WP5 • Storylines of rare and unseen events
WP5 will develop methods for the production of quantitative storylines for plausible but as-yet unseen events focusing primarily on persistent heat and cold waves and meteorological and agricultural drought under climate change. The WP will combine Machine-Learning-based and data-driven methods with methods based on numerical climate models and physical process understanding to generate ensembles of unseen events.
(i) advanced statistical methods (Stochastic Weather Generators: SWG) trained with ML on reanalyses, observations and model simulations,
(ii) physics-based methods “boosting” dynamical global climate models by re-initializing large ensembles for selected most intense extreme conditions for past, present, and future climates.
WP5 will further develop and apply ML-based dynamical adjustment techniques (e.g. quantile regression forests) on very rare extreme events to disentangle the underlying dynamical and thermodynamical drivers of very rare events and assess how they differ from rare events that occurred in the observational record. Finally, the role of climate change on the physical drivers of unseen events will be assessed.
Ultimately, this methodological development in WP5 will allow producing sets of storylines for plausible unseen events in present-day and future climate in collaboration with WP6 and WP7.