Particulate nitrogen (PN) dynamics in the oligotrophic northern South China Sea (around the SouthEast Asian Time-series Study (SEATS) station) was explored by examining the isotopic compositions of suspended PN in the top 200 m over 3 years and sinking PN collected by sediment traps. The PN inventory (IPN) in the upper 100 m is larger than in the lower 100 m, exhibiting stronger seasonality. Both layers reveal significant seasonality in meanδ15NPN, yet, the mean in the upper 100 m (2.0 to 5.3‰) is consistently smaller than that in the lower 100 m, implying the occurrence of vertical biological fractionation and/or an addition of 15N-depleted N from the atmosphere. Theδ15NPN surges in winter, when the mixed layer is deeper, indicate an intensified nitrate supply from thermocline, during which relatively stronger downward transfer efficiency was inferred by a small IPN gradient. The largest vertical gradient in IPN appeared during intermonsoon periods, corresponding with weak vertical mixing, low δ15NPN, and high N* values. N fixation is likely the cause for the intermonsoon δ15NPN lows. The δ15NPNvalues of trapped material at 374 m and 447 m range from 3.3 to 7.3‰ with a flux-weighted mean of 5.6‰ resembling theδ15NO3of upwelled sources. By using a mass-isotope balance model under the assumption of no atmospheric N deposition, we obtained an N fixation input of ∼20 ± 26 mmol N m−2 yr−1. This value accounts for only ∼5–10% of the new production on an annual basis.
 Nitrogen fixation has long been perceived as a dominant allochthonous source of newly fixed nitrogen to the ocean [Karl et al., 1997; Montoya et al., 2004]. The input of newly fixed nitrogen is one of the primary controls on the oceanic autotrophic productivity and organic matter sinking fluxes, thereby affecting the biological pump, which is intimately coupled with the climate system through its modulation on the atmospheric CO2 [Altabet et al., 2002; Falkowski, 1997]. Yet the nitrogen dynamics in the surface ocean associated with nitrogen fixation are still neither well understood nor well quantified [Capone et al., 1997], because nitrogen fixation is episodic and patchy and poses severe obstacles to direct observations [Mahaffey et al., 2005]. Acquiring isotopic information of suspended and sinking particles over different seasons helps to pin down when, how and to what extent N fixation may contribute to nitrogen inventory in the deep ocean.
 High water temperature (above 20°C), low availability of reactive nitrogen, a stratified water column and presence of bioavailable iron are preferred environmental factors for nitrogen fixation [Karl et al., 1997]. The geographical setting of the South China Sea (SCS), the second largest marginal sea in the world [Liu et al., 2010], offers favorable environmental features for nitrogen fixation [Chao et al., 1996; Wong et al., 2002]. Among the relatively small number of measurements on the rates of nitrogen fixation in the world ocean, the rates observed in the SCS were among the highest values reported [Capone et al., 1997, and references therein], and comparably high rates were also observed in the adjacent areas, like the Kuroshio Current (KC) and East China Sea [Chen et al., 2008; Shiozaki et al., 2010].
 The surface circulation and the biogeochemistry in the SCS are subject to the forcing of the East Asian monsoons [Liu et al., 2002]. Interannual variations in the SCS circulation are mostly associated with modulation of the monsoon strength by El Niño–Southern Oscillation events [Chao et al., 1996]. Total concentrations of nitrate plus nitrite in the mixed layer rise above the detection limit to up to ∼0.7 μmol kg−1 in the winter, but drop below the detection limit in the rest of the year [Chou et al., 2006; Wong et al., 2007a]. However, drawdown of CO2 continues as indicated by the decline of dissolved inorganic carbon (DIC) for a period of time with nondetectable nitrate in the surface water. The continuing net carbon uptake is likely sustained by the supply of new nitrogen derived from nitrogen fixation [Chou et al., 2006]. Also, a distinctly high-nitrate anomaly (N*, up to ∼2μM) with a corresponding low δ15NO3−in the upper nitracline at the SouthEast Asian Time-series Study (SEATS) station (115.61°E, 18.01°N) (Figure 1) has been interpreted as a result of remineralization of biomass resulting from nitrogen fixation [Wong et al., 2002]. However, whether this is a cumulative signal of regional N fixation or fresh input from diazotrophs in the SCS remains unclear. Nevertheless, it is widely accepted that the ecosystem in the SCS is limited by bioavailable nitrogen. In winter, nutrients are mainly supplied diapycnally from the subsurface waters, where the nutrient reserves have N/P ratio lower than the Redfield ratio [Wu et al., 2003], as cooling and wind stress are intensified. By contrast, when mixing is weak, nitrogen fixation could play an important role in supplying bioavailable nitrogen, which will be taken up by nondiazotrophic phytoplankton in time [Mulholland, 2007].
 Apart from monsoons, the episodic tropical cyclone (typhoon) may also pump up nutrients into the mixed layer. These events can contribute as much as 20–30% of annual new production in the SCS, given an average of 14 typhoons traversing the SCS each year [Lin et al., 2003]. Moreover, increasing quantities of anthropogenic-sourced atmospheric nitrogen deposition to the open ocean may serve as a remarkable external nitrogen supply in oligotrophic marginal seas, such as the SCS [Duce et al., 2008, and references therein]. Strong seasonality in stratification versus various nutrient supply pathways in different seasons provides us an opportunity to explore formation, transportation and transformation of particulate nitrogen in this dynamic marine system.
