Abstract: Coastal floods often coincide with large waves, storm surge and tides. Thus, joint probability methods are needed to properly characterize extreme sea levels. This work introduces a statistical downscaling framework for multivariate extremes that relates the non-stationary behavior of coastal flooding events to the occurrence probability of daily weather patterns. The proposed method is based on recently-developed weather-type methods to predict extreme events (e.g., significant wave height, mean wave period, surge level) from large-scale sea-level pressure fields. For each weather type, variables of interest are modeled using Generalized Extreme Value (GEV) distributions and a Gaussian copula for modelling the interdependence between variables. The statistical dependence between consecutive days is addressed by defining a climate-based extremal index for each weather type. This work allows attribution of extreme events to specific weather conditions, enhancing the knowledge of climate-driven coastal flooding.

Authorship

Download reference