Using 228Th/228Ra Disequilibrium to Constrain Particulate Organic Carbon Export From the Upper Twilight Zone of the Northern South China Sea

We utilized the 228Th/228Ra and 234Th/238U disequilibrium methods to estimate the export of particulate organic carbon (POC) from the upper twilight zone in the northern South China Sea during four cruises from August 2009 to May 2011, covering an entire seasonal cycle of spring, summer, autumn, and winter. A significant finding of this study is that 228Th/228Ra disequilibrium method is highly sensitive for tracing POC export in the twilight zone, as demonstrated by the comparison of the “Minimum Detection Limit” (MDL) of various methods. We thus further conclude that 228Th/228Ra disequilibrium is suitable for the research of the twilight zone by comparing the POC export fluxes derived from 228Th/228Ra and 228Ra‐NO3− disequilibrium. The mesopelagic transfer efficiency (Teff), defined as the ratio of POC flux at 500–100 m, varied from 8% to 130%. The Teff derived from >53 μm 234Th (228Th) was higher than that based on 1–53 μm. Large particles were observed to sink rapidly and exit the twilight zone. The Teff was lower at the sampling station influenced by the Kuroshio waters in winter and spring, where the exported organic matter was relatively labile and prone to remineralization in the twilight zone. Conversely, the Teff was relatively higher at other stations, indicating that the exported organic matter was refractory and underwent comparatively less degradation in the twilight zone.


Introduction
The transport of particulate organic carbon (POC) from the euphotic zone to the deep ocean plays a crucial role in the ocean's biological pump, which has implications for atmospheric CO 2 levels (Volk & Hoffert, 1985).Global estimates suggest that the biological pump exports over 10 gigatonnes of carbon per year from surface waters (Boyd & Trull, 2007;Laws et al., 2000).However, this export significantly diminishes in the mesopelagic waters known as the "twilight zone," which extends from the euphotic zone to about 1,000 m (Buesseler et al., 2007).The twilight zone is responsible for about 90% of organic material mineralization, making it crucial for the transfer of biogenic carbon from the surface to the deep ocean (Tréguer et al., 2003).Understanding the mechanisms governing carbon export and its attenuation in the twilight zone is essential for assessing the impact of climate change on carbon sequestration in the ocean.
Despite the significant attenuation of particle export in the twilight zone, the depth region has been understudied for a long time.Technical challenges, such as sampling a large volume of water, have hindered our understating of particle transport efficiency in the twilight zone.Recent advances in sediment trap techniques have started to fill this research gap.Two notable projects, VERTIGO (VERtical Transport in the Global Ocean) and MedFlux, have investigated the processes controlling particle transport efficiency in the twilight zone.The VERTIGO project utilized a Neutrally Buoyant Sediment Trap (NBST) to observe regional variability in particle export attenuation (Buesseler et al., 2007).The MedFlux project used an Indented Rotating Spherical Sediment Trap (IRS) to measure carbon and other chemical fluxes in sinking particles of various velocities, testing the "ballast In order to assess the temporal variability and controls on POC export rates in the upper twilight zone of the South China Sea, we conducted four survey cruises from 2009 to 2011, covering all seasons.We present results of dissolved, size-fractionated particulate and total 234 Th and 228 Th, and 228 Ra measurements from the upper 500 m of the South China Sea.We utilized the disequilibrium between 228 Th ( 234 Th) and its parent nuclide, 228 Ra ( 238 U), to estimate POC export in the upper 500 m in the South China Sea.The basis of the 228 Th ( 234 Th) method for quantifying the upper ocean POC export is relatively straightforward: the deficit of 228 Th ( 234 Th) with respect to 228 Ra ( 238 U) in seawater is mainly due to removal by scavenging onto sinking particles.POC export rates can be calculated by multiplying the 228 Th ( 234 Th) flux by the POC/ 228 Th (POC/ 234 Th) ratio on sinking particles at the desired depth horizon.Furthermore, we compared POC fluxes derived from various methods, including 234 Th/ 238 U disequilibrium, 228 Th/ 228 Ra disequilibrium and 228 Ra-based nutrient budgets (Cai, Huang, Chen, Guo, et al., 2002;Ku et al., 1995).This comparison provides an opportunity to examine the applicability of 228 Th/ 228 Ra disequilibrium as a tool for tracing POC export and offers insights into the processes controlling POC export attenuation in the twilight zone of the northern South China Sea.

