G 3. AN ELECTRONIC JOURNAL OF THE EARTH SCIENCES Published by AGU and the Geochemical Society

Transcription

1 Geosystems G 3 AN ELECTRONIC JOURNAL OF THE EARTH SCIENCES Published by AGU and the Geochemical Society Article Volume 9, Number 4 10 April 2008 Q04016, doi: /2007gc ISSN: In situ calibration of Mg/Ca ratio in planktonic foraminiferal shell using time series sediment trap: A case study of intense dissolution artifact in the South China Sea Kuo-Fang Huang Department of Earth Sciences, National Cheng Kung University, No. 1 University Road, Tainan, Taiwan 701 Also at Earth Dynamic System Research Center, National Cheng Kung University, Tainan, Taiwan Cheng-Feng You Department of Earth Sciences, National Cheng Kung University, No. 1 University Road, Tainan, Taiwan 701 (corresponding author: cfy20@mail.ncku.edu.tw) Also at Earth Dynamic System Research Center, National Cheng Kung University, Tainan, Taiwan Hui-Ling Lin Institute of Marine Geology and Chemistry, National Sun Yat-Sen University, No. 70 Lian-Hai Road, Kaohsiung, Taiwan 804 Ying-Tzeng Shieh National Taiwan Museum, Taipei, Taiwan 100 [1] This study examines Mg/Ca-temperature equations through the use of planktonic foraminifers collected from continuous time series sediment traps at four water depths in the South China Sea. This deployment provides an opportunity to refine partial dissolution artifacts and to identify responses of shell chemistry to changes in ambient seawater conditions. Paired Mg/Ca and d 18 O measurements on shells are strongly correlated with seasonal variability in terms of temperature at specific habitat depths. However, partial dissolution can significantly alter Mg/Ca ratios even at depths well above the calcite lysocline or carbonate compensation depth. Of the three species studied, Neogloboquadrina dutertrei is the most sensitive species to dissolution relative to the other surface-dwelling species (Globigerinoides ruber and Globigerinoides sacculifer). Differences between the Mg-derived temperature equation here and those from previously established core top or sediment trap calibrations can be attributed to various degrees of dissolution and lateral advection, thus highlighting the need for in situ empirical calibration for different ocean basins. Components: 11,515 words, 9 figures, 3 tables. Keywords: in situ Mg/Ca calibration; sediment trap; partial dissolution. Index Terms: 4924 Paleoceanography: Geochemical tracers; 3030 Marine Geology and : Micropaleontology (0459, 4944); 1065 Geochemistry: Major and trace element geochemistry. Received 14 April 2007; Revised 19 January 2008; Accepted 8 February 2008; Published 10 April Huang, K.-F., C.-F. You, H.-L. Lin, and Y.-T. Shieh (2008), In situ calibration of Mg/Ca ratio in planktonic foraminiferal shell using time series sediment trap: A case study of intense dissolution artifact in the South China Sea, Geochem. Geophys. Geosyst., 9, Q04016, doi: /2007gc Copyright 2008 by the American Geophysical Union 1 of 20

2 1. Introduction [2] Reconstruction of past sea surface temperatures (SSTs) in the tropical/subtropical ocean is a fundamental element for understanding the distribution of heat and moisture, as well as the evolution of the monsoon system, in the past. Moreover, it is also influential in driving long and short-term climatic oscillations, because of the perceived importance of dynamical interaction between the atmosphere and the surface oceans in the tropics [Lea et al., 2000; Rosenthal et al., 2003]. The South China Sea (SCS) is located at the western equatorial Pacific, and adjacent to two major climatically sensitive areas, the Western Pacific Warm Pool (WPWP) and the Indonesian archipelago. The SCS, therefore, becomes one of the most important settings in response to seasonal variability of Asian monsoon intensity. However, the potential influence of the East Asian Monsoon (EAM) on SST variations in the tropical ocean remains unclear and this can be investigated by high-resolution and accurate downcore records of SSTs in the SCS during the Quaternary [Oppo and Sun, 2005; Steinke et al., 2005]. [3] Previous studies based on laboratory and field work have highlighted that uptake of Mg by planktonic foraminifers from ambient seawater is positively correlated with water temperature (9 ± 1%, per C) [Hastings et al., 1998; Lea et al., 1999; Elderfield and Ganssen, 2000; Rosenthal and Lohmann, 2002; Anand et al., 2003]. Mg/Ca in planktonic foraminiferal calcite is, therefore, becoming a well-established temperature proxy for past surface oceans. Additionally, paired