Hydrographic, bio-optical, and biogeochemical data from the NSCS were simultaneously acquired during the Opening Research Cruise of the R/V Shiyan 3, operated by the South China Sea Institute of Oceanology, Chinese Academy of Sciences, from August 16 to September 6, 2008. Figure 1 shows a map of the investigated station locations. The NSCS shelf can be sub-divided into the inner shelf, the middle shelf and the outer shelf corresponding to water depths of <40 m, 40–90 m and 90–120 m, respectively (Wong et al., 2015). Open ocean generally refers to water depths greater than 120 m.

Vertical profiles of temperature and salinity were acquired using a ship-launched SeaBird CTD and sampling rosette. The ocean mixed layer is the top layer of the water column, where the temperature and salinity are vertically well mixed. The mixed layer depth (MLD) was defined using the threshold method with a finite difference criterion based on a near-surface reference value. The reference depth is set at 10 m to avoid much of the strong diurnal temperature cycle that occurs in the top few meters of the ocean. The chosen temperature criterion is 0.5°C absolute difference, as recommended by Kara et al. (2000).

In situ profiles of inherent optical properties (IOPs) were measured using a ship-deployed profiling package at 48 stations, as shown in Fig. 1. The WETLabs ac9, which consists of dual, 25 cm path length flow tubes for each path, was used to measure the spectral absorption and attenuation coefficients at nine wavelengths (412, 440, 488, 510, 532, 555, 650, 676 and 715 nm). The ac9 data were corrected for temperature- and salinity effects using the coefficients derived by Sullivan et al. (2006). The proportional method (Zaneveld et al., 1994) was the applied to correct for scattering errors in the absorption measurements. We did field calibration before and in the middle of the observation period to monitor the stability of the ac9. Here, a_pg(λ) and c_pg(λ) denote the non-water spectral absorption and attenuation coefficients, respectively. All notations are as shown in Table 1.

Apparent optical profile data were collected at about 16 stations during the daytime. A freefall spectroradiometer (Satlantic, Inc.) was used to measure the vertical profiles of downwelling spectral irradiance (E_d(z, λ)) and upwelling spectral radiance (L_u(z, λ)) at seven wavelengths (412, 443, 490, 520, 555, 620 and 683 nm) away from the ship’s shadow. Photosynthetic active radiation (PAR (quanta/(cm²·s)) was calculated from E_d(z, λ) spectra with the ProSoft software provided by Satlantic Inc. The depth of the euphotic zone, Z_eu, defined as the depth where irradiance was reduced to 1% of its surface value, was calculated based on the vertical profile of PAR.

Water samples were collected using Niskin bottles at selected depths (0 m, 10 m, 25 m, bottom in the coastal ocean, and 0 m, 25 m, 50 m, 100 m in the open ocean) to determine the Chl, the particulate absorption coefficient (a_p(λ)), and the particulate organic carbon (POC) concentration.

Water samples of roughly 0.5–1 L were collected at standard depths within the euphotic zone to be measured for Chl. Filters were flash-frozen and initially stored in liquid nitrogen and then stored at –80°C prior to analysis in the laboratory. Pigments were extracted with acetone at –20°C for about 24 h under dark conditions, and then analyzed with a Turner Designer 10 fluorometer (Parsons et al., 1984). Light absorption by suspended particles (a_p(λ)) was determined using the quantitative filter technique (QFT) (Mitchell et al., 2003; Lin et al., 2014). POC data were obtained using a method generally consistent with the Joint Global Ocean Flux Study (JGOFS) protocols (Knap et al., 1996; Wang et al., 2011).

Daily sea level anomaly (SLA) data, based on a merged product from multiple satellite missions (T/P and ERS-1/2, followed by Jason-1/2 and Envisat) and distributed by Archiving, Validation and Interpretation of Satellite Oceanographic data (AVISO, http://www.aviso.oceanobs.com/) were used to describe the background physical properties of the seawater. These have been mapped to global Mercator grids with (1/4)°×(1/4)° spatial resolution. Maps of sea surface temperature (SST) with a 6 km×6 km spatial resolution were provided by the Group for High-Resolution Sea Surface Temperature (GHRSST) Project. Daily satellite chlorophyll data were collected from version 3.0 of the Ocean Colour Climate Change Initiative (OC-CCI) and a merged MERIS, MODIS-Aqua, SeaWiFS and VIIRS products available at http://www.oceancolour.org/).