Study Sites and Sampling
Samples in this study were collected from the northern South China Sea from 18 July to 16 August 2009 (summer), 6-30 January in 2010 (winter), 26 October to 24 December 2010 (autumn) and 30 April to 24 May 2011 (spring).The sampling locations are shown in Figure 1 and the detailed sampling information is given in Table 1.

234 Th and 238 U Analyses
Analysis for total 234 Th in 4-L samples was based on the small volume MnO 2 co-precipitation technique (Buesseler et al., 2001;Rutgers van der Loeff et al., 2006) following the procedure modified by Cai et al. (2006b).Particulate samples collected on the 53 and 10 μm pore size Nitex screens using in situ pumps were re-suspended by ultrasonication in filtered seawater and re-collected on 47-mm 1.0 μm QMA filters.The 142-mm QMA filters as well as the 47-mm QMA filters were dried overnight, and a 22-mm subsample was cut from each filter and prepared for beta counting of 234 Th.The remaining QMA filter material was used for the determination of natural 228 Th.
Beta counting for the total and the size-fractionated particulate 234 Th samples was performed on board with gas-flow proportional beta counters (RisØ GM-25-5).Each sample was beta-counted 5 times over a period of >150 days in order to follow the decay of 234 Th.After beta counting, the particulate samples were then used for POC measurements.The total 234 Th samples, meanwhile, were processed for recovery assessment of 230 Th on the MnO 2 precipitates using Inductively Coupled Plasma Mass Spectrometry (ICP-MS).Briefly, we dismounted the samples and spiked with 229 Th solution, adding 10 mL 8 mol L −1 HNO 3 to digest samples.Samples were purified through classic anion column according to the procedure for 228 Th analyze and then measured using ICP-MS.All 234 Th data were recovery-corrected and decay-corrected to the time of collection and reported with a propagated uncertainty that includes the standard uncertainty associated with fitting the decay curve of 234 Th and the 1 sigma counting uncertainty from the recovery measurements.Uranium-238 activities were estimated from salinity measurements using the relationship of 238 U (dpm L −1 ) = 0.0686 × salinity × density (J.H. Chen et al., 1986).In the open ocean, this relationship is thought to hold within ±<3% (Pates & Muir, 2007).

228 Th and 228 Ra Analyses
To minimize the error induced by the 228 Th ingrowth correction, MnO 2 cartridge samples were processed immediately after the cruise. 230Th was added as a spike and the cartridges were leached according to the protocol of Cai et al. (2006a).Thorium nuclides in the leachate were co-precipitated with Fe(OH) 3 .The supernatant was separated from Fe(OH) 3 precipitate by centrifuging and saved for 228 Ra analysis (Luo et al., 1995).Thorium isotopes in the Fe(OH) 3 precipitate were isolated via ion-exchange chemistry according to the classical procedure by Anderson and Fleer (1982).As for the analysis of size-fractionated particulate 228 Th, 230 Th was added as a yield tracer to the remaining QMA filters mentioned above.Subsequent radiochemical purification of Th isotopes was also according to Anderson and Fleer (1982).Finally, Th was extracted into a 0.25 mol L −1 Theonyl trifluoroacetone (TTA)/benzene solution, and evaporated onto a stainless steel disc.The discs were counted by alpha spectrometry in ultra ion-planted detectors (Octête™ PC) until a minimum of 500 counts of 228 Th alpha was reached.The 228 Th peak was corrected for the underlying 224 Ra activity by the analysis of the dominant 224 Ra alpha at 5.7 MeV.Blank activity in the QMA filter was determined to be <2% of the particulate 228 Th activities.Dissolved 228 Th activities were corrected for the ingrowth from 228 Ra according to where λ 1 , λ 2 are the decay constants for 228 Ra and 228 Th, respectively, A 0 is the 228 Ra activity on the first MnO 2 cartridge at the time of sample collection, and t is the time elapsed from sample collection to the separation of 228 Ra and 228 Th.The cartridge collection efficiencies for 228 Th were determined through the comparison of 234 Th activities on the first MnO 2 cartridge and those based on the small volume MnO 2 co-precipitation technique (Cai et al., 2006a).All the dissolved and particulate 228 Th activities were decay corrected to the time of collection and reported with an error propagated from ±1σ counting statistic, the uncertainty associated with cartridge collection efficiency (for the dissolved phase) as well as the error in yield tracer calibration.

