Active and passive microwave remote sensing of soil moisture

This study focuses on the development of a consistent methodology for soil moisture inversion from Synthetic Aperture Radar (SAR) data using the Integral Equation Model (Fung et al., 1992) without the need to prescribe time-varying land-surface attributes as constraining parameters. Specifically, the dependence of backscatter coefficient on the soil dielectric constant, surface roughness height and correlation length was investigated. The IEM was used in conjunction with an inversion model to retrieve soil moisture using multi-frequency and multi-polarization data (L, C and X-Bands) simultaneously. The results were cross-validated with gravimetric observations obtained during the Washita '94 field experiment in the Little Washita Watershed, Oklahoma. The average error in the estimated soil moisture was of the order of 3.4%, which is comparable to that expected due to noise in the SAR data. The retrieval algorithm performed very well for low incidence angles and over bare soil fields, and it deteriorated slightly for vegetated areas, and overall for very dry soil conditions.; The IEM was originally developed for scattering from a bare soil surface, and therefore the vegetation effects are not explicitly incorporated in the model. We coupled a semi-empirical vegetation scattering parameterization to our multi-frequency soil moisture inversion algorithm. This approach allows for the explicit representation of vegetation backscattering effects without the need to specify a large number of parameters. The retrieval algorithm performed well for vegetated conditions when a land-use based vegetation parameterization was used. The explicit incorporation of land-use in the parameterization scheme is equivalent to incorporating the effect of vegetation structure in the soil moisture estimates obtained using the SAR observations.; ESTAR images of brightness temperature obtained during the same period were inverted independently for soil moisture. The results at individual sampling sites were first compared against gravimetric soil moisture observations for Washita '94, and the RMS errors for both applications were between 3% and 4%. Subsequently, we investigated the use of high resolution SAR-derived soil moisture fields to estimate sub-pixel variability in ESTAR derived fields. The effect of sub-pixel variability of various land surface properties (namely soil moisture, soil texture, soil temperature, and vegetation). The results demonstrated the linear scaling behavior of ESTAR based soil moisture estimates.; We also investigated the problem of consistency between the two systems. Estimated and observed brightness temperature fields were compared and analyzed to establish the aggregation kernel inherent to ESTAR, i.e., how the instrument actually processes/integrates sub-pixel variability. The scaling properties of both SAR and ESTAR at all frequencies were investigated and the results indicated that both sensors demonstrated fractal behavior. The results suggested that the two systems can be used to complement each other, and there is a potential to downscale ESTAR observations for high resolution soil moisture estimation, using only one SAR frequency (e.g. L-band).