The inf luence of turbulent mixing on the subsurface chlorophyll maximum layer in the northern South China Sea*

Chenjing SHANG , Changrong LIANG , Guiying CHEN , Yongli GAO

1 Shenzhen Key Laboratory of Marine Bioresources and Eco-environmental Science, College of Life Science and Oceanography,Shenzhen University, Shenzhen 518060, China

2 State Key Laboratory of Tropical Oceanography, South China Sea Institute of Oceanology, Chinese Academy of Sciences,Guangzhou 510301, China

3 Southern Marine Science and Engineering Guangdong Laboratory (Guangzhou), Guangzhou 511458, China

4 Equipment Public Service Center, South China Sea Institute of Oceanology, Chinese Academy of Sciences, Guangzhou 510301,China

Abstract We present observations from deployments of turbulent microstructure instrument and CTD package in the northern South China Sea from April to May 2010. From them we determined the turbulent mixing (dissipation rate ε and diapycnal diff usivity κ), nutrients (phosphate, nitrate, and nitrite), nutrient f luxes, and chlorophyll a in two transects (A and B). Transect A was located in the region where turbulent mixing in the upper 100 m was weak ( κ～10-6-10-4 m 2/s). Transect B was located in the region where the turbulent mixing in the upper 100 m was strong ( κ～10-5-10-3 m 2/s) due to the inf luence of internal waves originating from the Luzon Strait and water intrusion from the Western Pacif ic. In both transects, there was a thin subsurface chlorophyll maximum layer (SCML) (>0.25 mg/m 3) nested in the upper 100 m. The observations indicate that the eff ects of turbulent mixing on the distributions of nutrients and chlorophyll a were diff erent in the two transects. In the transect A with weak turbulent mixing, nutrient f luxes induced by turbulent mixing transported nutrients to the SCML but not to the upper water. Nutrients were suffi cient to support a local SCML phytoplankton population and the SCML remained compact. In the transect B with strong turbulent mixing, nutrient f luxes induced by turbulent mixing transported nutrients not only to the SCML but also to the upper water, which scatters the nutrients in the water column and diff uses the SCML.

Keyword: turbulent mixing; diapycnal diff usivity; nutrients; nutrient f lux; chlorophyll a

1 INTRODUCTION

Subsurface chlorophyll maximum layers (SCMLs)are ubiquitous in the ocean, and they have a signif icant contribution to the water column biomass and primary production (Cullen, 2015). Depth, thickness, and intensity are the three main factors that characterize the SCML (Taguchi, 1980; Li et al., 2012; Gong et al.,2014). These factors are aff ected by hydrological dynamic including turbulent mixing, upwelling,mesoscale eddies and circulation (Vandevelde et al.,1987; Kononen et al., 1998; Huisman et al., 2006;Ledwell et al., 2008; Lu et al., 2010; Wang and Goodman, 2010; Williams et al., 2013a; Hu et al.,2014). Nutrients in the ocean are generally stored in deep layer while nutrients in the upper layer are scarce due to the consumption by phytoplankton. Nutrients need to be transported upward from deep layer to support the phytoplankton communities in the upper layer. In the absence of nutrients, the growth of phytoplankton will be inhibited and the chlorophylla(Chla) will decay. Hydrological processes play an important role in the nutrient supply.

Turbulent mixing in the ocean varies in time and space. In recent years, more and more studies focused on the inf luence of turbulent mixing on the nutrient supply of phytoplankton communities (MacIntyre and Jellison, 2001; Sharples et al., 2007; Hales et al.,2009; Schafstall et al., 2010; Tanaka et al., 2012;Tweddle et al., 2013; Williams et al., 2013b). For example, Hales et al. (2009) observed high vertical turbulent nutrient f luxes in the euphotic zone at the New England shelf break front. The average nitrate f luxes there were up to 6×10-5mmol/(m2·s). Schafstall et al. (2010) reported the tidal-induced mixing and diapycnal nutrient f luxes in the Mauritanian upwelling region. Nitrate f luxes at the base of the mixed layer over the continental slope reached a mean value of 12×10-5mmol/(m2·s). A study from Tanaka et al.(2012) revealed that a high Chl-aregion along the shelf break in the south eastern Bering Sea was sustained by vertical turbulent f luxes. An observation from Wang and Goodman (2010) indicated that the thickness and intensity of the SCML in Monterey Bay were modulated by turbulent mixing. These studies indicated that nutrient f lux induced by turbulent mixing is an important dynamic factor for redistributing nutrients and supporting primary productivity.