228
Ra in the aforementioned supernatant were measured by analyzing the 228 Th daughter approximately 1 year after the separation of 228 Ra from Fe(OH) 3 precipitation (Lepore & Moran, 2007;Luo et al., 1995).Radium collection efficiency on the MnO 2 cartridge pair was assumed to be unity, though the collection efficiency on a single MnO 2 cartridge has been demonstrated to be lower than 100% in the GEOTRACES radium inter-calibration program (organized by M. Charette).With this assumption, the final 228 Ra activities were calculated as the sum of 228 Ra measurements on the first and second MnO 2 cartridges.In order to verify this assumption, in parallel to the deployment of the in-situ pump, we have collected two large-volume (100 L) samples from the surface mixed layer at SEATS and TS1 in August 2007 using the onboard intake system. 228Ra activities using the MnO 2 cartridge method at SEATS0807 and TS10807 were 124 ± 7 and 107 ± 6 dpm m −3 .The samples were then passed through a column of MnO 2 -coated acrylic fiber at a flow rate of <1 L min −1 to quantitatively remove Ra isotopes. 228Ra activity of the samples was measured using a well-type germanium detector (ORTEC, GWL-120-15-S).Results showed that the 228 Ra activities in the surface mixed layer at SEATS0807 and TS10807 were 122 ± 18 and 140 ± 13 dpm m −3 respectively (W.Chen et al., 2010), in excellent agreement with the measurements using the MnO 2 cartridge method.

POC and Analyses
POC concentrations of the size-fractionated particle samples that have been beta-counted were determined with a PE-2400 SERIES II CHNS/O analyzer according to the JGOFS protocols (Knap et al., 1996).All samples were treated via acid fuming to remove the carbonate phase.Each sample was corrected for a C blank.The C blank of the filter was less than 6 μg C, which on average accounted for ∼5% of the POC on the QMA filters.Based on replicate analyses, the precision for the POC determination was better than 10%.We thus assigned an error of ±10% to POC measurements when propagating the uncertainty associated with POC/Th ratios (see Section 4.2).NO 2 − and NO 3 − were analyzed onboard according to classical colorimetric methods with a Technicon AA3 Auto-Analyzer (Bran-Lube).The detection limits for NO 2 − and NO 3 − were 0.02 and 0.07 μmol L −1 , respectively.

Results
Size-fractionated 234 Th, 228 Th as well as 238 U and 228 Ra activities in the upper 500 m at the sampling stations are presented in Table 1.Listed in Table 2 are size-fractionated organic carbon contents and nitrate concentrations in the ambient seawater.

Depth Distributions of 234 Th
Overall, the deficit of total 234 Th with respect to 238 U was confined to the upper 100 m (Figure 2).On average, the total 234 Th/   Note that this phenomenon is much more prominent compared to the small size particles.Below 150 m, 234 Th activities on the medium and large particles were low and uniform (Figure 2).

