AICE Speaker Series: Pierre Gentine (Columbia)

Guest Speaker: Pierre Gentine, Maurice Ewing and J. Lamar Worzel Professor of Geophysics; Professor of Earth and Environmental Sciences; Professor at the Climate School; LEAP Director

Agenda
3:00pm – 3:45pm:  Presentation
3:45pm – 4:00pm:  Q&A
4:00pm – 4:30pm: Reception

Title: Lost in latent land

Abstract: Machine learning and artificial intelligence have transformed weather and climate modeling, delivering unprecedented predictive accuracy. Yet, the extent to which these models advance scientific understanding remains uncertain. In this work, I show how carefully designed latent space representations can reveal new insights into the terrestrial water and carbon cycles, as well as key atmospheric processes such as convection. These latent spaces also provide a powerful framework to characterize complex stochastic dynamics, including turbulence, and can be integrated with data assimilation techniques to further enhance model performance.

Bio: Pierre Gentine investigates the continental hydrologic cycle using multi-scale modeling and big data (machine learning, remote sensing, high-resolution turbulent simulations) in the context of rising CO2 concentrations. Gentine hopes to answer questions such as what will be the future of droughts and extreme dryness/precipitations with a changing climate, and how will they impact ecosystems? Pierre Gentine received his undergraduate degree from SupAéro, the French National Aeronautical and Space Engineering School in Applied Mathematics in Toulouse, France. He obtained a MSc and PhD in civil and environmental engineering from Massachusetts Institute of Technology (MIT) in 2006 and 2010, respectively. He joined the faculty of the Department of Applied Mathematics and Applied Physics at Columbia Engineering in 2010. He is the recipient of the NASA, DOE, and NSF Early Career Award, as well as American Geophysical Union Macelwane medalist.