Incidence Angle Normalization of Radar Backscatter Data

Abstract

The National Aeronautics and Space Administration's (NASA) proposed Soil Moisture Active Passive (SMAP) satellite mission ( ~ 2014) will include a radar system that will provide L-band multi-polarization backscatter at a constant incidence angle of 40 °. During the pre-launch phase of the project, there is a need for observations that will support the radar-based soil moisture algorithm development and validation. A valuable resource for providing these observations is the NASA Jet Propulsion Laboratory Uninhabited Aerial Vehicle Synthetic Aperture Radar (UAVSAR). However, SMAP will observe at a constant incidence angle of 40 °, and UAVSAR collects data over a wide range of incidence angles (25 °-60°). In this investigation, a technique was developed and tested for normalizing UAVSAR data to a constant incidence angle. The approach is based on a histogram matching procedure. The data used to develop and demonstrate this approach were collected as part of the Canadian Soil Moisture Experiment 2010 (CanEx-SM10). Land cover in the region included agriculture and forest. Evaluation was made possible by the acquisition of numerous overlapping UAVSAR flight lines that provided multiple incidence angle observations of the same locations. Actual observations at a 40° incidence angle were compared to the normalized data to assess performance of the normalization technique. An optimum technique should be able to reduce the systematic error (Bias) to 0 dB and to lower the total root mean square error (RMSE) computed after correction to the level of the initial residual error (RMSEres) present in the data set. The normalization approach developed here achieved both of these. Bias caused by the incidence angle variability was minimized to ~ 0 dB, whereas the residual error caused by instrument related random errors and amplitude fluctuations due to ground variability was reduced to approximately 3 dB for agricultural areas and 2.6 dB for forests; these values were consistent with the initial RMSEres estimated using the un-corrected data. The residual error can be reduced further by aggregating the radar observations to a coarser grid spacing. The technique adequately adjusted the backscatter over the full swath width irrespective of the original incidence angle, polarization, and ground conditions (vegetation cover and soil moisture). In addition to providing a basis for fully exploiting UAVSAR (or similar aircraft systems) for SMAP algorithm development and validation, the technique could also be adapted to satellite radar systems. This normalization approach will also be beneficial in terms of reducing the number of flight lines required to cover a study area, which would eventually result in more cost-effective soil moisture field campaigns.