228
Ra activities in the surface mixed layer at the three sampling stations varied between 33.6 and 111.6 dpm m −3 (Table 1), slightly lower than those reported in the southern South China Sea and the western South China Sea (Cai, Huang, Chen, Liu, et al., 2002;W. Chen et al., 2010;Nozaki & Yamamoto, 2001).This is probably due to the intrusion of 228 Ra-depleted waters from the West Pacific (Nozaki et al., 1998) to the study region.Notably low 228 Ra concentrations were measured at A1 in winter (2010 01) and spring (2011 05), with surface 228 Ra activities of 39.7 and 35.5 dpm m −3 respectively.This was mostly due to the Kuroshio waters intrusion.Figures 3a and 3b show that the T-S diagrams were closer to the Kuroshio waters in winter and spring comparing with another seasons.Depth profiles of 228 Ra showed a sharp decrease between 25 and 75 m at all the sampling stations (Figure 4).For 228 Ra, such a vertical gradient in the thermocline is a common feature in the open ocean.High 228 Ra activity in the surface mixed layer is considered to be caused by the diffusion of 228 Ra from shelf sediment or desorption from river-born suspended particulate matters during estuarine mixing (Moore, 1969;Rutgers van der Loeff et al., 2003).Compared to the upper 100 m, 228 Ra activities below 150 m generally declined more slowly with depth.At 500 m, 228 Ra activities converged at a level of ∼4-6 dpm m −3 (Table 1 and Figure 4).Dissolved 228 Th activity varied between 2.2 and 16.6 dpm m −3 at the sampling stations (Table 1).Like 228 Ra, it generally decreased with depth (Table 1). 228Th activity on small size class (1-10 μm) particles varied between 0.40 and 1.73 dpm m −3 .Except at A1, depth distributions of small size particulate 228 Th showed a slight reduction between 75 and 200 m (Table 1).This depth pattern can be explained as the result of the change of 228 Th supply and particle removal rates in the water column. 228Th activities on medium and large particles fell in the range of 0.04-0.52 and 0.07-0.62dpm m −3 .On average, they accounted for 2.7% and 4.0% of the total 228 Th, respectively.The larger fraction of 228 Th on particulate phases relative to 234 Th is in accordance with previous finding of the significant 234 Th decay during particle coagulation and aggregation (Cai, Dai, Chen, et al., 2006).
To eliminate the effect of the change in supply rate, we plotted the 228 Th/ 228 Ra ratio against depth (Figure 5).Size-fractionated 228 Th/ 228 Ra ratios, including the dissolved phase, tended to follow a similar depth pattern.Overall, 228 Th/ 228 Ra ratios on all the phases increased steadily with depth, except at around 300 m at station A1 in summer and autumn, where reductions in 228 Th/ 228 Ra ratio were observed.These reductions were mainly due to the increase in 228 Ra.Similar to 234 Th, total 228 Th in the upper 100 m was in deficit with respect to 228 Ra.Below 200 m, 228 Th generally reached secular equilibrium with 228   228 Th was also in deficit with respect to 228 Ra.Overall, the deficit of 228 Th relative to 228 Ra was much larger compared to 234 Th versus 238 U.This is due to the fact that 228 Th has a longer half-life, such that the supply rate of 228 Th in the water column is lower than its scavenging rate onto sinking particles.

Depth Distributions of Nitrate and POC
Nitrate was depleted in the upper mixed layer at all the sampling stations (Figure 4).Below the upper mixed layer, it increased steadily with depth, to around 30 μmol L −1 at 500 m.Bottle POC concentrations at the sampling stations ranged from 0.35 to 4.88 μmol L −1 (Table 2).POC concentration on the small size particles ranged from 0.11 to 1.94 μmol L −1 , whereas those 10 of 29 measured on the medium and large size particles varied between 0.008 and 0.155 μmol L −1 , and between 0.010 and 0.327 μmol L −1 respectively.On average, POC concentrations in the medium and large particle size classes accounted for 13% and 11% of the total POC respectively.For all particle size classes, POC concentrations were high in surface and low in bottom generally (Figure 6).Bottle POC showed a maximum in the depth of 50 m, where occurred a Chl a maximum (Figure 7).We saw the POC concentrations seasonal variations.The euphotic zone was more evident than the twilight zone.
The depth distributions of NO 3 -based POC flux showed that it increased from ∼0 mmolC m −2 d −1 in the surface to a subsurface maximum between 50 and 100 m.

