Online Program Home
My Program

Abstract Details

Activity Number: 151 - Beyond the VAR: Advances in Spatial and Spatio-Temporal Modeling for Climate and Environmental Data
Type: Invited
Date/Time: Monday, July 29, 2019 : 10:30 AM to 12:20 PM
Sponsor: Section on Statistics and the Environment
Abstract #300064 Presentation
Title: Spatial Extreme Value Analysis
Author(s): Eric Gilleland*
Companies: National Center for Atmospheric Research
Keywords: Extreme-Value Analysis; Spatial Statistics; Spatial Extremes; Propinquity Model; Latent Process Model; Max-stable processes

A challenging topic in statistics concerns inferring about extreme events; not necessarily the upper quartile event, but those way out in the tails of the distribution. The problem is essentially one of having to extrapolate far beyond the range of the data, and while such a procedure is generally not recommended, the importance of the problem demands it be done anyway. Extreme-value theory (EVT) provides justification for utilizing certain distributions in this context, namely, in the univariate context: the generalized extreme value (GEV) for maxima taken over very long stretches of time (e.g. annual) and the generalized Pareto (GP) for excesses over a very high threshold. The latter approximates the tail of the former, and both encompass three types of distributions according to the value of their shape parameters. A third, point process, characterization ties the two approaches together. Particularly for environmental topics, incorporating spatial information into any such analysis is important, but although multivariate methods exist in EVT, analyzing spatial data for extremes is a tricky, and active, area of research. This talk will review these methods.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2019 program