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.
Haonan Qi, Bin Lu, Yimian Hu et al.· Proceedings of the 32nd ACM...· 0 citations
Climate change and anthropogenic activities have exacerbated hypoxic conditions in the global ocean, posing a serious threat to marine ecosystems. Quantifying changes in dissolved oxygen levels is crucial for understanding the impact of these factors on marine life and earth sustainability. However, dissolved oxygen observations are severely sparse, limiting comprehensive analysis. Here we present Jingwei, a 4D spatiotemporal graph transfer learning model that reconstructs ocean oxygen levels globally over the past six decades on a 1◦ x 1◦ grid, covering depths of 0-5500 meters. Jingwei simultaneously leverages intra-profile knowledge transfer using a pattern bank from simulations and captures inter-profile correlations through zoning-varying message passing among observations. It significantly reduces reconstruction error by 27.26% compared to CMIP6, demonstrates high consistency with cruise surveys, and provides satisfactory visual quality. Furthermore, Jingwei produces interpretable results, effectively identifying vertical profile patterns and distinguishing fine-grained spatial distributions. Jingwei provides cartography and quantitative analysis of oxygen minimum zones (OMZ) evolution since 1960. We foresee Jingwei revolutionizing observation-based ocean modeling and deepening our understanding of the breathless ocean. Alongside, we have released an open-source online platform (https://jingwei.acemap.info/ https://jingwei.acemap.info/), providing data visualization, resource sharing and ongoing updates.
Bin Lu, Luyu Han, Ze Zhao et al.· Proceedings of the 32nd ACM...· 0 citations
The first oceanic 4D sparse observation reconstruction dataset, named OceanVerse, is presented, providing a novel large-scale dataset that meets the MNAR (Missing Not at Random) condition, supporting more effective model comparison, generalization evaluation and potential advancement of scientific reconstruction architectures.
Bin Lu, Jingjing Shen, Ze Zhao et al.· Proceedings of the 32nd ACM...· 0 citations