GreenSurge: An efficient additive model for predicting storm surge induced by tropical cyclones

Abstract

Storm surge is one of the main components of sea level beyond coastal flooding induced by intense storm events such as tropical cyclones (TCs). This component can be estimated using dynamic numerical simulations that consider both the inverse barometer effect induced by pressure gradients and wind setup. However, the dynamic approach can be computationally demanding and time-consuming, particularly for being included in early warning systems of resource-constrained communities. In this study, we introduce as an alternative, a novel additive hybrid model known as GreenSurge. This model relies on the generation of a library of sea-level responses to unitary wind sources from any direction, along with the assumption of a linear dynamics framework for the summation of the spatial and temporal sea-level responses, facilitating the efficient reconstruction of storm surge at regional-to-local scales. To showcase the capabilities of GreenSurge, we have implemented the method in the Pacific Island of Tongatapu (Tonga) to predict the storm surge induced by several TCs and compare its capabilities against dynamic numerical simulations and available tide gauge data. Given its similar accuracy (errors less than 10% of the maximum storm surge value) and higher computational efficiency when compared with dynamic hydrodynamic models, GreenSurge has proven to be a great alternative for reconstructing historical time series, feeding coastal flooding models, or even analysing climate change scenarios.