Influence of a Decaying Cyclonic Eddy on Biogenic Silica and Particulate Organic Carbon in the Tropical South China Sea Based on 234Th-238U Disequilibrium

Eddies play a critical role in regulating the biological pump by pumping new nutrients to the euphotic zone. However, the effects of cyclonic eddies on particle export are not well understood. Here, biogenic silica (BSi) and particulate organic carbon (POC) exports were examined inside and outside a decaying cyclonic eddy using 234Th-238U disequilibria in the tropical South China Sea. For the eddy and outside stations, the average concentrations of BSi in the euphotic zone were 0.17±0.09 μmol L-1 (mean±sd, n = 20) and 0.21±0.06 μmol L-1 (n = 34). The POC concentrations were 1.42±0.56 μmol L-1 (n = 34) and 1.30±0.46 μmol L-1 (n = 51). Both BSi and POC abundances did not show change at the 95% confidence level. Based on the 234Th-238U model, BSi export fluxes in the eddy averaged 0.18±0.15 mmol Si m-2 d-1, which was comparable with the 0.40±0.20 mmol Si m-2 d-1 outside the eddy. Similarly, the average POC export fluxes were 1.5±1.4 mmol C m-2 d-1 and 1.9±1.3 mmol C m-2 d-1 for the eddy and outside stations. From these results we concluded that cyclonic eddies in their decaying phase have little effect on the abundance and export of biogenic particles.


Introduction
Mesoscale eddies significantly affect biogeochemical processes in the upper ocean especially in oligotrophic oceanic settings [1][2][3][4]. Three types of eddies, i.e., cyclonic, anticyclonic and modewater eddies, are reported [5], all of which supply new nutrients to the euphotic zone either through uplift of isopycnals [6] or through horizontal advection between the eddy center and edge [7]. Commonly, cyclonic eddy-induced nutrients stimulate phytoplankton growth, the production of particulate organic carbon (POC) and biogenic silica (BSi) at the early life-stage of the eddy [8][9]. However, POC export shows little variation although an increase of BSi export is commonly observed [3][4][10][11]. In these case studies, diatom groups make up the

Study area
The SCS is located in Southeast Asia, with a shallow mixed layer of <50 m throughout the year. The euphotic zone of >1% light is usually 75-100 m [20], resulting in a rapid consumption of nutrients in the upper 100 m. Hence, the SCS shows typical oligotrophic characteristics [21] similar to the Sargasso Sea and the North Pacific gyre. To date, very limited nutrients have been reported in the SCS basin, especially in the southern SCS. The primary production (PP) is usually less than 46 mmol C m -2 d -1 [22] with high values in seasons showing a deeper mixed layer [23]. Model simulations indicate that physical processes, such as eddies and monsoon-forced circulation, to a large degree, affect PP in the SCS [19][20]24]. However, the response of biogenic particles (including their abundance in and export out of the euphotic zone) to these physical processes has been poorly understood in the SCS. The in situ investigation identified involves only the influence of an anticyclonic eddy on the export of POC and BSi [7], which shows enhancement of POC export, contrary to the general concept.

Sample collection
Samples were collected from 7 to 18 November 2010 on board R/V SHIYAN 3. Stations were situated mainly in the tropical SCS (Fig 1), covering the 6-17°N and 109-118°E area. Transects I and II, mainly along 6°N and 10°N, involved 15 stations and, in addition, a south-north oriented Transect III along 113°E included 10 stations. Seawater samples were collected mainly in the euphotic zone (normally~1, 25,50,75, and 100 m) using CTD (conductivity, temperature and depth) Rosette integrated sampling bottles. Generally, 6-8 L of seawater was filtered through a pre-combusted (450°C for 4 h) quartz fiber filter with 1 μm pore-size (QMA, Whatman) to collect particles for POC and particulate 234 Th measurements. Four liters of filtrate was used to determine the dissolved 234 Th. BSi samples were collected only from 15 selected stations. Usually, 2 L of seawater was filtered through a 0.8 μm polycarbonate membrane filter to concentrate the BSi.

