Detecting Changes in Spatial Extremes with Attribution to Both Anthropogenic Influences and Natural Modes of Climate Variability

• Activity Number: 240 - Making the Case for Professional Climate Statisticians
• Type: Topic-Contributed
• Date/Time: Wednesday, August 11, 2021 : 10:00 AM to 11:50 AM
• Sponsor: Section on Statistics and the Environment
• Author(s): Mark Risser* and Likun Zhang and Chris Paciorek
• Companies: Lawrence Berkeley National Laboratory and Lawrence Berkeley National Laboratory and University of California, Berkeley

Abstract

In spite of the diverse literature on spatial extreme value analysis, characterizing the extremes of an environmental process like daily precipitation for a large network of monitoring stations over a heterogeneous spatial domain remains a challenging statistical problem. Here, we compare and contrast two methods that are scalable to high-dimensional, heterogenous spatial data sets, namely conditional independence-based GEV methods that utilize bootstrapping and hierarchical Bayesian methods for block maxima. Both approaches are used to detect long-term trends and year-to-year changes in precipitation extremes over the contiguous United States for a large network of several thousand weather stations. Locally, in spite of significant noise and the fact that changes due to natural climate variability are a similar order of magnitude, we are able to detect statistically significant anthropogenic influence on seasonal precipitation extremes. Human-induced climate change generally results in larger and more frequent extreme events, although there are also important areas where the opposite is true.