Laure Zanna (Courant Institute, NYU)

Increasing greenhouse gas emissions cause ocean temperatures and sea levels to rise, as the oceans absorb most of the excess energy in the climate system. Climate simulations are essential for understanding and predicting global and regional ocean warming. However, uncertainty remains regarding the causes and pace of future ocean warming due to inadequate representations of unresolved processes, such as ocean turbulence or clouds, in global climate models. Improving these unresolved processes' representations (parametrizations) is crucial in reducing uncertainty in climate projections.

We propose to rethink the parametrization problem in climate models by leveraging different data sources and machine learning advances. In our work, we discover new equations to describe unresolved ocean turbulence processes, such as energy exchange between spatiotemporal scales, and their impact on large-scale flow. Our ongoing efforts to incorporate data-driven ocean parametrizations have simultaneously exposed new challenges in climate modeling and revealed exciting opportunities for delivering reliable climate projections. The next decade holds great promise for climate modeling in the age of data, computing, and artificial intelligence.