Coastal upwelling is an important dynamic process in the marginal sea that transports low-temperature, high-salt, and high-nutrient subsurface water to the sea surface (García-Reyes and Largier, 2010; Shu et al., 2018a; Shu et al., 2018b). Upwelling is important for the global carbon cycle and has a significant impact on nearshore ecosystems and fisheries resources (Ryther, 1969; Barth et al., 2007; Shu et al., 2011; Lachkar, 2014). The South China Sea (SCS) is a semi-enclosed marginal sea in the northwestern Pacific Ocean, with a total area of 3.5 million km² and an average depth of 2000 m. It is connected to the East China Sea to the northeast, the Pacific Ocean and the Sulu Sea to the east, and the Java Sea and the Indian Ocean to the southwest. There is a broad shelf in the SCS that depths shallower than 200m and isobaths roughly parallel to the continental coastline. The southwest monsoon during summer induces offshore transport in the northern and western SCS, driving upwelling in three typical areas, i.e., near the coast of southeast Vietnam, east of Hainan Island, and east of Guangdong province, China (Xie et al., 2003; Wang et al., 2020).

Wyrtki (1961) first noticed the coastal upwelling in the northern SCS (NSCS). A distinct area of low-temperature upwelling exists along the eastern coast of Guangdong ( Figure 1 , Box A). Southwesterly alongshore wind stress drives offshore Ekman transport and creates a pressure gradient perpendicular to the shore, resulting in a strong coastal current, bottom friction drives onshore movement and transports deep cold water upwards (Gan et al., 2009; Wang et al., 2012; Wang et al., 2013; Gan et al., 2015; Wang et al., 2015; Hu and Wang, 2016; Shu et al., 2018a). A positive local wind stress curl can lead to the enhancement of coastal upwelling in northern SCS (Jing et al., 2009; Xie et al., 2016; Hong and Zhang, 2021). The upwelling along the eastern of Guangdong (UEG) is also modulated by the Pearl River plume, which enhanced the upper off-shore current in the upper water column, thereby amplifying the inshore upwelling (Shu et al., 2011; Chen et al., 2019). The upwelling along the eastern coast of Hainan (UEH) is another typical coastal upwelling in the NSCS ( Figure 1 , Box B). During the upwelling season, as a result of the energetic offshore Ekman transport driven by the prevailing southwesterly monsoon, the coastal upwelling nearly completely dominates the area within 100km offshore of the east coast of Hainan Island (Xie et al., 2003; Jing et al., 2015; Lin et al., 2015). The upwelling along the southeastern coast of Vietnam (UEV)is located between 10–15° N along the WSCS coast ( Figure 1 , Box C1 and C2), and the main mechanism is also thought to be offshore Ekman transport induced by the southeasterly monsoon (Loick et al., 2007; Bombar et al., 2010; Loisel et al., 2017; Lu et al., 2018). Meanwhile, the Ekman pumping associated with an eastward wind jet also plays an important role in the upwelling cooling east of Vietnam (Xie et al., 2003). During summer monsoon season, abundant SST fronts contribute to the development of air-sea coupling and affects wind stress curl anomalies, which in turn influence SST and upwelling strength through Ekman pumping (Yu et al., 2020).

There is significant interannual variability of coastal upwelling in the SCS, which is usually considered to be linked to the changes of atmospheric and oceanic circulation (Hu and Wang, 2016; Nadh et al., 2016). ENSO can largely impact on the interannual variability of mesoscale air-sea interaction, modifying local winds and thus modulating the upwelling intensity (Jing et al., 2011; Wang et al., 2012; Yang et al., 2015; Shen et al., 2018; Yu et al., 2020). In the region of UEG, alongshore wind stress is a major factor to the interannual variability of upwelling, and the large-scale current also plays an important role (Shu et al., 2018a). There is a 3-year cycle in the interannual variability of UEH (Xie et al., 2016). In addition to the Ekman transport caused by alongshore wind stress, previous studies have confirmed that wind stress curl significantly contributes to the interannual variability of UEH (Jing et al., 2011; Xie et al., 2012; Wang et al., 2013; Lin et al., 2015);. In WSCS, the interannual variabilities in the southern and northern UEV have different responses to the summer monsoon (Ngo and Hsin, 2021). Strengthening the northward monsoon is favorable to the southern UEV but suppresses the northern UEV (Thai et al., 2022).

