Aug 2026· Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2· pp. 11843-11853· 0 citations· 20 references
Abstract
Ocean vertical velocity plays a crucial role in influencing heat exchange, water mass movement, and biogeochemical processes between the surface and deep waters. Due to the difficulty in directly measuring subsurface vertical velocity w, a promising approach is to diagnose w from high-resolution surface observations, like horizontal surface velocity and sea surface height anomaly, via remote sensing. However, both existing traditional dynamic methods and classic machine learning algorithms struggle to provide accurate estimates due to their inability to capture the multi-scale spatiotemporal structures and intense vertical fluctuations. To address these challenges, combined with theoretical ocean dynamics, we propose TriSEFormer, a novel approach that leverages frequency-embedded attention mechanism from a tri-dimensional (3D) frequency perspective. Specifically, TriSEFormer comprises cascaded TriSE blocks, each consisting of a neural dynamics cell following a refinement attention module. The 3D spectral transformation within the neural dynamics cell decomposes the attention-enhanced surface embedding into different spectral components, improving the understanding of multi-scale turbulent structures. Meanwhile, we propose depth-encoded vertical weights to selectively modulate both real and imaginary parts, enhancing the vertical representation compared to dynamic estimation. Extensive experiments on one ideal simulation and three regional ocean datasets demonstrate that TriSEFormer outperforms all baselines, achieving up to a 9.9% improvement, particularly enhancing deep-ocean w diagnosis from 300 m to 2100 m in the Indian Ocean. Code is available at https://github.com/JessiQi25/TriSEFormer.
Submesoscale dynamics strongly influence the upper ocean, regulating mixing, air–sea exchange, and vertical heat transport. The recent Surface Water and Ocean Topography mission provides unprecedented high‐resolution observations of sea surface height (SSH), yet linking these surface measurements to subsurface ocean dynamics remains challenging. We develop a theoretical framework for diagnosing key mixed layer (ML) properties from surface‐observable states. We show that horizontal density anomalies induced by mixed layer eddies produce surface imprints that can be effectively captured by spatially filtered SSH. The filtered SSH is integrated into the ML Eddy parameterization to infer the effects of submesoscale restratification. A potential energy budget accounting for the mixing–restratification competition in the ML is diagnosed from surface buoyancy flux, wind stress, and the SSH gradient, enabling reconstruction of the mixed‐layer depth. Vertical eddy heat flux can be further reconstructed from the SSH gradient. This framework offers a promising approach for diagnosing interior submesoscale processes using surface observations.
Yidongfang Si, Leah Johnson, A. Bodner· Geophysical Research Letters· 0 citations
The 3-D ocean temperature fields are fundamental to ocean dynamics, ecosystems, and climate, yet observing them at high resolution remains a major challenge. A critical gap exists between extensive satellite surface coverage and sparse subsurface in situ measurements—lacking the ability to map the 3-D context around individual profiles. This study introduces and validates a novel paradigm: directly inferring the synoptic 3-D ocean temperature field from a single vertical profile. We propose the profile-to-3D-field network (P3DFNet), a lightweight deep learning model that transforms an instantaneous profile into a 3-D field (∼100 km × 100 km, 0–1000 m); its efficient architecture enables near-real-time deployment on resource-constrained platforms. A key methodological contribution addresses an overlooked issue: conventional root-mean-square error (RMSE) is sensitive to depth-layer configuration, impeding fair cross-study comparisons. We argue for a domain-averaged perspective and propose volume-averaged RMSE (VARMSE) as a robust, discretization-invariant metric, using its squared form as the training loss for evaluation alignment. Developed using HYbrid Coordinate Ocean Model (HYCOM) reanalysis data from 1992 to 2008 for training and 2009 to 2010 for validation, P3DFNet outperforms climatology and neighbor estimations on a held-out HYCOM test set (2011–2012), achieving VARMSE reductions of 59% and 9% while extending the area below the neighbor's average error by 36%. Critically, this advantage is confirmed against independent in situ observations (2014–2024), where P3DFNet reduces error by 44% and 17% versus baselines. This work provides the first proof-of-concept for the profile-to-3D-field paradigm, demonstrating that a single profile contains recoverable information about its spatial context. By transforming point measurements into synoptic fields, this approach augments ocean observing systems and creates new opportunities for integrating sparse in situ data with satellite observations.
