Precipitation observations from high frequency spaceborne polarimetric synthetic aperture radar and ground-based radar: theory and model validation

Abstract

Global weather monitoring is a very useful tool to better understand the Earth's hydrological cycle and provide critical information for emergency and warning systems in severe cases. Developed countries have installed numerous ground-based radars for this purpose, but they obviously are not global in extent. To address this issue, the Tropical Rainfall Measurement Mission (TRMM) was launched in 1997 and has been quite successful. The follow-on Global Precipitation Measurement (GPM) mission will replace TRMM once it is launched. However, a single precipitation radar satellite is still limited, so it would be beneficial if additional existing satellite platforms can be used for meteorological purposes. Within the past few years, several X-band Synthetic Aperture Radar (SAR) satellites have been launched and more are planned. While the primary SAR application is surface monitoring, and they are heralded as "all weather'' systems, strong precipitation induces propagation and backscatter effects in the data. Thus, there exists a potential for weather monitoring using this technology. The process of extracting meteorological parameters from radar measurements is essentially an inversion problem that has been extensively studied for radars designed to estimate these parameters. Before attempting to solve the inverse problem for SAR data, however, the forward problem must be addressed to gain knowledge on exactly how precipitation impacts SAR imagery. This is accomplished by simulating storms in SAR data starting from real measurements of a storm by ground-based polarimetric radar. In addition, real storm observations by current SAR platforms are also quantitatively analyzed by comparison to theoretical results using simultaneous acquisitions by ground radars even in single polarization. For storm simulation, a novel approach is presented here using neural networks to accommodate the oscillations present when the particle scattering requires the Mie solution, i.e., particle diameter is close to the radar wavelength. The process of transforming the real ground measurements to spaceborne SAR is also described, and results are presented in detail. These results are then compared to real observations of storms acquired by the German TerraSAR-X satellite and by one of the Italian COSMO-SkyMed satellites both operating in co-polar mode (i.e., HH and VV). In the TerraSAR-X case, two horizontal polarization ground radars provided simultaneous observations, from which theoretical attenuation is derived assuming all rain hydrometeors. A C-band fully polarimetric ground radar simultaneously observed the storm captured by the COSMO-SkyMed SAR, providing a case to begin validating the simulation model. While previous research has identified the backscatter and attenuation effects of precipitation on X-band SAR imagery, and some have noted an impact on polarimetric observations, the research presented here is the first to quantify it in a holistic sense and demonstrate it using a detailed model of actual storms observed by ground radars. In addition to volumetric effects from precipitation, the land backscatter is altered when water is on or near the surface. This is explored using TRMM, Canada's RADARSAT-1 C-band SAR and Level 3 NEXRAD ground radar data. A weak correlation is determined, and further investigation is warranted. Options for future research are then proposed.