Following the classification of Wong et al. (2015), average hydrographic and IOPs in these hydrographic sub-divisions (inner shelf, middle shelf, outer shelf and the open ocean) in summer 2008 are summarized in Table 2. The relationship between potential temperature (θ, °C), and salinity (S) at observed stations across the shelf sea are shown in Fig. 4. The θ-S relationship in inner shelf waters is obviously different from that of the open ocean. Two stations collected in the inner shelf waters were located in a center of upwelling (with lower temperature and high salinity in the surface waters) and in the plume waters of the Zhujiang River Estuary (with high temperature and low salinity in the surface waters). In contrast, the water over the outer shelf was warm and saline, with a mean temperature of 28.7°C and a salinity of 33.56 in the mixed layer. The density of outer shelf water (mean σ_θ=21.10 kg/m³) is even higher than that of the open ocean (σ_θ=20.82 kg/m³). The middle shelf water was generally found to be a mixture of the inner- and outer shelf waters, with σ_θ=20.54 kg/m³ and S=32.96. The extensions of θ-S relationships in the inner shelf converged with those further offshore at a salinity of roughly 34.2, a temperature of 22.03°C and a density of 23.74 kg/m³. This point of convergence, which likely represents the source water of the inner shelf, was traced to a depth of about 40–50 m in the middle and outer shelf. Except for some specific points with low salinity in upper surface waters, the θ-S relationships in the outer shelf waters match well with those observed in the open ocean, which converged at a salinity of 34.5 and a temperature of 19.7°C.

Superimposed on these general hydrographic characteristics, coastal upwelling, river discharge in shelf waters, and meso-scale eddies in the open ocean are major additional sub-regional features in the study area. Intrusions of water from the Kuroshio Current, as well as the passage of typhoons, may also result in different vertical structures of hydrological properties. During the survey period (16 August–6 September 2008), the SST was generally high in most regions of the NSCS (>30°C), with the exception of two cold centers off the coast (Fig. 5a). One is located in the northeast part of the NSCS along the coast of Guangdong Province with the minimum value around 27.5°C, which is generally noted as the eastern Guangdong coastal upwelling (Wang et al., 2014). This upwelling event could also be detected in the SLA map, which displays significantly lower values in this region (Fig. 5b). The other cold center, which is relatively weaker (minimum SST = 29°C), appears to the northeast of Hainan Island, and is usually noted as the Hainan northeastern upwelling (Lin et al., 2016). Based on in situ data, obvious shoreward and upward tilting of the isotherms from the middle shelf to the inner shelf were observed in the eastern Guangdong coastal upwelling. The open ocean of the NSCS is mainly characterized by three anticyclonic eddies (ACE), detected by satellite sea level anomaly maps (SLA) (Fig. 5b). Two of them were located across the 18°N transect, with in situ sampling stations cutting through the centers of these ACEs. In the northeast part of SCS, there is another ACE across the Luzon Strait, with high SLA and high SST. Surface Chl exhibits obviously higher values along the coast of the NSCS (Fig. 5c), especially in the coastal upwelling regions and the Zhujiang River Estuary. In the offshore area, where water depth is greater than 120 m, Chl exhibits much lower values, with the spatial pattern largely matching that of SLA. Extremely low Chl values could be observed at the centers of identified ACEs. During the survey, the central part of the NSCS also exhibited high Chl as affected by the passage of typhoon Nuri (Ye et al., 2013).