 From the perspective of paleoceanographic research, the nitrogen isotopic composition of sedimentary organic matter has been widely used to infer past changes in biogeochemical dynamics at a given core site [Altabet, 1996]. However, there have been few, if any, studies, in which the seasonal variations of nitrogen isotopic composition of suspended particles in the upper water column were examined; let alone a detailed mechanistic investigation of the processes responsible for the nitrogen isotopic composition of organic matter during its transportation to depths. The SEATS project [Tseng et al., 2005; Wong et al., 2007b] allowed us to conduct seasonal sampling of the suspended particles; at the same time, the sediment traps deployed in the subsurface layer at ∼500 m in the northern SCS have provided samples of sinking particles for comparison. In this study, we compare seasonal changes in isotopic composition (δ15NPN) of suspended and sinking particles with related parameters, such as the mixed-layer depth, nitrate concentration and nitrate anomaly (N*). This study advances our understanding of the seasonal nitrogen supply to the euphotic zone, and dynamical isotopic compositions between suspended and export particles in the upper 400 m, which benefits future modeling work and paleostudies in the South China Sea.
2. Materials and Method
2.1. Sample Collection
 Water samples were collected for suspended particulate matter on nine cruises between May 2004 and January 2007 at the SEATS Station (Figure 1 and Table 1) using a CTD-rosette sampling system with 20 L GoFlo bottles for sampling on board R/V Ocean Researcher I. On the FR1 cruises, 10 L GoFlo bottles were used (Table 1). Nine to 20 samples were collected on 1–3 hydrocasts depending on the individual cruise schedule. Salinity and temperature were recorded during each cast with a SeaBird model SBE9/11 conductivity-temperature-depth (CTD) recorder. The mixed-layer depth (MLD) was defined as the layer of water holding a density gradient less than 0.125 kg m−4 [Suga et al., 2004].
Table 1. Sampling Cruises for the Collection of Suspended Particulate Nitrogen
Number of Samples
6 May 2004
6 Aug 2004
22 Jan 2005
30 Mar 2005
30 Jul 2005
25 Dec 2005
5 Jul 2006
20–22 Oct 2006
16 Jan 2007
 About 10 L of seawater was filtered through a precombusted (500°C for 4 h) Whatman quartz filter (25 mm diameter, 1.0 μm pore size) to collect suspended particulate matter (SPM). After the filtration, the filters were folded and wrapped in aluminum foil and stored at −20°C until analysis.
 We used time series sediment traps (Technicap PPS 3/3, France) to collect sinking particles. Two traps were deployed near the SEATS Station (Figure 1). Trap M1S was deployed at ∼374 m (119.28°E, 21.31°N) during September 2001 through March 2002. Trap M2S was situated at ∼447 m (117.30°E, 19.00°N) during December 2001 through May 2002. The sampling interval was 15 days. Before deployment, all collecting cups were filled with a buffered (5% or 1.7 M) formalin solution (7.5 < pH < 8.0) for sample preservation. A detailed description of pretreatment for trap samples has been given by K. K. Liu et al. . Briefly, samples were stored below 4°C before being transported to the laboratory. Large particles including zooplankton and other swimmers were screened out by a 1 mm Teflon sieve and by hand picking. The sinking particles were then filtered onto a preweighted 47 mm Millipore filter (0.45 μm pore size), which was subsequently washed with deionized water to remove salts. The desalted sample was dried at 40°C for 48 h and then the sample was scraped off from the filter, ground to powder and stored in an autodesiccator until it was analyzed.
2.2. Isotopic and Chemical Analyses
 Filters with suspended particles were freeze dried. Filters were acidified with 1 mL of 1 N HCl solution to remove carbonate after picking off visible zooplankton. This acid pretreatment has been applied in previous studies [e.g., Casciotti et al., 2008; Loick et al., 2007]. All filters were dried at 60°C in an oven for 48 h. The sediment trap samples were treated with 2 mL of 1 N HCl solution to remove carbonate and then dried at 60°C for 48 h. The decarbonated samples were analyzed for total nitrogen contents and their isotopic compositions in a continuous flow elemental analyzer (Carlo-Erba EA 2100) and isotope ratio mass spectrometer (Thermo Finnigan Deltaplus Advantage) system. Through this analysis, we obtain both particulate organic carbon (POC) and PN contents with precision <1% for the standard measurement.
 The nitrogen isotopic ratio (15N/14N) was calibrated to working standards and is presented by δ notation:
 Data were converted to the reference atmospheric N2 with a δ15N = 0 ‰. The long-term reproducibility of nitrogen isotope determination for the international (USGS 40, L-glutamic acid: −4.5 ± 0.1 ‰), working (Merck Acetanilide: −1.5 ± 0.2 ‰) standards and field samples is better than 0.2 ‰.
 Dissolved nitrate and nitrite (N + N) concentrations were measured by the standard pink azo dye method with a flow injection analyzer, and phosphate was determined manually by the standard molybdenum blue method [Wong et al., 2007a, and references therein]. The precisions of the determinations of (N + N) and phosphate were ±0.3 and ±0.01 μM, respectively. As for the N anomaly, N*, we followed the definition by Deutsch et al.  as N* = [NO3−] − 16 × [PO43−] + 2.9 (unit in μM).
3.1. Isotopic Compositions of Suspended Particulate Nitrogen
 A T-S (temperature-salinity) diagram (Figure 2) for observations on all the sampling occasions indicates that the water masses in the upper water column at the SEATS Station were influenced by the intruding Kuroshio Water (KW) as reported previously [Wong et al., 2007b]. The T-S field (box inFigure 2) of KW was located at a water depth of ∼250 m. The consistent T-S characteristics suggest the water column below 250 m was uniform with no seasonality.