The "Minimum Detection Limit" of Different Methods for Determining POC Export in the Twilight Zone
Before comparing the POC fluxes derived from different methods, we need to examine the uncertainty associated with each individual method.This is particularly important for 234 Th-derived POC export, as in an oligotrophic sea where 234 Th deficit is small, the uncertainty with 234 Th flux tends to be large since 234 Th flux is determined from the difference of two large numbers.In this regard, Cai et al. ( 2010) introduced a concept of the "Minimum Detection Limit" (MDL) to POC export studies using 234 Th/ 238 U disequilibrium.In doing so, the MDL is defined as 3σ uncertainty with 234 Th flux calculation multiplied by POC/ 234 Th ratio at the export horizon, that is, Using this definition, we have calculated an average MDL of 0.5-1.8mmolC m −2 d −1 for 234 Th method at the export horizon of 100-500 m (Table 4).As shown in Table 4, the MDL of 234 Th method tended to increase with depth, and below 200 m the MDL of 234 Th method >1 mmolC m −2 d −1 .This suggests that 234 Th/ 238 U disequilibrium is not a sensitive method for tracing POC export in the twilight zone.
We can also apply the concept of the MDL to 228 Th method in a manner similar to 234 Th method.This resulted in an average MDL of 0.1 mmolC m −2 d −1 at the export horizon of 100-500 m, which is much lower than the MDL of 234 Th method (Table 4).At 500 m, the MDL of 228 Th method was about an order of magnitude lower than that of 234 Th method.The low MDL of 228 Th method is predominantly due to the small uncertainty associated with 228 Th flux calculation.This makes 228 Th/ 228 Ra disequilibrium an extremely sensitive method for tracing POC export in the twilight zone.Indeed, even at the deepest export horizon (500 m), the signals of POC export were still about an order of magnitude higher than the MDL of 228 Th method (Table 4).For the coupled 228 Ra-NO 3 − method, it may not be appropriate to use a concept of the MDL as the uncertainty associated with the NO 3 − -based POC flux depends mainly on the standard error from the regression of depth profile of 228 Ra.Compared to the uncertainties associated with the measurements of 228 Ra, 228 Th and nitrate concentrations, the standard error from the regression of 228 Ra profiles was generally high and more variable.3, the NO 3 -based POC fluxes were of 0.7-6.0mmolC m −2 d −1 at the export horizon of 100-500 m, and the POC fluxes derived from 228 Th/ 228 Ra disequilibrium varied from 0.3 to 5.2 mmolC m −2 d −1 .The results were consistent between the two methods at the export horizon of 100-500 m only a few stations are not (Figure 12).We thus further concluded that 228 Th/ 228 Ra disequilibrium is suitable for the research of the twilight zone.The NO 3 -based POC fluxes were higher than the 228 Th-derived POC fluxes at station A10 in summer and at A1 in winter and spring.It is most likely due to  3).On the contrary, 228 Th is probably not in a SS as its parent 228 Ra has large seasonal variability.For instance, 228 Ra had higher-than-normal concentration in the summer of 2009, whereas the 228 Th didn't have enough time to reach transient equilibrium with 228 Ra, yielding an overestimated disequilibrium between 228 Ra and 228 Th as well as 228 Th-derived POC flux.The opposite trend was observed in summer 2009, where the 228 Th-derived POC fluxes were consistently higher than the 234 Th-derived ones.
A likely cause of the difference in POC fluxes derived from the two methods is the assumptions of a SS condition and negligible effect of physical processes.These assumptions can be directly tested by comparing the residence times of 234 Th and 228 Th.Kaufman et al. (1981) proposed that under SS conditions and when physical processes (advective and diffusive terms) can be ignored, the residence time of 234 Th should equal the residence time of 228 Th.In this circumstance, 234 Th/ 238 U and 228 Th/ 228 Ra ratios can be related by where R 234 and R 228 are the daughter-parent activity ratios 234 Th/ 238 U and 228 Th/ 228 Ra, respectively.Conventionally, Equation 11 is illustrated as a line of concordant scavenging of 234 Th and 228 Th, and the steady-state assumption is evaluated by comparing the measured daughter-parent activity ratios to the concordant line (Kaufman et al., 1981;Lepore & Moran, 2007).When our 234 Th/ 238 U and 228 Th/ 228 Ra data are treated in the same manner, we found that most data points cluster around the line of concordant scavenging (Figure 14).Meanwhile, from the line of concordant scavenging, we noted that when 234 Th/ 238 U ratio is >0.8, 228 Th/ 228 Ra ratio becomes very sensitive to the change of 234 Th/ 238 U ratio.As such, for an oligotrophic ocean where 234 Th/ 238 U ratio is typically >0.8, the steady-state assumption can not be readily evaluated by comparing the measured daughter-parent activity ratio to the concordant line given that a small error in the measured 234 Th/ 238 U ratio would lead to considerable change in the concordant 228 Th/ 228 Ra ratio.In order to circumvent this difficulty, we propose to calculate the concordant 234 Th flux at the export horizon of interest from the measured 228 Th/ 228 Ra ratios using Equation 11, and then compare it to the measured 234 Th flux (Figure 15).Since 234 Th flux integrates 234 Th deficit from the surface to a certain depth, any small difference between the concordant 234 Th/ 238 U ratio and the measured 234 Th/ 238 U ratio that can not be readily seen from the line of concordant scavenging would be magnified in the 234 Th flux calculation.Indeed, the plot of concordant 234 Th flux versus measured 234 Th flux clearly demonstrates that most of the data points fall below the 1:1 line (Figure 15).The extreme case is in summer 2009, when the concordant 234 Th flux can be >2 times higher than the measured flux.The higher concordant 234 Th      sediment traps at water depths >1,000 m in the central South China Sea.Their results were back-calculated using the "Martin Equation Martin et al. (1987)" to derive a POC export estimate at 100 m, which ranged from 1.0 to 3.3 mmolC m −2 d −1 during the deployment period.In general, our results are in good agreement with these previous studies.According to the recent compilation of POC export flux in the world oceans, our results are also consistent with those derived from the oligotrophic North Pacific and the Sargasso Sea (Cai et al., 2010).This suggests that the northern South China Sea is an oligotrophic sea characterized by low shallow POC export.
In this study, the POC export fluxes at 500 m derived from different methods varied between 0.8 and 2.9 mmolC m −2 d −1 (Table 4).There are no historical measurements of POC export at 500 m in the South China Sea for direct comparison.In VERTIGO, POC export fluxes from 150 to 500 m were measured using NBST at two contrasting sites in the Pacific: ALOHA and K2 (Buesseler et al., 2007).At the oligotrophic ALOHA, POC export fluxes at 500 m were determined to be 0.3 mmolC m −2 d −1 .At the more eutrophic K2, POC export fluxes at 500 m were higher, varying between 1.1 and 2.4 mmolC m −2 d −1 .Our results are consistent with these values.
The degree of flux attenuation in the twilight zone can be expressed as mesopelagic transfer efficiency (T eff ) or the ratio of POC flux at 500 to 100 m (Buesseler et al., 2007).It should be noted that when calculating the T eff , we used the POC flux at 100 m based on 234 Th/ 238 U disequilibria, while 228 Th/ 228 Ra disequilibria at 500 m.The mesopelagic transfer efficiency (T eff ) was determined to be 8%-130% in the South China Sea (Table 5).Some  values were greater than 1, may due to the POC fluxes at 100 and 500 m based on different methods, respectively.The >53 μm 234 Th ( 228 Th)-derived T eff was higher than the results based on 1-53 μm.It implied that the large particles were fast-sinking and move out of the twilight zone.The T eff was low at A1 affected by the Kuroshio waters in winter and spring, where the organic matter exported is relatively labile and is prone to remineralization in the twilight zone.The T eff was relatively high in other stations because the organic matter exported is refractory and under-goes comparatively less degradation in the twilight zone.

