International Journal of Applied Earth Observations and Geoinformation (Dec 2022)
Multi-resolution soil moisture retrievals by disaggregating SMAP brightness temperatures with RADARSAT-2 polarimetric decompositions
Abstract
Mapping soil moisture (SM) at high spatial resolution assists to trigger important agricultural management, such as irrigation, to enhance crop yields. This study investigates disaggregation of SMAP brightness temperature (TB) using RADARSAT-2 polarimetric decompositions to retrieve high-resolution SM. Compared to Sentinel-1 backscattering coefficients used in the SMAP baseline active–passive SM retrieval algorithms, the RADARSAT-2 surface scattering power Ps with a reduced vegetation influence was hypothesized to be more relevant to disaggregate the SMAP TB. Different polarimetric decompositions were evaluated to extract an optimal Ps, followed by an incidence angle normalization. Then, the optimal Ps parameter was aggregated to the same spatial resolution as the SMAP TB to develop empirical relationships between Ps and TB. Furthermore, the airborne TB data collected by Passive Active l-band Sensor (PALS) were analyzed in terms of the Ps across multiple spatial resolutions, to account for the scale effect on the Ps/TB relationships. Finally, the τ-ω emission model was used to retrieve SM at multiple spatial resolutions (10 km, 1 km, 500 m, 100 m, and 50 m). The impacts of spatial resolution on retrieval accuracy were analyzed to determine the best spatial resolution for SM retrievals. The results indicated that the An polarimetric decomposition with the de-orientation provided the highest surface scattering powers, which may benefit the SM estimation. In contrast to the traditional cosine algorithms, the incidence angle normalization of Ps with span resulted in a temporally decreasing surface scattering power, because of the increasing vegetation attenuation as the crop grows. The sensitivity of TB to Ps decreases as the resolution scale varies from 36 km to 50 m. The SM retrievals across multiple resolutions obtained marginal differences in retrieval accuracy. Although slightly better results were obtained with 1 km spatial resolution which is close to the nominal size of agricultural fields in the study area (R = 0.68–0.8 and RMSE = 0.039–0.062 m3/m3), the retrievals at 50 m spatial resolution (R = 0.63–0.76 and RMSE = 0.046–0.067 m3/m3) capture the spatial heterogeneity of SM within and across different fields which could be very helpful for the precision agriculture.