After the 1980s, global warming entered an accelerated phase (Varela et al., 2016). In 1990, Bakun firstly proposed the hypothesis that increased thermal differences between land and sea will increase atmospheric pressure gradient forces and summer winds, thereby enhancing coastal upwelling in the eastern boundary areas, such as the coasts of California, Humboldt, Iberian/Canary, and Benguela (Bakun, 1990; Bakun et al., 2015). Previous studies have reported an overall increase in the upwelling intensity along the coast of the eastern boundary (Narayan et al., 2010; Seo et al., 2012; Sydeman et al., 2014; Varela et al., 2016; Wang et al., 2015; Quilfen et al., 2021);. The SCS is one of the sensitive regions in response to global warming (Fang et al., 2006; Park and Choi, 2017; Thompson et al., 2017; Liu et al., 2020). A positive trend of SST front has been found along the Chinese coast and the Vietnam coast (Yu et al., 2019), which may suggest an enhanced coastal upwelling. Coastal upwelling in the SCS has exhibited a significant long-term trend (Liu et al., 2009; Liu et al., 2013; Su et al., 2014; Xie et al., 2016; Hong and Zhang, 2021). Su et al. (2014) showed that the intensity of UEH weakened during 1960–2006. Using an upwelling index defined as the difference in sea surface temperature (SST) between offshore and nearshore waters, Xie et al. (2016) also found that the UEH weakened between 1982 and 2012. However, based on reconstructed SST data from 1906–1993, Liu et al. (2009) showed an intensified upwelling trend for UEH in the last century, which was consistent with the strengthened summer winds observed in the last hundred years. Hong and Zhang (2021) also suggested an intensified upwelling trend in UEG and UEH during 1979–2019 and thought that the wind stress curl was the main mechanism for the intensified upwelling. Therefore, the long-term trend of UEH is inconclusive and might be influenced by the length of time, the method of defining the upwelling index, and the chosen study area. Compared with the study in the UEH, fewer studies have been conducted on the long-term trends of UEG and UEV.

In this study, we used the SST-based topographic position index (TPI) to study coastal upwelling trends in the SCS from 1982–2020. The remaining paper is organized as follows. In Section 2, we describe the data and methods; Section 3 presents the long-term trends of the UEG, UEH, and UEV from 1982 to 2020; Section 4 discusses the potential mechanisms that cause upwelling trends, and Section 5 presents our conclusions.

Previous studies have shown that coastal upwelling in the SCS is closely related to the southwestern summer monsoon (Xie et al., 2003; Gan et al., 2009; Shu et al., 2018a; Shu et al., 2018b). In this study, we explored the potential causes of the enhanced upwelling along the coast of the SCS using wind data products from the CCMP, ERA5, and ORAS5. It should be noted the northern and southern regions of the UEV has opposite responses to summer winds (Thai et al., 2022). Therefore, we divided the UEV into two regions: southern UEV (SUEV) and northern UEV (NUEV). The correlation between the alongshore wind stresses from all three datasets (CCMP, ERA5, and ORAS5) and WTPI was approximately -0.3 in the region of NUEV, further confirming that local wind was not responsible for the NUEV.

Alongshore wind stress is thought to be the main mechanism driving coastal upwelling in the SCS (Shu et al., 2011; Wang et al., 2012; Hu and Wang, 2016; Shu et al., 2018a, Shu et al., 2018b). Figure 6 shows the long-term trends of coastal wind stress from CCMP, ERA5, and ORAS5 in four typical upwelling regions in the SCS from 1982 to 2020. The three wind datasets presented inconsistent trends along the coast of the SCS ( Figure 6 ). The ORAS5 data showed an enhanced trend in coastal wind stress in all three upwelling regions, except NUEV ( Figures 6A3, B3, and D3 ), which was consistent with the intensified upwelling in the SCS over the past 40 years ( Figures 3 , 4 ). The wind stress from CCMP had an increasing trend in the areas of SUEV, whereas it presented a decreasing trend was observed in the area of UEG and did not show a significant trend in the area of UEH ( Figures 6A1, B1, D1 ). In other words, the alongshore wind stress from the CCMP can explain the intensified upwelling in the SUEV region. The long-term trends in alongshore wind stress from ERA5 increased slightly in the UEG region, whereas a slight decrease was observed in the UEH region and did not appear in the SUEV region ( Figures 6A2, B2, D2 ). Thus, the long-term variability in alongshore wind stress from ERA5 did not support intensified upwelling during 1982–2020.

Ekman pumping driven by wind stress curl is another important mechanism for coastal upwelling (Xie et al., 2003; Jing et al., 2009; Jing et al., 2011; Wang et al., 2013). In the UEG, UEH, and SUEV, the climatological wind stress curl was significantly positive. The SUEV is one order of magnitude larger than the UEG and UEH (Shen et al., 2018). Notably, around 13–14° N in the NUEV region, the CCMP, ERA5, and ORAS5 showed a negative wind stress curl ( Figures 7A–C ). The same decreasing trends were observed for all datasets that did not favour intensified NUEV ( Figures 8C1–C3 ). The wind stress curl trend from ORAS5 supports upwelling enhancement in all three upwelling regions, except NUEV ( Figures 8A3, B3, and D3 ). In agreement with the results obtained from ORAS5, the wind stress curl trend from CCMP was favourable to the upwelling enhancement in all three regions of the UEG, UEH, and SUEV ( Figures 8A1, B1, D1 ). Therefore, the long-term trends in wind stress curl obtained by both ORAS5 and CCMP can explain the enhanced trends in the three typical coastal upwelling zones, except NUEV. In contrast, the wind stress curl from the ERA5 data did not show a significant supporting trend in any of the three regions to explain the long-term trend in upward flows.

