Elevated particulate organic carbon export flux induced by internal waves in the oligotrophic northern South China Sea

To understand the biogeochemical response to internal waves in the deep basin of the northern South China Sea (NSCS), particulate organic carbon (POC) export fluxes were quantified for the first time during the passage of large internal waves using drifting sediment traps attached with hydrographic sensors. Results revealed large variations in temperature, nitrate and chlorophyll a (Chl a) concentrations during and after internal waves, suggesting that cold nutrient-replete waters may be brought to the euphotic zone in the dissipation zone during and after the passage of internal wave packets, resulted in phytoplankton flourished. Most importantly, POC export fluxes (110.9 ± 10.7 mg C m−2 d−1) were significantly enhanced after internal waves compared to non-internal wave area (32.6–73.0 mg C m−2 d−1) in the NSCS. Such elevated POC fluxes may be induced by downward flourished biogenic particles, particle aggregation or converged particles from mixed layer triggered by internal waves.

temperature, salinity, and Chl a concentration were observed at stations K1 and K2 (Fig. 2). For example, at station K1, when the CTD sensors were brought to 146 m from 160 m, water temperature increased to 21.0 °C from 19.7 °C during 5~7 am on 6 th August (Fig. 2a). Such a short-period (2 hours) variation of temperature, salinity and sensors (including CTD sensors and the frame, ~30 kg) vertical displacement reveals that sediment traps and hydrographic sensors experienced internal waves at station K1. After that, a remarkable warm water signal (21~22 °C) was observed at depth of 160 m during 10~12 am on 6 th August, suggesting warm water might be carried down to deep depths. Following the warm shock, the CTD sensors were up and down from 160 to 130 m, and measured large variations of temperature (ranging from ~18 to 22 °C, see Fig. 2a) during 1~6 pm at station K1, suggesting that high turbulent mixing occurred and influenced biological activity. The repeated high vibration of changes in temperature, salinity and sensors vertical displacement (i.e. CTD depth in Fig. 2) during such a short period within 5 hours evidenced that nonlinear internal waves induced turbulent processes at station K1. Namely, the observed hydrographic changes at station K1 should experience the dissipation zone affected by internal waves.
Reversely, hydrographic settings (variations of temperature, salinity, and depth) at station K2 only showed short timescale variations excluding a strong internal wave event occurred at 5:50 am, suggesting that internal waves indeed affected station K2. However, the internal wave event observed at station K2 near Dongsha Atoll was likely a non-linear internal wave because its turbulent processes showed a large pulse of variations in temperature and salinity (Fig. 2, see the time series between the vertical dashed lines) following by immediately decaying within 1 hour to 2 hours. Clearly, the hydrographic settings varied largely at station K2, but they quickly restored to their normal conditions, i.e. likely belonging to the transmission zone affected by internal waves.
Similar rapid temperature increase and decrease events in the NSCS were reported by other researchers 14,23 , which can be explained by upward or downward pumping of subsurface layer water. Furthermore, the images of Synthetic-aperture radar (SAR) and Terra MODIS can be used to verify internal wave packets passing stations K1 and K2 7 ( Fig. 1 and Fig. 3). Figure 1 shows the stations K1 and K2 are in the internal-wave active zones observed by SAR. Figure 3 displays a snapshot of Terra MODIS images that observes the internal waves on August 4, 2016, during the experiment at the station K2. Overall, the time series of temperature, salinity, and Chl a provide a record of internal waves affecting hydrographic settings in water column at our study area, but the magnitude of such events (and their impact on marine organisms) is difficult to be quantified.
Changes in water column properties after internal waves. Vertical profiles of temperature, nitrate, Chl a, and POC concentrations at stations K1 and K2 show conspicuous changes in the euphotic zone ( Fig. 4) after the passage of internal waves. Surface nitrate concentrations during or after the internal waves periods were below the detection limit (<0.1 μmol L −1 ) (Fig. 4). However, nitrate inventories integrating from surface to the subsurface Chl a maximum depth (0-75 m) were 18.34 and 46.78 mmol m −2 (Table 1 and Fig. 4) during and after internal waves at station K1, respectively. Similar case was observed at station K2 with nitrate inventories of 64.97 and 72.60 mmol m −2 during and after internal waves, suggesting internal waves may affect nutrient supply in the euphotic zone.
