The coefficient of variation (CV) of the satellite Chl-a data was used to evaluate the Chl-a variability in summer, which is calculated as the ratio of the standard deviation of the monthly mean Chl-a in summer to the summer mean Chl-a. According to the definition, a larger CV means a high variability of Chl-a. To examine the effect of wind, SST, and Changjiang River discharge (CRD) on summer phytoplankton biomass in the Yellow Sea, we performed correlation analyses on the monthly mean surface Chl-a concentrations, wind, SST, and CRD in summer (June, July and August). As the monthly mean wind direction basically remains the same in summer (e.g., Lin et al., 2019b), we used the monthly mean wind speed to quantify the wind intensity. Distributions of Chl-a in June, July, and August all show similar patterns and weak monthly variability during summer (see Figure S1 in the supplementary material), indicating the basically identical environmental background for the three summer months in the Yellow Sea. To further avoid the potential effect of the intrinsic monthly variation in one summer on the correlation, the anomalies of monthly mean values were used in the correlation analysis, which was calculated by subtracting the corresponding climatological monthly mean from the monthly mean values. Using the least squares method, we analyzed the correlations between the anomalies of wind, SST, CRD, and Chl-a concentration in June, July, and August of 2002-2020. Moreover, to understand the effect of the climate variability on the interannual variability of the summer Chl-a, i.e., PDO and El Niño-Southern Oscillation (ENSO), the correlation between the summer mean Chl-a concentration and the PDO and Niño 3.4 Indexes. As the strong signal of ENSO occurs in winter and there are usually several month lags between the variabilities of the climate and coastal environment (e.g., Zhai et al., 2021), the mean Niño 3.4 Index in winter (January to March) was used in the correlation analysis. The correlation coefficient (R value) was used to quantify the degree of correlation, and the positive and negative R values denote positive and inverse correlations, respectively. The statistical significance (P value) was assessed using an F-test.

In this study, the wind-driven upwelling theory was adopted in our analysis to understand the response of the western Yellow Sea to summer winds. The 2D analytical solution derived by Estrade et al. (2008) suggests that a wind-induced upwelling cell should be concentrated in the region of

where h is the water depth and D = π 2 A υ f is the thickness of the Ekman layer ( A _υ is the vertical viscosity coefficient and f is the Coriolis parameter).

According to the wind-driven Ekman transport, the magnitude of wind-driven coastal upwelling can be estimated as (e.g., Estrade et al., 2008):

where L is the cross-shore width of the upwelling region (~140 km on average according to the observation in the western Yellow Sea), ρ ₀ is the reference density (1025 kg/m³) and τ is the sea surface alongshore wind stress. The wind stress related to wind speed is formulated as:

where ρ _a is the air density (~1.22 kg/m³), C _d is the drag coefficient (0.0013) and V → w i n d is the alongshore wind speed. As the summer wind direction is basically parallel to the coastline of the western Yellow Sea ( Figure 1C ), the summer mean wind speed was directly used here to estimate the upwelling intensity in the western Yellow Sea. Based on Equations (3) and (4), the average magnitude of wind-induced upwelling in the western Yellow Sea can be estimated using:

Based on the upwelling velocity and nutrient concentrations of the YSCWM, the average nutrient fluxes to the upper layer induced by wind-driven upwelling can be estimated using:

where F is the upwelling nutrient flux, C is the average nutrient concentration in the YSCWM and w is the upwelling velocity.

The satellite data showed that the relatively high variability (high CV) of summer Chl-a occurred in the continental slope region where the water depth ranged between 20 m and 60 m ( Figure 3A ). Meanwhile, there was a significant positive correlation between summer winds and satellite Chl-a concentrations at the water depth of 20-60 m in the western Yellow Sea ( Figure 3B ). Whereas, in the eastern Yellow Sea, the wind-Chl-a correlation was negative ( Figure 3B ). In contrast, the spatial correlations between SST ( Figure 3C ), CRD ( Figure 3D ) and Chl-a were much weaker than the wind-Chl-a correlation. The location overlaps of the high Chl-a variability and the strong wind-Chl-a correlation ( Figures 3A, B ) suggest the effect of the summer wind on the variability of summer Chl-a in the southwestern Yellow Sea.

