How unlikely was the storm Hans? Reusing extended range forecasts to anticipate unprecedented extremes

The storm Hans struck Eastern Norway in August 2023 with two days of intense rain, triggering what may become the country’s most expensive weather-related disaster. Remarkably, the flooding affected large rivers that normally peak during the snowmelt season from April to June. We answer the two related research questions of a) could we have anticipated a storm of Hans’ magnitude, and b) could the storm have coincided with the April to June snowmelt, causing a compound event. We leveraged the UNSEEN methodology (UNprecedented Simulated Extremes using ENsembles), using 3.5 years of extended-range weather forecasts and their 20-year reforecasts from ECMWF as well as ERA5 reanalysis. UNSEEN uses large ensembles of model simulations to assess rare but plausible extremes, which enables the assessments of return periods not possible with the short historical record. We found that while Hans is at the high end of what could be expected, the larger sample reveals a continuum of weaker but still previously unprecedented events throughout the year that remain capable of causing severe impacts. The most extreme rainfall events are most probable between July and September, when snowmelt does not amplify flooding. However, two of the most extreme simulated events occurred in May, indicating that a physically consistent compound disaster — heavy rain coinciding with snowmelt — is possible under the right conditions. Together, these results show how data-driven counterfactuals can help anticipate unprecedented extremes.