Mg/Ca and d 18 O measurements on foraminifers are of great merit in reconstructing both SSTs and d 18 O in seawater [Mashiotta et al., 1999]. Through this approach, the past changes in regional sea surface salinity (SSS) can be estimated directly [Schmidt et al., 2004]. These results provide important information related to potential linkages of abrupt climatic changes between high latitudes and tropical regions (e.g., Heinrich events and paleo-el Niño or La Niña) in a centennial-millennial timescale [Koutavas et al., 2002; Stott et al., 2004]. [4] Besides temperature effects, a growing body of studies indicate that foraminiferal Mg/Ca can be altered by post-depositional dissolution [Rosenthal and Lohmann, 2002] and physiological processes [Rosenthal et al., 1997; Bentov and Erez, 2005, 2006]. Such artifact and bio-mineralization processes would significantly affect accurate estimates of Mg/Ca-based temperature records. The current state of the art for planktonic Mg/Ca calibration heavily relies on culture experiments [Nüernberg et al., 1996; Lea et al., 1999; Russell et al., 2004] and core top specimens [Elderfield and Ganssen, 2000; Rosenthal et al., 2000; Dekens et al., 2002]. In this study, we propose a third method to establish Mg/ Ca calibration through the application of planktonic foraminifers from the modern water column, collected by time series sediment traps deployed in the SCS. These trap materials will bridge the gaps between controlled laboratory culture experiments and field core top calibrations, and can be particularly helpful to validate possible correlations between shell chemistry and natural variability of environmental factors that can be measured precisely [Anand et al., 2003; Pak et al., 2004]. [5] We employ the sediment trap arrays deployed at various depths in the northern SCS. This design enables us to systematically refine artifacts of partial dissolution in the SCS, and to evaluate directly shell compositional changes in response to ambient seawater conditions. Furthermore, these results will be compared with core top calibrations by Hastings et al. [2001] and Dekens et al. [2002] to elucidate possible relationships between core top and sediment trap data. Our main objectives are (1) to document intra-annual (seasonal) variations of Mg/Ca and d 18 O in foraminiferal shells; (2) to establish an in situ Mg/Ca-SST empirical equation for the SCS; (3) to evaluate partial dissolution effects on foraminiferal Mg/Ca in the SCS; (4) to assess other environmental parameters that would affect shell Mg/Ca; and (5) to compare in situ empirical calibration in SCS with available equations from core top and sediment trap materials worldwide. 2. Previous Paleoceanographic Studies [6] Since the sedimentation rates in the SCS are extremely high ( cm/ka, an order of magnitude higher than other parts of the Pacific) with widespread distribution of carbonate sediments, it provides an ideal region for high-resolution paleoceanographic reconstruction [e.g., Wang et al., 1995]. Previous studies have shown growing interests in the late Quaternary oceanographic environments of the SCS, such as paleo-sst, productivity, deep water circulation, carbon cycles, monsoon variations, and climatic interactions between the continent and the ocean during the glacial-interglacial periods [Thunell et al., 1992; Miao et al., 1994; Huang et al., 1997; Wang et al., 1999; Pelejero et al., 1999; Oppo and Sun, 2005]. Among these 2of20

3 Geosystems G 3 huang et al.: mg/ca ratio in planktonic foraminiferal shell /2007GC studies, determination of past temperature (T) and salinity (S) in the surface ocean at low latitudes is the most essential element for understanding changes in atmosphere/ocean interaction, as well as relative role of the Inter-Tropical Convergence Zone in the Asian monsoon system. Thus, accurate estimates of SST and SSS in the past ocean are particularly important for a detailed evaluation of climatic sensitivity in response to solar radiation and monsoon strength variation. [7] Although considerable efforts have been dedicated to study the temporal paleo-sst variations in the SCS, results of marine faunal assemblage, oxygen isotope, unsaturated alkenone, and foraminiferal Mg/Ca data display a large discrepancy in temperature changes (1 6 C) from the Last Glacial Maximum to the late Holocene [Wang and Wang, 1990; Pelejero et al., 1999; Kienast et al., 2001; Chen et al., 2005; Oppo and Sun, 2005]. This can be attributed partly to a lack of understanding in oceanographic, ecological and geochemical processes affecting these temperature proxy indicators. To evaluate these potential interferences, we focus to refine geochemical processes that affect shell Mg/Ca due to partial dissolution, which can alter significantly the Mg/Ca ratios in planktonic foraminifers [Brown and Elderfield, 1996; Rosenthal and Lohmann, 2002]. 