Dr Joshua Dorrington, University of Bergen

14/05/2026 ore 14:30

CNR-ISAC, Bologna – meeting room online

Abstract

Forecast skill for extreme rainfall in the extended range (days 7-21) is generally poor. Nevertheless, ensemble forecasts can show meaningful skill for the occurrence of daily heavy rainfall, sufficient to inform low-regret preparatory actions as part of an early-warning system. However the operational utility of such probabilistic predictions is reduced by interpretability challenges: monitoring tens of ensemble members for high risk scenarios, parsing their narrative, and understanding their uncertainties in an operational setting is often infeasible. To address this, we have developed a framework based on the monitoring of ‘flow-precursors’, regionally specific synoptic weather patterns that favour heavy precipitation. These precursors reduce the forecast to a small number of extreme-relevant scalar indices with clear physical interpretations. These can be used to identify elevated risk of extremes 10-15 days ahead of time, used to improve ML postprocessing, and can be used to identify and interpret extreme-relevant predictability barriers in real time.

Bio

Joshua Dorrington is a Marie Curie postdoctoral fellow at the University of Bergen, where he are working on understanding the narrative of extreme precipitation in large ensemble climate models and high resolution regional simulations. Previously at KIT (Germany) and University of Oxford, he also has a strong interest in the predictability of European weather and using dynamical perspectives to add value to weather forecasts.

Remote Connection