Elevated carbon flux in deep waters of the South China Sea

We measured particulate organic carbon (POC) fluxes from the euphotic zone into the twilight zone and deep waters (>1000 m) that occurred between the shelf and the basin in the South China Sea (SCS) and at the SouthEast Asia Time Series Station (SEATS) using floating sediment trap arrays. Additionally, selected sinking particles were imaged by scanning electron microscope (SEM) to reveal particle morphology and composition. Results showed large variations in POC fluxes with elevated values (32–104 mg-C m−2 d−1) below the euphotic zone and a trend towards lower values in the deep SCS. Vertical POC fluxes measured in deep waters between the shelf and the SCS basin were much higher than those estimated by Martin’s attenuation equation. These elevated POC fluxes in deep waters were attributed to lateral particle transport as opposed to enhanced settling out of the euphotic zone. SEM images of sinking particles at 150 m show abundant marine biogenic detritus, while those in deep waters contained a higher proportion of lithogenic material. A great deal of the spatial variability in POC fluxes across the twilight zone and deep waters of the SCS cannot be represented by current biogeochemical models.

Scientific RepoRts | (2019) 9:1496 | https://doi.org/10.1038/s41598-018-37726-w and a steep topography in the NSCS all facilitate the lateral transport of biogenic and terrestrial carbon across the shelf 22,23 . This transport may also occur sporadically between the continental shelf and deep ocean basin but data about this phenomenon is very scarce.
In this study, we measured carbon fluxes from upper layers (150 m), twilight zones (150 to 1000 m) to deep waters (greater than 1000 m) at the border between the shelf and the basin, and in deep waters in the SCS using our floating sediment trap array. We estimated POC fluxes, measured trace metal concentrations and obtained scanning electron microscope (SEM) images of the sinking particles. We then attempted to discuss the origin and processes of the laterally transported POC observed in the boundary between the shelf area and the open ocean of the SCS.