Figure 3 presents the vertical profiles of PN concentration and δ15NPN value, as well as N + N concentrations (shown as NO3−), from the surface down to a depth of 200 m. The PN concentration profiles often showed a subsurface maximum in the upper 50 m, where the N + N concentration was at or below the detection limit in all but one case in May 2004, during which the maximal PN concentration ranged from 0.22 to 0.36 μM with an exceptionally high value of 0.6 μM. Below the subsurface maximum, all profiles showed a clear decreasing trend in PN concentration to 200 m (ranged from 0.06 to 0.14 μM). In the mixed layer, the nitrate and phosphate (not shown) concentrations were mostly below the detection limit, but increased to near 15 and 3 μM, respectively, at 200 m in all cases (Figure 3). The δ15NPN values varied from 0 to 8.0 ‰ in the euphotic zone. Our observed δ15NPN values are consistent with those reported for PN in the western SCS [Loick et al., 2007]. All except two δ15NPN profiles (August 2004 and October 2006) exhibited an increasing trend with depth below the PN maximum.
 To reveal the monsoon forcing, Figure 4 shows the time series data of the observed MLD and the daily wind speed in the vicinity of the SEATS Station (from the Reanalysis Project data of NCEP at http://www.esrl.noaa.gov/psd/data/reanalysis/reanalysis.shtml). The cyclic curve is derived by using all available MLD data collected at SEATS (see Figure 5a) including period of 1999–2003 reported by Tseng et al. . Similarly, we presented the depth integrated PN (IPN), which represents the PN inventory, and the concentration-weighted averageδ15NPN values with empirically fitted cyclic functions in Figure 4. We estimated separately for the water column at 0–100 m (hereafter top 100 m) and 100–200 m (lower 100 m). The depth of 100 m, the bottom of the euphotic zone and the lower bound of seasonal mixed layer, was set for integration since recycled production will be mainly occurring in the euphotic zone.
 The variation of the MLD (Figure 4b) mimics the seasonal pattern of the wind (Figure 4a): deepening of the mixed layer reaching 100 m occurred in winter while the strong northeast monsoon prevailed, subsequent shoaling in spring when the wind subsided, and slight deepening in summer when the southwest monsoon picked up. The high frequency variability of the wind speed was missing in the sparsely spaced time series observations, partially attributed to the damping effect of the water. The more pronounced deepening of the MLD corresponded to the winter monsoons, which are in general stronger and more persistent than the summer monsoons. Another episodic event in summer is typhoons (Figure 1), which can cause strong vertical mixing in the SCS to 200 m temporarily [Lin et al., 2003]. Thus such a cyclic MLD curve in fact represents a mean seasonal pattern of nonevent periods since all cruises were conducted under nontyphoon conditions.
 The IPN of the top 100 m shows a cyclic behavior (Figures 4c and 5b) varying between 14 and 42 mmol N m−2 with higher values in spring and autumn (Table 2) corresponding to the shallowest MLD. Compared to the top 100 m, the IPN in the lower 100 m showed considerably lower values and narrower variability (7–14 mmol N m−2). Nevertheless, the seasonal variation in the subsurface layer broadly agrees with that in the top 100 m.
Table 2. The Concentration-Weighted Averageδ15NPN and Integrated PN (IPN) of the Top 100 m and the Lower 100 m Observed From May 2004 to January 2007a
Mean δ15NPN (0–100 m)
Mean δ15NPN (100–200 m)
IPN (0–100 m) (mmol N m−2)
IPN (100–200 m) (mmol N m−2)
Values obtained from Station 5 in September 1981 [Saino and Hattori, 1987], which was located near the SEATS Station (see text), were compiled.
 The concentration-weighted averageδ15NPN in the top 100 m (Figures 4d and 5c and Table 2) varied between 2.0 and 5.3 ‰ with a mean value of 4.2 ± 1.0 ‰ of all cruises. The low values appeared mainly in spring and fall when the MLD was at its minimum. The δ15NPN values in the lower 100 m showed a similar pattern of seasonal variation but shifted consistently to a range of higher values (4.0–7.8 ‰) (Table 2). Such consistent increase in δ15NPN values in the subsurface layer is similar observations in the Pacific [e.g., Saino and Hattori, 1987; Casciotti et al., 2008] and the Atlantic [e.g., Altabet, 1988] oceans. To ensure the integrated δ15NPN in the top 100 m is representative, we also integrated δ15NPN values for water column above the nitracline, which is shown in Figure 5c for comparison. No significant difference can be observed in between the two depth integrations. Hereafter, we still use 100 m to separate the upper and lower water columns.
 POC concentrations were linearly correlated with PN concentrations at the SEATS Station with a regression slope of 5.88 (R2 = 0.92, not shown), which is close to the Redfield C/N ratio of 6.63. According to the observed C/N ratios and reported δ13CPOC values [K. K. Liu et al., 2007], we may conclude that the suspended particles in upper 200 m at the SEATS Station were mainly of marine origin with negligible contribution from terrestrial sources.
3.2. Isotopic Compositions of Sinking Particulate Nitrogen
 Nitrogen isotopic compositions of sinking particles and their PN fluxes collected by sediment traps are shown in Figure 6a. For comparison, also plotted are the C/N ratios determined for the trap samples (Figure 6b).
 The PN fluxes measured at 374 m of the M1S site varied between 2.7 and 16.7 mg m−2 d−1, and those observed at 447 m of the M2S site are slightly higher ranging from 7 to 66 mg m−2 d−1 with an unusually high peak in April 2002 deserving further discussion (see below). The fluxes were relatively high in February and March at both sites.
