Soil moisture estimation using combined multifrequency SAR data: A comparison between two inversion models using simulation

Although high-resolution microwave synthetic aperture radar (SAR) sensors possess all-weather capability for mapping soil moisture from spaceborne platforms, continuous temporal and spatial monitoring of this important hydrological parameter has been relatively limited. However, the recent launch of operational SAR sensors aboard various satellites have made possible synoptic soil moisture monitoring a reality. Such systems operate over a wide range of frequencies, look angles, and polarization combinations, and thus show synergistic advantages when combined for estimating soil moisture patterns. Two soil moisture inversion algorithms have been developed using as inputs radar backscattering data at L, S, and C bands in the microwave frequency range. These models have been tested using radar image simulation with speckle added. It is observed that the neural network algorithm yields superior results in mapping actual soil moisture patterns over the linear statistical inversion technique, although both models show comparable errors in soil moisture estimation. We infer that using statistical estimation errors alone for comparison purposes may lead to erroneous conclusions regarding the advantages of one soil moisture inversion algorithm over another.