Retrieving near-surface soil moisture from Radarsat SAR data

On the basis of six Radarsat synthetic aperature radar (SAR) images acquired over the Paddle River Basin, Alberta, and 1350 field soil moisture data collected from nine selected sites of the basin in the same days, this study demonstrated the feasibility of retrieving near-surface soil moisture (5-10 cm depth) from Radarsat SAR data. Between two retrieval algorithms, a linear regression and the theoretical integral equation model (IEM) of Fung et al. [1992], the results indicate that IEM performed consistently better than the regression model. However, the performance of both models generally deteriorates when the basin is nearly saturated, which is the case for September 10 and 27, 1996, with a basin saturation level of 82 and 75%, respectively. With respect to Radarsat's configuration and IEM's surface parameters, it is found that radar backscattering sigma(0) is highly sensitive to the surface rms height, especially for a relatively smooth surface and a Gaussian correlation function. From the semivariogram of measured soil moisture and sigma(0) it is found that the homogeneity assumption of spatial interpolation as a function of distance lag breaks down only at a distance lag of about 200-275 m. Apparently, spatial interpolation methods based on this assumption can only interpolate soil moisture meaningfully for single land use plots of several hundred meters in size.