Soil moisture retrieval from heterogeneous surfaces by 1.4 GHz multi-angle SMOS observations using ´a priori knowledge´ of surface cover fractions

A recently published feasibility study on the consequences of surface heterogeneity for parameter retrieval from multi-angle L-band SMOS observations indicated that even small portions of forest (>10%) may have a dramatic effect on the retrieval of pixel-average soil moisture, resulting in retrieval errors exceeding the required precision of 4% by volume formulated for the SMOS mission. These retrievals were done without a priori knowledge of surface temperature and without a priori knowledge of the within-pixel forest cover fraction. This study demonstrates that inclusion of a priori knowledge of the within-pixel forest fraction allows the retrieval of the soil dielectric constant (or the soil emissivity) of the non-forested area with a precision better than the requirement defined for SMOS. It was also found that a bias of plusmn2 K in the assumed thermodynamic surface temperature has a relatively small effect on the retrieval precision