Conclusion
POC export and the mechanisms that control its attenuation in the twilight zone have long been understudied despite the fact that the most marked attenuation of particle export occurs in this depth region.This is predominantly due to a lack of reliable methods for determining POC export, which have in turn hampered our ability to understand the controls on particle transport efficiency through the twilight zone.In this study, we have attempted a variety of methods including 234 Th/ 238 U disequilibrium, 228 Th/ 228 Ra disequilibrium and 228 Ra-based nutrient budget approach to constrain POC export flux in the upper twilight zone of the South China Sea.We showed that the average POC export fluxes at 100 m in the study region varied between 1.2 and 3.5 mmolC m −2 d −1 .This estimate is in good agreement with previous studies using 234 Th/ 238 U disequilibrium or sediment trap technique.At 500 m, the average POC export fluxes were constrained to be in the range of 0.8-2.9mmolC m −2 d −1 .In the meantime, the mesopelagic transfer efficiency (T eff ) was determined to be 8%-130%.The high T eff in the South China Sea contrasts with previous suggestion that pico-plankton-dominated ecosystem is associated with a low efficiency of POC transport through the twilight zone.The >53 μm 234 Th ( 228 Th)-derived T eff was higher than the results based on 1-53 μm.The large particles were fast-sinking and move out of the twilight zone.The T eff was low at A1 affected by the Kuroshio waters in winter and spring, where the organic matter exported is relatively labile and is prone to remineralization in the twilight zone.The T eff was relatively high in other stations because the organic matter exported is refractory and under-goes comparatively less degradation in the twilight zone.
A major conclusion in this study is that 228 Th/ 228 Ra disequilibrium is an extremely sensitive method for tracing POC export in the twilight zone.This conclusion is based on the comparison of the "Minimum Detection Limit" (MDL) of various methods.We have demonstrated that the MDL of 228 Th/ 228 Ra method can be as low as 0.1 mmolC m −2 d −1 .This MDL is lower than or comparable to the signal of POC export in most of the bathypelagic region of the world oceans.This implies that 228 Th/ 228 Ra disequilibrium has a great potential of being applied to tracing POC export in the twilight zone and below.
POC export fluxes above 500 m derived from 234 Th/ 238 U and 228 Th/ 228 Ra disequilibria were generally inconsistent.We have demonstrated that this inconsistency is predominantly due to the different time scales over which the two methods are applied.The steady-state assumption is harder to be fulfilled for the calculation of 228 Th-derived POC fluxes, which should not be ignored.By comparing the POC export fluxes derived from 228 Th/ 228 Ra and 228 Ra -NO 3 -disequilibrium, we found that the results are consistent while only a few are not.We thus further conclude that 228 Th/ 228 Ra disequilibrium is suitable for the research of the twilight zone.Currently, the bottleneck in the application of 228 Th/ 228 Ra method is the difficulty in analyzing the low activities of 228 Th and 228 Ra in seawater.Meanwhile, with the advent of a RaDeCC system (Moore & Arnold, 1996), it is now possible to conduct a non-destructive analysis of 228 Th and 228 Ra in seawater through the measurement of 224 Ra.We are thus foreseeing prosperity in the application of 228 Th/ 228 Ra disequilibrium to studying carbon cycles in the twilight zone.

