Soil Moisture Retrieval Using Time-Series Radar Observations Over Bare Surfaces

A time-series algorithm is proposed to retrieve bare surface soil moisture and rms height using two copolarized (HH and VV) L-band backscattering coefficients (sigma(0)). The retrieval approach inverts a forward model for radar scattering from an isotropic bare surface. Because real-time inversion of a complex forward model is often computationally impractical, the inversion is implemented using a precomputed lookup table representation of s0 obtained from numerical Maxwell model in 3-D simulations. The retrieval process assumes that surface roughness properties are constant during the time-series interval, so that only a single rms height estimate is produced for the entire time series. The use of this rms height estimate as a constraint simplifies the associated soil moisture retrievals at each time step. A Monte-Carlo simulation of this algorithm with 0.7 dB radar measurement error (1-sigma) shows that retrievals using six time steps outperform a "snapshot" method (which retrieves rms height and soil moisture at each time step) by a factor of about two in rms soil moisture error. A second study using measured data having 6 to 11 time steps shows an rms error of 0.044 cm(3)/cm(3) for soil moisture with a correlation coefficient of 0.89 between retrieved and in situ data. Surface rms height estimates are also found accurate to 10 to 30% of in situ measurements. It is also shown that retrieval performance is not sensitive to errors in knowledge of the surface roughness correlation length for most of the bare surface conditions examined.