On the accuracy of soil moisture content retrieved at pixel, segment or field scale, from advanced-SAR data: a simulation study

In this work, the effects of SAR measurement errors as well as direct model errors on soil moisture retrieval from SAR data are investigated. In particular, the attention is focused on understanding under which conditions it is more convenient: a) feed the retrieval algorithm with accurate backscattering values (i.e. estimated at "field scale") then retrieve soil moisture estimate directly at "field" scale; b) use relatively noisy backscattering values, estimated at smaller scales (i.e. "segment scale"), to retrieve soil moisture estimates at "segment" scale and subsequently average the obtained soil moisture estimates at "field" scale. The adopted soil moisture retrieval algorithm is based on a regularized Neural Networks appropriately trained by IEM model. The SAR synthetic data simulates SAR data acquired by ERS and ENVISAT satellites. The performance of the inversion method is -given as a function of the SAR configuration and noise level.