Figure 1 .
Figure 1.Map of the South China Sea.Sampling stations are indicated by solid circles.Note that A1 and SEATS were sampled during all seasons, and A10 was occupied in August 2009.
a maximum in the euphotic zone.Below 100 m, particulate234 Th activities on 1-10 μm size class were relatively uniform.Particulate 234 Th activities on medium (10-53 μm) and large (>53 μm) size class particles varied between 0.002 and 0.080 dpm L −1 , and between 0.005 and 0.129 dpm L −1 .The depth profiles of particulate 234 Th on medium and large size class particles showed a maximum at depth below the nutricline.

Figure 2 .
Figure 2. Depth distributions of size-fractionated 234 Th/ 238 U ratios in the upper 500 m at station A1, SEATS in four seasons and at A10 in summer.Vertical dashed lines represent radioactive equilibrium between total 234 Th and 238 U. Note that 238 U is conservative in the ocean and almost homogenous regardless the variations of salinity in the open ocean.234Th is produced via alpha-decay by238 U at a known rate.

Figure 3 .
Figure 3.The vertical distributions of potential temperature and salinity at station A1, SEATS in four seasons and at A10 in summer in the northern South China Sea.

Figure 4 .
Figure 4. Depth profiles of 228 Ra and nitrate in the upper 500 m at station A1, SEATS in four seasons and at A10 in summer.

Figure 5 .
Figure 5. Vertical profiles of size-fractionated 228 Th/ 228 Ra ratios in the upper 500 m at station A1, SEATS in four seasons and at A10 in summer.Vertical dashed lines represent radioactive equilibrium between total 228 Th and 228 Ra.

