A spatial hierarchical model with temporal effects for analyzing extreme rainfall in Taiwan
Conference
Format: IPS Abstract
Keywords: bayesian hierarchical model, climate change, spatial smoothing, spatio-temporal analysis
Session: Invited Session 11A - Advanced Statistical Methods for Climate Data Analysis
Thursday 5 December 3:30 p.m. - 5 p.m. (Australia/Adelaide)
Abstract
The PoT-GEV model integrates the generalized extreme value distribution with the peaks over threshold method and is well-suited for extreme value analysis. Originally designed by Olafsdottir et al. (2021) to fit block maximum data, it can assess trends in the frequency and intensity of extreme events simultaneously. Our research enhances this model by incorporating a spatial hierarchical structure with temporal effects. We integrate spatial correlations using a latent spatial Gaussian process applied to the PoT-GEV parameters and include temporal covariates to capture time effects. To improve computational efficiency, we have replaced traditional Markov Chain Monte Carlo techniques with the Laplace approximation method. We demonstrate the efficacy of our proposed methodology through extensive simulation studies covering various scenarios. Additionally, we illustrate the practical application of our model by analyzing rainfall data from Taiwan.