"Data-driven parameterization of ocean surface boundary layer mixing"
Junhong Liang
Department of Earth, Ocean and Atmospheric Science (EOAS)
Florida State University
Wednesday, Apr 8, 2026
- Colloquium - 499 DSL Seminar Room
- 03:30 to 04:30 PM Eastern Time (US and Canada)
Abstract:
In ocean and climate models, turbulence is not explicitly resolved and must be parameterized. The simulation of upper-ocean physical states, therefore, depends critically on mixing parameterizations for ocean surface boundary layer turbulence. Existing schemes are formulated with algebraic formulas based on physical principles with a few tunable empirical parameters. They cannot fully represent the diversity of conditions in the real ocean, contributing to biases in ocean and climate simulations. In this talk, I will present our recent efforts to develop data-driven parameterizations of turbulent mixing for use in ocean models. In our approach, neural networks are trained on a large ensemble of large-eddy simulations that explicitly compute upper-ocean turbulence. The resulting data-driven parameterizations are more accurate than traditional schemes, while remaining efficient and numerically stable. We have implemented this approach in the Hybrid Coordinate Ocean Model (HYCOM) and demonstrated improved simulations of upper-ocean states in the Gulf of Mexico.