 The δ15N values of sinking PN (Figure 6a) collected at 374 m of the M1S site ranged from 3.3 to 6.6 ‰ with a flux-weighted average of 5.2 ± 1.1 ‰, while those at 447 m of the M2S site ranged from 4.0 to 7.6 ‰ with a flux-weighted average value of 5.5 ± 1.3 ‰. The overall distribution of theδ15N values of sinking PN showed a broad peak occurring from mid-January to mid-March. The peak roughly corresponded to the relatively higher values of the PN fluxes except for the aforementioned unusual peak (Figure 6a).
 Three exceptionally high C/N ratios of >14.2 observed at the M2S site indicating a relatively stronger terrigenous signal [K. K. Liu et al., 2007]. Following K. K. Liu et al. , the δ15N values of three samples are corrected by a simple mixing model based on the C/N ratio (6.63 for marine and 22 for terrestrial) and the δ15N value (δ15N of 5.0‰ for the terrestrial, which is the mean value of the surface sediment in the northern shelf of the SCS reported by Kienast ). The corrected δ15N values (black solid circles in Figure 6a) were shifted by 0.6 to 2 ‰. The flux-weighted average of theδ15N value in sinking PN collected at 447 m of the M2S site was shifted from 5.5 to 5.6 ‰. The correction shows insignificant influence on the overall pattern of isotopic variation of sinking PN. Nevertheless, the corrected values should better represent δ15N of marine sourced organics, therefore, were used for further discussion of particle dynamics and bioavailable nitrogen sources in the upper water column.
 It is interesting to note that the isotopic composition of the sinking PN appears to follow the seasonal trend of the suspended PN in the lower 100 m with a 1 month lag (Figure 6a). Conceivably the suspended PN in the subsurface layer represents the particulate materials that sink out of the surface layer and disintegrate to become the suspended particles, while the lag time could be attributed to reaggregation in the subsurface layer and subsequent downward transport to the sediment traps several hundred meters below. It is also noteworthy that most of the data points of the sinking particles fall below the curve of the suspended particles, suggesting that the δ15NPN in the lower 100 m represent the upper limit of the isotopic composition of the sinking PN.
4.1. Controls of Vertical Distribution of Particulate Nitrogen Isotope
 The SPM in the upper open ocean is mainly derived from phytoplankton. The nitrogen isotopic composition in SPM is controlled by many factors, among which the origin of the nitrogen used during its synthesis is perhaps the most important. Other factors include the isotopic fractionation during its formation and the subsequent degradation and alteration processes.
 At the SEATS Station, the practically absent nitrate in the surface layer all year round implies its rapid and essentially complete consumption by phytoplankton. Since the near complete consumption of nitrate causes little or no isotopic fractionation, the δ15N of the freshly produced SPM should reflect the isotopic compositions of its nitrogen sources [Altabet, 1988; Gaye et al., 2009].
 The δ15NNO3 of the subsurface water, which is the main nutrient reserve for the euphotic zone, plays a critical role in controlling the δ15N of suspended PN in the overlying surface layer. In the case of the northern SCS, the supply of 15N-replete nitrate from the subsurface nutrient reserve should be the major N source driven by various nutrient pumping processes, such as basin-wide active upwelling (90 m yr−1 [Qu et al., 2006]), monsoon-driven localized upwelling and the prevailing northeasterly monsoon in winter and episodic tropical cyclones in summer [Liu et al., 2002; Lin et al., 2003; Liu et al., 2010].
 The mean δ15NPN value (4.2 ± 1.0 ‰) in top 100 m is higher than δ15NPNreported for other low-latitude oligotrophic oceans (e.g., −0.9 to 0.8 ‰ in the Sargasso Sea [Altabet, 1988], and 0.5–1.1 ‰ at ALOHA Station [Casciotti et al., 2008]). Aside from the higher δ15NPN values, δ15NNO3 at 200–300 m at SEATS is also higher than the mean δ15NNO3 value (2.65 ± 0.32 ‰) near Bermuda [Knapp et al., 2005] or the reported 1.9 to 6.1‰ near Hawaii [Casciotti et al., 2008] at depth of 200–300 m. Unfortunately, we have only one reported δ15NO3 value of 6.2‰ below the permanent thermocline near SEATS Station [Wong et al., 2002]. Nevertheless, δ15NPN values in top 100 m are consistently lower than δ15NNO3 in subsurface layer at all three oligotrophic regions.
 On the other hand, isotope fraction during mineralization may account for the downward increasing trend in δ15NPN value. The mean of isotopic fractionation factor between top and lower 100 m is 2.2‰, which resembles estimates from lab experiments [Sigman et al., 2009, and references therein]. Preferential turnover of light nitrogen sourced from sinking particles and dissolved organic nitrogen may result in such a vertical pattern in δ15NPN [Altabet, 1988; Saino and Hattori, 1987; Knapp et al., 2011].
 Near Bermuda, particles collected by traps deployed at 100 m had higher δ15NPN values than the suspended PN from 0 to 100 m by 3.5 ‰ on average [Altabet, 1988]. Similarly, the δ15NPN of 150–500 m trapped particles were higher by 1.0–4.4 ‰ near Hawaii [Karl et al., 1997]. In our case, the particles collected by traps deployed at 374 m and 447 m had flux-weighted meanδ15N values of 5.2–5.6 ‰, which are higher than those in the top 100 m by 1.6 ‰ on average. Like most oligotrophic oceans, regenerated ammonia may play an important role to support photosynthesis at SEATS [Chen, 2005]. It has long been proposed that the regenerated ammonia via zooplankton grazing is depleted in 15N, which is corroborated by the production of 15N-enriched fecal pellets observed in laboratory experiments [e.g.,Checkley and Entzeroth, 1985]. Moreover, recent findings showed that preferential uptake and subsequent transport by specific phytoplankton community [Fawcett et al., 2011] may also induce such a downward increasing pattern.
