Discrimination of Water Surfaces, Heavy Rainfall, and Wet Snow Using COSMO-SkyMed Observations of Severe Weather Events

Abstract

An automatic method to distinguish water surfaces (either flooded or permanent water bodies) from artifacts caused by heavy precipitation and wet snow is designed to improve flood detection accuracy in X-band synthetic aperture radar (SAR) images. The algorithm implementing the proposed method, mainly based on image segmentation techniques and on the fuzzy logic, consists of two principal steps: 1) detection of regions (or segments) of low-radar backscatter that appear dark in a SAR image, and 2) classification of each detected segment. Ancillary data, such as a local incidence angle map, a land cover map, and an optical image (helpful to detect wet snow), are also used. Through the fuzzy logic, the algorithm integrates different rules for the detection of dark areas, as well as for their classification based on radiometric, geometrical and shape features extracted from the segmented SAR image and on the ancillary data. The algorithm is tested on the COSMO-SkyMed imagery of the severe weather event that hit Northwest Italy on November 2011. A comparison with measured data, provided by the weather radars belonging to the Italian radar national network, and with the ground precipitation, forecasted by a numerical weather prediction model routinely used within the framework of the EUMETSAT Hydrology Satellite Application Facility project, indicates that the algorithm produces reliable classification maps, being able to distinguish the rainfall signature on X-band SAR images from that of flooded areas.