Synthesis for Extreme Event Attribution: Methodological Review and New Approaches

Probabilistic attribution of extreme weather events to anthropogenic climate change is attracting growing medial and societial attention, increasing the demand for precise and reliable estimation of changes in frequency or intensity of extreme weather events. Numerous rapid extreme event attribution studies rely on the “synthesis” of multiple lines of evidence, including observational data products and climate model ensembles, prominently within the World Weather Attribution (WWA) framework and related tools (e.g., KNMI Climate Explorer).

Classically, as in [Philip et al. 2020], the anthropogenic influcence on a particular extreme event is expressed via a single summary statistic, either a probability ratio, measuring the change in probability of such an extreme event, or a change in intensity. Both emerge from modeling annual (seasonal, monthly) maxima as realizations from a nonstationary generalized extreme value (GEV) distribution with the smoothed global mean surface temperature (GMST) as a covariate. Current synthesis procedures [Otto et al. 2024] operate at the level of estimated summary statistics and their associated uncertainties. We review this approach and propose an alternative parameter-level synthesis, in which estimates of the nonstationary GEV model parameters are combined prior to inference on attribution-relevant statistics. A large-scale simulation study demonstrates this alternative to have favorable statistical properties, mitigating issues such as infinite estimates and miscalibrated confidence intervals encountered in existing approaches. The findings are illustrated using case studies of extreme weather events, primarily heavy precipitation and heat waves.

References:

[Otto et al. 2024] Otto, F. et al. (2024). “Formally combining different lines of evidence in extreme-event attribution”. In: Advances in Statistical Climatology, Meteorology and Oceanography 10.2, pp. 159–171. doi: 10.5194/ascmo-10-159-2024.

[Philip et al. 2020] Philip, S. et al. (2020). “A protocol for probabilistic extreme event attribution analyses”. In: Advances in Statistical Climatology, Meteorology and Oceanography 6.2, pp. 177–203. doi: 10.5194/ascmo-6-177-2020.