3. Modern Hydrography of the SCS [8] The climatic variation and water mass circulation in the upper ocean of the SCS are mainly modulated by seasonal prevailing monsoons. Typically, during the summer wet season, May to October, the SCS is dominated by a warm monsoon and prevailing southwest winds. In contrast, a winter dry monsoon presented from December to February is known to link with the Southern Oscillation Index [Zhang et al., 1997]. These annually reversing monsoon winds cause a clockwise water mass circulation in summer, and then gradually switch to a reversal counter-clockwise gyre during wintertime [Wyrtki, 1961] (see auxiliary material 1 Figures S1a and S1b). Local upwelling is also forced by wind-driven circulations, such as of east Vietnam coast in the summer and northwest Luzon and north Sunda shelf during the winter seasons [Shaw et al., 1996; Liu et al., 2002] (Figure S1). 1 Auxiliary materials are available in the HTML. doi: / 2007GC [9] The modern SST distribution in summer is rather uniform (ranging from 28 to 29 C) and show a small north-south gradient. Average salinity is relatively low (approximately 33 psu), mainly resulted from fresh water inflow throughout the coastal region. In contract, the average surface temperature is 18 C and 27 C in northern and southern SCS, respectively and shows a northsouth gradient in winter time [Wyrtki, 1961]. The winter SSS is relatively stable due to rather constant regional precipitation. Modern monthly temperature and salinity variations at different depths are also shown in Figures S1c and S1d, the seasonal variability in temperature and salinity is 5 6 C and %, respectively. [10] The intermediate water occupies depths at 250 to 1000 m in the SCS, and its physical properties are similar to the North Pacific Intermediate Water (NPIW) [Wyrtki, 1961; Nitani, 1972]. This has been attributed to the intrusion of NPIW, which is characterized with a salinity minimum at 490 m. The SCS deep water is fed possibly by the North Pacific Deep Water, on the basis of hydrographic observation and T-S characteristics at 3500 m [Nitani, 1972]. The present-day calcite lysocline and carbonate compensation depth (CCD) in the SCS is located approximately at 3000 m and 3800 m, respectively [Miao et al., 1994] whereas calcite saturation depth (W calcite < 1) is at 2500 m [Chou et al., 2007, and references therein]. Although sediment trap arrays were deployed at depth shallower than lysocline or CCD in the SCS, Chou et al. [2007] also reported carbonate dissolution in shallow water based on TCO carb 2 measurements (i.e., excess TCO 2 produced by carbonate dissolution) in the time series SEATS station ( N, E), in adjacent to the study area. 4. Materials and Methods 4.1. Sediment Trap Deployment [11] Three dominant species of planktonic foraminifers (Globigerinoides ruber white, Globigerinoides sacculifer without sac-like chamber, and Neogloboquadrina dutertrei) were collected from continuous trap arrays deployed at the northern SCS (M1s, M2 and M3; see Figure 1 and Table 1 for details) to examine relationships between planktonic shell chemistry and seawater properties. Three mooring sites were selected at the northern SCS basin and these trap cups were deployed at four different depths to evaluate dissolution artifacts that may bias metal/ca ratios. The collecting 3of20

4 solution (7.6 < ph < 8.0) to prevent degradation or growth of organic matters. Operational details on trap types, deployment and retrieving procedures were described by Lin et al. [2004]. Owing to the original trap design for radioisotope study, only discrete but representative intra-annual specimens can be obtained for foraminiferal trace element/ca ratio (TE/Ca) analyses. Figure 1. Location map of sediment traps (M1S, M2, and M3), ODP Sites 1144 and 1145, and MD in the SCS. The red line represents the 100-m isobath, showing the approximate position of coastline during the glacial low sea level. Mekong and Molengraaft rivers on the Sunda Shelf are also indicated. Modern upwellings occurred in southeast Vietnam (blue shade) and northwest Luzon Island (green shade) in summer and winter, respectively [Shaw et al., 1996]. interval for each cup was set for 15 days except for those were low in