Integrated Chl a inventories at stations K1 (34.9 vs. 69.9 mg m −2 ) and K2 (32.4 vs. 41.5 mg m −2 ) also showed significant difference during and after internal waves (Table 1 and Fig. 4c,g). As mentioned above, the elevated Chl a at station K1 might be due to the enhanced vertical mixing by internal waves, and transferring cold and nutrient-replete subsurface water to the surface euphotic zone to support biological activity. The enhanced Chl a inventory after internal waves may be related to nutrient supply in the euphotic zone. However, it is difficult to explain that phytoplankton can flourish double during a short time period at station K1 (<12 hours). One possibility is that there are many internal waves passing the study area according to SAR images (Fig. 1), and the enhanced Chl a may be supported by the enhancement of vertical mixing via earlier internal waves which bring cold and nutrient-replete water into the euphotic zone before the deployment at station K1 on August 6 ( Fig. 4), evidenced by the Terra MODIS image that shows internal waves at station K1 on August 4 (Fig. 3). Namely, the enhanced phytoplankton biomass affected by the internal wave packets should have a time lag or due to shipboard location shift. Moreover, injected CDOM from the deep water to the euphotic zone may cause partial contribution to Chl a concentration 16 , but it is difficult to estimate CDOM's influence on Chl a in this study. Alternately, phytoplankton transported by internal waves from deep depths or other areas may partly explain Chl a inventory changes. Furthermore, the vertical variations of Chl a at stations K1 and K2 may be partly caused by the photo-adaptation of phytoplankton assemblages when they experienced varied irradiance at different depths 24,25 . Enhanced POC flux and potential mechanisms. POC flux at station K1 was 110.9 ± 10.7 mg m −2 d −1 (Table 1) which is much larger than previously reported POC flux (32.6-73.0 mg m −2 d −1 ) in the NSCS without influence of internal waves and extreme atmospheric events [26][27][28] (Table 1). The measured POC flux this study  is based on the same method as previous reports of Huang 27 and Hung et al. 26,29 , thus the possible bias in POC export method to our results should be insignificant. Here, we propose several mechanisms to explain elevated POC flux at station K1. First, in longer time scale, POC export flux was regulated partly by POC stock in euphotic zone 30 , which was strongly related to Chl a concentration as they are both originated from biological production. We diagnosed if POC export was affected by biogenic phytoplankton biomass (e.g. Chl a values). In the time series record (Fig. 2a), Chl a concentration increased from <0.10 to 0.25 μg L −1 at ~160 m during warm shock period from 10~12 am on 6 th August. Areal Chl a stock at K1-2 (69.91 mg m −2 ) after internal waves is two folds of that at K1-1 (34.86 mg m −2 ). As mentioned above, flourished phytoplankton in water column might result from nutrient supply from subsurface, CDOM effect or phytoplankton shift from other areas or depths. According to our measured Chl a values by a fluormeter after extraction with acetone, they matched well with CTD-fluorescence-derived Chl a values (Fig. 4c,g) so that the effect of CDOM may not be important here. Instead, nutrient-stimulated phytoplankton biomass (i.e. POC) increase may likely play a role contributing some of POC flux. This can be interpreted as follows: internal waves enhance vertical mixing that entrains cold and nutrient-replete water to the upper euphotic zone in the dissipation zone, stimulates phytoplankton growth (the amount of nutrient will be estimated in later section), and then fuels zooplankton and biological activity.
Secondly, one may note that the water in the subsurface layer (80-150 m) at station K1-2 was much warmer than K1-1 (Fig. 4a), consistent with time series temperature graph in which warmer water was carried into deep depths during 10~12 am. As a consequence, phytoplankton may be killed by sudden temperature change either cold shock or warm shock. According to the investigations, phytoplankton cannot survive after sudden temperature change 31 . Therefore, these dead phytoplankton may enhance POC flux partly.