The northern South China Sea (SCS) is characterized by unevenly distributed turbulent mixing which is related to internal waves (St. Laurent,2008; Tian et al., 2009; Liu and Lozovatsky, 2012;Yang et al., 2014; Shang et al., 2017). Numbers of internal waves generated in the Luzon Strait (Fig.1)propagate westward to the northern SCS and undergo nonlinear interactions during the propagation,providing a large amount of energy for the turbulent mixing and creating an uneven distribution of turbulent mixing in the northern SCS (Zhao et al.,2004; Zhao and Alford, 2006; Alford et al., 2015; Xie et al., 2018). An observation reported by St. Laurent(2008) indicated that the column integrated dissipation levels in the shelf-break region reached 50 mW/m2,an order of magnitude larger than levels typical in the open ocean. Study from Liu and Lozovatsky (2012)showed that the level of averaged pycnocline dissipation to the north of 20°N was two times larger than that to the south of 20°N. Measurements conducted by Shang et al. (2017) indicated that strong turbulent mixing was mainly limited in the region close to the Luzon Strait. The averaged thermocline dissipation rate and diapycnal diff usivity in the region close to the Luzon Strait (the east of 115.5°E) were almost one order of magnitude higher than those in the region far away from the Luzon Strait (the west of 115.5°E).Chlais also unevenly distributed in the northern SCS. Sea surface Chladuring May 2010 was high in the region (the east of 115.5°E) close to the Luzon Strait compared to the region (the west of 115.5°E)far away from the Luzon Strait (Fig.1). The sea surface Chl-adata are monthly MODIS-Aqua data obtained from ERDDAP (https://coastwatch.pfeg.noaa.gov/erddap/index.html). Many studies have reported the nutrient supply and distribution of phytoplankton in the SCS (Liu et al., 2002; Gan et al.,2010; Han et al., 2013; Wang and Tang, 2014; Li et al., 2016). However, these studies mainly focused on the eff ects of upwelling and coastal currents on the nutrient supply and the distribution of phytoplankton.The inf luence of the unevenly distributed turbulent mixing on the nutrient supply and the distribution of phytoplankton is not fully understood. It is unclear whether and how the turbulent mixing contributes to the uneven distribution of the sea surface Chla. In this study, microscale shear, conductivitytemperature-depth (CTD), f inescale velocity, Chlaand nutrient data obtained from two transects in the northern SCS are used to investigated the eff ects of turbulent mixing on the vertical distributions of nutrients and Chla. This study is primarily the biophysics of the northern SCS. We investigate the distribution of turbulent mixing using microscale shear data and explore the dynamic mechanism of turbulent mixing using CTD and f inescale velocity data. The eff ects of turbulent mixing on the vertical distribution of nutrients and Chlaare studied with nutrient f luxes.

2 MATERIAL AND METHOD

Fig.1 Spatial distribution of sea surface Chl a with stations (circles)

Fig.2 Time series of barotropic tidal velocity predicted from TPXO (a); wind speed at a height of 10 m during the observation at each station (b)

