Spatial Bayesian hierarchical modeling of precipitation extremes over a large domain

TL;DR

This paper introduces a Bayesian hierarchical model for analyzing spatial precipitation extremes over large areas, effectively handling missing data and providing detailed risk maps for infrastructure planning.

Contribution

The paper presents a novel Bayesian hierarchical framework with a composite likelihood approach for large-scale spatial extremes modeling, accommodating missing data and capturing spatial dependence in GEV parameters.

Findings

Return levels vary spatially and seasonally.

The model efficiently processes large datasets with missing data.

Provides detailed maps of precipitation extreme risks.

Abstract

We propose a Bayesian hierarchical model for spatial extremes on a large domain. In the data layer a Gaussian elliptical copula having generalized extreme value (GEV) marginals is applied. Spatial dependence in the GEV parameters are captured with a latent spatial regression with spatially varying coefficients. Using a composite likelihood approach, we are able to efficiently incorporate a large precipitation dataset, which includes stations with missing data. The model is demonstrated by application to fall precipitation extremes at approximately 2600 stations covering the western United States, -125E to -100E longitude and 30N to 50N latitude. The hierarchical model provides GEV parameters on a $1/8$th degree grid and consequently maps of return levels and associated uncertainty. The model results indicate that return levels vary coherently both spatially and across seasons, providing information about the space-time variations of risk of extreme precipitation in the western US, helpful for infrastructure planning.