Extreme Event Attribution

Extreme event attribution examines whether human-caused climate change altered the likelihood or intensity of a particular class of extreme weather event. It compares the observed, human-influenced climate with a counterfactual climate in which anthropogenic influence is absent or reduced, while accounting for natural variability.

Constructing factual and counterfactual climates

Researchers first define the event by its location, duration and measurable threshold, such as maximum temperature or seasonal rainfall. They then run many simulations because chaotic natural variability can produce different weather outcomes under identical external conditions.

• The factual ensemble includes observed or estimated anthropogenic forcings, such as greenhouse gases and aerosols, along with natural forcings.
• The counterfactual ensemble represents a climate without human influence, generally using natural forcings and removing the estimated anthropogenic climate signal.
• Depending on the experimental design, researchers use coupled climate models or atmosphere-only models with adjusted sea-surface temperature and sea-ice boundary conditions.

Estimating anthropogenic influence

The frequency or intensity of the defined event is compared across the two ensembles. If its probability is p1 in the factual climate and p0 in the counterfactual climate, the probability ratio, p1/p0, measures how much more or less likely human influence made it.

• The fraction of attributable risk, calculated as 1 minus p0/p1, estimates the proportion of the event risk associated with anthropogenic influence.
• An intensity-based analysis compares the magnitude of events having a similar probability or return period in the two climates.
• Observations are used to evaluate whether models adequately reproduce the event and its relevant physical processes.

Interpretation and limitations

Attribution is a probabilistic statement, not a claim that climate change was the sole cause of an event. Results are conditional on the event definition, model design, forcing estimates and representation of relevant processes.

• Uncertainty increases for very rare events because observational records and simulated samples may be limited.
• Model bias and poorly represented circulation, convection or land-atmosphere feedbacks can affect conclusions.
• Attribution of the physical hazard does not by itself attribute disaster losses, which also depend on exposure and vulnerability.