Séminaire
Explaining and predicting dynamical drivers of regional climate across timescales using causal representation learning
Fiona Spuler (Leipzig Univ.)
Séminaire du LMD à l’ENS
Description
Large-scale modes of variability in the climate system and their teleconnections are key sources of predictability and modulate future trends in regional extremes. However, existing approaches do not easily bridge the influence of these large-scale drivers to local extremes and so offer limited insight into how internal variability and forced trend uncertainty modulate extremes across models and observations. Here, we present approaches based on causal representation learning, and emerging field of machine learning research that learns a small set of causally meaningful representations from high-dimensional data. We apply these approaches to identify the dynamical drivers of regional extremes and to predict and attribute their changes across timescales. In particular, we study subseasonal drivers of precipitation over Morocco, seasonal teleconnections influencing precipitation over East Africa, and the attribution and prediction of decadal to long-term changes in Western Mediterranean droughts and heatwaves over Germany.
Informations supplémentaires
Lieu
École normale supérieure – PSL
24 rue Lhomond – aile Erasme
salle Claude Froidevaux – E314
Lien visio
https://visio.numerique.gouv.fr/mjz-cvyy-lea