Soil water estimation with SAR backscatter as a function of soil properties, instrument noise, and incidence angle

The purpose of this study was to estimate soil water from single and multi-frequency/polarization, C-Band hh and vv, and L-Band hh and vv, synthetic aperture radar (SAR) backscatter considering soil properties, instrument noise, and incidence angle. Two studies with simulated backscatter and an application with measured SAR backscatter acquired by an airborne instrument were employed to test and compare two algorithms, Brent's and Newton-Raphson, for estimating soil water.;A noise simulation found the best dielectric constant predictor from a single band/polarization combination was determined to be C-Band vv, from a two band/polarization combination the C-Band vv, and L-Band hh pairing, and the best three band/polarization grouping was C-Band hh, C-Band vv, and L-Band hh. The Newton-Raphson and Brent's algorithms performed similarly, but the Newton-Raphson procedure had more invertible cases. The best band/polarization combinations for dielectric constant estimation were used for an incidence angle simulation and the application.;Findings from the simulations and application revealed dielectric constant estimation error increases with the value of the dielectric constant. The increasing estimation error is due to a loss of sensitivity of simulated backscatter to dielectric constant at the upper end of its range. The simulations revealed that even minimal amounts of added noise resulted in dielectric constant estimation errors in the range of 6 to 7. When added noise is positive an overestimation of dielectric constant results, and negative added noise produces an underestimation of dielectric constant. The simulations also revealed that noise and incidence angle can interact to either generate considerable dielectric constant estimation error or cancel each other, with no or little influence on the estimation outcome.;The application, using data from the Washita '92 instrument campaign, found that under dry to moderately dry conditions field averaged dielectric constant could be estimated with an error less than 1 for 4 of 12 and 6 of 12 cases with the single band/polarization Brent and two band/polarization Newton-Raphson algorithms, respectively. The four band/polarization was nearly as good as the single band Brent's algorithm. The three band/polarization combination estimated dielectric constant with great error and is not recommended for use. Field measured soil roughness values were used in the single and two band/polarization inversions. If the soil roughness data had been used in the three and four band/polarization inversions then their results probably would have improved.