Learning native eddy-resolving ocean dynamics outperforms coarse emulation and downscaling
Abstract
Eddy-resolving ocean simulations are essential for representing mesoscale dynamics that regulate ocean circulation, energy transport, and climate variability. However, they are costly to generate, store, and share. Artificial intelligence (AI) may help, but it remains unclear whether high-resolution ocean fields can be faithfully represented by evolving ocean states at coarse resolution followed by spatial downscaling, or whether dynamical evolution needs to be learned directly from native high-resolution simulations. We show that directly learning dynamical evolution on a 0.1∘ grid provides a more faithful representation of mesoscale variability than evolving ocean states on a 1∘ grid followed by downscaling reconstruction. We developed a full-depth 1∘ emulator and a native 0.1∘ sea surface temperature (SST) emulator, both of which remained stable over long-term free-running simulations without significant drift, providing reliable baselines for this comparison. Direct learning at 0.1∘ preserves mesoscale energy, spectral characteristics, and frontal gradients, whereas the coarse-resolution evolution followed by downscaling approach recovers large-scale patterns but fails to reproduce key mesoscale signatures. We show that the learned mesoscale statistics remain consistent across diverse hardware platforms, supporting the reproducibility and portability of AI-based ocean simulations. These results suggest that downscaling of low-resolution large-scale states alone is insufficient to represent ocean mesoscale variability, supporting the need to learn dynamical evolution directly in a high-resolution state space. These findings highlight the continued importance of high-resolution physics ocean models as a foundation for AI-based ocean simulation, providing dynamically resolved training data for physically faithful high-resolution emulators.
Related
- Deep learning-based downscaling of PM2.5, NO2 and O3 over Europe across emission-reduction scenarios
- Generalizable Neural Downscaling of Earth System Model Wind Fields via Continuous Dynamics Modeling
- A dynamical downscaling approach for the improved representation of sub-annual coastal sea level variability: the importance of shelf and slope ocean dynamics
- GRACE Downscaling and Machine Learning Models for Groundwater Prediction: A Systematic Review
- A spatial downscaling framework for vegetation optical depth based on geographically weighted stacking ensemble learning
BibTeX
@misc{liu2026learning,
title = {{Learning native eddy-resolving ocean dynamics outperforms coarse emulation and downscaling}},
author = {Zilu Liu and Tao Zhang and Hailong Liu and Kai Xu and Qinghui Lin and Maoxue Yu and Jiangfeng Yu and Jingwei Xie and Zhao Jing and Minghua Zhang and Lixin Wu},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.14645},
eprint = {2610.14645},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.14645},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - UNPB
TI - Learning native eddy-resolving ocean dynamics outperforms coarse emulation and downscaling
AU - Zilu Liu
AU - Tao Zhang
AU - Hailong Liu
AU - Kai Xu
AU - Qinghui Lin
AU - Maoxue Yu
AU - Jiangfeng Yu
AU - Jingwei Xie
AU - Zhao Jing
AU - Minghua Zhang
AU - Lixin Wu
PY - 2026
DA - 2026/10/10
PB - arXiv
DO - 10.48550/arxiv.2610.14645
UR - https://arxiv.org/abs/2610.14645
AB - Eddy-resolving ocean simulations are essential for representing mesoscale dynamics that regulate ocean circulation, energy transport, and climate variability. However, they are costly to generate, store, and share. Artificial intelligence (AI) may help, but it remains unclear whether high-resolution ocean fields can be faithfully represented by evolving ocean states at coarse resolution followed by spatial downscaling, or whether dynamical evolution needs to be learned directly from native high-resolution simulations. We show that directly learning dynamical evolution on a 0.1^∘ grid provides a more faithful representation of mesoscale variability than evolving ocean states on a 1^∘ grid followed by downscaling reconstruction. We developed a full-depth 1^∘ emulator and a native 0.1^∘ sea surface temperature (SST) emulator, both of which remained stable over long-term free-running simulations without significant drift, providing reliable baselines for this comparison. Direct learning at 0.1^∘ preserves mesoscale energy, spectral characteristics, and frontal gradients, whereas the coarse-resolution evolution followed by downscaling approach recovers large-scale patterns but fails to reproduce key mesoscale signatures. We show that the learned mesoscale statistics remain consistent across diverse hardware platforms, supporting the reproducibility and portability of AI-based ocean simulations. These results suggest that downscaling of low-resolution large-scale states alone is insufficient to represent ocean mesoscale variability, supporting the need to learn dynamical evolution directly in a high-resolution state space. These findings highlight the continued importance of high-resolution physics ocean models as a foundation for AI-based ocean simulation, providing dynamically resolved training data for physically faithful high-resolution emulators.
ER -