Extreme Water Level Predictions on the Nova Scotian Coastline using a Bayesian Hierarchical Model

Abstract

The modelling of extreme coastal water levels is crucial when it comes to flood risk preparations, management, and defense design in Nova Scotia. Given the increasing frequency of flooding events in this region, it is essential to estimate their potential magnitude to mitigate their potentially devastating consequences. In extreme value analysis, return levels are quantifications of risk and represent the level that is expected to be exceeded once on average in a given time period. The challenge is that these extreme events are rare, and the data are limited and often incomplete. To address this challenge, a Bayesian hierarchical extreme value model is developed where information on extreme events is shared spatially across locations. By incorporating atmospheric covariates as physical drivers of extreme water levels, this research aims to establish a well informed Bayesian Hierarchical Model that enhances estimation precision and enables return level predictions for ungauged locations across the Nova Scotian coastline.