Downscaling SMOS/SMAP soil moisture product using high resolution Radarsat-2 SAR data: a case study in southern Ontario

Abstract

Current soil moisture products derived from SMOS and SMAP satellites have very coarse resolutions (~40km). This limits the soil moisture products for applications where a resolution of 1-10 km is generally needed (e.g. hydro-climatological applications). To obtain high resolution soil moisture, an algorithm was developed to downscale SMOS/SMAP products by using multi-temporal dual-polarized (HH and HV) C-band SAR images. Recently the use of full-polarized (FP) SAR data was reported to improve soil moisture retrieval even under dense vegetation canopy, but FP SAR data has the disadvantage of narrow swath, which limits its applications for soil moisture estimates over large areas. With the Radarsat Constellation Mission (RCM) becoming operational in 2018, it will provide compact polarimetric (CP) configurations that has a wider swath and hence is a possible alterative to the FP data. A new downscaling algorithm has been developed to improve soil moisture retrieval by using RCM-CP data, which is simulated from the Radarsat-2 FP imagery. Comparing to previous algorithm, the new algorithm can enhance the removal of vegetation effect by using a normalized scattering based empirical model, in which the conditions of vegetation are characterized by the RCM-CP derived vegetation parameter. In addition, a new mathematical model, rather than complicate wavelet transform, is used to better account for scale change. The new algorithm is validated with in-situ soil moisture data from Southern Ontario collected in the summer of 2017. When compared to results obtained from a previous algorithm there is a promising improvement in soil moisture estimation. Since the RCM will provide a short revisit time (<4 days), this study demonstrates the potential for fusing RCM data with SMOS/SMAP soil moisture (2-3 days repeat cycle) to continuously map soil moisture at higher resolutions over large areas, even under dense vegetation.