A0965
Title: INLA-based Bayesian spatiotemporal models for downscaling hurricane wind speeds on the U.S. north Atlantic coast
Authors: Concepcion Ausin - Universidad Carlos III de Madrid (Spain) [presenting]
Michael Wiper - Universidad Carlos III de Madrid (Spain)
Ali Sarhadi - Georgia Tech (United States)
Abstract: The aim is to develop a Bayesian spatiotemporal modeling framework to assess wind-related risk and return periods of major tropical cyclones along the U.S. North Atlantic coast, accounting for nonstationary and warming climate conditions. The analysis is based on a high-resolution downscaled dataset derived from a large ensemble of synthetic storm tracks, generated to be consistent with large-scale circulation patterns from climate models. A binomial generalized linear model is employed, and the probability of extreme wind events is estimated at each location over time using the integrated nested Laplace approximation (INLA), which offers substantial computational efficiency over traditional MCMC methods. Both parametric and semi-parametric structures are incorporated to capture spatial and temporal variations in wind intensity distributions, enabling robust projections of future wind-related hazards across coastal sites.
