Empirical modelling to estimate surface soil moisture at field scale in Sardinia, Italy: Comparison between optical and SAR data

Surface soil moisture is an important variable in hydrological modeling and is critical in agricultural and other applications. There is thus a strong interest in assessing the potential of remote sensing for providing regular spatio-temporal soil moisture observation. This study analyzes the correlation between 18 ENVISAT ASAR images (C-band, HH and VV polarized, ascendant or descendant mode, incidence angle from 15.0° to 31.4°) and surface soil moisture measurements over six small (<;5 hectares) bare fields in Sardinia (Italy) taken over the period from 2005 to 2009. When comparing the backscatter coeffficient of variations and corresponding field data, it was found that images taken with a VV polarisation in a descendant mode were more correlated to soil moisture variations, with an R² of 85.03% and a variance of 0.001113. When applying a cross validation technique to estimate soil moisture on the same six fields (only those with more than 5% relative humidy) we obtain an R² of 75,87% (measured versus retrieved soil moisture) for ENVISAT imagery and an R² of 51.71% for RADARSAT-2 imagery.