Previous studies have also shown that different wind data present different, even opposite, long-term trends (Narayan et al., 2010; Seo et al., 2012; Sydeman et al., 2014; Varela et al., 2016; Wang et al., 2015; Quilfen et al., 2021). Using CCMP as a reference, we compared the wind datasets using Taylor diagrams. The high STD of the alongshore wind stress and wind stress curl of ORAS5 (except for the wind stress curl in the UEG region) indicated a high degree of dispersion in ORAS5 ( Figure 9 ). The RMSD and STD of ORAS5 were higher than that of ERA5, indicating that the coastal wind stress and wind stress cyclicity of ERA5 were closer to those of CCMP ( Figure 9 ). There is a problem that in climate reanalysis, which the interaction between any bias in numerical atmospheric models and evolving observational systems could introduce spurious climate signals (Hersbach et al., 2018). For multi-satellite gridded products, it is also challenging to achieve the accuracy required for climate research (Bourassa et al., 2019). The systematic and time-dependent differences between wind products also indicate the need for further research on the temporal stability of these widely used wind datasets (Alvarez et al., 2017; Quilfen et al., 2021). Therefore, a better understanding of the gaps between wind datasets and the selection of appropriate wind datasets are important and require continuous effort.

Because of the differences in the long-term trends in the alongshore wind stress and local wind curl between the three datasets, it is difficult to draw a direct conclusion on the mechanisms of intensified coastal upwelling during 1982–2020 in the SCS. Considering the CCMP is based on satellite data, we regarded it as a reference to discuss the potential mechanism of intensified upwelling. In the NSCS, long-term trends for wind stress curl in the UEG and UEH regions from the CCMP were responsible for the enhanced coastal upwelling ( Figures 8A1 and B1 ), which was also confirmed by Hong and Zhang (2021) and Xie et al. (2016). Xie et al. (2016) pointed out wind stress curl is a more important factor than alongshore wind stress influencing the upwelling in the area of UEH. Hong and Zhang (2021) found that the strengthening of the wind stress curl was an important factor for the enhancement of coastal upwelling along the coast of the NSCS.The enhancement of long-term trends in both alongshore wind stress and wind stress curl from CCMP was favourable to the intensified SUEV during 1982–2020, which was consistent with previous studies (Thai et al., 2022).

In addition, background flow is also an important cause, except for the local wind field contributing to the coastal upwelling intensity in the SCS (Wang et al., 2014; Lin et al., 2015; Shu et al., 2016; Shu et al., 2018a; Shu et al., 2018b). In the coastal area of Vietnam, the enhancement or weakening of NUEV is more influenced by the ocean’s own flow (Ngo and Hsin, 2021). Offshore currents close to the coast, controlled by the spatial distribution of sub-mesoscale and mesoscale dynamical processes, can influence NUEV. The southward or northward currents off the coast of Vietnam, caused by two opposing conditions, wind and offshore circulation, ultimately suppresses the NUEV (Thai et al., 2022). In the NSCS, UEG is not only related to alongshore wind stress, the alongshore currents modulated by topography and thermocline depth in slopes also significantly influence upwelling (Shu et al., 2018a). Non-local western boundary currents also significantly influence the strength of the UEH (Su et al., 2014). Exploring the effects of these factors on the long-term trends in coastal upwelling may require further research.

Using three SST datasets, OSTIA, ERA5, and ORAS5, this study assessed the long-term trends in three typical coastal upwelling zones (UEG, UEH, and UEV) in the SCS during 1982-2020. To remove the long-term trend of the basin-scale SSTs, a TPI was used to quantify the upwelling intensity. All three datasets showed an intensified trend in upwelling intensity in the UEG, UEH, and UEV regions between 1982 and 2020. Among the three datasets, the largest WTPI values were obtained from the OSTIA data calculations. These were -1.12 x 10⁵°C  km ² /10 years for UEG, -2.12 x 10⁴°C  km ² /10 years for UEH and -1.04 x 10⁴°C  km ² /10 years for UEV.

The Ekman transport derived from alongshore wind stress and the Ekman pumping derived from wind stress curl were considered the main mechanisms influencing inshore upwelling. The long-term trends in alongshore wind stress and wind stress curl were calculated using three types of wind data: CCMP, ERA5, and ORAS5. Based on the CCMP and reports from previous studies, the enhancement of UEG and UEH during 1982–2020 may be more related to the wind stress curl. North and south Vietnam had different responses to summer winds, and both coastal wind stress and wind stress curl are possible reasons for the intensified long-term trend of the SUEV. In contrast, in the NUEV region, neither alongshore wind stress nor wind stress curl was the main influencing mechanism. It may be more influenced by currents, which were not explored in this study. The stability of widely used wind data requires further research, and differences in calculated trends can occur with the selection of different wind datasets. It should be noted that we only considered the effect of wind on intensified coastal upwelling in the SCS. There may be multiple influencing mechanisms that require further investigation.

The original contributions presented in the study are included in the article/supplementary material. Further inquiries can be directed to the corresponding author.

YS and JY designed the research and guided the study. SL analyzed the data. SL and YS wrote the first draft of the manuscript. QJ and JZ contributed to the interpretation of the results. All authors contributed to the data analysis discussion and manuscript writing. All authors contributed to the article and approved the submitted version.