Results
Distributions of chlorophyll a and POC. A diagram of potential temperature (T) against salinity (S) is shown in Fig. 1. It is worth noting that the water characteristics of stations (Sts.) T1, T2, T3 and T4 appear to result from the mixing between the Kuroshio and the SCS 24 . Distributions of Chl a concentrations at all stations showed a subsurface Chl a maximum at 75-100 m, below which Chl a decreased with depth to a background value (<0.1 mg m −3 ) deeper than 200 m (Fig. 2). The vertical distributions of POC concentrations at the four stations were similar to the Chl a pattern, with a subsurface maximum suggesting biogenic sources of POC. However, elevated POC concentrations (30~70 mg m −3 ) were observed in the twilight zone (~150 to 500 m) at Sts. T2 and T3, perhaps due to the more recalcitrant nature of the particles in the twilight zone. Elevated POC concentrations were also observed in deep waters at Sts. T1 and T2 (~30 mg m −3 ) when compared to background values (~20 mg m −3 ) at adjacent deep-water sites (Fig. 2). Resuspension of material settled at Sts. T1 and T2 is unlikely since the deepest trap were at 1000 and 2000 m, respectively, while water depths were ~2700 and ~3200 m, respectively (Fig. 2). In addition, the sampling location was approximately 100 km away from Taiwan where the influences of terrestrial materials are expected to be insignificant during non-typhoon conditions. POC fluxes in the northern SCS. POC export fluxes from the euphotic zone (i.e. 150 m) were approximately 33 (St. T1) to 36 mg-C m −2 d −1 (St. T2), which is slightly lower than the value (50 ± 7 mg-C m −2 d −1 ) reported for the NSCS 25 as a whole. Interestingly, elevated POC fluxes (32 and 23 mg-C m −2 d −1 at 500 and 1000 m) were found at St. T1 which did not exponentially decrease with increasing depth based on Martin's attenuation curve (Fig. 3a). A similar trend of elevated POC fluxes (46, 24 and 35 mg-C m −2 d −1 at 500, 1000 and 2000 m) was also observed at St. T2 (Fig. 3b). An analogous distribution profile of POC fluxes was observed at St. T3 ( Fig. 3c) (Fig. 3d). Although lower than values at Sts. T1~T3, these values were slightly higher than fluxes previously measured at SEATS when particles were collected by moored sediment traps (deployed for six months or twelve months) 23 . Our elevated fluxes (collected time ~1 day) at St. T4 are reasonable since sinking particles collected by moored traps (collected time: ~six months) will be degraded, which can be inferred from Hung et al. 26 assessment about the extent of degradation of sinking particles collected by sediment traps.
The distribution of POC concentrations in deep waters at Sts.T2 and T3 (Fig. 2) showed anomalously high values situated 2500 m above the seafloor suggesting the trap-collected particles from deep waters were less likely affected by resuspended bottom sediments and potential lateral POC transport occurring at Sts. T2 and T3. Similar lateral transport of sinking material is reported at the depth of 500 to 1500 m in the southern ECS 27 .   either through its decomposition to dissolved inorganic carbon (DIC) or its conversion to dissolved organic carbon (DOC) 1 . In this study, however, we found elevated POC fluxes below 500 m which provide evidence for the large-scale lateral advection thought to occur in these deep waters. The measured carbon export fluxes to the deep waters at Sts. T1 and T2 were approximately 23-32 and 24-46 mg-C m −2 d −1 , respectively. It is common practice to predict the attenuation of POC flux with increasing depth using Martin's equation 6 . We used b values previously measured in the NSCS 25,28 of 0.65 (lower value) and 1.10 (upper value) to estimate the POC flux attenuation. This calculation led to POC fluxes of 15-9 and 9-4 mg-C m −2 d −1 at water depth of 500 and 1000 m at St. T1, and 17-10, 11-5 and 7-2 mg-C m −2 d −1 at water depth of 500, 1000 and 2000 m at St. T2, respectively (Fig. 3). In addition to using the b values described above, we also estimated POC fluxes at Sts.   marginal seas, such as the SCS, cannot be adequately described by a traditional particle attenuation function (e.g. Martin's equation) under events of lateral transport. According to Hung and Gong 29 and Hung et al. 30 , the e-ratio (POC flux/primary production (PP)) in open ocean water is around 0.1 (0.05~0. 16). If we use e-ratio 16% (maximum value) and previous annual PP values (280 mg-C m −2 d −1 ) 14 , the estimated POC fluxes (b = 0.65 and 1.10) 25,28 at 500 and 1000 m would be 12-20 and 6-13 mg-C m −2 d −1 , respectively. The estimated POC fluxes are much lower than observed values (32-46 and 23-24 mg-C m −2 d −1 at 500 and 1000 m, respectively) suggesting POC fluxes in deep waters cannot be accounted for by biogenic particles settling from the upper ocean.
The average POC fluxes in deep waters were 27 and 35 mg C m −2 d −1 at Sts. T1 and T2, respectively. Assuming the affected area was 10 6 m 2 (=100 km × 10 m, based on the steep topography in the NSCS) during each episodic event, it follows that the integrated vertical flux of POC in deep waters at Sts. T1 and T2 was 27 × 10 3 g-C d −1 (on September 5 th , 2012) and 35 × 10 3 g-C d −1 (September 26 th , 2012). The lateral POC flux can be estimated to be 28 × 10 8 g-C d −1 (=POC concentration (~17 mg-C m −3 ) at 1000 m × average flow rate at 1000 m) if we assume an eastward flow of 1.9 Sv (m 3 s −1 ) at depth of 1000 m 11 . Hence, the lateral POC flux is several orders of magnitude higher than the integrated vertical flux estimated by sediment traps. It is worth noting, however, that this lateral transport likely occurs sporadically, when currents strong enough to resuspend settled material contact the continental slope seafloor. Such resuspension events may not last for the full 24 h period implied by our calculation. More studies are clearly needed to better understand real lateral carbon transport in deep waters in the SCS.
We have good reasons to believe that some POC can be transported to the ECS and on to the western Pacific via the eastward flowing SCSIW, and that this transport occurs at depths of 350 to 1300 m 11 . Other deep lateral POC flux can be stored in deep waters (>1300 m), potentially trapping the carbon for centuries, if not millennia, given the long residence time of the deep seawater 31 . In this way, the lateral carbon transport plays an important role for carbon sequestering in deep waters of the NSCS.

