Assessing the performance of speckle filtering techniques in interpreting soil properties of arable lands using Radarsat-2 data by the example of test fields in Saratov Povolzhye

Radar data have great potential for soil studies. However, their interpretation is complicated due to presence of speckle. Using Saratov Povolzhye as a test region we analyzed the performance of adaptive (Gamma Map, Refined Lee, Frost) and non-adaptive (Median) filtering techniques in speckle suppression, preservation of original information and assessed their influence on the possibility of soil features interpretation using Radarsart-2 data. Gamma Map filter was founded to be more effective in speckle suppression for vertical-horizontal and vertical polarizations regardless of the soil surface conditions at the time of image acquisition. Applying this filter with 5×5 window size allowed modelling of organic matter content and particles 0.05-0.01 mm in size with overall accuracy of over 70% for open soil surface and 60% for covered surface. Lower filtering window size (3×3) appeared to be more suitable for mapping 1-0.25 mm sized particles and slope when soil surface is open. In case of granulometric composition and parent material, the best results for the test region were obtained when applying Refined Lee filter for vertical-horizontal polarization and open surface with overall accuracy of the models of 63-65% of the models. The considered results are applicable only for the studied radar data, acquisition time and test region. At the same time, the findings can be used to organize remote monitoring of properties of soil surface layer of the test fields which is important for land use.