Figure 6 .
Figure 6.The vertical distributions of particulate organic carbon in different size classe particles in the upper 500 m in the northern South China Sea.
Th-derived POC fluxes (hereafter referred to as 1-53 μm234 Th-derived POC flux) fell in the range of 0.1-9.7 mmolC m −2 d −1 .In comparison, when the POC/Th ratios on the >53 μm size class particles were used, the 234 Th-derived POC fluxes (hereafter referred to as >53 μm 234 Th-derived POC flux) fell in the range of 0.1-6.2mmolC m −2 d −1 .The vertical distribution of 234 Th-derived POC flux generally displayed a subsurface maximum at around 50-100 m (Figure10).POC export flux at 100 m was lower than that at 500 m.This indicates that POC remineralization took place in the twilight zone.It is also worthwhile to note that in the upper 200 m, the 1-53 μm 234 Th-derived POC fluxes were significantly higher than the >53 μm 234 Th-derived POC fluxes.Below 200 m, applying POC/Th ratio on either size class particles essentially resulted in a comparable POC export estimate (Figure10).The vertical profiles of POC fluxes based on >53 μm 234 Th were similar to those based on 1-53 μm234 Th.The 228 Th-derived POC export fluxes based on the POC/Th ratios on the 1-53 μm size class particles (hereafter referred to as 1-53 μm228 Th-derived POC flux) ranged from a low of 0.1 mmolC m −2 d −1 to a high of 7.1 mmolC m −2 d −1 (Table3).The highest 1-53 μm 228 Th-derived POC flux was observed at 50 m at A1 in summer (2009 08).In comparison, when applying the POC/Th ratios on the >53 μm size class particles, the resultant 228 Th-derived POC export fluxes (hereafter referred to as >53 μm 228 Th-derived POC flux) were significantly lower, ranging from 0.1 to 3.6 mmolC m −2 d −1 .The highest >53 μm 228 Th-derived POC flux was also observed at SEATS in summer (2009 08).Meanwhile, it occurred at 100 m, deeper than the depth where the highest 1-53 μm 228 Th-derived POC flux was observed.Similar to 234 Th-derived POC flux, the vertical distribution of 228 Th-derived POC flux generally displayed a subsurface maximum (Figure11).Below the subsurface maximum, POC export flux gradually declined.Exceptions to this depth pattern were found at SEATS in winter (2010 01) for both the 1-53 μm and the >53 μm 228 Th-derived POC flux, which were relatively constant throughout the upper 500 m.Like 234 Th-derived POC flux, in the upper 200 m, the 1-53 μm 228 Th-derived POC fluxes were significantly higher than the >53 μm 228 Th-derived ones.Below 200 m, the two estimates were in better agreement.The vertical profiles of POC fluxes based on >53 μm 228 Th were similar to those based on 1-53 μm 228 Th.

Figure 7 .
Figure 7.The vertical distributions of Chl a in different seasons in the northern South China Sea.
ratio on 1-53 μm size class was calculated from the sum of POC and 234 Th ( 228 Th) measurements on the 1-10 and 10-53 μm size class particles.bThe unit of Th-based POC export is mmolC m −2 d −1 ; The 1-53 and >53 μm results represent POC export flux derived by using the POC/Th ratios on the 1-53 and >53 μm size classes, respectively.c NO 3 flux denotes POC flux derived by converting the upward diffusion flux of nitrate using a C/N ratio of 6.6.
comparison, we have further converted the nitrate fluxes into the downward fluxes of POC using a Redfield ratio of 6.6.The resultant NO 3 -based POC flux fell in the range of 0-12.1 mmolC m −2 d −1 (Table

Figure 8 .
Figure 8. Plots of steady-state (steady state) 234 Th flux versus depth in the upper 500 m at the sampling stations.

Figure 9 .
Figure 9. Plots of steady-state (steady state) 228 Th flux versus depth in the upper 500 m at the sampling stations.

Figure 11 .
Figure 11.Vertical profiles of 228 Th-based particulate organic carbon (POC) flux in the upper 500 m at station A1, SEATS and A10.The solid and open circles denote 1-53 μm 228 Th-based POC flux and >53 μm 228 Th-based POC flux, respectively (See text for the definition).

Table 2
Size-Fractionated Particulate Organic Carbon (POC) Contents and Nitrate Concentrations in the Upper 500 m at Station A1, SEATS and A10 in the Northern South China Sea Ra except at A1 in summer (2009 08), where 228 Th deficit was 9 of 29 observed below 200 m.At around 300 m at A1 in autumn (2010 11),

Table 3
Thorium Flux, Particulate Organic Carbon (POC)/Th Ratio, POC Export Flux, NO 3 -Based POC Flux in the Upper 500 m at Station A1, SEATS and A10 in the Northern South China Sea

Table 3 Continued
DENG AND LIN Note.For the comparison purpose, the depth horizon was selected at 100, 150, 200, 300, and 500 m.The associated uncertainty represents one standard deviation for the all sampling stations in four seasons, which represents the spatial and temporal variability in POC export flux.

Table 4
Average Particulate Organic Carbon Fluxes Derived From Different Methods and Their "Minimum Detection Limit" (Unit: mmolC m −2 d −1 ) for the Sampling Stations in the South China Sea