Abstract

Rising seas increase the vulnerability of coastal cities due to higher extreme sea levels, flooding, and coastal erosion. Understanding the links between sea-level extremes caused by storm surges is crucial for informing flood risk decisions. This paper proposes a Bayesian hierarchical spatio-temporal model to estimate extreme sea levels at both gauged and ungauged locations in Ireland. The model captures spatial and temporal dependence of annual extreme sea-level data, considering coastlines as physical barriers affecting spatial dependence. Data from the GESLA (Global Extreme Sea Level Analysis) database was used, consisting of 26 tide-gauge locations in Ireland and 21 from the West coast of Great Britain (GBR), taking advantage of the proximity of the islands and the longer GBR records to share information across the Irish Sea. The model was fitted using the INLA (Integrated Nested Laplace Approximation) Bayesian framework, employing the SPDE (Stochastic Partial Differential Equation) approach for spatial/barrier modelling. Annual maxima (1953-2021) were modelled using the GEV (Generalised Extreme Value) distribution as a function of spatial coordinates and coastline barriers. The spatial dependence between locations was modelled using the Matérn covariance function, while a first-order random walk process captured temporal dependence. A prediction mesh was built around the coasts to predict sea level extremes at ungauged locations. The aim was to estimate indicators of possible extreme sea-level rise over the years, including spatio-temporal effects, the probability of a sea-level above a threshold, and return levels. These indicators can help policymakers in identifying critical areas and making informed decisions. The results indicate an increasing trend in sea-level, with higher risk areas identified on Ireland’s East coast. The spatial/barrier model offers a more realistic approach compared to traditional spatial models by accounting for coastline effects on spatial dependence.