An approach to validate soil moisture derived from passive microwave sensors using SAR as an interface

Abstract

Validation of soil moisture derived from space-borne passive microwave sensors remains a challenging task owing to the large variability in soil moisture within the coarse-resolution cell offered by passive sensors. In this article, nine spatial/temporal synthetic aperture radar (SAR) data sets from Radar Satellite-1 (Radarsat-1) and Environmental Satellite-1 (Envisat-1) Advanced Synthetic Aperture Radar (ASAR) were used as an interface to upscale the field-level soil moisture to soil moisture at a 25 km × 25 km grid, the scale at which the Advanced Microwave Scanning Radiometer for the EOS (Earth Observing System) (AMSR-E) level-3 soil moisture product is available, leading to a total of 68 such grids where there is a pair of SAR-derived and corresponding AMSR-E-derived values. Out of the these 68 grid cells, each of size 25 km × 25 km, pair data from nine dates, data corresponding to 51 grid cell pairs (from six dates) were used for the development of the approach and data corresponding to 17 grid cell pairs (from three dates) were used for validating the approach. It is observed that while the spatial variation in SAR- and AMSR-E-derived soil moisture is broadly similar, there exists a bias in the difference between AMSR-E- and SAR-derived soil moisture. The root mean square (RMS) difference between AMSR-E soil moisture and SAR-derived soil moisture at 25 km over the model development sample of 51 grid cells, each of 25 km × 25 km, was observed to be 4.2 (%g cm–3). The study shows that the bias is significantly affected by the AMSR-E vegetation/roughness parameter, g, which is an indicator of the prevailing vegetation/surface roughness conditions. This dependence of the bias on g was exploited to arrive at modified soil moisture estimates at 25 km. The RMS difference between the modified AMSR-E soil moisture at 25 km and the SAR-derived soil moisture was observed to be 2.13 (%g cm–3) over a validation sample of 17 grid cells each of size 25 km × 25 km, comprising extreme, dry and wet, and soil moisture conditions. This demonstrative study provides the feasibility of calibrating and improving the soil moisture product from coarse-resolution space-borne passive sensors with the help of soil moisture derived using SAR. The approach is demonstrated for validating passive radiometer-derived soil moisture over bare/crop-covered conditions through an experiment involving 68 25 km × 25 km grid cells spread over nine temporal/spatial Radarsat and Envisat-1 satellite SAR data sets and synchronous ground soil moisture measurements over farmers’ fields, covering a large range of soil/crop conditions prevailing during the nine months of SAR data acquisition. Hence the results reported in this manuscript are encouraging.