Physical and biogeochemical measurements were conducted from April to May 2010. The locations of the stations are shown in Fig.1. The stations were divided into two transects (A and B). Transect A was located in the region far away from the Luzon Strait and was conducted from 22 to 23 May, 2010 (Fig.2a,during neap tide). Transect B was located in the region close to the Luzon Strait and was conducted from 26 to 28 April, 2010 (Fig.2a, during neap tide). The barotropic tide data in Fig.2a were obtained from the global inverse tide model (TPXO) (Egbert and Erofeeva, 2002) at 20°N/116°E. The bias in the arrival of spring-neap tides in diff erent stations was shorter than 2 h. The observation of the two transects was conducted in sunny days. The wind speed at a height of 10 m was smaller than 8 m/s at each station(Fig.2b). The wind speed data come from the European Centre for Medium-Range Weather Forecasts (http://apps.ecmwf.int/datasets/data/interim-full-daily/levtype=sfc/). From April to May 2010, there was no high wind event occurring in the observed region and no tropical cyclone passing through the SCS (http://112.124.12.97/publictyphoon/). Transect A includes six stations(A1-A6) and transect B includes nine stations (B1-B9). One CTD cast was made at each station to collect hydrological and nutrient data. Temperature and salinity data were documented with the Sea-Bird Electronic 911 Plus (calibrated in January 2010).Water samples were collected with Niskin bottles from diff erent depths for nutrient extraction. The extraction method has been described by Hu et al.(2014). Sea-water from each depth was pre-f iltered through a Whatman GF/F and decanted into a 100-mL polycarbonate bottle, frozen immediately and stored at -20 °C prior to analysis in the laboratory. According to the standard colorimetric techniques (Kirkwood et al., 1996), the concentrations of nitrate (NO3), nitrite(NO2), and phosphate (PO4) were analyzed with a f low injection analyzer (Quickchem 8500, Lachat Instruments, USA). Continuous time series of f inescale velocity at 5-min intervals and 16-m vertical spacing between 38 and 982 m were obtained from a shipboard 38-kHz acoustic Doppler current prof iler.

At all stations except station A4, microstructure data were collected with the Turbulence Ocean Microstructure Acquisition Prof iler (TurboMAP)(Wolk et al., 2002). At each station, one microstructure prof ile was conducted right after the CTD cast.TurboMAP is a quasi-free-falling instrument equipped with microstructure shear sensor, temperature sensor,f luorescence sensor, pressure sensor, and turbidity sensor. The parameters collected by TurboMAP include turbulent parameters (microscale shear), biooptical parameters (f luorescence), and hydrographic parameters. The sinking rate of the prof iler was 0.5-0.7 m/s. The Chl-aconcentration from the f luorescence sensor of TurboMap was calibrated by the bottle sampling. The dissipation rate (ε) was estimated with the observed microscale shear (∂u/∂z) using the following isotropic formula:

wherevis the kinematic viscosity, 〈 〉 denotes the spatial average, andΨ(k) is the microscale shear spectrum.k1andk2are the integration limits. The lower integration limitk1is set to 1 cpm and the upper limitk2is the highest wavenumber that is not contaminated by vibration noise. An integrated software application TMToolsTMdeveloped by Alec Electronics Co., Ltd. was used to derive the dissipation rate.

Examples of microscale shear and their corresponding dissipation spectra are shown in Fig.3.The dissipation spectra are approximately consistent with Nasmyth’s spectra (Nasmyth, 1970) within the integration range (between the two dashed vertical lines). Distinct peaks associated with high wavenumbers (beyond the upper integration limit)were caused by instrument vibrations. The weak microscale shears at depths of 65 m and 71 m correspond to weak dissipations (ε～10-9W/kg), and strong microscale shears at depths of 67 m and 69 m correspond to strong dissipations (ε～10-8W/kg).Diapycnal diff usivity (κ) was calculated based on the dissipation rate and stratif ication (Osborn, 1980):

3 RESULT

3.1 Hydrographic condition

Fig.3 Examples of microscale shear (a) at specif ied depth segments (z) and corresponding dissipation spectra (b-e) from station B6 (collected on April 27, 2010)

Fig.4 Relationship of potential temperature (θ0) versus salinity with potential density (unit in kg/m 3)contours overlaid

Intrusion of water from the Western Pacif ic can inf luence the water properties of the SCS.Measurements and models (Shaw, 1991; Wu and Hsin, 2012) have conf irmed that there is a strong intrusion of water from the Western Pacif ic into the SCS through the Luzon Strait. The potential temperature versus salinity curves of the two transects and the Western Pacif ic are given in Fig.4. Data of the Western Pacif ic (19.5°N-22.0°N, 121.5°E-123.5° E)were obtained from the World Ocean Database 2013.The potential temperature versus salinity curve of the Western Pacif ic shows a reversed S shape with one salinity minimum and one salinity maximum. The maximum salinity layer (at potential density between 22.5 and 25.5 kg/m3) corresponds to the high-salinity North Pacif ic Tropical Water (NPTW), and the minimumsalinitylayer(atpotentialdensitybetween 25.5 and27.5 kg/m3)corresponds tothelow-salinity North Pacif ic Intermediate Water (NPIW) (Qu et al.,2000). The NPTW mainly occupies the water column in the upper 200 m, and the NPIW mainly occupies the water column below. The salinity of transect B was close to the NPTW value in the maximum salinity layer. However, the salinity of transect A was signif icantly smaller than that of the NPTW. These observations indicate that the inf luence of the NPTW intrusion on the water properties of transect B was stronger than that of transect A. A reversed trend was found in the minimum salinity layer. The minimum salinity in the Western Pacif ic is smaller than that of the two transects. Small salinity diff erence between transects A and B and large salinity diff erence between the Western Pacif ic and the two transects suggest that the inf luence of the NPIW intrusion on the water properties was weak in both transects.

