Robust Ocean Subgrid-Scale Parameterizations Using Fourier Neural Operators
This work addresses errors in long-term climate projections for climate scientists and modelers, though it appears incremental as it builds on existing neural operator approaches.
The paper tackled the problem of approximating small-scale ocean processes in climate simulations, which are computationally expensive to resolve directly, by developing parameterizations using Fourier Neural Operators, resulting in improved accuracy and generalizability compared to other methods.
In climate simulations, small-scale processes shape ocean dynamics but remain computationally expensive to resolve directly. For this reason, their contributions are commonly approximated using empirical parameterizations, which lead to significant errors in long-term projections. In this work, we develop parameterizations based on Fourier Neural Operators, showcasing their accuracy and generalizability in comparison to other approaches. Finally, we discuss the potential and limitations of neural networks operating in the frequency domain, paving the way for future investigation.