Using high-resolution in situ data from gliders, satellite data of sea level anomaly and geostrophic currents, we presented the detailed structure of an anticyclonic eddy during spring 2015 in the northern South China Sea. The impact depth of the anticyclonic eddy reached about 1000 m and had a maximum temperature anomaly of about 3°C at ∼120 m and maximum salinity anomaly of more than 0.3 psu in the mixed layer. The maximum geostrophic velocities perpendicular to the glider path were about 0.3 m s⁻¹ at 100 m. The estimated radius was about 72 km and the translation velocity was about 5.2 cm s⁻¹. The rotational speed of the eddy was estimated to be 0.35 m s⁻¹. The high temperature and large salinity of the anticyclonic eddy indicated it did not originate from the South China Sea locally. The analysis of water mass indicated the character of the eddy water was similar to Kuroshio water, and the time evolution of the sea level anomaly and surface geostrophic velocity anomaly further validated that it originated from the Kuroshio intrusion as a loop current to the southwest of Taiwan.

Using in situ data, satellite observations, and model outputs, we analyzed the interannual variability in coastal upwelling intensity in the northern South China Sea. Comparing coastal upwelling observed from three cruises during the summers of 2008 and 2016, we found that coastal upwelling was stronger during 2016 compared to 2008, although the local upwelling favorable wind was stronger in 2008. The stronger near‐bottom cross‐shelf current and shallower thermocline in the slope resulted in stronger upwelling intensity during the summer of 2016. The topographic position index (TPI), which is defined by the sea surface temperature difference between one center cell and its neighbors, was used to quantify the interannual variability in upwelling. Stronger (weaker) upwelling intensity occurred during the summers of 2007, 2008, 2011, 2015, and 2016 (2004, 2009, 2012, and 2014) when the local wind was more favorable (less favorable) to coastal upwelling. The correlation coefficient between the area‐weighted TPI and alongshore wind speed was −0.60, thereby confirming that local wind is the primary dynamical factor controlling the interannual variability in upwelling intensity. The correlation coefficient between the area‐weighted TPI and the eastward boundary current transport averaged between the 75‐ and 100‐m isobaths on the shelf was −0.42, indicating that the interannual variability in large‐scale circulation in the northern South China Sea also contributes to the interannual variability in upwelling intensity. The anomalously shallow thermocline in the summer of 2016 was likely associated with the strong 2015–2016 El Niño event through planetary wave propagations.

Using sea surface temperature (SST) data from satellite observations and reanalysis, we revealed the long-term trend of coastal upwelling in the South China Sea (SCS) during 1982–2020. An upwelling index (topographic position index, TPI), defined as the SST difference between the center cell and its neighbors, was used to quantify the upwelling intensity. There are three typical regions (eastern Guangdong, eastern Hainan, and eastern Vietnam) of coastal upwelling in the SCS, in which the upwelling intensity showed an increasing trend during 1982–2020. For all three regions, the upwelling intensity trend reached ~0.2 °C per 10 years in the past 40 years or so. A satellite wind dataset, CCMP, and reanalysis products, ERA5 and ORAS5, were used to discuss the causes of intensified upwelling. Based on the CCMP, the enhanced upwelling in eastern Guangdong and eastern Hainan may be more related to the strengthening of the wind stress curl. Both alongshore wind stress and wind stress curl contribute to the enhanced upwelling in southern Vietnam, and the increasing trend of upwelling in northern Vietnam is not significantly related to the wind field. This study could contribute to the understanding of coastal upwelling response to the long-term climate change in a marginal sea under the background of global warming.

Amid global energy demands and environmental concerns, the exploration of renewable energy sources has become critically important. This study presents a comprehensive analysis of wind energy dynamics over a 35-year period (1988–2022) using CCMP and ERA5 datasets. It focuses on the spatial and temporal characteristics of wind energy within the region, particularly its distribution. The wind field in Sri Lanka has significant seasonal and regional characteristics. Despite seasonal variations, the overall wind activity remains moderate, with speeds generally above 5.4 m/s and annual average wind power densities exceeding 150 W/m², reaching up to 200 W/m² in certain areas. The two primary wind directions in the waters near Sri Lanka are WSW and NE. The study identifies periods of high stability, particularly from June to August, where effective wind speed occurrence (EWSO) exceeds 80%. A ‘W’-shaped pattern in monthly variations corresponds with changes in wind power density (WPD), highlighting optimal periods for wind energy extraction. Additionally, estimated reserves and technically exploitable wind energy resources (WERs) suggest that wind energy development off Sri Lanka is feasible, with potential capacities of approximately 2.00 GW and 1.60 GW, respectively. The overall coefficient of variation (CV) is small, indicating stable wind energy conditions. This analysis not only provides a scientific basis for evaluating WERs near Sri Lanka but also offers valuable insights for strategic planning and development in the renewable energy sector.

Sea surface height (SSH) derived from satellite altimetry is essential for oceanographic research and marine monitoring. Although artificial intelligence (AI) models show considerable potential in forecasting, their application in oceanography remains constrained by several limitations. To address these challenges, we propose a set of physics-informed methods to improve SSH prediction based on neural networks in the South China Sea (SCS). The key strategies include: (1) incorporating land mask information to mitigate artifacts induced by the presence of land in marine data; (2) introducing a geostrophic constraint into the loss function; and applying latitude-dependent weighting to this constraint to account for the breakdown of geostrophic balance near the equator. On the test dataset, the physics-informed SimVPv2 (Phys-SV) model achieves an RMSE of 0.0173 m, a 13% improvement over the baseline SimVPv2 (Base-SV). The PredRNNv2 (PR) model also benefits significantly from the inclusion of land mask input, with RMSE reduced by 12% (from 0.0280 m to 0.0246 m). To the best of our knowledge, this study is the first to identify the artifact issue in AI models caused by land points in ocean data and to reveal the limitations of directly concatenating heterogeneous oceanic variables as model inputs.