4.2. Seasonal Variations of the Suspended PN and Its Isotopic Composition
 The annual cycle of δ15NPN variations in the two layers somewhat resembled that of the MLD variation (Figure 5) with the lowest δ15NPN values during periods of the shallowest MLD indicating that the seasonal pattern of particulate isotopic composition might be associated with nutrient supply via diapycnal mixing.
 The shallowest MLD reflects weak diapycnal mixing, thus, the regenerated production should be the highest to maintain the depression of the δ15NPN values in the top 100 m. This can be supported by the observation that f ratio (integrated new production/integrated primary production) values [Chen, 2005] in spring (0.18–0.36) and autumn (0.15–0.20) were lower than winter (0.47). Interestingly, the two lowest δ15NPN values coincide with the IPN peaks of top 100 m. It is highly unlikely that the regenerated production in spring and autumn would cause IPN peaks. It appears some other mechanisms must be added to support the observed intermonsoonal IPN highs.
 In winter the high f ratio is consistent with the physical conditions in the SCS. The f ratio of 0.47 implies that nearly half of the primary production is fueled by nitrate uptake apparently resulting from the strong supply of nitrate by regional upwelling and wind driven mixing [Liu et al., 2002]. The relatively high mean δ15NPN values (4.4–5.1 ‰) in the top 100 m in winter (Figure 5b) apparently originated from the new nitrate pumped up from the subsurface layer, where the δ15NNO3 ranges from 4.8 to 6.2 ‰ [Wong et al., 2002]. When compared with the persistently strong winter monsoons, the summer monsoons are usually weak but could be spiked with occasional tropical cyclones. The relatively wide range of the summer δ15NPN values (3.3–5.3 ‰) suggest different conditions exist for summer. The two high values were likely induced by episodic strong mixing events, such as caused by typhoons. Now the remaining question will be the origin of the additional isotopically light nitrogen to increase IPN and decrease the δ15NPN in spring and autumn.
4.3. Origin of Isotopically Light Nitrogen
 Several distinct mechanisms may provide 15N-depleted nitrogen to the surface waters in oligotrophic regions. First, fixation of atmospheric N2 (δ15N = ∼0 ‰) by marine diazotrophs is considered a significant source of new nitrogen [e.g., Karl et al., 1997; Montoya et al., 2004; Mahaffey et al., 2005]. Little fractionation occurs during N2 fixation, resulting in a δ15N of −2 to 0 ‰ for the newly fixed nitrogen [Carpenter et al., 1997], which could lower the 15N/14N ratio of particulate and dissolved species in the upper water column. Second, atmospheric precipitation generally contains inorganic nitrogen with distinctively low δ15N values [Hastings et al., 2003; Jia and Chen, 2010] and high atmospheric deposition (∼1000 mg N m−2 yr−1 or even higher) had been reported for Asian marginal seas [Duce et al., 2008].
 Quite a few studies provide geochemical evidence of widespread occurrences of N2 fixation in the SCS [e.g., Wong et al., 2002; Chou et al., 2006]. Positive N*, corresponding to lowered δ15NO3−, indicates the significant contribution from the remineralization of organics formed by N2 fixation [Wong et al., 2002]. By means of a box model of nitrogen budget at the SEATS Station, Chou et al.  concluded that N2 fixation by unicellular cyanobacteria as well as by Trichodesmium and Richelia could provide the additional bioavailable nitrogen required for net community production.
 At the SEATS Station over the 3 year observation periods our own records reveal an upward increasing trend in N* values from <0 at ∼500 m up to ∼4 μM at ∼100 m (Figure 7). It is conceivable that nitrogen fixation plays a role in introducing 15N-depleted nitrogen to the SPM pool at the SEATS Station from May 2004 to January 2007, particularly, the highest N* had occurred in May 2004 corresponding to the peak IPN and lightestδ15NPN values. The shallow MLD during the intermonsoon seasons resulted in weakened vertical transport of subsurface nitrate and favored the occurrence of nitrogen fixation. This presents a plausible mechanism to explain the depressed δ15NPN values in spring and autumn.
 Moreover, the proposed onset of nitrogen fixation in spring was in phase with the atmospheric deposition of land-derived dusts and aerosols in the northern SCS, which could have supplied the dust-carried iron to the P-replete (5–20 nM) but low available iron (∼0.2–0.3 nM) condition in the surface water [Wu et al., 2003]. The exceptionally high N* observed in May 2004 (Figure 7) reflects an exceptionally strong diazotroph bloom, which was attributable to a strong atmospheric fallout event in early April of 2004 [Hsu et al., 2008]. On the other hand, the biomass burning prevalent in surrounding landmasses of the SCS in August [Lin et al., 2007] could produce aerosols that may possibly enhance nitrogen fixation in early autumn as the summer monsoon subsides.
 We admit that the direct atmospheric bioavailable N input might contribute, yet, it is difficult to validate due to the lack of field observations on fluxes and isotopic compositions of fixed nitrogen in wet and dry depositions in SCS proper. We will discuss the potential of atmospheric N input in section 4.6.
4.4. Fate of Nitrogen Fixation Signals
 It is widely accepted that nitrogen fixation by marine organisms, such as Trichodesmium [Carpenter et al., 1997], occurs mostly in the upper water column near the surface. However, little is known about the fate of the marine diazotrophs that carry out this process. To assess the effect of nitrogen fixation on the marine nitrogen and carbon cycles, it is crucial to appreciate whether the N fixation signal is mostly exported to the deep water or remineralized in the upper layer.
