The study aims to detect ground standing water in cropland during the spring/early summer season in eastern South Dakota, USA. The goal is to develop a reliable and accurate method that can distinguish ground surface open water from vegetation, which is often mistakenly identified as water. To achieve this, the study utilized Sentinel-1 synthetic aperture radar (SAR) data due to its high reliability, short revisit time, and free availability. A total of 159 sites were selected and surveyed, including 78 water sites and 81 non-water sites, located between Brookings, SD and Sioux Falls, SD, USA. The SAR data were preprocessed at both VV and VH polarizations for both water and non-water sites. In previous work, we used maximum likelihood estimation (MLE) of the density functions with a shifted Rayleigh distribution. In this paper, a Neyman-Pearson test for SAR data classification is developed using the Rayleigh priors at the dual-polarization. The developed method demonstrates good performance in distinguishing between water and non-water sites, providing an alternative approach to ground water detection that is important for precision agriculture, hydrologic and environmental studies.
|