Spatiotemporal Modeling of Extreme Events

• Activity Number: 583
• Type: Invited
• Date/Time: Thursday, August 7, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Brian Reich*+ and Sam Morris
• Companies: North Carolina State University and North Carolina State University
• Keywords: spatial ; Bayesian ; extreme ; air pollution ; meteorology ; environmental

Abstract

Spatial analysis of extreme events often reduce the data to the yearly maximum at each spatial location, and assume a max-stable spatial process for the maximums. Reducing the data set to yearly maximums may discard important information in large observations below the maximum. In this paper, we propose a new spatiotemporal spatial model of all observations that exceed a predefined threshold. Our model is based the spatial t distribution, which provides asymptotic spatial dependence and computationally-convenient full conditional distributions. We also employ a spatial partitioning prior to eliminate long-range spatial dependence. The new method is applied to study the relationship between meteorological factors and extreme ambient air pollution events in the Southeast United States.