 Two different pathways for the transfer of the N fixation signals were proposed by Mulholland : the N fixation signals could be transferred to higher trophic levels and transformed into large-size particulates through the food chain, thus, to enhance the sinking flux. Alternatively, the nitrogen fixers and their phytodetritus could remain mostly in the surface layer for decomposition and recycling because nonsymbiotic diazotrophs are usually not regarded as important contributors to the direct export flux of organic matter from the euphotic zone because of their small size or high buoyancy [Capone et al., 1997].
 For our investigation, nitrogen isotope signatures are used to trace the passage of the N fixation signals. Aforementioned data from Sta. Five of Saino and Hattori are also included for discussion. The seasonal variation of the concentration-weighted averageδ15NPN from the lower 100 m mimics that of the top 100 m as shown in Figure 5c implying that the surface isotopic signals have been transferred to the subsurface layer. Since the IPN inventory in the top 100 m reveals stronger seasonal patterns compared with IPN of 100–200 m, the downward transfer efficiency varies seasonally and the efficiency can be evaluated from the deviation in IPN between the two layers. From Figures 5b and 5c, we notice that the δ15NPN for both layers correlates negatively with IPN in the top 100 m. The highest IPN values correspond to the lowest mean δ15NPN values and may imply high contribution from N2 fixation. A high IPN probably suggests that the phytodetritus have a longer residence time in the upper water column due to sluggish sinking of the diazotrophs from the surface layer to the subsurface layer.
 Diazotrophs in the SCS include all forms of N2 fixers, filamentous (e.g., Trichodesmium), unicellular and symbiotic cyanobacteria [Chen et al., 2008; Chou et al., 2006]. Aside from the symbiotic diazotrophs, the rest do not make it into the sinking particles readily. Trichodesmium, likely the most abundant among the three types, are grazed by harpacticoid copepods, which appear tolerant to the often toxic Trichodesmium species [O'Neil and Roman, 1994]. An important species of such copepods, Macrosetella gracilis, was found in significant amounts in the surface Kuroshio water and the abundance of nauplii was found positively correlated to the Trichodesmium chlorophyll concentration (T. Suzuki et al., personal communication, 2011). Since the Kuroshio water intrudes into the SCS periodically [Shaw, 1991], it is very likely that these specialized copepods also exist in the SCS in quantities that are capable of gazing on Trichodesmium colonies intensively. Such grazing could quickly convert the newly fixed nitrogen into ammonia either by fast metabolism or by sloppy feeding [O'Neil et al., 1996]. The latter could also produce slow-sinking phytodetritus instead of fast-sinking fecal pellets. The released ammonium could induce quick growth of picoplankton, which is the dominant phytoplankton in the SCS [Chen, 2005; H. B. Liu et al., 2007].
 In Figure 8, we plot observed IPN values for both upper and lower layers against the δ15NPNin top 100 m layer. The two negative trends converge at the high-endδ15NPN during which upwelling and the primary productivity are intensified. The phytoplankton composition may be regarded as the background community dotted with modest blooms of diatoms, which occur when nutrient pumping is more effective [Chen, 2005; Liu and Chai, 2009]. “Normal” food chain activity is in operation such that the primary producers are grazed down and converted into fecal pellets that sink out of the euphotic zone relatively quickly. This explains the reported high primary productivity yet lower particle inventory in euphotic zone in winter. By contrast, at the high end of IPN the upper 100 m holds more 14N-replete SPM due to weaker downward transfer efficiency.
 The downward transport of the N fixation signals is further explored by examining the sediment trap data. The δ15N values of sinking particles collected by sediment traps at 374–447 m below surface showed significant seasonal contrast over the range from 3.3 to 7.6 ‰, with high values in winter and low values in autumn and spring. As mentioned above, the pattern is somewhat similar to the temporal variation of δ15NPN in the overlying subsurface layer (100–200 m), except that occurrences of the high values lagged behind those in the overlying water column by a month or so (Figure 6). Since the δ15N values of sediment trap samples have been adjusted to remove the strong terrigenous signals, the seasonal contrast most likely originated from seasonal variation of N2 fixation. In other words, the signals of N2 fixation have been transferred to the sediment trap samples, despite the rather low transfer efficiency in the surface layer. This is supported by the observation that Trichodesmium specimens were found in our trap at the depth of 374 m of the M1S site [Chen et al., 2003]. The average flux of Trichodesmium, estimated to be 3.6 × 103 trichomes m−2 d−1 during the deployment period, however, was rather small. No trichomes were ever detected in samples from the deeper traps illustrating the limited downward transport of Trichodesmium. However, once the diazotrophs are broken down and decomposed, the signal of the regenerated nitrogen should be passed on to other phytoplankton that may undergo the normal food chain sequence so that the signal will be transferred to the deeper water column.
4.5. Assessment of N Fixation Rate Using Isotope Balance
 To estimate the relative contributions of nitrate from upwelled deep water and nitrogen fixation in supplying newly combined nitrogen to the new production, we followed a similar approach adopted in other oligotrophic oceans [e.g., Karl et al., 1997; Bourbonnais et al., 2009]. In steady state, the export of sinking PN through 410 m should balance to the sum of the upwelled nitrate and newly fixed nitrogen. Thus, the mass and isotope balances are formulated:
where ΦEX, ΦNO3, and ΦN2 represent, respectively, the fluxes of the sinking PN at 410 m, the upwelled nitrate and N2 fixation, and δ15NEX, δ15NO3− and δ15N2 denote the respective nitrogen isotopic compositions.
