Retrieval of Soil Moisture Using Sliced Regression Inversion Technique

A soil moisture (SM) retrieval algorithm, rooted in electromagnetic theory, is proposed and validated against synthetic data with good retrieval accuracy. The proposed Sliced Regression (SR) algorithm generates piece-wise linear fits to a data set obtained from a forward model, and then retrieves the soil moisture by applying a linear least-squares approach. This algorithm is compared against a widely employed semi-empirical inversion algorithm (SMART), and it is observed that better results are returned from the former. Performance of the proposed retrieval algorithm on single band and dual band (L and S-bands) retrieval is studied; and it is found dual band retrieval returns better retrieval results for SM, along with other soil surface parameters. The proposed algorithm is employed on synthetically generated data-sets (generated using the Integral Equation Method as the forward model for ground scattering) for SM retrieval scenario considering all the primary influencers on backscatter into the model. To measure SM for vegetated terrain, vegetation is modelled as a collection of randomly oriented dielectric cylinders with radii and length specific to the vegetation. The overall backscatter is calculated as the sum of individual contributions from the ground, vegetation and their interactions. Having parameterized the vegetation forward model as above, we extend the SR to retrieve SM in the presence of vegetation.