foraminiferal abundances (i.e., M1s, which deployed during a winter season [see Lin et al., 2004]), inadequate low flux or cup array malfunction (i.e., level 3 at M3). In addition, these cups were filled with a 0.5% buffered formalin 4.2. Foraminiferal Cleaning and Analysis [12] About individuals ( mm in size) of foraminiferal shells were hand-picked under a stereoscope to minimize ontogenetic effects. In order to pick statistically sufficient numbers of foraminiferal shells for obtaining reliable data, some smaller size fractions were necessary as relatively high abundances of these species presented in the northern SCS. These foraminiferal tests were gently crushed to facilitate removal of remnant cytoplasma from the interior of shell chamber. Then a series of physical and chemical cleaning protocols modified from Boyle and Keigwin [1985/1986] and Anand et al. [2003] were applied to remove potential contamination phases. For crushing shell samples, foraminiferal tests were ultrasonically cleaned three times with distilled water, twice with methanol, and finally rinsed with distilled water several times to remove adhering detrital grains and fine clay materials. Reduction and DTPA reagents (so called Cd or Ba cleaning method [Boyle and Keigwin, 1985/1986; Lea and Boyle, 1993]) were used to remove Fe-Mn oxides and barite accumulation within these shells. For trap samples, more care is necessary to clean all adsorbed contamination due to the presence of organic matter not commonly found in down-core sediments and lengthy exposure to organic-rich materials inside the collecting cups [see also Anand et al., 2003; Pak et al., 2004]. Organic matter Table 1. Sediment Trap Deployments in the Northern South China Sea That Provided Samples Used in This Study Trap ID Latitude Longitude Water Depth, m Trap Depth, m Sampling Date Season M1s N E 2943 T2: 925 Dec 2001, Jan 2002, winter-spring T3: 1925 T4: 2702 and Mar 2002 M N E 3743 T1: 240 Oct 2001 and Jan 2001 autumn-winter T2: 1240 T3: 2240 T4: 3240 M N E 2626 T1: 597 May 2002, July 2002, spring-summer T2: 1126 T3: 1726 T4: 2326 and Aug of20

5 presented in the inner chambers was intensively oxidized in 50% buffered hydrogen peroxide-sodium hydroxide solution in a 90 C water bath for at least 30 minutes followed by a dilute acid (0.001N HNO 3 ) polish to remove adsorbed metals. After weak acid leaching, distilled water was introduced again to remove all surface-adsorbed and residue contaminants, and finally the calcite fragments were transferred into new acid-leached microcentrifuge vials prior to analysis. The remaining shell calcites were then dissolved in 0.065N HNO 3 containing internal standards of Sc and Y. All dilute nitric acids used were prepared with a Vycor sub-boiling system and other chemical reagents were purified by PFA two-bottle equipments or commercial ultrapure-grade reagents. Sample preparation procedures were processed under a class-10 working bench to eliminate the trace element blank derived from HEPA air-filter. [13] High-resolution ICP-MS (Finnigan Element 2) was used to measure multiple elemental ratios simultaneously at the National Cheng Kung University. The foraminiferal TE/Ca ratios were determined at the same detection mode to eliminate possible intensity bias and to quantify intensity ratios using external matrix-matched standards adapted from Rosenthal et al. [1999]. In the analysis, 24 Mg, 25 Mg, 43 Ca and 44 Ca were determined both by low (m/dm = 300) and medium mass resolution (m/dm = 4000) to avoid spectral interferences. A series of matrix-matched standards were prepared from high-purity ICP standard solutions by a gravimetrical method and were standardized for TE/Ca by standard addition. Long-term reproducibility of Mg/Ca is evaluated on the basis of routine runs of the consistency standards and matrix-matched standards, and analytical precision is ±1.04%. Internal standards of 1 ppb Sc and Y were added into sample solutions to monitor the instrumental stability (1 hr < 0.5% and 8 hrs < 2.0% variations) during measurement. In addition, 27 Al, 55 Mn and 56 Fe were also measured by low and medium resolution to monitor detrital contaminants, such as fine clay and Fe-Mn coatings. Al/Ca, Mn/Ca and Fe/Ca were mostly lower than 40 mmol mol 1, well below thresholds for likely contamination (>100 mmol mol 1 ). Stable oxygen and carbon isotopic data of shell calcites ( mm forg. ruber and mm for G. sacculifer) separated from the same samples were adopted from Lin et al. [2004] to compare with Mg/Ca results. A systematic comparison of two