Microwave scattering models for nonuniform forest canopies

A bistatic microwave canopy scattering model is developed based on the existing Michigan Microwave Canopy Scattering model (MIMICS). The new model is termed Bi-MIMICS and can accommodate a bistatic scattering simulation capability in anticipation of future spaceborne radar systems. This is followed by the development of the multi-layer microwave canopy scattering model (Multi-MIMICS), a first-order radiative transfer model that is developed to simulate bistatic microwave scattering coefficients from nonuniform forest canopies. Multi-MIMICS accounts for vertical heterogeneity of a mixed species forest and provides a significantly enhanced representation of actual complex forest structures compared to the conventional two-layer models. Multi-MIMICS allows overlapping layer configurations and includes a tapered trunk model applicable to forests of mixed species and/or mixed growth stages. A fully polarimetric first-order transformation matrix is provided and forest canopies are characterized as discrete random media. Multi-MIMICS simulates SAR bistatic scattering coefficients based on input dimensional; geometrical and dielectric variables of forest canopies. Multi-MIMICS is parameterized using field data representing several configurations of mixed species forest in Queensland, Australia, with each containing a diversity of species, structural forms and growth stages. The resulting simulations represented a considerable improvement over those generated using MIMICS with the same source data and are successful simulations of the backscattering coefficients, as indicated by the close correspondence with airborne SAR data. The potential retrieval of forest biomass and other vegetation parameters can be studied by integrating the simulated radar response at multiple frequencies, polarizations and observation angles.; For the case of open forests, texture information carried by SAR images of non-uniform canopies reveals the horizontal spatial variation of the scene, a distinct characteristic from the vertical inhomogeneity studied in Bi-MIMICS and Multi-MIMICS. We investigate the formation and properties of texture in SAR images and evaluate different texture measurement models. A blind deconvolution approach is presented to retrieve the texture correlation length and is shown to result in notable improvements in its estimation. A coherent SAR texture simulator is also developed to simulate SAR images of surface targets with horizontal spatial variations.