Subsurface maximum layers of phytoplankton were generally observed in most offshore regions of the South China Sea. Based on the estimated profiles of Chl from middle shelf waters to open ocean, Z_SCM was found to vary from 25 m to 84 m, with the magnitude of SCM between 0.184 mg/m³ and 0.811 mg/m³. For most stations, variations in Z_SCM followed fluctuations of the isopycnal, σ_ɵ=23 kg/m³, and the isotherm, θ=23.38°C (Figs 9a, b). However, there are two abnormal stations that exhibit large deviations between Z_SCM and the isopycnal, σ_ɵ=23 kg/m³. One is Sta. 402 (21°N, 120°E), located to the southwest of Taiwan Island, near the Luzon Strait (Fig. 1), where Z_SCM and the depth of the isopycnal (σ_ɵ=23 kg/m³) are at 37 m and 119 m, respectively. As illustrated in the θ-S relationship, this area could be influenced by an intrusion of water from the Kuroshio Current through the Luzon Strait, as suggested by higher subsurface salinity and temperature in the upper water column. Another abnormal station is 609, located in the central part of NSCS, with Z_SCM and depth of the isopycnal (σ_ɵ=23 kg/m³) at 43 m and 83 m, respectively. This station was observed after the passage of typhoon Nuri (Ye et al., 2013). In both abnormal stations, the vertical hydrological properties showed unique structures with shallow density surface rising toward the surface, and the deep density surface dipping downward. These vertical structures are similar with the mode-eddy reported by McGillicuddy et al. (1999) and McGillicuddy (2016), which is composed of a lens-shaped disturbance that raises the seasonal pycnocline and lowers the main pycnocline.

Excluding these two anomalous stations, Z_SCM varies consistently with the depth of isopycnal (σ_ɵ=23 kg/m³) and the isotherm (θ=23.38°C), as shown in Figs 9a and b, with the determination coefficients (R²) for linear regression being 0.76 and 0.74, respectively. The magnitude of SCM (Chl_max) in the open ocean was much lower than that in the shelf waters. With the deepening of the SCML, Chl_max exhibits a decreasing trend (Fig. 9c). The thickness of the SCML ranges between 3.8 m and 54 m and exhibits a weak correlation with Z_SCM (R²=0.36, N=38). Z_SCM lies primarily between the mixed layer depth and Z_eu, and a linear relationship was observed between Z_SCM and Z_eu, with R²=0.76 (N=11).

Based on high-resolution in situ measurements, the column-integrated biomass over Z_eu (<Chla>_Zeu) could be quantitatively analyzed. The range of <Chla>_Zeu is wide, from 4.24 mg/m² in the very oligotrophic waters to 30.34 mg/m² at the coastal upwelling station to the east of Guangdong Province. Large scatters were observed for the relationship <Chla>_Zeu and Chl_surf, especially for open oceanic waters (Fig. 10a). With a deepening of the SCML, the <Chla>_Zeu tended to decease, exhibiting a much closer power law relationship with depth, and R² reached a maximum of 0.73 (Fig. 10b). A close relationship between <Chla>_Zeu and the ratio of Z_SCM and Z_eu (Z_SCM/Z_eu) was also observed, with an R² of 0.72 (Fig. 10c).

Bulk variabilities of absorption and attenuation coefficients with Chl and POC are consistent with previous studies in the SCS, which indicate the importance of phytoplankton and its accompanying materials (Wang et al., 2008; Cui et al., 2016). Physical processes play an important role in determining the spatial variations observed in our profiles of bio-optical properties in the NSCS. Coastal upwelling and river discharge are two important features of the inner and middle shelf waters of the NSCS, which supply nutrients that support phytoplankton growth by Ekman pumping and freshwater discharge, respectively. These results are consistent with previous studies in NSCS (Lu et al., 2010; Lin et al., 2014; Xu et al., 2018). Terrestrial material could also been exported from the Zhujiang River plume or the upwelling process which resulted higher attenuation coefficient (c_pg(650)) in the upper layer of inner shelf water (Cui et al., 2016). Conspicuous shoaling of the isotherms from the middle shelf to the inner shelf in coastal upwelling areas could induce resuspension of particles and the transport of detrital matter within the bottom boundary layer. This process would be consistent with the high attenuation coefficients observed below the SCML that increased rapidly towards the bottom, as well as the decreasing trend of Chl/c_pg(650) in the middle and outer shelf waters. Similar characteristics of attenuation coefficient profiles have been investigated in the bottom boundary layer of a continental shelf in the Mid-Atlantic Bight (Boss et al., 2001). These patterns observed in bio-optical properties are consistent with the summertime hydrography and nutrient dynamics of the NSCS as discussed by Wong et al. (2015). Vertical mixing within the NSCS shelf-sea area is the primary mechanism for bringing nutrients from the sub-surface of the adjoining ocean basin up to the mixed layer of the shelf, supporting primary production. Based on our survey data, coastal upwelling was found to be the primary mechanism contributing to vertical mixing in northeast and northwest part of the NSCS (Lin et al., 2014).

