This website was set up to support the project A conditional approach to extreme event attribution (paper forthcoming).

The frequency and intensity of extreme weather events is widely believed to be increasing as a result of human-induced climate change, but quantifying these changes is hard. This research aims to answer three questions, (a) characterizing the probability of an extreme event, (b) how global climate change has affected that probability (the attribution question), (c) how such probabilities will change in the future under various emission scenarios. I use only public data sources: daily temperature and precipitation data from weather stations, regional means from gridded data products, and climate models from CMIP6. I use extreme value theory to analyze how extremes at individual stations and spatially depend on regional climate averages, and develop a Bayesian hierarchical approach to link regional climate averages to climate models. The entire analysis is Bayesian, thus allowing realistic projections of uncertainty as well as estimates of extreme event probabilities. The approach is illustrated using data from three extreme weather events: the British/French heatwave of July 2022, the Northwest USA/Canada heatwave of 2021, and extreme rainfall events in 2017 from Hurricane Harvey. Comparisons will be made with made with other approaches, in particular those from the World Weather Attribution group.

A draft paper and supporting computer code will be posted as soon as they are available. For a recent talk I have given based on a prliminary version of the work, see my slides or the accompanying video. Meanwhile, preliminary data files and results can be accessed from the three regions used to support the analysis:

Pacific Northwest
Northern Europe
Gulf of Mexico

Update (June 29, 2023). My talk at the EVA 2023 meeting is here .

Update (July 5, 2023). My talk at the IDAG meeting is here . Same material, slight rearragement of previous talk.

Return to Richard Smith's page