Machine Learning for Climate Modeling: Parameterizing Sub-Grid Turbulent Fluxes for the Ocean Surface Boundary Layer

Abstract: Sub-grid turbulence is challenging to resolve in climate models; therefore, it is parameterized. Traditionally, turbulent parameterizations have relied on physics-based and equation-based approaches. However, ad hoc and uncertain components in these parameterizations introduce uncertainty in future climate predictions. Recently, data-driven techniques have emerged as an alternative for modeling sub-grid fluxes. I will demonstrate the use of machine learning to model vertical turbulent fluxes in the ocean surface boundary layer and its impact on reducing biases in NOAA’s Geophysical Fluid Dynamics Laboratory ocean climate model.

I will show how neural networks, trained to predict the eddy diffusivity profile from high-fidelity yet computationally expensive turbulence schemes, enhance the vertical mixing scheme in the climate model. These networks replace ad hoc components while maintaining the conservation principles of the standard ocean model equations. The enhanced scheme outperforms its predecessor by reducing biases in the mixed-layer depth and modestly improving tropical upper-ocean stratification in ocean-only global simulations. Furthermore, simplified equations that can replace the neural networks show similar improvements but with lower computational cost and better interpretability. They point to structural deficiencies in the baseline parameterization. This work is one of the first successful applications of machine learning to improve a sub-grid parameterization of turbulent mixing in ocean climate models.

Bio: Aakash Sane is a postdoctoral research associate at Princeton University and an affiliate of NOAA’s Geophysical Fluid Dynamics Laboratory (GFDL). His research focuses on small-scale mixing in the upper ocean and its influence on weather and climate. By integrating traditional physics-based modeling with machine learning, he advances the understanding and simulation of oceanic processes.

His recent work leverages machine learning to improve vertical mixing schemes in ocean climate models, enhancing the representation of physical processes and reducing biases. His research highlights how data-driven approaches can complement or replace missing physics in climate models, leading to more accurate simulations.

Aakash earned his Ph.D. in Engineering (Physical Oceanography) from Brown University in 2021 under the guidance of Dr. Baylor Fox-Kemper. His doctoral research involved developing a regional ocean model for Narragansett Bay to improve oceanic forecasting. Before transitioning to oceanography, he studied interfacial fluid dynamics, using theory and experiments to measure the surface tension of flowing soap films, uncovering novel dependencies on film thickness and flow rate.

Beyond research, Aakash actively contributes to open-source climate modeling efforts and serves as a reviewer for leading journals in oceanography and climate science.

*This seminar will be held in person and online*

Topic: IACS Seminar Speaker: Aakash Sane, Princeton University

Time: Mar 31, 2025 01:30 PM Eastern Time (US and Canada)

Join Zoom Meeting

https://stonybrook.zoom.us/j/97764942108?pwd=MzCWupCe3L9mKdrgfO2bJg3GBbvXuf.1

Meeting ID: 977 6494 2108

Passcode: 519324