Combined active and Passive microwave remote sensing of Soil Moisture for vegetated surfaces at L-band

摘要

Combined active and passive microwave remote sensing of vegetated surfaces is of great interest and importance given the increasing number of active and passive satellite microwave missions and datasets available for studies in land surfaces for application in hydrology and terrestrial ecology [1]. For many years, passive microwave retrieval algorithms for satellite missions such as AMSR-E, SMOS, and SMAP have been based on the omega (ω)-tau (π)-h model [2, 3], which is derived from the zeroth order solution of radiative transfer equation. Since the zeroth order solution ignores the Phase matrixterm, it is only valid when the omega is small. Using the physical scattering model of branches and leaves, the calculated omega at L-band is in the range of 0.2 to 0.6 and is not small. Thus the small omega used in omega-tau-h model is an effective parameter rather than a physical parameter. In modeling the rough surface effects, the omega-tau-h model only includes the coherent wave specular reflection as represented by h [4] while ignoring the bistaic scattering. Thus the omega-tau-h model uses a small h such as 0.1. The physical h, as calculated by numerical solutions of Maxwell's equations in 3D simulations (NMM3D) [5], is much larger and can be as large as unity at L-band. Thus the omega-tau-h model uses effective small omega and effective small h both of which are much smaller than the physical calculated values. For active remote sensing, we previously used the distorted Born approximation [6] and NMM3D (NMM3D-DBA)where the coherent reflectivity and rough surface scattering are calculated by NMM3D [7]. The model was used to calculate the V V and HH backscatter at L-band for grass [8], wheat and canola fields. The active model has been validated with Soil Moisture Active Passive Validation Experiment 2012 (SMAPVEX12) data [9] and shown good soil moisture retrieval results from the radar data compared with the ground measurements [8]. The distorted Born approximation is valid because the optical thickness at L-band is small. The distorted Born approximation is consistent with the first order radiative transfer theory except it includes backscattering enhancement in the double-bounce term. In this paper, we study the active and passive microwave remote sensing using NMM3D-DBA for both active and passive where NMM3D for rough surface bistatic scattering is used. The active model NMM3D-DBA is extended to calculate bistatic scattering and an integration of the bistatic scattering over the hemispherical solid angle is used to calculate emissivity. Thus the active and passive microwave remote sensing models are founded on the same theoretical basis and use the same physical parameters such as crop density, plant height, stalk orientation, leave radius, surface roughness, and so on. The vegetation canopy is modeled as a layer of uniformly distributed dielectric cylinders and disks representing stalks and leaves, respectively [6]. The distorted Born approximation is derived from Foldy-Lax equation with first-order iteration using the half-space Green's function and T-matrix [10]. The attenuation through the vegetation layer is accounted for by the imaginary part of the effective propagation constant calculated using Foldy's approximation [10]. The total bistatic scattering is expressed as the sum of three scattering mechanisms: volume scattering, double bounce scattering and surface scattering. Data-cubes which are lookup tables with three axes: vegetation water content (VWC), root mean square (RMS) height of an isotropic surface and soil permittivity directly related to the soil moisture [8], are then generated for both active and passive. From the active and passive data-cubes, the β parameter describing the linear relation between brightness temperature and co-polarized backscatter [11] is also derived for various rough surface and vegetation conditions. The data-cubes are useful for retrievals of detailed soil and vegetation characteristics such as soil moisture [8]. The model results are validated by coincidental airborne Passive Active L-band Sensor (PALS) data and Uninhabited Aerial Vehicle Synthetic Aperture Radar (UAVSAR) data taken during the SMAPVEX12 field campaign [9]. Time-series retrieval of the soil moisture [8] is also performed by inverting the data-cubes assuming that RMS height does not change over time whose results are then compared with the ground measurements.