Moreover, to better understand how much nutrient supply influenced by internal waves, we simply use a diffusion model to estimate upward nutrient flux in the NSCS. Generally, moderate POC flux of 32.6-73.0 mg m −2 d −1 are reported in the oligotrophic water of NSCS in summer (June and September) under non-typhoon conditions [26][27][28]30 , it means it needs new nitrogen (in dissolved inorganic form) of 298-912 μmol N m −2 d −1 if a C/N ratio of 6.6 is applied 21 . The sources of dissolved inorganic nitrogen in the water column could come from atmospheric (including dry and wet) deposition, phytoplankton fixation, and diffusion from subsurface. The amount of atmospheric deposition and phytoplankton fixation approximately account for 56~204 μmol m −2 d −1 (Table 2), on average 130 μmol m −2 d −2 , which is much lower than new nitrogen value needed (298-912 μmol m −2 d −1 ) ( Table 2). If we use nitrate gradient of 0.06 mmol m −4 below mixed layer 27 , diffusivity rate of 4.9 × 10 −4 m 2 s −1 during internal waves condition 11 , the upward diffusion flux could be 2540 (0.06 mmol m −4 × 4.9 × 10 −4 m 2 s −2 × 86400 × 1000) μmol N m −2 d −1 triggered by internal waves. Using this estimated flux we next assume that the effective diffusion time is 6 hours (Fig. 2a) during the dissipation zone affected by internal waves, the daily nitrate diffusion flux will be 635 μmol N m −2 d −1 (Table 2). Nitrate diffusive flux triggered by internal waves is much larger than the maximum nitrogen fixation 32,33 , and atmospheric deposition reported [34][35][36] (Table 2), supporting a new production of 50.3 mg C m −2 d −1 . Overall, our filed observation shows that enhanced vertical mixing induced by internal waves transport considerable nutrients into the upper euphotic zone, and stimulate phytoplankton growth and other biological activity. But POC export flux conveyed by sinking particles, collected by sediment traps, may have a time lag from phytoplankton peak since it is difficult to understand how many internal waves passed the study area before our field cruise.
Finally, the POC flux of station K2 was 54.3 ± 7.7 mg m −2 d −1 which is much less than the flux of station K1 but is close to POC flux (32.6-73.0 mg m −2 d −1 ) at stations H1, H2 and H3 during non-typhoon conditions in summer. We also observed strong influence of internal waves which transport POC downward at station K2 (Fig. 4h). There were also Chl a increase in K2-2 (Fig. 4g), indicated possibly biological production. Less POC export at station K2 may be explained by the sake of the transmission zone influenced by internal waves since the entrained nutrient at station K2 is much lower than station K1. Alternatively, it may be caused by tilting of sediment traps during rapid rise of trap (~32 m within 10 min), but it is difficult to quantify possible trapping efficiency when traps are tilt.

Materials and Methods
Sinking particles were collected at ~150 m (below euphotic zone) using a drifting sediment trap array 29,37 . The trap deployment periods were 14 hours atstationK1 (6 th August), and 24 hours at station K2 on (4 th August), respectively (Fig. 1). Swimmers evident with a microscope on the filters were carefully removed using forceps. Attached to the trap, a conductivity-temperature-depth sensor (CTD, Ocean Seven 310 Multi parameter Probe) was deployed at stations K1 (~160 m) and K2 (~180 m) to record the hydrological properties of water column, respectively. In addition, we collected seawater samples from water column twice for each trap station aboard RV OR-3. Here, K1-1 and K1-2 represent sampling time at 9 am and 5 pm on 4 th August at station K1, respectively. K2-1 and K2-2 represent sampling time at 5 am and 11 am on 6 th August at station K2, respectively. A SBE9/11 plus CTD sensor was used to measure hydrographic conditions. Concentration of POC in suspended and sinking (trap collected) samples were measured using an elemental analyzer (Elementa, Vario EL-III, Germany), according to Hung et al. 26 . Concentration of Chl a was measured with a Turner Designs 10-AU-005 fluormeter after extraction with 90% acetone using non-acidification method 37 . Nitrate concentrations at stations H1, H2 and H3 were measured according to Gong et al. 38 . Concentrations of nitrate at stations K1 and K2 were calculated using linear regression result between nitrate and temperature 39 . Integrated nitrate (In-Nitrate, 0-75 m), Chl a (In-Chl a, 0-150 m), and POC (In-POC, 0-150 m) stocks were calculated using trapezoidal rule. A PAR scalar quantum irradiance sensor (Chelsea Technologies Group Ltd, UK) was used to calculate the euphotic depth (EZ) which was defined as the depth of 0.1% surface light penetration.