Conference Agenda

The ocean 3-D temperature and salinity is an important physical parameter of the marine environment, closely related to ocean physical processes. In situ observation can only obtain sparse temperature and salinity data with spatial and temporal distribution, while satellite remote sensing can only obtain rich observation data of the ocean surface, but cannot detect the subsurface. How to obtain 3D temperature and salinity information of the subsurface based on rich ocean surface remote sensing observation data, and provide high-resolution spatiotemporal seawater temperature and salinity data, is of great significance for ocean monitoring and forecasting, marine science research, carbon cycle and global climate change research.

In response to the problem of obtaining the monthly average subsurface 3D temperature, this study proposes a subsurface three-dimensional temperature field reconstruction model based on multi-source sea surface remote sensing data - CNN Swin Temporal Parallel ConvLSTM (CSTP-ConvLSTM). The model achieves multi-scale modeling of local details, global dependencies, and nonlinear evolution through spatial dual branch structure and temporal enhancement mechanism, improving the accuracy of seawater temperature above 1000 meters; The reconstruction experiment results show that CSTP-ConvLSTM significantly outperforms traditional methods and single deep learning models in overall performance. The evaluation based on ARMOR3D and EN4 independent datasets validates the advantages of the model in terms of spatiotemporal consistency and physical rationality. This study provides an effective approach for obtaining high-resolution 3D temperature fields and lays a methodological foundation for a deeper understanding of the dynamic processes and climate effects of the thermocline. A high-resolution 3D temperature field intelligent reconstruction study based on generative adversarial networks was carried out to address the problem of obtaining daily subsurface 3D temperature fields. A reconstruction model of ocean 3D temperature fields was constructed based on traditional conditional generative adversarial networks and generative adversarial networks that consider seasonal changes. The reconstruction of daily subsurface 3D temperature fields based on SLA, SST, SSS, and sea surface wind data was achieved. The experimental results of 3D temperature and salinity field reconstruction in the South China Sea and Pacific Ocean showed that the reconstruction results were in good agreement with reanalysis data, and could reflect the seasonal variation characteristics of typical vertical sections and the temperature field characteristics of typical ocean phenomena. Aiming at the difficulty of obtaining high-resolution daily 3D temperature and salinity, a spatiotemporal fusion model based on graph attention mechanism was developed to reconstruct the subsurface 3D temperature and salinity fields using a spatiotemporal graph attention network (STGAT). This model achieved the reconstruction of the 3d temperature and salinity based on sea surface satellite remote sensing data such as SLA, SST, SSS, and sea surface wind. The reconstruction experiment of the Northwest Pacific t3d temperature and salinity shows that the STGAT model reconstruction maintains good consistency with reanalysis data, and can fully capture complex and variable ocean dynamic processes, providing a new method for obtaining high-resolution and high-precision ocean 3D temperature and salinity data.

The Lofoten Basin (LB) is located along the advective path of Atlantic Water (AW) is the most eddy active part of the Nordic Seas. A main feature of the LB circulation is the spinning up off eddies from the Norwegian Atlantic Current, which in turn transports warm AW into the basin. The AW transported into the basin interior is found to penetrate deep down to 1000 m, and not surprisingly, the LB is the largest heat reservoir in the Nordic Seas. The eddy shedding from the slope current to the deeper areas of the basin, transports heat to the basin interior and thus mediates the gradual cooling of the slope current on its way toward the Arctic. This is of great importance since AW transported to the Arctic Ocean through the Nordic Seas plays a major role in the global climate system. Furthermore, this also links to the “Atlantification”, a commonly used term for the increasing influence of AW in the Arctic Ocean, having a large impact on the Arctic marine ecosystem.

The overall objective of LoWa is to quantify the impact of ocean eddies on the heat transport across the LB affecting Atlantification. To address the overall objective LoWa uses a holistic approach, thanks to the outstanding opportunities offered by the: available satellite constellations with multi-sensor covariability/synergies; in-situ observational networks, gliders and ARGO drifters; co-location of in-situ and remote-sensed data and data driven Artificial Intelligence (AI) methods. Such an approach also extends the possibility to not only investigate ocean mesoscale eddies but also the sub-mesoscale eddies that are even less studied in comparison to mesoscale eddies. We will report on all ongoing activities in the LOWA project.