Fig.5 Left panel: distributions of temperature (a), salinity (b), squared buoyancy frequency (c), and f inescale shear variance(d) for transect A; right panel: the same as left but for transect B

In addition to water intrusion, the SCS is also characterized by energetic internal waves. These waves originate from the Luzon Strait and have a strong impact on the velocity and temperature f ields of the SCS (Zhao et al., 2004; Zhao, 2014; Alford et al., 2015). Transect A was located in the region where internal wave activities are weak while transect B was located in the region where numbers of internal waves pass (Zhao et al., 2004, cf. their Fig.1). To further investigate the hydrologic condition of these two transects, we show the distributions of temperature and salinity in Fig.5. The temperature in transect A(Fig.5a) shows a rapid temperature change with increasing depth in the upper 50 m while the temperature in transect B (Fig.5e) remains uniform in the upper 50 m. Similar patterns are observed in salinity. A rapid salinity change is found in the upper 50 m of transect A (Fig.5b) while the salinity in the upper 50 m of transect B (Fig.5f) remains relatively uniform. In the upper 50 m, the temperature of transect A was higher than that of transect B but opposite in the salinity. Water intrusion contributes to the diff erence in hydrological conditions between the two transects. High-salinity NPTW intruded into the SCS through the Luzon Strait and was mixed with the local water of transect B, which resulted in the high salinity of transect B in the upper layer (Qu et al., 2000).

Using the temperature and salinity data, we estimated the stratif ication which is shown in Fig.5c& g. A comparison of these two transects shows that the surface mixed layer in transect A (<10 m) was thinner than that in transect B (～45 m). Below the surface mixed layer is a thermocline with strong stratif ication. Here, we roughly def ine the top of the thermocline (namely the bottom of the surface mixed layer) as the depth at whichN2=1×10-4/s2and the bottom of the thermocline as the depth at whichN2=2×10-4/s2. The thermocline of transect A was mainly limited in the upper 100 m while the thermocline of transect B was found at a depth between 45 and 125 m. The thermocline stratif ication of transect A was stronger than that of transect B.Stratif ication of transect A between 15 and 35 m reached 7×10-4/s2. The deep surface mixed layer and weak thermocline stratif ication of transect B might be caused by the NPTW intrusion and internal waves.Intrusion waters could change the salinity f ield by mixing with the local waters and internal waves could enhance the turbulent mixing among the waters,which weakens the stratif ication of transect B. Figure 5d & h show the distribution of f inescale shear variance for transects A and B, respectively. The f inescale shear variance of transect B was stronger than that of transect A, especially those at depth of 50-150 m where the levels ofS2were two to three times higher than that of transect A.

Fig.6 Left panel: distributions of ε (a) and κ (b) for transect A; right panel: the same as left but for transect B