As the Pacific climate variabilities (especially PDO and ENSO) have important effects on the East Asian Monsoon system (e.g., Chen et al., 2013; Kim et al., 2014), they could indirectly influence the Chl-a in the Yellow Sea, which is demonstrated by the negative correlation between the summer mean Chl-a and both PDO and ENSO in the western Yellow Sea ( Figures 4A, B ). As shown in Figure 4 , the region with the strong negative correlation between the summer Chl-a and the climate variabilities concentrated in the western Yellow Sea and basically corresponds to the region with the strong positive wind-Chl-a correlation ( Figure 3B ). Although the effect of ENSO and PDO on the East Asian monsoon could be complex and not be a simple linear correlation (e.g., Kim et al., 2014), the PDO/ENSO positive (negative) phase over the past nineteen years basically corresponded to the negative (positive) anomaly of the summer wind speed of the Yellow Sea ( Figure 4C ). Thus, the negative PDO/ENSO-wind and positive wind-Chl-a correlations in the western Yellow Sea could result in the inverse correlation between the PDO/ENSO and Chl-a, implying that the wind variability could be the pivotal factor driving the climatic variability of the summer Chl-a in the western Yellow Sea. In addition, PDO/ENSO could also influence SST and CRD besides wind. However, as shown in the correlation analysis ( Figures 3C, D ), the effects of SST and CRD on the summer Chl-a were much weaker than that of wind in summer (the potential cause will be discussed in Section 4.3). Thus, it is more likely that the PDO/ENSO influences the summer Chl-a variability by modulating the summer wind.

Based on two summer field surveys in C15 (weak wind time) and C18 (strong wind time), we analyzed the variations of sea temperature, salinity, nutrients along the three transects ( Figure 1A : 34°N, 35°N and 36°N) to understand the effect of wind on the southwestern Yellow Sea. As shown in Figure 5 , significant thermoclines and stratification occurred during the C15 and C18, with water depth > 20 m, high salinity (> 32), and low temperature (<12°C) occurring below the thermoclines, showing the edge position of the YSCWM. Compared with the T-S characteristics during weak wind observation (C15), high salinity water tongue (~32–33) during strong wind observation (C18) expanded more toward the coastal side and tilted upward to the surface waters ( Figures 5A , and 6 ), indicating the occurrence of upwelling along the continental slope of the southwestern Yellow Sea under a strong wind condition.

The spatial patterns of DIN, DIP, and DSi ( Figure 7 ) along the three transects matched well with the patterns of water temperature and salinity ( Figure 5A ). The average DIN, DIP, and DSi concentrations in the YSCWM were similar during the observations in the C15 and C18 and larger than 6.0 μM, 0.32 μM, and 6.5 μM, respectively, which were 2-3 times higher than those in the upper waters ( Table 1 ). Similar to the water temperature and salinity pattern, during the C18, high nutrient water expanded more to the coastal side along the seabed slope and tilted upward to the upper waters compared with their characteristics during the C15. The results reflect that the upwelling can transport nutrients from the YSCWM to the upper waters.

The relatively high Chl-a concentrations responded to the upwelling during the C18 ( Figure 8 ). In comparison, the Chl-a concentrations along the three transects in C18 were about twice of that in C15 ( Table 1 ). In addition, relative to Chl-a in C15, the spatial scale of high Chl-a in C18 was denser and covered a wider range, including the edge of the YSCWM ( Figure 8 ).

In summary, the observation results of the two cruises show that the stronger upwelling along the slope in the southwestern Yellow Sea occurred during the cruise with the stronger wind, which pumped up more nutrient from the bottom water and resulted in a higher Chl-a in the upper layer. Although other factors (e.g., tide and stratification) could contribute to the different results of the two cruises, the wind is considered as the main factor inducing the strong upwelling and the higher Chl-a in the C18, which is further discussed in Section 4.