234
Th analyses 234 Th in the <1 μm fraction (i.e. dissolved 234 Th) was co-precipitated using the MnO 2 method [25]. In order to obtain a stable 234 Th recovery, we examined the co-precipitation conditions in detail, including filter membrane, pH value and amount of MnO 2 , using calibrated Th yield tracers [26]. We found a mean recovery of 234 Th of 95.7±1.0% (mean±sd, n = 5). In this study, dissolved 234 Th was concentrated using co-precipitation conditions [26]. The pH value of the filtrate was adjusted to 9.0 and a solution containing KMnO 4 and MnCl 2 was added quantitatively while stirring. The resulting suspension containing precipitated MnO 2 was left to stand for more than 6 h to allow the MnO 2 particles to grow, since large particles benefit filtration [27]. MnO 2 carried 234 Th was finally filtered on a QMA membrane filter and dried at 60°C. All the dried samples were counted using a low-level beta counter with a counting efficiency of 41.2% until the net counting errors were less than ±5%. 150 days later, a second counting was conducted for quantifying other beta emitters in order to remove their contribution to the first obtained counters [25]. The activity concentrations of 234 Th were calculated based on the counting efficiency, recovery of Th, and net 234 Th counts (excluding blank and background), and corrected to sampling time ( Table 1). The errors for the 234 Th data in Table 1 represent the propagated errors from the statistical count errors of 234 Th obtained from two measurements. 238 U activities were calculated from salinity based on the widely used relationship between 238 U (dpm L -1 ) and salinity from the dataset [28]. Salinity was measured with the calibrated CTD.

POC and BSi analyses
POC was determined using the particulate 234 Th sample. Such a strategy can best support the accuracy of the 234 Th-based POC flux [29]. Reviews indicate that the influence of particle size on the C/Th ratio is complicated [30][31]. Because of limited shiptime, >53 μm particles were not collected in the present study. And >1.0 μm particles were used to quantify the POC flux as previously used in the SCS [7]. The QMA filters were acid-fumigated with concentrated HCl (12 mol L -1 ) for 48 h to remove carbonate. After drying at 60°C, the POC content and the 13 C abundance in POC (δ 13 C) were determined with a Perkin Elmer CHN analyzer connected to a Finnigan MAT DELTA plus XP mass spectrometer [32]. The standard used for POC and δ 13 C was IAEA-C8. The procedural carbon blank including filter and tin cup was less than 3 μg, accounting for less than 8% of the bulk POC. Based on the replicate analyses, the uncertainties of the POC contents were better than 10%. The double wet-alkaline digestion method [33] was adopted to determine BSi contents. Briefly, particles on polycarbonate membranes were digested with 0.2 mol L -1 NaOH. Silica concentration was measured through molybdosilicate blue spectrophotometry. Then particles were rinsed with silica-free water and dried. A second digestion was conducted with the same protocols to quantify the mineral-derived silica [33]. The BSi contents were calculated based on the sequential leaching. The procedural blank of BSi, including regents and membrane, was less than 0.03 μmol L -1 . Two reference materials (still pond and R64) used for inter-laboratory comparison [34] were used for BSi determination. Our results, 2.79±0.26% (mean±sd) for the still pond sample and 5.27±0.24% for the R64 sample, were comparable with the inter-laboratory averages of 2.82±1.17% and 6.49±2.09%.

Sea-level anomaly (SLA)
To examine the possible influence of a cyclonic eddy on biological particle abundance and export, stations in the eddy were separated from those located outside the eddy based on the SLA with weekly resolution obtained from the Global Delayed-Time merged SLA Data (Fig 1). SLA represents the variations of the sea surface height (SSH) relative to a mean sea surface (MSS). The MSS is the mean of the SSH from 1993 to 1999. All samples were collected from 7 to 18 November. Based on the SLA contours on 10 and 17 November, the cyclonic eddy, lying in the southwest part of the sampling area, was at its decaying stage, as was confirmed by the temporal evolution of SLA from 13 October to 1 December, 2010 (not shown). The eddy was elongated extending from 6°to 12°N, and the core located around 11°N, 111°E. The maximum SLA was much lower than 0 cm at the eddy center on 10 November, and the area with SLA decreased on 17 November. Stations 33, 50, 52, 57, 58, 60 and 63 were located in the eddy, and were evidently influenced by the eddy on both 10 and 17 November. Stations 40,42,44,46,48,54,55,72 and 75 were ordinary stations (Fig 1). Stations 36 and 69 were sampled on 17 November and, based on the SLA, they were regarded as outside stations. Since Stations 22 and 23 were located in the other eddy north to our studied eddy (Fig 1), they were not included in either the eddy or ordinary station groups. Such a separation was confirmed by the detailed distributions of hydrologic parameters and silicate concentration (Figs 2-4).