3.2 Distributions of ε and κ

Transect A had weak f inescale shear but strong stratif ication, while transect B had strong f inescale shear but weak stratif ication. Finescale shear and stratif ication are important factors aff ecting turbulent mixing in the ocean (MacKinnon and Gregg, 2003,2005; Shang et al., 2017; Liang et al., 2019). To investigate the eff ects of f inescale shear and stratif ication on the turbulent mixing, we show the distributions ofεandκin Fig.6. In both transects, the upper 20 m was occupied by strong dissipations with values ofεreachingO(10-8) W/kg (Fig.6a & c).However, dissipations of transect B were stronger than that of transect A below 20 m. The averageεbelow 20 m of transect B was 1.92×10-8W/kg, which is three times larger than that of transect A. Strong dissipations of transect B might be caused by internal waves generated in the Luzon Strait (Liu and Lozovatsky, 2012; Alford et al., 2015; Shang et al.,2017). The diapycnal diff usivity shows diff erent distributions in transects A and B (Fig.6b & d).Diapycnal diff usivity of transect A has a clear hierarchical structure. A weak diapycnal diff usivity layer withκof 10-7-10-6m2/s occupies the water column between ～20 and 50 m. This weak diapycnal diff usivity layer was mainly due to the strong stratif ication between ～20 and 50 m (Fig.5c). Strong stratif ication can suppress shear instability and weaken the diapycnal mixing (Polzin et al., 1996;MacKinnon and Gregg, 2005; Liang et al., 2019).Below the weak diapycnal diff usivity layer is a slightly enhanced diapycnal diff usivity layer,occupying the water column between ～50 and 100 m.Values ofκin this layer were 10-6-10-5m2/s, almost one order of magnitude larger than that of the upper layer. Diapycnal mixing below 100 m was weak(κ～10-7-10-6m2/s). There is no hierarchical structure in the diapycnal diff usivity of transect B. Strong diapycnal mixing almost occupied the upper 100 m and values ofκwere one to three orders of magnitude larger than that of transect A. Strong turbulent mixing is generally related to the shear instability of internal waves that depends on the f inescale shear and stratif ication (Polzin, 1996). Transect A had strong stratif ication but weak f inescale shear (Fig.5c & d),while transect B had weak stratif ication but strong f inescale shear (Fig.5g & h), which indicates that the water body in transect B is more prone to shear instability than that of transect A.

Fig.7 Left panel: distributions of Chl- a concentration (a), NO 2+NO 3 concentration (b), and PO 4 concentration (c) for transect A; right panel: the same as in the left panel but for transect B

3.3 Distributions of Chl a and nutrient concentrations

Figure 7a & d shows the distribution of Chl-aconcentration for transects A and B, respectively. The distribution of Chl-aconcentration shows a sandwich structure in both transects. A low Chl-aconcentration layer with a concentration lower than 0.25 mg/m3occupied the upper ～50 m. A high Chl-aconcentration layer with a concentration higher than 0.25 mg/m3was nested in the water column between ～50 and 100 m. This layer is known as the SCML. Here, we def ine the boundaries of the SCML as the depths at which the Chl-aconcentration is equal to 0.25 mg/m3.Below the SCML is another low Chl-aconcentration layer with a concentration lower than 0.25 mg/m3.The SCML features of transect B are diff erent from that of transect A. On the continental shelf(0 km170 km), the maximum Chl-aconcentration in transect A was higher than that of transect B and the SCML of transect A was more compact than that of transect B. In addition, the surface Chl-aconcentration in transect B was higher than that in transect A.

Figure 7b-c and 7e-f shows the distributions of the nitrate and nitrite (NO2+NO3) concentration and phosphate (PO4) concentration for transects A and B.In transect A, nitrate and nitrite are evenly distributed in the horizontal direction, and there are obvious nutricline in the vertical direction. The (NO2+NO3)concentration was lower than 2.5 mmol/m3in the upper 50 m but higher than 12.5 mmol/m3below 100 m. The layer at depth between 50 and 100 m was a nutricline. The (NO2+NO3) concentration in the nutricline increased rapidly with increasing depth,from ～2.5 mmol/m3at 50 m to ～12.5 mmol/m3at 100 m. The nutricline almost coincides with the SCML. A diff erent pattern is found in the transect B.The distribution of nitrate and nitrite was scattered.No nutricline was found in this transect. Water column above 75 m was occupied by nitrate and nitrite with concentration lower than 7.5 mmol/m3, and water column below 75 m was occupied by nitrate and nitrite with concentration higher than 7.5 mmol/m3.Overall, transect B had more nitrate and nitrite than transect A above 75 m but less nitrate and nitrite than transect A below 75 m. Similarly, a clear nutricline was found in the phosphate distribution of transect A(Fig.7c), but no nutricline was found in that of transect B (Fig.7f).