G. Zheng, Xuan-Wei Wan, Lizhang Zhou et al.· IEEE Journal of Selected Top...· 0 citations
Abstract. Real-time and accurate three-dimensional ocean temperature–salinity (T–S) field are of great significance for a deeper understanding of ocean dynamics and prediction skill improvement of numerical models. However, current ocean observations, especially those below the sea surface, still suffer from significant limitations in temporal and spatial resolution. Several neural network methods using multi-source satellite data for underwater temperature and salinity reconstruction have been proposed, achieving real-time temperature and salinity reconstruction, but their biases relative to in-situ observations are still significant. This study focuses on the northwestern Pacific region (0–40° N, 120–160° E) and proposes an attention-enhanced three dimensional U-Net++ model, which reconstructs daily T–S fields (26 layers, 1/4° resolution, 5–2000 m depth) using real-time available sea surface temperature (SST) and sea surface height (SSH) data. The model introduces cross-scale feature aggregation and selective information gating, allowing it to emphasize temporally coherent surface features most relevant to subsurface variability, while suppressing noise propagation and over-smoothing. By integrating 26 consecutive days of SST and SSH as inputs, the model effectively alleviates the underdetermined problem of mapping limited surface observations to full-depth structures. In addition, a two-stage transfer learning strategy is employed: the model is first pretrained using monthly SST/SSH data and the gridded Argo data to learn observation-dominated low-frequency spatiotemporal patterns, and then fine-tuned using daily SST/SSH data and the high-resolution reanalysis to capture the meso-scale dynamic processes. Evaluation results show that the reconstructed T–S fields agree better with in-situ T–S profiles from World Ocean Database than previous studies, both during the validation period and in long-term statistical analyses, suggesting that the proposed approach is reliable and accurate for subsurface ocean field reconstruction. The reconstructed T–S field is available at https://doi.org/10.57760/sciencedb.31950 (Wang et al., 2025).
Hao Wang, Linlin Zhang, Shuguo Yang et al.· Earth System Science Data· 2 citations
The advanced 2-D sea surface height (SSH) observations from the surface water and ocean topography (SWOT) satellite have achieved a global average geoid resolution better than 13 km, as estimated from spectral analysis of adjacent single-cycle observations. However, in regions with complex oceanic dynamics (the ocean off southwestern Argentina and east of Japan), SWOT observations show significant temporal variability across cycles. This variability greatly hampers the satellite’s ability to detect and precisely measure steady-state marine geoid signals. Even with multicycle data stacking, resolution remains poor in some regions, sometimes underperforming conventional satellite altimetry. To solve this issue, this study employs a multiresolution coherent spatio-temporal scale separation (mrCOSTS) method to analyze sea surface height anomaly (SSHA) data from 26 SWOT cycles, aiming to isolate stable signal components and reconstruct the geoid. The approach notably improves geoid resolution in regions with intricate ocean features, with the maximum enhancement reaching up to 76%. This resolution reflects the spatial scales of features reproducibly resolved by two independent SWOT geoid estimates, rather than a direct improvement in absolute marine geoid accuracy. The method effectively suppresses ocean dynamic signals while preserving the static oceanic component, and it also contributes to improving the recovery of the marine gravity field in regions characterized by complex ocean dynamics. Furthermore, in regions with relatively high resolution, the method achieves additional subtle improvements, which are of considerable significance for obtaining a stable static field.
Sihai Zhao, Shengjun Zhang, Xiangxue Kong et al.· IEEE Transactions on Geoscie...· 0 citations
The Surface Water and Ocean Topography satellite mission now delivers global sea surface height (SSH) observations at scales fine enough to resolve submesoscale eddies (<50 km). At these scales, the traditional geostrophic approximation, commonly used to infer surface currents from SSH, no longer holds. Here, we present a new dynamical framework that diagnoses ageostrophic currents and, in particular, divergent motions directly from SSH. The framework is trained and validated using a high‐resolution numerical simulation of a western boundary current system, where submesoscale eddies are the most energetic. This approach highlights the unique capability to reveal vertical motions in the upper ocean from SSH, allowing for diagnosing transport of heat, carbon, oxygen, and nutrients between the surface and the interior of the ocean.
H. Torres· Geophysical Research Letters· 0 citations
This study presents a novel cyclostrophic balance correction method for estimating submesoscale ocean surface currents in the Northern Arabian Sea using surface water and ocean topography (SWOT) mission altimetry. High-resolution Ocean Color Monitor (OCM-3) data from the EOS-06 satellite reveal fine-scale eddies and filaments with high chlorophyll-a concentrations (>0.4 mg m−3), spatially coherent with geostrophic current patterns from SWOT. At these scales, the geostrophic assumption is invalid; therefore, we introduce a curvature-based cyclostrophic correction that accounts for enhanced centripetal accelerations. Validation against high-resolution model simulations shows that our approach is in better agreement with model outputs than uncorrected and previously published corrected fields, particularly in regions with strong vorticity and strain. When applied to SWOT data, the corrected velocities demonstrate spatial correspondence with chlorophyll patterns and suppress spurious gradients. Probability density functions of normalized vorticity and strain also match theoretical expectations, emphasizing the potential of SWOT for advancing submesoscale ocean dynamics.
N. Agarwal, Aditya Chaudhary, J. M. et al.· IEEE Journal of Selected Top...· 0 citations