A model for downscaling SMOS soil moisture using Sentinel-1 SAR data

Abstract

A model for downscaling SMOS (Soil Moisture Ocean Salinity) soil moisture products is developed by using multi-temporal dual-polarized (HH+HV) C-band SAR data. In this model, the effect of vegetation on soil moisture retrieval from SAR data is minimized by using the water-cloud model (WCM), in which vegetation contribution is quantified using the backscatter coefficient of HV polarization. The wavelet transform is used to fuse high resolution Sentinel-1A SAR backscatter with low resolution SMOS soil moisture, where the difference in spatial heterogeneity between scales is also accounted for. The influence of soil surface roughness is eliminated by using multi-temporal data. The multi-temporal SMOS soil moisture and dual-pol Sentinel-1/SAR data are the only inputs of this downscaling model. The model is tested in southern Ontario, Canada to downscale 40 km resolution SMOS soil moisture to 1.25 km and 2.5 km resolutions. The downscaled results show good agreements with the in-situ soil moisture collected in May and July of 2016 with an unbiased root-mean-square-error (RMSE) of 0.045 m3/m3 and 0.047 m3/m3 and a coefficient of determination (R2) of 0.54 and 0.70 at 1.25 km and 2.5 km resolutions respectively. The results suggest that the model can be applied for C-band at regional scales to provide continuous soil moisture mapping at higher resolutions. The high revisit frequency of the up-coming Radarsat Constellation Mission (RCM) combined with its large areal coverage characteristics are ideal for the generation of downscaled products.