With the continuous improvement in computer science and HPC systems on one hand, and the availability of more accurate and frequent observations on the other hand, especially satellite-based ones, data-driven models, namely AI-based and machine learning models, are becoming more efficient and accurate in simulating the Earth system. These developments opened a paradigm shift between considering physical-based models and data-driven ones, or even both combined in the so-called hybrid systems.
Machine Learning for Land Modeling (ML4LM) aims at exploring the extent and the role that machine learning would play for better land surface studies, especially identifying the main areas where it could be applied and providing tools and data to the land surface modeling community. It is a project of the GLASS Panel, which coordinates the evaluation and intercomparison of the latest generation of land models and their applications to scientific queries of broad interest.
ML4LM Science Committee
Chaired by Souhail Boussetta (ECMWF)
Gab Abramowitz, Climate Change Research Centre at UNSW, Sydney, Australia
Andrew Bennett, Hydrology & Atmospheric Sciences, University of Arizona, Tucson, U.S.A.
Souhail Boussetta, Land surface modeling and L-A coupling, ECMWF
Nuno Carvalhais, Max Planck Institute for Biogeochemistry, Germany
Pierre Gentine, Learning the Earth with Artificial intelligence and Physics (LEAP), Columbia University, U.S.A.
Jana Kolassa, Data assimilation, Parameter estimation and machine learning of land Surface models, ECMWF
Sungmin O, Spatiotemporal variability in land surface processes, Deep learning for land, Kangwon National University, South Korea
Jon Page, The University of Oulu, Finland
Ewan Pinnington, Machine learning of land surface, land data assimilation, and ensemble perturbations, ECMWF
Nina Raoult, Parameter estimation and machine learning of land surface models, ECMWF
Marouane Temimi, Stevens Institute of Technology, New York, New York, U.S.A.
ML4LM Webinar Series
The ML4LM webinar series gathers eminent scientists to share their experience in these combined fields. Visit the webinar page for the 2026 schedule. The 2025 schedule with links to presentations is available here.
Recordings for previous webinars are available here or linked to below. Find links to individual presentations below.
Register for March’s presentation at https://gmu.zoom.us/webinar/register/WN_U4Rb2S6AQEierrAieBEQRQ! Prof. Sungmin O (Kangwon National University, Korea) will present “Machine Learning for Land Modeling: Lessons from Hydrologic Benchmarking”.