The observation results showed a stronger upwelling along the slope in the southwestern Yellow Sea during the C18. However, the two summer cruises C15 and C18 were carried out in August and July, respectively, and have about 20 days interval. Beyond wind, other environmental factors, such as tides and stratification, could also be different and influence the coastal upwelling. The upwelling induced by tidal mixing fronts and its contribution to the nutrient fluxes in this region has been reported by Lü et al. (2010) and Wei et al. (2016). The numerical studies by Huang et al. (2018) and Liang et al. (2018) suggested the existence of a relation between the upwelling intensity induced by the tidal mixing fronts and the tidal phase. Their study indicated that the spring tide phase (larger tidal range) corresponds to a stronger upwelling in the southwestern Yellow Sea than the neap tide phase (smaller tidal range). Our analysis showed that the tidal range during C15 was slightly larger than that during C18 (see Figure S2 in the Supplementary Material ), but the observations demonstrated a stronger upwelling during C18, implying that the stronger upwelling observed in C18 than C15 should not be induced by the difference in the tidal phase. Meanwhile, the intensity of stratification could also affect the intensity of the coastal upwelling to some extent (e.g., Jacox and Edwards, 2011). However, the temperature sections showed no significant difference in the stratification intensity between the two cruises ( Figure 5 ). Hence, we suspected that the stronger upwelling observed during C18 could be mainly caused by wind since the upwelling-favorable wind during C18 was much stronger than that during C15 ( Figure 2 ). Our speculation was further confirmed using the wind-driven upwelling theory (Method section: Equation 2) as the following analysis.

According to Estrade et al. (2008), wind-driven upwelling in coastal waters should occur in the region of 0.5D < h < 1.25D. According to the model results of Lin et al. (2020), the order of A _υ in the Yellow Sea in summer was ~10^-2 m²/s, and f was ~10^-4 1/s. Thus, the thickness of the Ekman layer D can be calculated as ~50 m for the southern Yellow Sea during summer. Finally, the wind-driven upwelling in the Yellow Sea should occur at 25 m < h < 62.5 m according to the wind-driven upwelling theory, which is highly consistent with the upwelling location inferred from the observed nutrient distribution in C18 (see the box region in Figure 7A ). Meanwhile, the satellite-derived high wind-Chl-a correlation also was found exactly in this region ( Figure 3B ), indicating the important role of the wind-driven upwelling on the variability of summer Chl-a.

Moreover, the analytic analysis by Estrade et al. (2008) suggests that a steep sea-bottom slope can induce more intense upwelling than a gentle slope under the same wind conditions. In this study, the slope of the seafloor at the 34°N transect was steeper than that of the other two transects ( Figure 5 ). Higher nutrient concentrations in the upper layer of the upwelling region in the 34°N transect indicate more intensified upwelling at 34°N than at 35°N and 36°N ( Figure 7 ), which agrees with the wind-driven theory. The average wind speeds during the week before C18 and C15 were 6.0 m/s and 2.7 m/s, respectively. The two field survey months generally showed southerly wind direction, which is upwelling-favorable for the western Yellow Sea ( Figure 2 ). According to Equation (5), the upwelling velocities for 6 m/s and 2.7 m/s winds were estimated as ~0.47 × 10^-5 m/s and <0.1 × 10^-5 m/s, respectively. The significant difference in the upwelling intensity estimated by the wind-driven upwelling theory could well explain the observed difference between the two cruises.

