Methods of flood extent mapping using SAR imagery in the Zambezi (Caprivi) Region, Namibia

Abstract

Flood extent mapping is an important disaster response tool made possible by synthetic aperture radar (SAR). The ability to rapidly assess ground conditions during flood events, independent of cloud cover or time of day, makes SAR imagery ideal for this task. This study compares and evaluates two methods of flood extent mapping using ENVISAT ASAR imagery of the Zambezi (formerly Caprivi) Region of Northern Namibia between 2006 and 2010. The area of interest is near the confluence of the Zambezi and Chobe rivers, and experiences severe flooding almost annually. This flooding affects thousands of people and, apart from washing away homes, settlements and crops, presents a disease risk in the form of Cholera and Malaria. The aim of this study is to evaluate the relative efficiency and accuracy of two flood extent mapping techniques using SAR data in Zambezi Region. These methods are SAR backscattering thresholding and flood extent delineation using the active contour model. The use of thresholding, specifically, has been shown to be a rapid method for flood mapping. In this study we evaluate the accuracy of this method and demonstrate different ways in which to determine optimal threshold values. The active contour model, also referred to as the snake algorithm, is an image processing framework for delineating an object outline from a possibly noisy 2D image. We use this model to delineate flood extents and perform a comparison with the results achieved from different methods of simple thresholding. Both methods are compared to independent observations from Landsat imagery across the 2006-2010 time period, where available. This paper presents the first results from the various flood mapping techniques, with specific attention to the effect of incidence angle, orbit direction and temporal distance to flooding event on the flood maps. Future research will incorporate ALOS PALSAR-1 imagery to investigate the effect of polarization and wavelength, as well as to investigate the use of interferometric coherence in analyzing flood extent change over time.