Assessing the UNSEEN Flood-Relevant Winter Extreme Precipitation Over the Island of Ireland in the Present Climate

Ireland’s winters are getting wetter, with more frequent heavy precipitation events increasing flooding risk across the Island. Extreme precipitation is a key driver of flooding in northwestern Europe; however, observational records are relatively short and represent only a single realisation of the climate state. As a result, they are inadequate for sampling low-likelihood, high-impact flood-relevant extreme precipitation events and for quantifying plausible maxima of such extremes. In this study, we quantify plausible maxima for flood-relevant winter precipitation under the current climate. We apply the UNprecedented Simulated Extremes using Ensembles (UNSEEN) approach to the flood-relevant winter precipitation indices (Rx1day, Rx5day, and Rx30days), using daily winter observations, the ECMWF SEAS5 seasonal prediction systems, and the CANARI Single Model Initial-condition Large Ensemble (SMILE) over the Island of Ireland. These indices are consistently derived across observations, pooled SEAS5 winter ensembles (ensemble member x lead times), and the CANARI SMILE. Model fidelity for CANARI and ensemble independence, stability, and fidelity for pooled SEAS5 are assessed to ensure that both models realistically represent extreme precipitation. Preliminary results indicate that both SEAS5 and the CANARI sample the physically plausible Rx1day and Rx5day extremes that exceed the maximum observed in the current climate, while neither system produces UNSEEN values exceeding the observed maximum Rx30day.  The CANARI large ensemble passes the fidelity test without bias correction, whereas the SEAS5 passes the fidelity test after applying simple multiplicative mean scaling bias correction. For CANARI, plausible maxima are approximately 18.01% higher for Rx1day and 20.77% higher for Rx5day than observed maxima, while Rx30day plausible maxima are approximately 8.70% lower than the highest observed Rx30day. For SEAS5, plausible maxima exceed observations by approximately 3.05% for Rx1day and 17.68% for Rx5day, while Rx30day plausible maxima are approximately 17.74% lower than the highest observed. These results highlight the limitations of observational records in sampling extreme tails and indicate that CANARI SMILE captures a broader range of internal climate variability than the initialised SEAS5 seasonal prediction system. They also show that UNSEEN ensembles are more effective at sampling short-duration precipitation extremes (Rx1day and Rx5day) than longer-duration accumulation precipitation extremes (Rx30day). Our study highlights the value of combining the UNSEEN approach with both seasonal prediction systems and SMILEs to better understand unprecedented flood-relevant precipitation extremes in the current climate.