Possible mechanisms affecting lateral advection. Several mechanisms, including typhoon-induced
hyperpycnal currents 32 , mesoscale eddies 33 , earthquakes 34 and episodic events 28,35 such as "benthic storms" associated with current interaction with steep topography, may contribute to lateral transport in the SCS. For example, the funnel-like topography of a canyon off southwest Taiwan induces sediment transport that is mainly driven by flood events in the summer and tidal action in the winter season 32 . Part of our sampling was undertaken in September, i.e. during the typhoon season. Under those conditions, large quantities of suspended riverine material containing some lithogenic particles mixed with biogenic particles are delivered directly into the deep sea by high-energy rivers 22 . Our measurements showed the Al-normalized flux (Fig. 5) as well as metal concentrations (Fe, Al, Mn, Ti, U and V) (Fig. 6) increasing with increasing depth. Moreover, percentages of Fe and Al in sinking particles below 500 m were close to those in the earth crust 36 , indicating that sinking particles in the deep waters were partially from terrestrial sources 37 (Fig. 6). Indeed, many researchers also reported that typhoon-induced floods and hyperpycnal flows play a key role in the transport of terrestrial materials and carbon to the ocean 23,38,39 . Mulder et al. 40 described that hyperpycnal processes involve transporting river material, do not entrain seafloor sediment and directly transport to the marine environment by a turbulent flow (hyperpycnal turbidity current). Typhoons are not only physical agents for enhanced delivery of terrestrial material to the sea. Geological agents can also trigger gravity flows such as turbidity currents that eventually leave deposits on the seafloor and form large-scale deep-water depositional systems 40,41 . For instance, typhoon Tembin was a category-2 typhoon (sustained winds: 30 m s −1 ) which traveled at a speed of 3.0~5.5 m s −1 from 22-29 August, 2012. The typhoon affected the study area (Sts. T1 and T2) for several days and dumped high amounts of rainfall, peaking at ~580 mm in Taiwan and resulted in flooding in southern Taiwan (a similar event was also recorded for typhoon Soulik, 11-13 July, 2013). It is likely that typhoon-induced hyperpycnal turbidity currents drive lateral particle transport from the shelf to basin area in the SCS. The lower salinity water (blue curve in Fig. 1b) found in the upper waters of St. T1 about 9-10 days after the passage of typhoon Tembin attests to the amount of rainfall and flooding that took place. Similar typhoon-induced heavy rainfall and flooding brought to open ocean after typhoons have been reported by Lee et al. 42 and Hung et al. 43 .
Furthermore, frequent earthquakes may be another trigger factor facilitating lateral POC transport 21 . For example, it is summarized that earthquakes often reinforce the effect of typhoons and vice versa, thus substantially enhancing the sediment load in fluvial systems 44,45 , which may frequently increase hyperpycnal concentrations. But it is difficult to estimate its real impact since the seas around Taiwan have more 10,000 earthquakes per year (www.cwb.gov.tw).
Some researchers also reported that mesoscale eddies can carry particles from nearshore to offshore 20,33 . The mesoscale eddies can even carry materials to the open ocean, as evidenced by satellite image observations, but it is not evident that these materials can enter deep waters. We cannot definitely distinguish all factors based on current data set, since resuspended particles may occur in the coast or on the continental shelf, and then be re-located elsewhere due to episodic events. Therefore, the elevated POC fluxes in deep waters probably can be caused by typhoons, earthquakes, eddies, seasonal and interannual variability in currents and/or cooling effects as well as deep layers sediment resuspension 28,33,35,46 . One may ask whether the deep sea POC flux is unique to the northern South China Sea or whether it applies to the whole South China Sea. For example, Chen et al. 10,11 used a box model to show that the eastward flow of the SCSIW (350~1300 m) can transport high-nutrient waters across the entire ECS and beyond. Lateral POC fluxes and POC concentrations could be measured directly but this would require intensive sampling and lots of ship time. Until now, very few cruises have captured lateral POC transport in the SCS, thus we are not sure if the lateral contribution to the POC export flux is present in all parts of the deep (>500 m) SCS. Such deep water lateral transport in the SCS definitely requires further study.

Materials and Methods
Four cruises aboard the R/V Ocean Researcher III (OR-III) and V (OR-V) were conducted in each of the following periods: Sep. 5-7 (T1), Sep. 25-28 (T2), 2012 and Jul. 14-17 (T3) and Sep. 12-18 (T4, OR-V), 2013. The first cruise was conducted approximately a week after a category 2 typhoon Tembin, and the third cruise was conducted ~2 days after typhoon Soulik (category 2). Sinking particles were collected at 150, 500 and 1000 m at St.T1 (21.7°N 120.0°E, water depth ~2700 m) located at the boundary between the shelf and the basin using a drifting sediment trap array 13,25 . At the second cruise, the sediment trap was also deployed at the boundary between the shelf and the basin (St.  (Figs 1 and 3). The deployment period was approximately 24 to 36 hours (less than 24 hours if bad weather conditions occurred). Swimmers on the filter were carefully removed using forceps under a microscope. POC flux measurement was determined according to Hung et al. 25 . A SBE 9/11 plus CTD sensor was used to assess hydrographic conditions. Seawater samples were collected to measure POC and Chl a concentrations. Organic carbon in suspended as well as sinking (trap collected) materials were assessed based on Hung et al. 25 using an elemental analyzer (Elementa, Vario EL-III, Germany). Chl a was analyzed with a Turner Designs 10-AU-005 fluorometer after extraction with 90% acetone following by the non-acidification method 47 . Contents of metal elements in sinking particles were determined by quadrupole-based inductively coupled plasma mass spectrometer (Hsu et al. 48 ) after total digestion of the particles with a combination of HF, HNO 3 and HClO 4 . Selected bulk sinking particles were filtered on polycarbonate filters and imaged by scanning electron microscopy (SEM) according to Hung et al. 49 .