Hydrologic parameters and silicate distributions
Temperature and salinity data are presented in Table 1, as well as the particulate components (i.e. POC and BSi) and the activity concentrations of 234 Th. Along Transect I, the distribution of temperature and salinity showed the uplift of deep water introduced by the cyclonic eddy (Fig 2). Within the eddy, cold and salty water intruded into the euphotic zone. The mixed layer was compressed to less than 25 m based on the definition (ΔT = 0.8°C) for the mixed layer depth (MLD) [35], much shallower compared with around 50 m at outside stations. At the eddy center, temperatures were much lower than those of the surrounding water, while salinities were much higher. The vertical distribution of silicate concentrations also revealed the cyclonic eddy (Fig 2). In the upper 25 m, all waters showed low silicate concentrations. The silicate concentrations increased to around 0.5 μmol L -1 at 75 m at the eddy center; however, they were around 0.3 μmol L -1 at 75 m for the ordinary stations. For Transect II, stations in the western area (including 57, 58, 60 and 63) showed cyclonic upwelling characteristics (Fig 3). The MLD was less than 25 m in the eddy, while it reached up to 50 m at other stations. Transect III also presented the cyclonic eddy at Station 33 (Fig 4). Cold water (T < 25°C) was observed in the upper 75 m around 11.5°N and high silicate concentrations revealed the intrusion of deep water into the upper 75 m.

Th/ 238 U disequilibria
The activity concentrations of total 234 Th at all stations varied from 0.74 to 2.83 dpm L -1 (Table 1), which was comparable to the values obtained in spring [36]. In order to assess the difference between the eddy and ordinary stations, student's t-test was used to check the two data groups in terms of a specific parameter as listed in Table 2. The sample sizes for statistical analysis are presented in parentheses in Table 2. No difference was observed for the 234 Th deficits between the eddy and ordinary stations at the 95% confidence level. The ratios of particulate 234 Th to dissolved 234 Th, indicating its partitioning between particle and seawater, averaged 0.68±0.75 (n = 35) for the eddy and 0.74±0.74 (n = 53) for the ordinary stations (Table 2). Obviously, particles did not result in difference in the partitioning of 234 Th. The average 234 Th/ 238 U ratio for all eddy stations was 0.85±0.20 (mean±sd, n = 35) and 0.79±0.18 (n = 53) for ordinary stations. Statistically, there was no difference between the eddy and outside stations (Table 2). Thus, the deficits of 234 Th relative to 238 U, via scavenging to particles and successive sinking, seemed to show little difference. Such a scenario indicated that the particle dynamics in terms of 234 Th scavenging were similar in the studied eddy and ordinary water.

POC and BSi
For the eddy stations, the average POC concentration was 1.42±0.56 μmol L -1 , comparable to that of 1.30±0.46 μmol L -1 at ordinary stations ( Table 2). The average BSi concentration of 0.17±0.09 μmol L -1 in the eddy was also similar to that in the surrounding water which had a mean of 0.21±0.06 μmol L -1 . Statistical analysis did not show any discernible difference in the BSi concentrations between the eddy and ordinary stations at the 95% confidence level. The average ratio of BSi concentrations in the eddy to those outside the eddy was 0.8±0.5.

Eddy influence on BSi and POC abundance
Statistically, no difference in the POC concentrations was observed between the eddy and reference stations at the 95% confidence level ( Table 2). The PP rates obtained using the 14 C technique at the eddy stations 60 and 63 were 0.078 mmol C m -3 d -1 [37]. At ordinary stations (36,69,72 and 75), the PP rates ranged from 0.042 to 0.084 mmol C m -3 d -1 with a mean of 0.064±0.023 mmol C m -3 d -1 [37]. It seemed that there was little difference in the PP rates. Similarly, the eddy did not change the BSi abundance in the euphotic zone ( Table 2). The comparable abundance of BSi in the present study was different from a few reports on an Atlantic mode-water eddy [4,38], where increased BSi was observed within the eddy.
The eddy age would explain the difference in BSi variability between our study and the references. At the early life-stage, cyclones usually result in an increase of biogenic particles including POC and BSi [39]. However, at the decaying life-stage, a decaying biological response is reported, and even a negative NCP in an older cyclone in the Sargasso Sea [16]. In the present study, the temporal variability of SLA revealed that the eddy was in its decaying phase (Fig  1). Diatoms in the eddy probably did not show a discernible response to the eddy.
The δ 13 C signals, usually showing different values either for various phytoplankton species or for different PP rates in terms of specific species [40], also lends support to the small variability in the phytoplankton community and PP rates (Table 1). In the eddy, the δ 13 C values varied from -25.15 to -21.14‰, averaging -22.91±0.85‰ (mean±sd). At ordinary stations, the δ 13 C values ranged from -25.57 to -21.53‰ with an average of -23.22±1.00‰. Obviously, there was no difference, indicating little influence of the eddy on the PP at its decaying life-stage. Table 2. Comparison of some parameters between the eddy and ordinary stations. E/O is the ratio of a specific parameter in the eddy to the surrounding water. The errors represent the standard deviation for data used to calculate the means.

