Searcharxiv⌕ Search
Home/Archive/arXiv

arXiv · 2610.14645

Learning native eddy-resolving ocean dynamics outperforms coarse emulation and downscaling

Zilu Liu·Tao Zhang·Hailong Liu·Kai Xu·Qinghui Lin·Maoxue Yu·Jiangfeng Yu·Jingwei Xie·Zhao Jing·Minghua Zhang·Lixin Wu
View original source ↗Read source document ↗Export metadata ↓

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$^\circ$ grid provides a more faithful representation of mesoscale variability than evolving ocean states on a 1$^\circ$ grid followed by downscaling reconstruction. We developed a full-depth 1$^\circ$ emulator and a native 0.1$^\circ$ 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$^\circ$ 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.

Explore related subjects

physics.ao-ph

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Zilu Liu, Tao Zhang, Hailong Liu, Kai Xu, Qinghui Lin, Maoxue Yu, Jiangfeng Yu, Jingwei Xie, Zhao Jing, Minghua Zhang, Lixin Wu. 2026-10-10. Learning native eddy-resolving ocean dynamics outperforms coarse emulation and downscaling. https://arxiv.org/abs/2610.14645

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

From Sea Surface Height to Velocity - A Machine Learning Approach

Inferring upper-ocean velocity from altimetric sea surface height (SSH) remains challenging below 50~km, scales now resolved by the Surface Water and Ocean Topography (SWOT) mission. Conventional methods rely on geostrophic or cyclogeostrophic balance. Here, we demonstrate that deep learning can predict the full velocity vector from SSH and environmental data and recover derivative-based diagnostics, including spectral energy flux. We trained a U-Net encoder--decoder on hourly fields from the pre-SWOT Cape Basin extraction of the LLC4320 ocean simulation, with 1.5-km grid spacing ($11.8$--$15.8^\circ$E, $41$--$45^\circ$S). Inputs were SSH, sea surface temperature (SST), and wind stress; targets were horizontal velocities at the surface and 50~m, where wind influence is weaker. At 50~m, the best model achieves a speed RMSE of $0.049\,\mathrm{m\,s^{-1}}$, 29% lower than the geostrophic approximation. Predicted relative vorticity has a correlation of 0.93, compared with 0.75 for the geostrophic approximation. The model also recovers divergence with a correlation of 0.66, whereas geostrophic velocity is nondivergent by definition. Tests of training strategies yield four findings: SSH gradients outperform raw SSH as inputs; wind stress is the most useful covariate for total velocity prediction; SST benefits vary seasonally; and including vorticity and divergence in the loss improves accuracy. These results offer a path for transferring high-fidelity simulation physics to SWOT applications, where kilometer-scale SSH snapshots contain superposed balanced and unbalanced signals.

physics.ao-ph↗
arXiv2026

Machine learning emulators of dynamic global vegetation models for fast gridded carbon budget predictions

The land carbon sink explains most of the variability of the atmospheric CO2 growth rate, but the Global Carbon Budget estimates it with dynamic global vegetation models (DGVMs) run once a year. Emulators could provide it sooner if they reproduce each model outside its training period. Here we show that convolutional neural network emulators of 19 of these models, driven by ERA5 climate and their own past predictions, reproduce the global land sink of these models and its spread over two years not used in training. Global monthly net biome production (NBP) anomalies were reproduced with a median R^2 of 0.81. The error of the multi-model mean annual NBP was at most 0.23 GtC yr^-1. At the grid-cell level, the median R^2 of monthly anomalies was 0.30, partly because of errors that accumulate in the emulated NBP history.

physics.ao-ph↗
arXiv2026

RadarFill: Observation-Constrained Gap Filling of Systematic Blind Zones in 3-D Weather Radar Reflectivity Mosaics

Ground-based weather radar mosaics contain systematic blind zones, especially at low altitudes, where missing reflectivity lacks corresponding complete observations for direct verification. We present RadarFill, an observation-constrained conditional diffusion framework for gap filling in three-dimensional radar reflectivity mosaics. RadarFill conditions on partially observed reflectivity and an observation mask, enforces available observations throughout reverse sampling, and preserves valid measurements in the final field. For training and controlled evaluation, structured synthetic missingness is designed from altitude-dependent missing-data statistics and the vertical nesting of blind zones in an operational 3-D radar mosaic product. Across 4,384 held-out test times and representative convective and typhoon cases, RadarFill recovers principal precipitation structures and moderate-to-strong echoes under extensive low-level missingness and generally outperforms conventional interpolation and a learning-based reference model. For real blind-zone evaluation, RadarFill model is applied directly to Fujian Province and compared with 90 matched volume scans from an independent phased-array radar. Within the principal 1-2 km low-level blind zones, RadarFill yields lower reflectivity errors than conventional interpolation, and the gap-filled precipitation structures show broad agreement with independent observations. In two controlled nowcasting experiments, improvements in RadarFill-initialized fields persist into subsequent forecasts, with the largest benefits at low altitudes and short lead times. These results provide complementary evidence from controlled synthetic missingness, independent radar observations, and downstream nowcasting, supporting the potential of observation-constrained conditional generation for gap filling in real 3-D radar blind zones.

physics.ao-ph↗