Surface and Volume Scattering Model in Microwave Remote Sensing of Snow and Soil Moisture

Abstract

My research consists of four topics: 1) analytical Kirchhoff approach for modeling of Signals of Opportunities (SoOp) land applications at L and P band, 2) Scattering of rough surfaces based on the numerical solutions of Maxwell’s equations in 3D (NMM3D) from L to Ku band, 3) the bicontinuous dense media radiative transfer (Bic-DMRT) model for volume scattering of random media with dense aggregates, and 4) retrieval algorithm of snow water equivalent (SWE) using X and Ku band radar observations and its synergism with passive observations. The thesis first presents the analytical Kirchhoff solutions (AKS) for SoOp land applications at P and L band. Different from backscattering observations, the topography has significant effects on the signals in addition to the microwave roughness. The land surface is divided into three scales: microwave roughness, fine-scale topography, and coarse-scale topography (deterministic DEM data). The salient features of the AKS model are (i) analytical expressions are obtained for both coherent and incoherent waves, (ii) Monte Carlo simulations are not required, and (iii) analytical solutions are expressed in terms of the surface spectrum. Results of AKS have been validated with the Numerical Kirchhoff Approach (NKA) and CYGNSS satellite data. In addition to the SoOp problem, I have applied the NMM3D method to calculate backscattering of rough surface from L to Ku band with kh up to 15 for remote sensing of snow and soil moisture, where k is wavenumber and h is rms height. The NMM3D is based on the discrete dipole approximation (DDA) with half-space Green’s function and periodic boundary conditions (PBC). In the past, surface scattering models only have been limited to kh<3. NMM3D results show that the backscattering of rough surfaces increase with frequency/rms height and then saturate at high frequency/large rms height. Simulated backscatter are in good agreement with radar measurements. For snow volume scattering, the DMRT is developed for random media with densely packed aggregates. The size distribution of the bicontinuous model has been improved to generate dense aggregates, which are controlled by mean grain size 〈ζ〉 and aggregation parameter b. Results show that C band volume scattering of snow with dense ice aggregates at cross-pol can be larger than soil surface scattering, which interprets the recent finding that C band cross-pol Sentinel 1 data have sensitivity to snow depth band For SWE retrieval, a radar SWE retrieval algorithm using X and Ku band radar observations has been developed. In the algorithm, the parameterized DMRT model is first derived by regressions of the DMRT lookup table (LUT), which makes backscatter a function of the scattering albedo at X band and SWE. The surface scattering from the snow/soil interface is subtracted from radar observations to get snow volume scattering. Based on the obtained volume scattering and a priori information, a cost function is established to find SWE. The performance of the algorithm is evaluated with three sets of airborne SnowSAR data acquired in Finland and Canada. To improve the radar SWE retrieval, passive observations at 19 and 37 GHz are used to determine the a priori scattering albedo of snow. Furthermore, 13.3 GHz is introduced to form a three-channel (9.6, 13.3, and 17 GHz) algorithm which reduces the effects of background scattering from the snow/soil interface. The improved algorithm has been evaluated against the ground-based radar measurements.