Simulation and physical understanding of extreme events dynamics under climate change
Abstract: A large portion of climate change impacts comes from the changes in extreme events frequency and intensity rather than mean state evolution. However, unraveling the physical mechanisms behind extreme weather and climate events, and how they may evolve under anthropogenic forcings, is impaired by several issues: few theoretical models exist to explain extreme events dynamics, extreme events are imperfectly represented by numerical models and in general they are undersampled both in observations and in climate models simulations. These issues limit our capability to disentangle forced and unforced evolution in extreme events dynamics and to make reliable projections under global warming. In this talk I will present several methodological and conceptual advances to help address these issues. First, I will detail recent developments in rare events algorithms inspired by statistical physics to sample short and long extremes, with applications to heatwaves and droughts. Second, I will show with the example of the extreme precipitation that followed storm Boris (2024) how decomposing an extreme event into its dynamical and thermodynamical components allows to understand finely how it can evolve in a warmer world. Finally, using a statistical technique from machine learning I will highlight the limitations of climate models in representing the role of long-term natural variability and what consequences it has for the anticipation of megadroughts.
Attention, jour inhabituel pour un séminaire SPEC