4 DISCUSSION AND CONCLUSION

4.1 Inf luence of turbulent mixing on the nutrient and Chl- a distributions

Both transects have a high Chl-aconcentration layer nested in the water column between ～50 and 100 m. However, the SCML of transect A was more compact than that of transect B. In addition, the nutrient distributions (NO2+NO3) and (PO4) of transect B were more scattered than that of transect A.Turbulent mixing plays an important role in redistributing nutrients and microorganisms in the ocean (Hales et al., 2009; Schafstall et al., 2010;Wang and Goodman, 2010; Tanaka et al., 2012). To investigate the impact of turbulent mixing on the distribution of nutrients, we estimate the nutrient f lux induced by turbulent mixing, which is calculated as(Schafstall et al., 2010):

where - dC/ dzis the vertical gradient of the dissolved nutrient concentration in the sample (positive upward). To calculate the nutrient f lux, the nutrient concentration was f irst interpolated onto the diapycnal diff usivity grid. For simplicity, we designate- d(NO2+NO3)/ dzasNz, - d PO4/ dzasPz, the NO2+NO3f lux asΦN, and the PO4f lux asΦP. Note thatΦNandΦPare absolute values of the calculation.

Nutrient f luxes in transect A show a multi-layer structure (Fig.8a-b). The water column below ～100 m was occupied by weak nutrient f luxes (ΦN～10-8mmol/(m2·s) andΦP～10-9mmol/(m2·s)). Although this layer was rich in nutrients, the nutrient f luxes were weak.This is mainly due to the small vertical nutrient gradient (Fig.8c-d). There is no phytoplankton in this layer, therefore the nutrients in which cannot be consumed. A long-term eff ect of nutrient transport results in the evenly distributed nutrients. Above the weak nutrient f lux layer, a slightly enhanced nutrient f lux layer (ΦN～10-7-10-6mmol/(m2·s) andΦP～10-8-10-7mmol/(m2·s)) exists, occupying the water column between ～50 and 100 m. This layer coincides with the SCML and the nutricline (Fig.7b-c). Both the large vertical nutrient gradient (Fig.8c-d) and strong turbulent mixing (Fig.6b) contributed to the strong nutrient f luxes. In this layer, phytoplankton consumed nutrients, which formed a large vertical nutrient gradient. Strong nutrient f luxes indicate that nutrients were transported to the SCML from deep layer. The nutrient transport was suffi cient to support a local SCML phytoplankton population and kept the SCML compact along the transect (Fig.7a). However, the turbulent mixing was not strong enough to transport nutrients to the layer above the SCML. As one can see from Fig.7b-c that water near the top of SCML was occupied by low (NO2+NO3) and (PO4) concentrations.Another weak nutrient f lux layer (ΦN～10-8mmol/(m2·s) andΦP～10-9mmol/(m2·s)) occupied the water column between ～20 and 50 m. Weak nutrient f luxes in this layer were due to the lack of nutrients and weak turbulent mixing. The (NO2+NO3) and (PO4)concentrations were almost zero (Fig.7b-c) and values ofκwere smaller than 10-6m2/s (Fig.6b) in this layer. Weak nutrient f luxes indicate that few nutrients were transported to the layer above the SCML, which contributes to the shortage of nutrients in the layer above the SCML (Fig.7b-c). Without adequate nutrient supply, the growth of phytoplankton was inhibited. It can be seen from Fig.7a that the Chl-aconcentration in the layer above the SCML was almost zero. A low surface Chl-aconcentration in transect A was also evident in the spatial distribution of the sea surface Chlafrom satellites (Fig.1). A diff erent distribution of nutrient f luxes is found in transect B. The upper 100 m was occupied by strong nutrient f luxes and there was no multi-layer structure(Fig.8e-f). Values ofΦNandΦPwere one to three orders of magnitude larger than that of transect A.Strong nutrient f luxes were mainly due to the strong turbulent mixing, as evidenced by the observation that most values ofNzandPzwere smaller than 0.15 mmol/m3(Fig.8g-h), while values ofκcould beO(10-4) m2/s (Fig.6d). Strong nutrient f luxes transported nutrients not only to the SCML but also to the layer above the SCML. It can be seen from Fig.7 that nutrients in the upper 100 m of transect B were distributed more evenly than that of transect A, and no clear nutricline was found. The evenly distributed nutrients could aff ect the distribution of Chla. On the continental shelf (0 km

