Semi-empirical method for estimation of forest biophysical properties from multifrequency polarimetric SAR

Empirical attempts to invert radar backscatter data for estimates of forest biophysical properties such as aboveground standing biomass have been partly successful in certain circumstances: (1) mono-species forests such as pine plantations and where (2) net biomass is less than a frequency dependent saturation threshold. In general, these rely on the power-law relationship between radar backscattering coefficient and net biomass. This relationship saturates at a level of biomass that scales with wavelength and is also known to be dependent upon forest structural type. In addition, these empirical relationships do not explicitly account for seasonal effects on radar backscattering behavior. These effects are related to (1) structural changes due to phenological development and (2) dielectric changes due to either phenological development or changes in the state of water contained within the plants as a consequence of freezing or thawing. As a result, the purely empirical approaches do not work well when applied to forests of mixed structural composition or when applied to forests that are structurally or dielectrically dissimilar to those used to develop the inverse model. The authors present a more robust approach to forest biomass retrieval that incorporates two main elements: (1) a pre-classification of forest by structural type and (2) a structure dependent semi-empirical inversion for biomass that uses multifrequency and polarimetric SAR data in a recursive fashion