The wind-driven upwelling could increase the nutrient release from the YSCWM and thus rise Chl-a concentrations in the upper layer. Compared with the C15, the surface nutrient concentration in the C18 significantly increased ( Table 1 ), indicating the significant contribution of wind-driven upwelling to nutrient upward transport. Thus, both satellite and observation data showed the positive effects of summer winds on phytoplankton biomass at the continental slope region in the southwestern Yellow Sea. In general, cold water masses with rich nutrients appear in shelf seas (also referred to as the “cold pool”) in summer because of ocean stratification, e.g., Middle Atlantic Bight, North Sea, and Irish Sea (Houghton et al., 1982; Brown et al., 1999; Holt and Proctor, 2003). Some earlier studies have reported the significant impact of nutrients released from the cold pool on primary production on shelves (Dickey-Collas et al., 1997; Stabeno et al., 2007; Na et al., 2010). The turbulence entrainment and upwellings induced by tidal-mixing fronts are also important physical processes for transporting nutrients from the bottom to the upper water during summer (e.g., Wei et al., 2016). However, the tidal-mixing front is relatively stable with negligible monthly variability during summer (Pisoni et al., 2015; Lin et al., 2019b). Thus, wind with relatively strong variability becomes an important factor to drive the monthly and interannual variability of the summer chl-a ( Figure 3 ).

Based on Equations (5) and (6), the contribution of the wind-driven upwelling to the nutrient upward flux can be estimated. The climatological mean wind in the Yellow Sea during summer is ~4.5 m/s, and thus the mean wind-induced upwelling in summer was estimated at ~0.26 × 10^-5 m/s. According to Equation (6), the estimated mean nutrient fluxes DIN, DIP, and DSi (see Table 2 ) induced by wind-driven upwelling were 1345 ± 674 μmol/m²/d, 81 ± 45 μmol/m²/d, and 1460 ± 899 μmol/m²/d, respectively, accounting for ~30% of the total DIN and DSi supply and ~40% of the total DIP supply. These numbers are comparable to the upwelling fluxes induced by the tidal mixing front ( Table 2 ), which is supported by the finding in the numerical study by Wang et al. (2021) that the summer wind and tide contribute almost equally to the coastal upwelling system in the southwestern Yellow Sea.

As shown in Table 2 , the DIN fluxes contributed by the wind-driven upwelling are also comparable to the horizontal transport fluxes from nearshore water, but the contribution of the former to the DIP and DSi fluxes is much larger than that of the latter. Especially, the DIP nutrient flux by wind-driven upwelling was several times higher than the horizontal DIP transport flux from nearshore waters, which can sufficiently compensate for the phosphorus deficiency of the upper layer (Wang et al., 2003). Moreover, the long residence time of the nearshore water in summer (average ~91 days according to Lin et al., 2022) could also limit the effect of the nearshore nutrient on the phytoplankton in the central Yellow Sea. Thus, for the southwestern Yellow Sea, the nutrient supply from the bottom water could be more important than that from the supply from the nearshore water.

Wind can enhance not only upwelling but also vertical turbulent mixing, both of which can facilitate the upward transport of the bottom water (e.g., Williams et al., 2013). Based on observed data and a turbulent model in summer, Wei et al. (2002) calculated the average upward fluxes of DIN, DIP, and DSi driven by turbulent mixing in the Yellow Sea, which were 211 μmol/m²/d, 16.4 μmol/m²/d, and 284 μmol/m²/d, respectively. These numbers are approximately one order of magnitude smaller than those caused by wind-driven upwelling ( Table 2 ), implying that the importance of wind-induced mixing in pumping bottom nutrients is much weaker than that of the wind-driven upwelling, especially for the continental slope region, which can well explain why the strong correlation between Chl-a and wind only occurred at the upwelling region rather than the entire sea. This conclusion is also supported by the numerical study by Meng et al. (2020). Meng et al. (2020) analyzed the effect of wind on the variability of the water temperature in the Yellow Sea in summer and pointed out that the mixing played a more important role in the variability in the central Yellow Sea while the advection dominated the variability in the slope region. In addition, our results also showed a negative correlation between winds and Chl-a in the eastern Yellow Sea ( Figure 3B ), indicating the dominance of wind-driven downwelling on phytoplankton variability over wind-induced mixing, because the latter can enhance Chl-a concentrations and should lead to a positive wind-Chl-a correlation. Therefore, the effect of wind on the summer Chl-a variability in the southwestern Yellow Sea should be mainly related to the wind-driven upwelling process, and the wind-induced mixing could play a secondary role. In addition, we further analyzed the correlation between the southeasterly wind (the alongshore component of the wind which is the major contributor to coastal wind-driven upwelling) and Chl-a in summer. The correlation pattern of the southeasterly wind with Chl-a was very similar to that of the wind speed ( Figure S3 in the Supplementary Material ), further suggesting the dominant role of the wind-driven upwelling in the wind effect on the Chl-a variability.