Fig.8 Left panel: distributions of nitrate and nitrite f lux ( Φ N) (a); phosphate f lux ( Φ P) (b); vertical gradient of nitrate and nitrite concentration ( N z) (c); and vertical gradient of phosphate concentration ( P z) (d) for transect A; right panel: the same as the left panel but for transect B

Fig.9 Spatial distributions of curl-driven upwelling velocity (color) and wind stress (vectors) with stations (circles)

Strong turbulent mixing close to the Luzon Strait is generally related to the internal waves originating from the Luzon Strait (St. Laurent, 2008; Tian et al.,2009; Liu and Lozovatsky, 2012; Yang et al., 2014;Shang et al., 2017). However, internal waves generated in the Luzon Strait are periodic (Ramp et al., 2004; Alford et al., 2015), especially internal waves with tidal frequencies. Internal waves with tidal frequencies show strong spring-neap tidal cycle(Fig.2a). Transect A was conducted almost one month after the observation of transect B. The eff ects of spring-neap tidal cycle and wind condition need to be taken into account. Both transects were conducted during neap tide and under weak wind condition(Fig.2). There was no tropical cyclone passing through the SCS from April to May 2010 (http://112.124.12.97/publictyphoon/) and no high wind events before the observation of transect B. Therefore, the diff erence of turbulent mixing between the two transects was not caused by spring-neap tides or winds. In addition to the spring-neap tidal cycle, internal waves also show other periodic variations. Yet to investigate the eff ects of the time variability of internal waves on the turbulent mixing and the distribution of nutrient and Chla, long-term observational data (e.g., simultaneous observations of current, turbulent mixing, nutrients and Chlafor weeks in the upper 200 m at one location) are needed in the near future.

4.2 Inf luence of advection on the nutrient and Chla distributions

In addition to turbulent mixing, upwelling is another factor aff ecting the distributions of nutrients and Chla(Li et al., 2016). Unlike the turbulent mixing, the upwelling transports nutrients upward through advection,W×C, whereWis the vertical velocity (positive for upwelling) andCis the dissolved nutrient concentration in the sample. Spatial distributions of curl-driven upwelling velocity and wind stress during the observation period are shown in Fig.9. The upwelling velocity and the wind stress are from 3-day mean METOP-ASCAT data (https://coastwatch.pfeg.noaa.gov/erddap/index.html). These data are averaged corresponding to the transect time periods. During the observation of transect A, the wind direction was generally south on the west of the transect and northeast on the east of the transect.There was strong curl-driven upwelling in this transect. The upwelling at stations A3-A5 could be larger than 10-5m/s. The eff ects of the strong upwelling on the Chlaand nutrient distributions can be observed in Fig.7a-c. Both the SCML and nutricline were lifted up by the upwelling and the biggest uplift occurred at stations A3-A5 where the upwelling velocity was strongest. Evidence of uplift induced by upwelling was also found in the distributions of temperature and salinity (Fig.5a-b). Both the isotherm and isohaline were lifted up by upwelling at a distance between 100 and 300 km. These observations suggest that the upwelling mainly aff ects the large-scale distribution of nutrients and Chlarather than the f ine structure.

The wind direction was generally east during the observation of transect B and the velocity f ield was predominantly dominated by small downwelling. The eff ect of downwelling on the distributions of nutrients and Chlawas weak. There is no good correlation between the downwelling and the variations of the SCML and nutricline (Fig.7d-f), which suggests that the scattered distribution of nutrients and Chlain transect B was not due to the upwelling or downwelling. In addition to the vertical advection induced by vertical velocity, horizontal advection induced by horizontal velocity might also aff ect the distributions of nutrients and Chla. Unfortunately,our data are insuffi cient to investigate the eff ect of horizontal advection on the distributions of nutrients and Chla. To complete the investigation, more data or numerical simulations are needed in the near future.

5 DATA AVAILABILITY STATEMENT

The research data are available at Zenodo (http://doi.org/10.5281/zenodo.3864885).

6 ACKNOWLEDGMENT

We thank all the crew of the survey ship from the South China Sea Institute of Oceanology, Chinese Academy of Sciences. We are very grateful to Professor Yehui TAN for her advice and nutrient data,and to the High Performance Computing Division and HPC managers of Wei ZHOU and Dandan SUI from the South China Sea Institute of Oceanology for their help in the preparation of the paper.