Simultaneous emulation and downscaling with physically-consistent deep learning-based regional ocean emulators

TL;DR

This paper introduces a deep learning framework for high-resolution regional ocean emulation in the Gulf of Mexico, effectively downscaling and bias-correcting ocean surface variables while maintaining physical consistency over decades.

Contribution

It presents a novel physics-constrained deep learning model that autoregressively emulates and downscales regional ocean data, addressing stability and bias issues in long-term simulations.

Findings

Accurately emulates ocean surface variables over decades without unphysical drifts.

Successfully downscales from 8 km to 4 km resolution with bias correction.

Maintains both short-term predictive skill and long-term statistical accuracy.

Abstract

Building on top of the success in AI-based atmospheric emulation, we propose an AI-based ocean emulation and downscaling framework focusing on the high-resolution regional ocean over Gulf of Mexico. Regional ocean emulation presents unique challenges owing to the complex bathymetry and lateral boundary conditions as well as from fundamental biases in deep learning-based frameworks, such as instability and hallucinations. In this paper, we develop a deep learning-based framework to autoregressively integrate ocean-surface variables over the Gulf of Mexico at $8$ Km spatial resolution without unphysical drifts over decadal time scales and simulataneously downscale and bias-correct it to $4$ Km resolution using a physics-constrained generative model. The framework shows both short-term skills as well as accurate long-term statistics in terms of mean and variability.