 Instead of calculating for each data point, we use the average values during the deployment periods, which should better reflect the average steady state under which the mass and isotope balances are achieved. We assumed the δ15N values of the two influxes of nitrogen (δ15NO3− and δ15N2) to be 6.1 ‰ and −0.9 ‰, respectively, following those used by Liu et al. . The choice of 6.1 ‰ for the δ15N value of supplied nitrate is supported by observations of a ∼300 m lifting of North Pacific Intermediate Water (NPIW) when the Kuroshio Current intrudes the South China Sea [You et al., 2005]. This isopycnal uplifting induces a shallow thermocline, thus, the δ15N values of nitrate observed in the 200–300 m layer in the northern SCS [Wong et al., 2002] are almost identical to those observed at a depth of ∼600 m in the Kuroshio Current northeast of Taiwan [Liu et al., 1996]. For the sinking PN, the δ15N (δ15NEX) is assumed to be the flux weighted mean δ15N (5.6 ‰, horizontal arrow in Figure 6a) observed from the two traps deployed at 374 and 447 m in the northern SCS. The average sinking flux of PN collected by these two traps (11.5 ± 14.6 mg N m−2 d−1 or 300 ± 380 mmol N m−2 yr−1) is taken as the value for ΦEX. An independent approach based on the POC-ballast relationship is used to estimate the value of ΦEX [Armstrong et al., 2002; Kao et al., 2006] to avoid the bias due to incomplete annual coverage of the sediment trap sampling. Surprisingly, this independent approach gives nearly identical value (∼258 mmol N m−2 yr−1), suggesting the observed mean flux of sinking PN to be representative. By solving equations (2) and (3), the flux of N2 fixation is estimated to be 21 ± 26 mmol N m−2 yr−1.
 Compared to previously reported N2 fixation rate (5.4 ± 10.6 mmol N m−2 yr−1) based on Trichodesmium counts in the northern SCS [Chen, 2005], our estimated rate was 4 times higher. However, the wide range of individual observations (0.3 to 40 mmol N m−2 yr−1) indicates the nature of the sporadic and patchy occurrences of Trichodesmium. Our estimate is similar in magnitude within the range of actual occurrences in this region.
 This estimated N fixation rate represents ∼5–10% of the reported total new nitrogen requirement or 1.5–3.3% of the nitrogen needed for primary production [Chen, 2005]. This is lower than the estimates of the contribution to the nitrogen needs of primary production (3–13%) based on nitrate anomalies observed at the SEATS Station [Wong et al., 2007a]. In Wong et al. [2007a], they use N* as a proxy for the imbalance between N and P supply to the upper ocean of 150 m. This method represents a longer-term cumulative contribution from N fixation to nitrate inventory in the upper 150 m water column. By contrast, our estimate focuses on the fraction of the downward transfer of N fixation signal through the sampling water depth of 410 m. Results for N fixation input from this kind of box model are depth dependent. A higher fractional contribution will be derived if bottom was set at 100 m deep since a portion of N fixation input is recycled.
 In addition, Gaye et al.  reported that nitrogen fixation could contribute up to 20% of sinking PN through 2000 m in the SCS basin, which was significantly higher than our estimate. It is worth mentioning that the PN flux for the deeper trap is larger than the flux from shallow traps in Gaye et al. . Though terrestrial PN signals had been eliminated, they did not consider the export production induced by riverine nitrate discharge, which equals the amount of N fixation input on a basin scale [Liu et al., 2010]. Basing on our observations, we speculate that N fixation signal can reach such deep water giving a high fractional contribution in on a basin scale. We hypothesize that the lateral PN (marine PN carried 15N-depleted signal from riverine nitrate/ammonium) transport from shelf to deeper layer in winter may play a role, especially because the shallow shelf occupies >1/3 of the basin area receiving huge amount of freshwater input. On the other hand,δ15N values for the ∼400 m sediment trap in our study is higher than the δ15N values for materials collected at deep water depths [Gaye et al., 2009; J. Y. T. Yang, unpublished data, 2006]. Many studies have observed a decreasing trend in the δ15N of sinking PN with depth [e.g., Altabet et al., 1991]. Some unknown alteration processes (either vertically or laterally or both) might be responsible for this deviation in δ15N between deep and shallow sinking particles.
 Our estimate is nearly the same as that (20 mmol N m−2 yr−1) estimated by Liu et al.  for the northern SCS, similar to that (15 mmol N m−2 yr−1) reported for the BATS Station [Orcutt et al., 2001] and slightly lower than those (31–51 mmol N m−2 yr−1) reported for the Hawaii Ocean Time series Station [Karl et al., 1997]. Moreover, a similar range for the nitrogen fixation rate (5–42 mmol N m−2 yr−1) was observed in the upwelling region off Vietnam in the western SCS [Voss et al., 2006]. It is reasonable to assume a similar range of N fixation rates occurs in the entire SCS basin, resulting in a calculated nitrogen fixation flux of 0.56 Tg N yr−1.
 Growing evidence indicates that islands and marginal seas adjacent to China receive significant atmospheric N deposition (∼1000 mg N m−2 y−1 or even higher) sourced from inland China emission [Duce et al., 2008; Zhang et al., 2011]. Such a large atmospheric deposition, in fact, is ∼3x our estimate of new nitrogen from N fixation and of the same magnitude as the nitrate supply from upwelling. The reported annual mean values of δ15N for NO3− and NH4+ from coastal area of South China, 300 km north of the SEATS Station, were as low as −0.7 ‰ and −7.3 ‰, respectively [Jia and Chen, 2010]. If atmospheric 15N-depleted N deposition sourced from inland China can really reach SCS proper, the isotopically light nitrogen should be brought by the winter northeasterly monsoon wind to lower theδ15NPN values significantly at SEATS. However, the δ15NPN values in the winter monsoon season approach the δ15N value of upwelled nitrate. Alternatively, if N deposition from the atmosphere is high enough and follows the seasonal precipitation pattern [Liu et al., 2010] of monthly rainfall from May to September, then, the 15N-depleted atmospheric N deposition indeed broadly explains (except cyclone disturbance) the seasonally lowδ15NPN.
