Bayesian Spatio-Temporal Models for Hurricane Risk Assessment under Climate Change
We present a Bayesian spatio-temporal modeling framework to assess hurricane-driven wind-related risks along the U.S. South Atlantic and Gulf coasts. Our analysis leverages a high-resolution, downscaled wind speed dataset generated from a large ensemble of synthetic tropical cyclones using the CHIPS model. To estimate the marginal probability that a tropical storm reaches hurricane strength at each location and time, we apply binomial GLM models fitted using INLA. Our framework incorporates key environmental and spatial covariates, including the El Niño-Southern Oscillation, Sea Surface Temperature, wind shear, and a land/water indicator. A copula-based dependence approach is then applied to derive joint spatial exceedance probabilities across the domain. By integrating both marginal behavior and joint spatial dependence, the framework captures the space-time variability in wind intensity and enables coherent assessment of hurricane-induced wind hazards under a changing climate.
Palabras clave: Bayesian Spatio-Temporal Models Hurricane Risk Assessment Physics-based Models INLA Extreme Wind Probabilities Copula Models Climate Change Impacts