In the open ocean area, phytoplankton was the dominant factor influencing the vertical distribution of bio-optical properties (Figs 2 and 7d). The subsurface maximum of the a_pg(488) and c_pg(650) were corresponding to the subsurface maximum of Chl. Cui et al. (2016) has ever reported similar variability in the shelf and basin waters in the NSCS during summer. However, this is different from the vertical separation of the maxima of biomass (as reflected by c_p) and chlorophyll during stratified conditions observed in some oligotrophic system (e.g., Sta. ALOHA in the subtropical North Pacific Ocean), caused by algal photo-adaptation (Fennel and Boss, 2003 and references therein). The SCML is a nearly ubiquitous feature in stably stratified waters, and the depth of the SCML is found to be consistent with the top of the nitracline where phytoplankton growth is enhanced by the optimal combination of nutrient flux and irradiance (Cullen, 2015). In our study area, a combination of mesoscale eddies, intrusions of Kuroshio water, and occasional typhoons are responsible for the variability of the SCML in the open ocean observed during the survey period. A clear relationship was observed between the depth of the SCML (Z_SCM) and the fluctuation of thermocline depth as denoted by the isopycnal (σ_ɵ=23 kg/m³) and the isotherm at 23.38°C. This observation confirms the importance of physical forcing on regulating the SCML within the euphotic zone (Wong et al., 2015). These results were also consistent with those reported in previous studies of the NSCS, which indicated that the vertical distribution of nutrients generally follows isopycnal (and largely isothermal) surfaces in the upper water column (Chen et al., 2006; Li et al., 2016; Zhang et al., 2016). These effects could be easily explained based on the doming of the isopycnals and the nutricline, where nutrient pumping may have been caused by a cold eddy, or by the passage of typhoon Nuri (McGillicuddy et al., 1999; McGillicuddy, 2016; Ye et al., 2013; Xu et al., 2018). However, the relationship may not hold true for specific profiles with complex hydrological structure, for example, at Stas 402 and 609. A clear uplift of the SCM was observed at both stations, however, the isopycnal depth (σ_ɵ=23 kg/m³) relatively deeper than nearby stations. These phenomena might be related with the unique hydrological profiles caused by intrusion of Kuroshio waters or by the passage of typhoon Nuri, respectively. Shoaling density surfaces may lift nutrients into the euphotic zone, which are rapidly utilized by the biota, and the subsurface maximum layer of chlorophyll (SCML) could be lifted up (McGillicuddy et al., 1999; McGillicuddy, 2016). However, deepening density surface serve to push nutrient-depleted water out of the well-illuminated surface layers, which may result in the departures from general relationship between Z_SCM and the ispycnal at 23 kg/m³ (as well as the isotherm at 23.38°C). Driving mechanisms may be even more complicated in the case of anticyclonic eddies (ACEs) (Wang et al., 2018). Based on our observations, for those profiles located in the two ACEs across the 18°N transact, the SCML lies mainly between 64–84 m, with Z_SCM generally following the depth patterns of the isopycnals (Figs 9a, b). Since the optical instrument (ac9) was mainly deployed to depths <90 m, there were also some profiles where the SCML was too deep to be detected. Full-depth observations and additional studies are expected in future work.

