A probabilistic event-storyline approach to assessing projected changes in a high-impact Mediterranean storm track

An increase in precipitation extremes is one of the most robust signals of anthropogenic climate change. However, the latest IPCC assessment still reports low confidence in projected changes over the Mediterranean region. Despite this uncertainty, several Mediterranean cyclones—intense mid-latitude storms—have caused severe precipitation extremes and substantial economic damage in recent decades. The role of climate change in these events remains poorly quantified.

Here, we develop a probabilistic event-storyline approach and apply it to assess projected changes in a high-impact Mediterranean storm track. The storyline describes an autumn large-scale trough over the Iberian Peninsula, followed by cyclone development and northeastward propagation over the western Mediterranean Sea, leading to widespread daily extreme precipitation over the Italian Peninsula. This evolution was characteristic of two notable historical high-impact events: storm Adrian (Vaia) in October 2018 and the November 1966 storm that caused major flooding in Florence.  The probability of such events is decomposed into three conditional components: (i) the occurrence of the large-scale trough, (ii) the probability of northeastward-propagating Mediterranean cyclones given the precursor, and (iii) the probability of extreme precipitation given cyclone development. These probabilities are estimated using ERA5 reanalysis and a 17-member ensemble of the CMIP6 EC-Earth3 climate model under present-day and future (SSP2-4.5) climate conditions.

We show that EC-Earth3 provides a satisfactory representation of the Mediterranean autumn storm track, with ERA5-based conditional probabilities lying within the model ensemble spread. However, none of the ensemble members simulates a storm with a trajectory and intensity comparable to storm Adrian, highlighting the rarity of such events. Under SSP2-4.5, the ensemble projects no overall change in the frequency of events following this storyline. This result arises from a compensation between a strong reduction in the frequency of the large-scale precursor (risk ratio r ≈ 0.6), a moderate decrease in cyclone development given the precursor (r ≈ 0.8), and a strong increase in the probability of extreme precipitation conditional on storm development (r ≈ 2). The shallowing of large-scale troughs and the reduced frequency of deep cyclones counteract the expected thermodynamic intensification of precipitation over the Alpine region.

Overall, these findings highlight the need to explicitly account for dynamical changes when assessing future projections of cyclone-driven Mediterranean precipitation extremes. While based on a single large ensemble, the proposed framework can be extended by estimating individual conditional probabilities from different global and regional climate models, offering a pathway to integrate complementary sources of information.