OceanNet: A principled neural operator-based digital twin for regional oceans

Ashesh Chattopadhyay,Michael Gray,Tianning Wu,Anna B. Lowe,Ruoying He

2024-09-05

Abstract:While data-driven approaches demonstrate great potential in atmospheric modeling and weather forecasting, ocean modeling poses distinct challenges due to complex bathymetry, land, vertical structure, and flow non-linearity. This study introduces OceanNet, a principled neural operator-based digital twin for ocean circulation. OceanNet uses a Fourier neural operator and predictor-evaluate-corrector integration scheme to mitigate autoregressive error growth and enhance stability over extended time scales. A spectral regularizer counteracts spectral bias at smaller scales. OceanNet is applied to the northwest Atlantic Ocean western boundary current (the Gulf Stream), focusing on the task of seasonal prediction for Loop Current eddies and the Gulf Stream meander. Trained using historical sea surface height (SSH) data, OceanNet demonstrates competitive forecast skill by outperforming SSH predictions by an uncoupled, state-of-the-art dynamical ocean model forecast, reducing computation by 500,000 times. These accomplishments demonstrate the potential of physics-inspired deep neural operators as cost-effective alternatives to high-resolution numerical ocean models.

Machine Learning,Artificial Intelligence,Chaotic Dynamics,Atmospheric and Oceanic Physics

What problem does this paper attempt to address?

### Problems Addressed by the Paper This paper primarily introduces OceanNet, a digital twin model based on the Fourier Neural Operator (FNO), for simulating regional Sea Surface Height (SSH). The paper aims to address the following issues: 1. **Challenges in Ocean Modeling with Complex Topography**: - Ocean modeling presents unique challenges compared to atmospheric modeling, such as complex seabed topography, land boundaries, vertical structures, and nonlinear flow characteristics. The paper introduces OceanNet to tackle these challenges. 2. **Stability Issues in Long-Term Predictions**: - Data-driven methods often exhibit instability and non-physical characteristics in long-term forecasts. OceanNet mitigates the growth of autoregressive errors and enhances model stability over extended time scales by using FNO and the Predictor-Evaluator-Corrector (PEC) integration scheme. 3. **Capturing Small-Scale Features**: - Deep learning models typically suffer from "spectral bias," making it difficult to capture small-scale features in turbulence. The paper employs spectral regularization techniques to alleviate this issue. 4. **Cost of High-Resolution Numerical Models**: - High-resolution numerical models, while accurate, are computationally expensive. OceanNet demonstrates predictive capabilities comparable to existing high-resolution dynamic ocean models while reducing computational costs by 500,000 times, showcasing its potential as an efficient alternative. In summary, by developing the OceanNet model, the paper aims to provide an efficient and stable forecasting method for regional ocean dynamics, particularly excelling in specific application scenarios such as the circulation eddies and Gulf Stream meanders in the Gulf of Mexico.

OceanNet: A principled neural operator-based digital twin for regional oceans

Evaluation of Deep Neural Operator Models toward Ocean Forecasting

Four-dimensional temperature, salinity and mixed-layer depth in the Gulf Stream, reconstructed from remote-sensing and in situ observations with neural networks

Digital Twin Earth -- Coasts: Developing a fast and physics-informed surrogate model for coastal floods via neural operators

OceanCastNet: A Deep Learning Ocean Wave Model with Energy Conservation

Streamlining Ocean Dynamics Modeling with Fourier Neural Operators: A Multiobjective Hyperparameter and Architecture Optimization Approach

GLONET: Mercator's End-to-End Neural Forecasting System

Spatiotemporal Weighted Neural Network Reveals Surface Seawater Pco2 Distributions and Underlying Environmental Mechanisms in the North Pacific Ocean

Deep neural operators can predict the real-time response of floating offshore structures under irregular waves

Data-driven Global Ocean Modeling for Seasonal to Decadal Prediction

Samudra: An AI Global Ocean Emulator for Climate

Parameterizing Vertical Mixing Coefficients in the Ocean Surface Boundary Layer using Neural Networks

Estimating Ocean Surface Currents With Machine Learning

Ocean reanalysis data‐driven deep learning forecast for sea surface multivariate in the South China Sea

Efficacy of Feedforward and LSTM Neural Networks at Predicting and Gap Filling Coastal Ocean Timeseries: Oxygen, Nutrients, and Temperature

Circumpolar Transport and Overturning Strength Inferred From Satellite Observables Using Deep Learning in an Eddying Southern Ocean Channel Model

AI-GOMS: Large AI-Driven Global Ocean Modeling System

A Fast AI Surrogate for Coastal Ocean Circulation Models

Reconstructing and Forecasting Marine Dynamic Variable Fields across Space and Time Globally and Gaplessly

A comparison of data-driven approaches to build low-dimensional ocean models

SDPNet: A Novel DeepLearning Method for Ocean Surface Current Prediction