Parameters
Outside a (mean±sd) Eddy a (mean±sd)

Eddy influence on BSi and POC exports
The export fluxes of BSi and POC out of the euphotic zone were calculated using the 234 Th-238 U disequilibria. The mass balance of 234 Th is expressed as [29,41]: where dA Th /dt is the change rate of the total 234 Th (i.e. non-steady state, NSS), A Th and A U are the total activities of 234 Th and 238 U, λ is the decay constant of 234 Th (0.0288 d -1 ), P Th denotes the net export flux of 234 Th, and V is the sum of advection. In open oceans, the V term is usually minimal, as well as the NSS term [42].
In the present study, the influence of the NSS and V terms was evaluated. For the outside stations, the average activity concentration of 234 Th was 1.91±0.14 dpm L -1 . The difference between the average and 234 Th activity at each station varied from -0.20±0.16 to 0.25±0.16 dpm L -1 in the euphotic zone. Together with the sampling dates for all outside stations, the estimated dA Th /dt values ranged from -34±27 to 42±27 dpm m -2 d -1 . For the eddy stations, the NSS term corresponded to a range of -90±80 to 110±80 dpm m -2 d -1 . On the one hand, the NSS term did not seem to evidently influence 234 Th flux since it was around two orders of magnitude lower than the other terms, i.e. λA U and λA Th . On the other hand, errors in the NSS term reached up to~90%, indicating it was almost meaningless to our study. In general, at least 1-2 weeks between duplicate samplings are required to match the NSS model [42], and computing NSS terms with a sampling interval shorter than 10 days introduces massive errors [43]. Therefore, the NSS approximation was no better than just assuming an SS for short sampling intervals [42][43]. In the present study, sampling of the eddy stations was conducted within 4 d. Thus, the NSS term was not included in estimating the export fluxes of 234 Th.
The advection term V consisted of vertical and horizontal advection, thus the export flux of 234 Th could be calculated as: where v is the horizontal current velocity, which was estimated based on the spatial current pattern as shown in Fig 1. The variable ω denotes the vertical mixing velocity. In the SCS, the vertical mixing velocity within a mesoscale scale eddy over a month timescale is around 3.4 m d -1 [44]. Considering the comparable timescale of the eddy in our study, this value was adopted. The term @A Th /@z is the vertical gradient of 234 Th activity from the twilight zone to the euphotic zone [45][46]. Here, it was estimated from the variability in the 234 Th activity concentrations between 75 m and 200 m. The @A Th /@ H term is the horizontal gradient of the 234 Th activity based on its spatial pattern. The exports of BSi and POC were calculated using the proposed approach [29]: where P x is the flux of particulate component "X" (i.e. BSi or POC) out of the euphotic zone, and (X/Th P ) z is the ratio of BSi (or POC) to particulate 234 Th at the bottom of the euphotic zone.
For POC, both >53 and >1.0 μm particles were used to quantify the POC export. Because the >53 μm particle was not collected, the C/Th ratio in the >1.0 μm particle was used which was similar to the published POC calculation in the SCS [7,36]. The relationship between the C/Th ratio and particle size is complicated [30][31]. Sometimes, the C/Th ratios in the >1.0 μm particle are higher than those in >53 μm particle [47][48], but this showed the opposite scenario in other studies [49][50]. A comparable C/Th ratio in both >53 and >1.0 μm particles is also observed [46,51]. Although the influence of particle size on the C/Th ratio is unpredictable, below the euphotic zone the difference in the C/Th ratio between >53 and >1.0 μm particles is less than in the surface water [30][31]. In the SCS, the C/Th ratio is 1.4±0.1 μmol/ dpm for >53 μm particles below the euphotic zone (100-125 m), and is 1.6±0.1 μmol/dpm for >1.0 μm particles [52]. This phenomenon is attributed to different POC in and below the euphotic zone. In the upper euphotic zone, the majority of POC were freshly produced by the plankton, while POC at the bottom of the euphotic zone might experience coagulation during sinking. Consequently, POC was large size, resulting in comparable C/Th ratios between >53 and >1.0 μm particles at the bottom of the euphotic zone. It should be noted that we cannot conclude that the POC export here was comparable to those from >53 μm particles although the reported C/Th ratios were similar between the >53 and >1.0 μm particles [52]. The influence of particle size on the POC flux is illustrated via a comparison between our results and others published for the SCS.
