Bodner, A.S., Balwada, D., and Zanna, L. (2024)
Presented at:
Ocean Sciences Meeting 2024The parameterization of submesoscale (<10km) ocean surface flows is critical in capturing the effects of vertical buoyancy fluxes in the ocean mixed layer, with significance to ocean-atmosphere interactions and overall climate sensitivity. Here we present a data-driven approach for the submesoscale parameterization, utilizing information from the ultra-high-resolution submesoscale-permitting MITgcm-llc4320 simulation (LLC4320). The new parameterization is given by a Convolutional Neural Network (CNN) trained to predict the subgrid-scale mixed layer vertical buoyancy fluxes as a function of relevant coarse-resolution variables. In contrast to previous physics-based approaches, here the CNN predicts vertical fluxes that are directly computed from the LLC4320 data, where the submesoscales are resolved down to a resolution of approximately 2km. We review how to leverage the LLC4320 data to ensure our method captures submesoscale-relevant dynamics, compare with previous parameterization estimates, present results from training the CNN, and discuss future implementation in coarse-resolution General Circulation Models (GCMs).
