Products

The use of microwave observations has been highlighted as a complementary tool for evaluating land surface properties. Microwave observations are less affected by clouds, water vapor, and aerosol and also contain valuable soil moisture information. However, a critical limitation in microwave observations is the coarse spatial resolution attributed to the complex retrieval process. The objective of the current study is to develop an independent (from ground observations) downscaling approach that merges information from higher spatial resolution MODerate-resolution Imaging Spectroradiometer (MODIS) (similar to 1 km) with lower spatial resolution AMSR-E (similar to 25 km) to obtain soil moisture estimates at the MODIS scale (similar to 1 km). We compare the developed (UCLA) method against a range of previous published approaches. Various key factors (i.e., surface temperature, vegetation indexes, and albedo) derived from MODIS provide information on relative variations in surface wetness conditions and contribute weighting parameters for downscaling the larger AMSR-E soil moisture footprints. Evaluation of the various downscaled soil moisture products is undertaken at the SMEX04 site in southern Arizona. Results show that the UCLA downscaling technique, as well as the previously published Merlin method, significantly improves the limited spatial variability of the current AMSR-E product. Spatial correlation (R) values improved from -0.08 to 0.34 and 0.27 for the Merlin and UCLA methods, respectively. The evaluated triangle-based methods show poorer performance over the study domain. Results from the current study yield insight on the integration of multiscale remote sensing data in various downscaling methods and the usefulness of MODIS observations in compensating for low-resolution microwave observations.