Flood Extent and Volume Estimation using Multi-Temporal Synthetic Aperture Radar.

Abstract

Satellite imagery has the potential to monitor flooding across wide geographical regions. Recent launches have improved the spatial and temporal resolution of available data, with the European Space Agency (ESA) Copernicus programme providing global imagery at no end-user cost. Synthetic Aperture Radar (SAR) is of particular interest due to its ability to map flooding independent of weather conditions. Satellite-derived flood observations have real-world application in flood risk management and validation of hydrodynamic models. This thesis presents a workflow for estimating flood extent, depth and volume utilising ESA Sentinel-1 SAR imagery. Flood extents are extracted using a combination of change detection, variable histogram thresholding and object-based region growing. An innovative technique has been developed for estimating flood shoreline heights by combining the inundation extents with high-resolution terrain data. A grid-based framework is used to derive the water surface from the shoreline heights, from which water depth and volume are calculated. The methodology is applied to numerous catchments across the north of England that suffered from severe flooding throughout the winter of 2015-16. Extensive flooding has been identified throughout the study region, with peak inundation occurring on 29th December 2015. On this date, over 100 km2 of flooding is identified in the Ouse catchment, equating to a water volume of 0.18 km3. The SAR flood extents are validated against satellite optical imagery, achieving a Total Accuracy of 91% and a Critical Success Index of 77%. The derived water surfaces have an average error of 3 cm and an RMSE of 98 cm compared to river stage measurements. The methods developed are robust and globally applicable, shown with an additional study along the Mackenzie River in Australia. The presented methodology, alongside the increased temporal resolution provided by Sentinel-1, highlights the potential for accurate, reliable mapping of flood dynamics using satellite imagery.