Standing Water Detection Using Radar

Mapping of natural and man-made water bodies is a critical component of environmental studies, including climate change, agriculture, flood forecasting and groundwater recharge estimation. Moreover, standing water heavily impacts the ability to accurately estimate soil moisture from satellite, particularly from low resolution passive microwave data using a radiometer. NASA’s soil moisture dedicated mission, the Soil Moisture Active Passive (SMAP), will use 3km resolution L-band radar to downscale 36km resolution L-band radiometer observations to 9km. These brightness temperature estimates are subsequently interpreted for soil moisture content. Consequently, the 3km resolution radar that will be on-board the SMAP affords the opportunity to also detect dynamic standing water, and thus correct the radiometer signals accordingly prior to downscaling. The objective of this study was to develop and validate a standing water detection method, using radar observations collected from the Soil Moisture Active Passive Experiments (SMAPEx) undertaken near the township of Yanco in the Murrumbidgee Catchment in southeast New South Wales, Australia. There were three water detection methods put forward in this study, including: i) difference in backscatter at different polarizations; ii) direct backscatter threshold; and iii) the ratios between backscatter at different polarizations. These three methods were all applied to the data collected over a known flooding area, so as to compare the feasibility of each method and thus determine the criteria of water body detection. The direct backscatter threshold method had the best performance for detecting water bodies, which was subsequently applied to radar data over the entire SMAPEx site. In order to validate the capability of this detection method, the derived water map across the SMAPEx site was compared against a range of reference maps, including 1km resolution brightness temperature data, Land Remote Sensing Satellite (Landsat) derived farm dam maps, and aerial photographs taken during the regional flights of the study site area. To analyze the influence from incidence angle normalization on the backscatter, both normalized and nonnormalized data were tested in this study. It was found that the water map derived from non-normalized backscatter was more accurate, due to the residual normalization error when using the backscatter data at its highest spatial resolution (10m). Consequently, the standing water map had the best estimation of water locations when used at coarser resolution (100m) due to the reduction in speckle as compared to data at high resolution (10m). Moreover, some areas were mis-identified as water, especially in grasslands, meaning that the method needs to be used with caution, and that radar is going to be a challenging approach to identify standing water in vegetated fields.