The effect of nutrient input from CRD was considered here because of the high Chl-a variability near the northern CRD ( Figure 3D ). The Changjiang River is the third largest river in the world and delivers abundant nutrients into the sea. However, Yamaguchi et al. (2012) and Wang et al. (2014) indicated that the influence of the Changjiang River diluted water mainly concentrated in the region from the central East China Sea to the Jeju Island. Under the forcing of the southeasterly summer wind, the Changjiang River diluted water extended northeasterly to the Jeju Island due to the Ekman transport (Lie and Cho, 2016), which could explain its influence on the regions from offshore of the Changjiang River mouth to the east of Jeju Island. However, the correlation between CRD and Chl-a was relatively weak in the southwestern Yellow Sea and even negative in some places, suggesting the limited effect of CRD on the southwestern Yellow Sea. This might be related to the influence of tides and the southwards of tidal residual currents in the Jiangsu offshore water during summer (Xuan et al., 2016). Thus, from the perspective of the ocean current, the Changjiang River’s diluted water might have a relatively small effect on the phytoplankton in the southwestern Yellow Sea during summer. In addition, the correlation between CRD and Chl-a was not significant in the Changjiang estuary offshore region ( Figure 3D ), which was also found in Yamaguchi et al. (2012). The insignificant correlation between Chl-a and CRD could be due to the large nutrient stocks in the offshore region of the Changjiang estuary, which is sufficient for the phytoplankton. However, there was a strong light limitation for the phytoplankton in this region due to the high water turbidity (e.g., Li et al., 2021), and thus the light was more important for the phytoplankton growth in this region, which can explain the weak correlation between CRD and Chl-a in the Changjiang estuary offshore region.

This study also presented a weak correlation between Chl-a and SST in the Yellow Sea during summer. SST can influence phytoplankton by affecting the density stratification between the surface layer and the nutrient-rich bottom waters, which can modulate vertical turbulent mixing and supply the upper layer with bottom nutrients for phytoplankton utilization (Boyce et al., 2011). However, the cold YSCWM and the strong surface heating induced a significant vertical temperature difference of >14°C within a water layer thickness of just 20 m ( Figure 5A ). This produced a highly stable thermocline in the Yellow Sea during summer, which could not be easily affected by the SST variability (monthly ±1.1°C). Earlier biophysiological studies also mentioned the poor effect of the SST variability of ±1.1°C on the growth rate of phytoplankton (e.g., Eppley, 1972; Berges et al., 2002). This should explain why the change in SST has a limited impact on phytoplankton biomass in the Yellow Sea in summer. Some studies reported significantly inverse correlations between interannual SST and Chl-a in the Yellow Sea (Liu and Wang, 2013; Liu et al., 2019; Zhai et al., 2021). Lin et al. (2019a) suggested the importance of water turbidity and vertical average irradiance in influencing Chl-a variability during winter. Thus, the inverse interannual SST-Chl-a correlation might be related to the impact of SST on phytoplankton biomass in spring and autumn during the occurrence of phytoplankton blooms. Both spring and autumn phytoplankton blooms are related to water stability (e.g., Song et al., 2010; Shi et al., 2017; Lv et al., 2022). Due to the relatively weak thermocline in spring and autumn, the SST variability might have a more significant effect on water stability and thus influence phytoplankton biomass in spring and autumn. Nevertheless, the detailed mechanism of the SST effect on interannual Chl-a variability is beyond the scope of the present study and will be investigated in the following study.