Based on in situ measured data with high vertical resolution, a more accurate assessment of Chl integrated over the whole water column of the euphotic zone could be achieved to investigate ecological responses to different physical processes. Although a general trend could be detected from surface Chl concentrations, which showed that the total biomass within the euphotic zone decreases from the coastal ocean to open oceanic waters, it is difficult to estimate the total chlorophyll a inventory of the surface phytoplankton biomass for those oligotrophic waters in the open ocean of the SCS (Fig. 10a). This may introduce a large uncertainty in the estimation of primary production from ocean color, since only the pigment content within the upper penetration depth is accessible to remote sensing techniques. The relationship between <Chla>_Zeu and surface Chl is different from that developed for stratified waters in open ocean (Morel et al., 2001; Uitz et al., 2006). With deepening of the SCML, <Chla>_Zeu tends to decease, exhibiting a much closer power law relationship. This variability is consistent with that observed by Mignot et al. (2011), confirming that the upward flux of nutrients driven by physical processes contributes to phytoplankton growth within the euphotic zone. The magnitude (that is the intensity) of the SCM (Chl_max) is sensitive to changes in the depth of the SCML and tends to decrease with deepening of the SCML. Similar results have been reported in previous studies, where they were attributed to nitrate transport from nitrate-rich waters to the euphotic zone (Chen et al., 2006; Mignot et al., 2011; Gong et al., 2015).

Light attenuation could be another factor determining the vertical variability of phytoplankton. Previous studies have proposed that both the intensity and the location of the SCML is the result of a balance between the upward flux of nutrients and the downward attenuation of light (Fennel and Boss, 2003; Cullen, 2015). Considering the seasonal variability of the SCML, light attenuation could be a critical factor in regulating the depth of the SCML, as discussed by Gong et al. (2015) based on physical-biogeochemical modelling and by Mignot et al. (2011) based on in situ observed profiles. During our survey period, the euphotic zone depth (Z_eu) was generally deeper than the mixed layer depth. Light was therefore readily available and likely not a limiting factor of phytoplankton growth in the mixed layer. Z_SCM lies primarily between the mixed layer depth and Z_eu, and a close relationship between Z_SCM and Z_eu was observed with R²=0.76 (N=11). Due to the limit of in situ observations of underwater light field, the relationship between Z_SCM and Z_eu (Fig. 9d) could be improved in future. For waters with low Chl in the upper layer, the SCML deepens and becomes closer to the nitracline, which may reduce the phytoplankton biomass at the SCML. Light attenuation in the upper water column is the primary factor affecting vertical variations in phytoplankton growth rates (Mignot et al., 2011, 2014; Gong et al., 2014, 2015). The relative positions of the SCML and the euphotic zone depth may result in variability of total integrated phytoplankton biomass within Z_eu (Fig. 10c). Since the distribution of the SCML depth was primarily regulated by fluctuations in the thermocline, this relationship reflects the combined effect of physical forcing and light attenuation to phytoplankton growth in our study area.

Vertical profiles of Chl/c_pg(650) in the open ocean of the NSCS show similar patterns with vertical distributions of the Chl:C ratio observed in California Current pelagic ecosystem (Li et al., 2010). These patterns reveal the great advantage of optical measurements in quantifying phytoplankton dynamics in the ocean. Cui et al. (2016) also found similar correlations between c_p(660) and POC for data collected in the NSCS, with the exception of nearshore stations with high sediment resuspension. The Chl/c_pg(650) ratio is low at the sea surface as a result of the strong light intensity compared with deeper layers, and tends to increase with depth as a result of photoadaptation (Behrenfeld et al., 2005; Xing et al., 2014), exhibiting relatively constant values near the bottom of the euphotic zone. This is a common feature of open ocean waters characterized by strong stratification (e.g., Kitchen and Zaneveld, 1990; Fennel and Boss, 2003) and has been attributed to complex interactions between light, nutrients, temperature, and phytoplankton community structures (Li et al., 2010; Nencioli et al., 2010).