Real-time coastal flood hazard assessment using DEM-based hydrogeomorphic classifiers

Abstract

In the last decade, DEM-based classifiers based on height above nearest drainage (HAND) have been widely used for rapid flood hazard assessment, demonstrating satisfactory performance for inland floods. The main limitation is the high sensitivity of HAND to the topography, which degrades the accuracy of these methods in flat coastal regions. In addition, these methods are mostly used for a given return period and generate static hazard maps for past flood events. To cope with these two limitations, here we modify HAND, propose a composite hydrogeomorphic index, and develop hydrogeomorphic threshold operative curves for rapid real-time flood hazard assessment in coastal areas. We select the Savannah River delta as a test bed, calibrate the proposed hydrogeomorphic index on Hurricane Matthew, and validate the performance of the developed operative curves for Hurricane Irma. The hydrogeomorphic index is proposed as the multiplication of two normalized geomorphic features, HAND and distance to the nearest drainage. The calibration procedure tests different combinations of the weights of these two features and determines the most appropriate index for flood hazard mapping. Reference maps generated by a well-calibrated hydrodynamic model, the Delft3D FM model, are developed for different water level return periods. For each specific return period, a threshold of the proposed hydrogeomorphic index that provides the maximum fit with the relevant reference map is determined. The collection of hydrogeomorphic thresholds developed for different return periods is used to generate the operative curves. Validation results demonstrate that the total cells misclassified by the proposed hydrogeomorphic threshold operative curves (summation of overprediction and underprediction) are less than 20 % of the total area. The satisfactory accuracy of the validation results indicates the high efficiency of our proposed methodology for fast and reliable estimation of hazard areas for an upcoming coastal flood event, which can be beneficial for emergency responders and flood risk managers.