Session 4 Navigating the Risk of Rare Extreme events in the Era of Climate Change

An AI-driven approach extracted nearly 5,000 historical catastrophes to model rare, extreme events with limited precedent. Using counterfactual scenarios of the 2021 Bernd flood, the study recalibrated its return period from 1-in-23 to 1-in-40 years. Climate-conditioned modelling showed future losses could rise 8–90% by 2050–2075 under different emission and adaptation pathways, underscoring the need to evolve risk models beyond historical data.