Quantifying the role of individual flood drivers and their correlations in flooding of coastal river reaches

Abstract

Flooding in coastal river reaches is the result of complex interactions between coastal and inland drivers. Flood hazard assessments need to consider how these drivers interact in space and time, for which a standard method is currently lacking. A complex hydrodynamic model is required to reproduce the physics of the combined forcings and, at the same time, to fully explore the combinations of drivers that can occur in order to determine extreme flood frequencies. In this work, we explore the individual role of astronomical tide, storm surge and river discharge and their correlations in the extreme flood levels of a coastal river reach. We apply a computationally efficient surrogate model of a 2D shallow water model based on least squares support vector machines (LS-SVM) regression to reconstruct 10000 years-long time series of water levels in the reach. As input to the model, we consider an ensemble of synthetic time series of the flood drivers, which differ in the number of variables considered and in their correlations. Probabilities of exceedance of water levels are then computed and compared. The proposed methodology can give a better understanding of the flooding processes in a multivariable environment, as low-lying coastal urban areas typically are, and can provide guidance on where to focus modelling efforts when developing flood hazard assessments in such areas.