Besides particle size, the sampling technique also influences the POC/ 234 Th ratio [30,53]. A few studies suggest that bottle derived POC may overestimate the true POC concentrations owing to the adsorption of dissolved organic carbon (DOC) on the filter [54]. The increase of seawater volume may minimize the effect of DOC adsorption [36]. In the present study, POC was collected from 6-8 liters of seawater using a QMA filter with a diameter of 25 mm as in a previous study in the SCS [36], which corresponded to~300-400 liters of seawater being pumped through a 142 mm diameter QMA filter. Such a strategy could decrease the influence of DOC. Indeed, the average of the POC/ 234 Th ratios was 1.56±0.08 at the bottom of the euphotic zone in our study, consistent with that obtained using a 53 μm pore size Nitex screen in the SCS [52]. This ratio was also comparable to the sediment trap results in the lee of Hawaii (1.50±0.04 μmol dpm -1 ) [10]. Even though the POC/ 234 Th ratio was comparable to those from the sediment trap or in situ pumps, the POC fluxes in the present study should be conservatively regarded as the upper limit reported in the SCS [7,36].
We collected data related to the SCS to provide more information for comparison. Thus, the 234 Th-derived POC flux, which ranged from 1.7 to 5.7 mmol C m -2 d -1 in November at 6°N [55], was comparable to our results; the sediment trap obtained POC export value at 720 m, which was 2.5 mmol C m -2 d -1 in November at a station (9°23 0 N, 113°14 0 E) [56] near Station 36, appeared to coincide with our result of 2.1±0.4 mmol C m -2 d -1 (Table 3); the 234 Th-and   210 Po-based average POC flux was 3.8±4.0 mmol C m -2 d -1 in April-May [36,57]; and the POC export out of 100 m was 2.6 mmol C m -2 d -1 over a 4-yr timescale in the central and northern SCS [58]. These results indicated that the particle fluxes in our study were comparable with the published results.
The average 234 Th flux was 985±629 dpm m -2 d -1 inside and 1218±419 dpm m -2 d -1 outside the eddy ( Table 3). The average BSi flux in the eddy was 0.18±0.15 mmol Si m -2 d -1 (Table 3), comparable to 0.17±0.06 mmol Si m -2 d -1 observed in a mode-water eddy in the Sargasso Sea [4]. It was 0.40±0.20 mmol Si m -2 d -1 for the ordinary stations. Student's t-test indicated no difference between the eddy and ordinary stations in terms of BSi fluxes at the 95% confidence level (p = 0.077, Table 2). This result lent support to the little influence of the studied eddy on BSi export in its decaying phase. This scenario was different from observations in cyclonic eddies at their early life-stages. For example, three to four times higher BSi flux is observed in an eddy in the subtropical North Pacific [3,10,59]. A two-to three-fold increase of the BSi flux is also found in a mature stage of an eddy in the Sargasso Sea [4]. Even up to~20 times higher BSi export comparable to the Bermuda Atlantic Time-series Study site is observed in a mode-water eddy [38]. This difference between our results and published results suggested that the increase of BSi export occurred mainly at the early life-stage of the eddy.
The average POC flux was 1.5±1.4 mmol C m -2 d -1 within the eddy, and 1.9±1.3 mmol C m -2 d -1 in the surrounding waters respectively. Statistically, the eddy did not cause evident variability in the POC flux (Table 2). Similar results are found in mature eddies in the North Pacific [3,10] and in the Sargasso Sea [4], indicating that these subtropical cyclones may not be effective in exporting POC to the mesopelagic zone [59]. Conversely, these studies reveal increased exports of BSi, implying that this eddy may act as a silica pump [3,59]. However, our results illustrated little variation in either BSi or POC exports. Based on these observations, further research is needed to examine the influence of eddy life-stage on the decoupling of POC and BSi exports in the tropical SCS.