Skip to main content
U.S. flag

An official website of the United States government

Publication Date
1 June 2018

Evaluating and Improving Statistical Methods Used in Assessing Human Influence on Severe Weather

Print / PDF
Powerpoint Slide
Science

This work considers how best to use statistical methods to assess human influence on severe weather when climate scientists use global climate models to estimate probabilities of severe weather with and without human influence.

Impact

There are a variety of statistical methods appropriate for estimating probabilities of severe weather using climate model output, many of them developed for biomedical research and not known or used by climate scientists. This work provides guidance to climate scientists in choosing a good statistical method when assessing human influence on extreme weather and provides user-friendly software for using these statistical methods.

Summary

Event attribution in the context of climate change seeks to understand the role of human-related greenhouse gas emissions on extreme weather events. Climate scientists often use global climate models to estimate the probabilities of extreme events in the real world and in a hypothetical world in which greenhouse gases had not be emitted. Event attribution is then quantified by these two probabilities. While this approach has been applied many times in the last 15 years, the statistical methods used in such work have not made use of the full set of methods from statistical and biomedical research. We examined the methods used in statistical and biomedical research, evaluated them for use by climate scientists for event attribution, and propose good methods to use.  In particular, there are methods not yet used for event attribution that have important advantages over the methods that climate scientists currently use. We provide software for using the methods so that it is straightforward for climate scientists to use the methods in their future assessments.

Point of Contact
William D. Collins
Institution(s)
Lawrence Berkeley National Laboratory (LBNL)
Funding Program Area(s)
Publication