MSGSS
Bayesian Hierarchical Modelling of Extreme Storm Surges
Bayesian hierarchical models (BHMs) provide a powerful framework for modelling extreme events through space. The flexibility of BHMs makes them well-suited for spatial modelling, as they can incorporate relationships between geographically distributed data points while accounting for multiple sources of uncertainty. Spatial dependencies in flood risk across Nova Scotia can be incorporated within the model, borrowing from data-rich areas to improve predictions in data-scarce locations.
In this talk, we will derive the model definition for a spatial Bayesian hierarchical extreme value model for the prediction of extreme storm surges across Nova Scotia. Environmental variables from the Canadian Surface Reanalysis will be used as covariates for the model. Preliminary results and model diagnostics will be shown in R along with some discussion of other possible model formulations.