Development and validation of flood inundation models for estuaries

Estuaries, at the interface between catchment and coast, are vulnerable to flooding from the combination of riverine and marine inputs. High river flows generated from intense precipitation can occur synchronously with high tides and storm surges, amplifying flood hazard. In the United Kingdom 20 million people are estimated to live near estuaries, with estuarine flooding regarded as the costliest impact to these areas and second highest hazard to civil emergency. On-going global warming increases sea-levels and modifies hydroclimate variability, thus affecting river fluxes, tidal maxima and the intensify of storm surges. There is therefore a need for improved methods and tools to understand compound flooding events, their impacts and how they may change into the future. In the present paper we developed a validated flood inundation model for the Conwy estuary in North Wales, one of the flashiest catchments in Britain where flooding makes headline news at least once every year. The Caesar-Lisflood 2D hydrodynamic flow model was combined with a range of publicly available datasets to represent channel bathymetry, land elevation, location and heights of flood defences and the hydraulic roughness across the model domain. The model was forced with recorded time-series (15-minute resolution) of tidal oscillations and river discharge data and validated by comparing simulated water levels against observations from existing water level gauges within the estuarine channel. Flood predictions were validated against observed flood extents extracted from SAR imagery using the Google Earth Engine. Calibrated, ortho-corrected (GRD) C-band interferometric Synthetic Aperture Radar (SAR) images captured by the Sentinel-1 constellation of satellites using a dual-band cross-polarization (VH) was used. SAR images were filtered to remove speckle noise and Otsu’s method of thresholding was adopted to automatically extract inundated areas from each available image. Comparison of model-based simulated flood extents against their SAR-derived equivalents was used as a means to validate the flood inundation model.