Vegetation/Forest Effects in Microwave Remote Sensing of Soil Moisture

Abstract

This thesis includes (1) the distorted Born approximation (DBA) and an improved coherent model for vegetation-covered surfaces at L-band for data-cube based soil moisture retrieval; (2) a unified approach for combined active and passive remote sensing of vegetation-covered surfaces with the same input physical parameters; (3) Numerical Maxwell Model in 3D (NMM3D) simulations of a vegetation canopy comprising randomly distributed dielectric cylinders; and (4) a hybrid method based on the generalized T matrix of single objects and Foldy-Lax equations for NMM3D full-wave simulations of the realistic vegetation/forest with vector spherical, spheroidal and cylindrical wave expansoins. The main contributions and novelty of this thesis are NMM3D full-wave simulations of vegetation/forest canopy using the generalized T matrix of the single object and Foldy-Lax equations of multiple scattering among many objects. Before this work, the large-scale full-wave simulations of vegetation/forst such as many tree trunks were deemed very difficult. The NMM3D full-wave simulation results showed that the results of past models significantly overestimate attenuation in a vegetation/forest canopy. The NMM3D full-wave models predict transmissions that are several times greater than that of past models. A much greater microwave transmission means the microwave can better penetrate a vegetation/forest canopy and thus it can be used to retrieve soil moisture. The thesis starts with the DBA to compute the backscattering coefficients for various kinds of vegetation-covered surfaces such as pasture, wheat and canola fields. For the soybean fields, an improved coherent branching model is used. The novel feature of the analytic coherent model consists of conditional probability functions to eliminate the overlapping effects of branches in the former branching models. In order to make use of complex physical models for real time retrieval for satellite missions, the outputs of the physical model are provided as lookup-tables (data-cubes). By inverting the lookup-tables, time-series retrieval of soil moisture is performed. Next, the DBA is extended to calculate the bistatic scattering coefficients. Emissivities are calculated by integrating the bistatic scattering coefficients over the hemispherical solid angle. The backscattering coefficients and emissivities calculated using this approach form a consistent model for combined active and passive microwave remote sensing. In the analytical physical models mentioned above, as well as in another commonly used approach of the radiative transfer equation (RTE), the attenuation of the wave is accounted for by the attenuation rate per unit distance, which originates from the concept of an “effective medium”. Such a model is unsuitable for a vegetation canopy. Because of these issues, NMM3D full-wave simulations of vegetation are pursued. Firstly, the scattering of a vegetation canopy consisting of cylindrical scatterers is calculated. The approach for solving Maxwell’s equations is based on the Foldy-Lax multiple scattering equations (FL) combined with the body of revolution (BOR). For a layer of extended-cylinders distributed in clusters, the NMM3D simulations at C-band show very different results from DBA/RTE. The method FL-BOR is limited for rotationally symmetric objects such as cylinders and circular disks. To perform NMM3D full-wave simulations for realistic vegetation/forests, a hybrid method is used, which is a hybrid of the off-the-shelf techniques and newly developed techniques. The newly developed techniques are the three key steps of the hybrid method: (1) extracting the generalized T matrix of each single object using vector spheroidal/cylindrical waves, (2) vector wave transformations, and (3) solving FL for all the objects.