 However, the atmospheric deposition of isotopically light bioavailable nitrogen does not explain the observed negative correlation between downward transfer efficiency and δ15NPN in the surface layer, which is presumably caused by diazotrophs and associated ecosystem responses. Meanwhile, a higher abundance of diazotrophs was observed from spring to autumn, which indicates that the surface ocean is N limited. Based on the mass balance of the N budget, Liu et al. reported a dry-wet deposition of N to be ∼100 mg N m−2 yr−1, which is much smaller than model predictions and would account for around one third of our estimated N fixation flux. Further observations on the seasonal pattern of atmospheric deposition in the SCS at the basin scale are urgently needed.
 On the other hand, our model calculation is prone to uncertainties, which are directly related to the reliability of the values used in the calculation. One of the most sensitive variables is the mean δ15N value of the sinking particles as emphasized by Knapp et al. . In our case, the sediment trap samples covered 7 months of a year. Better temporal coverage should improve the estimation revealing seasonal variations in N fixation. However, the missing months are from May to September, which is the period of relatively low productivity, and, therefore, may not cause too big a bias as far as the flux-weighted mean value is concerned.
 As in most models, we neglect the effect of DON pool, which represents the largest fixed N pools in the oligotrophic surface layer [Knapp et al., 2005]. A significant flux of low δ15N-DON derived from N fixation could export out of surface waters [Mulholland et al., 2006], although, many studies argued that a large fraction of DON is recalcitrant [Knapp et al., 2005] and not responsive to the sporadic events of N fixation [Knapp et al., 2011]. On the other hand, we prescribed a steady state for this model which might not be realistic; accumulation of low δ15N-DON in surface might cause uncertainties. Apparently, DON concentration and its isotopic composition need to be well studied in order to construct a complete mass balance of fixed N pools in the SCS. Last, the contribution if any from lateral transport of SPM with unknown isotopic composition to sediment trap at ∼410 m might also cause uncertainties.
 We first present nitrogen isotopic data of SPM at SEATS Station in the SCS. The concentration-weighted averageδ15N values of the SPM (mean δ15NPN) in the upper 100 m is 4.2 ± 1.0 ‰, which are higher than those observed in other low-latitude oligotrophic oceans, ascribed probably to a stronger supply of15N-replete nitrate sourced from the deeper waters and driven by various nutrient pumping processes. The meanδ15NPNvalues in the upper 100 m show significant seasonal patterns strongly correlated to mixed-layer depth. Strong wind during the winter monsoon deepened the mixed layer and brought up more15N-replete NO3 into the euphotic zone from the thermocline, whereas lower δ15NPN values correspond to a shallow MLD in the intermonsoon seasons (spring and autumn). A persistent downward increasing in δ15NPN was observed throughout the observation period implying that some biologically mediated preferential downward transfer of isotopically heavy nitrogen, preferential ammonification of 14N in the top 100 m and/or an addition of 15N-depleted N from the atmosphere are required to facilitate such verticalδ15NPN gradient. A good correlation between the mean δ15NPN below 100 m and that in the upper 100 m was found implying that the nitrogen isotope signals in the euphotic zone have been transferred to the subsurface layer. The smallest gradient in IPN (decreasing downward) and δ15NPN (increasing downward), which implies a strong transfer efficiency, appeared in winter time with abundant nutrient supply. By contrast, the largest gradient in the IPN (weak transfer efficiency) was accompanied by low δ15NPN, high N* and relatively high N fixation activities. Accordingly, we suggest nitrogen fixation superimposed onto biological fractionation lowers the δ15NPN values in the euphotic zone and weakens the transfer efficiency during periods with shallow MLD due to sluggish sinking of the diazotrophs.
 On the other hand, the δ15N values of sinking particles collected by sediment trap at depths of 374 and 447 m (mean of 410 m) follow a similar temporal pattern relative to that of δ15NPN in the overlying water column with a lag time of 1 month. The δ15NPNvalues of trapped material range from 3.3 to 7.3‰ with a flux-weighted mean of 5.6‰, which is slightly deviated from theδ15NO3of the upwelled source. By using a one-dimensional mass and isotope balance model for upper 410 m, we obtained the N fixation input of ∼20 ± 26 mmol N m−2 yr−1 under the assumption of steady state and no atmospheric N deposition. This model derived N fixation input accounts for only ∼5–10% of the reported new production at SEATS on an annual basis. Similar values of N fixation flux were presented at BATS [Orcutt et al., 2001] and HOTS Station [Karl et al., 1997]. Further investigation on isotopic values are needed for nitrate and sinking PN in subsurface layer at SEATS Station to provide a more reliable estimate. Moreover, the atmospheric deposition could also contribute significant amounts of new N to the ocean [Duce et al., 2008], so a clear demand for future work is the measurement of δ15N in order to identify the relative importance of the different N sources to the marine ecosystem.
 Supported by Taiwan (NSC 98-2116-M-001-005, Academia Sinica Thematic Program AFOBi) and China (973 Program, 2009CB421200, and the program of Introducing Talents of Discipline to Universities, B07034).