Probabilistic storyline attribution using machine learning

2026-06-01

AI Summary

This study quantifies the contribution of anthropogenic climate change to extreme weather events by constructing counterfactual scenarios that replicate observed atmospheric circulation under varying climate forcings. For the first time, a distributional autoencoder (DAE) is integrated into a storyline attribution framework to model the conditional probability distribution of European temperature fields given specific circulation patterns and levels of global warming, using ERA5 reanalysis and climate model data. This approach generates physically plausible extreme temperature fields while preserving circulation consistency and enables full probabilistic comparisons. Applied to the 2003 European heatwave, the method shows that if a similar circulation were to recur in 2028 and 2053, the conditional mean temperatures would rise to 30.3°C and 32.1°C, with occurrence probabilities 2.1 and 3.2 times higher than in 2003, respectively.

Abstract

A fundamental goal in climate attribution is to estimate how forced climate change contributes to observed extreme weather events. The storyline attribution method compares an observed weather event, conditional on its atmospheric dynamic state (i.e., atmospheric circulation), in the current, 'factual' climate to an event with very similar circulation conditions in a hypothetical, 'counterfactual' climate. However, physical climate models cannot directly transfer these storyline counterfactuals across different climate forcing states. Statistical and machine learning techniques may overcome this limitation; yet, emulating circulation-conditional extreme events under different climate states is challenging. Here, we demonstrate distributional autoencoders (DAEs) as a versatile method for generating climate counterfactuals. They model the full distribution of spatially resolved European temperature fields conditional on the atmospheric circulation state and the mean global warming level. These distributions allow for deriving meaningful conditional probability ratios, which is a particular advantage of the DAE-based storyline approach. We train DAEs on fully coupled climate model simulations and we evaluate the modelled distributions across different factual and storyline-based counterfactual climate model simulations. In an illustrative case study, we revisit the 2003 European heatwave and we generate counterfactuals for a hypothetical `2003-like European heatwave' using ERA5 circulation, which we hypothesize to occur a quarter century (2028) and a half century (2053) after 2003. The conditional intensity would increase from 29.3 °C in 2003, to 30.3 °C and 32.1 °C in 2028 and 2053, respectively and conditional probability ratios would be 2.1 and 3.2 when compared to 2003.

Problem

• storyline attribution
• climate change
• extreme weather events
• counterfactual climate
• atmospheric circulation

Innovation

• distributional autoencoders
• storyline attribution
• climate counterfactuals
• conditional extreme events
• machine learning