Retrieval of soil moisture variations in agricultural fields through a new Bayesian change detection approach

A new change detection algorithm based on a Bayesian approach is developed and tested. The main objective of this approach is to exploit the changes in backscattering signals and relate them to soil moisture variations over agricultural fields under the hypothesis of both constant and variable roughness. The proposed methodology overcomes the limitations of the some change detection methods because it takes into account also possible changes in the radar signal due to roughness variability. The method is trained and tested on two data sets considering both C and L-band backscattering coefficients in relation to soil moisture and roughness measurements. The C-band dataset was acquired over bare soils while the L-band data set was acquired on vegetated fields and was exploited to understand the impact of vegetation in such approach. The results indicate that the approach is able to detect soil moisture changes both for C-and L-band data. In case of L band data, the presence of vegetation seems to determine backscattering dynamics reduction with respect to soil moisture changes.