Lost (and found) in latent land: applications to weather and climate — Pierre Gentine | HAICON26

Speaker: Pierre Gentine Affiliation: Columbia University / NSF Science and Technology Center "LEAP", USA Date: June 9, 2026 Time: 12:00–12:45pm Session: Keynote Talk Recorded at the Helmholtz AI Conference (HAICON26), June 9–11, 2026. More info: https://haicon.cc/conference-program/ About this talk: Lost (and found) in latent land: applications to weather and climate P. Gentine 1 1 NSF Science and Technology Center "LEAP" / Columbia University, New York, USA Machine learning and AI have been revolutionizing weather and climate modeling over the past few years. However, it remains unclear how much understanding has been gained from those models. Through the study and analysis of carefully chosen latent spaces, I will demonstrate how we can get new understanding on the terrestrial water and carbon cycle as well as on atmospheric processes, specifically on convection. Those latent spaces can also be used to better characterize complex stochastic processes such as turbulence and combined with data assimilation in order to achieve improved performance in those models.