Our ability to model the Earth’s climate is constrained by our ability to model the ocean: oceans function as the climate system's long-term memory. Today, the best models are physics simulation models, or AI models that use the outputs of these massive models. These models are limited in terms of resolution, speed of simulation, and accuracy.

‍

FOMO (Foundation Ocean Model + Observations) is Open Athena's collaboration with NYU and MIT to build a foundation model for the ocean that can accelerate climate research by providing simulation results hundreds or thousands of times faster than can be obtained with traditional modeling. This will enable scientists to simulate climate projections on a much wider range of scales, assess the possible impact of marine carbon dioxide removal, or provide high-resolution local predictions for sea level rise.

‍

FOMO builds on Samudra, which established the first whole-ocean AI simulation that uses realistic atmospheric information. Our goal is to scale Samudra to operate on much larger datasets and higher resolution inputs, including observational data like satellites and ocean buoys. This will enable more accurate simulations of the ocean, better climate projections, and fine-tuning of this foundation model to solve novel downstream tasks.