different cleaning methods ( Ba-cleaning and Mg-cleaning procedures) on the three planktonic species were also conducted to evaluate the effect of cleaning procedure on shell Mg/Ca (see auxiliary material Figure S2), a < 8% Mg/Ca decrease was found by involving the reductive step. This Mg/Ca bias was estimated by linear regression of data presented in auxiliary material Figure S2, excluding three samples associated with high Fe, Mn, Al and Ba after the cleaning procedures Estimates of Habitat Mean Temperature and Salinity [14] The habitat mean temperature (T H ) and salinity (S H ) where calcification occurred can be calculated using modern seawater T and S profiles and the specific habitat depths of three dominant planktonic foraminiferal species. To achieve this, it is necessary to obtain appropriate hydrographic data and calcification information for individual species. The life cycle of planktonic species is about 2 4 weeks, while monthly temperature and salinity information were inferred from CTD records during trap deployment periods. The hydrographic grid data ( E; N) of monthly mean T and S (between ) in the northern SCS were also available from the National Center for Oceanographic Research (NCOR) archives if no CTD record available. The two data sets show consistent variations in the intra and interannual SST records. [15] Calcification depth of the three planktonic species can be theoretically estimated by a systematic comparison of predicted equilibrium d 18 O c and measured foraminiferal d 18 O c. For the SCS, considerable works on S and d 18 O sw have indicated that the two parameters were strongly affected by monsoon precipitation and freshwater input [Su, 2000; Lin, 2000; Lin et al., 2003]. The modern S-d 18 O sw relationship, d 18 O sw = S, has been established for coastal seawater [Lin et al., 2003]. [16] On the other hand, paleotemperature equations described by O Neil et al. [1969] and Shackleton [1974] can also be re-arranged for equilibrium d 18 O c calculation: d 18 O c ¼ d 18 O sw 0:27 h þ 4:38 4:38 2 0:5 i 4 0:1 ð16:9 TÞ =2 0:1 ð1þ 5of20

6 Figure 2. Habitat depth determination for three planktonic species in the SCS: (a) G. ruber white; (b) G. sacculifer without sac; and (c) N. dutertrei. Measured d 18 O c in foraminiferal calcite plotted against the calendar month, together with predicted d 18 O c in equilibrium with seawater at various depths. The predicted d 18 O of calcite, in equilibrium with seawater, were then calculated using equation (1) and monthly mean temperature in seawater profiles. Comparison between foraminiferal d 18 O c and theoretically predicted d 18 O c at various depths (Figure 2) allow for estimating a mean calcification depth for individual species and the results are compiled in Table 2. The estimated species-specific calcification depths are in good agreement with depth ranges for living species by plankton tow and trap specimens from the SCS [Lin et al., 2004]. Then, the T H and S H can be derived by combining calcification depth information with monthly mean T and S profiles. This in turn provides the most reliable habitat T and S records for the three planktonic species. [17] To separate temperature signals from measured shell d 18 O c, salinity-corrected foraminiferal d 18 O c are selected to compare with Mg/Ca ratios. The salinity correction will involve d 18 O c and d 18 O sw, which are linearly correlated at depth of the trap arrays deployed. All d 18 O sw used in the following discussion were converted from SMOW to PDB using a correction constant of [18] Foraminiferal Mg/Ca results obtained from continuous trap materials over one-year deployment period are summarized in auxiliary material Table S1. Mg/Ca ratio vary from to mmol mol 1, to mmol mol 1 and to mmol mol 1 for G. ruber, G. sacculifer and N. dutertrei, respectively. In addition to seasonal SSTs variability, the observed wide ranges of Mg/Ca for intraspecies could be associated with biological activity. Planktonic foraminifers commonly migrate vertically throughout their life cycle and often live at deeper depth as they get mature, resulted in heterogeneous distribution of Mg/Ca ratio within a signal shell and/or interspecies [Brown and Elderfield, 1996; Elderfield and Ganssen, 2000; Eggins et al., 2003; Sadekov et al., 2005]. Nevertheless, species-specific Mg/Ca calibration for individual species from the same core can provide valuable information of thermal structure variability in water column. [19] There are several remarkable features clearly documented in this study: (1) Strong intra-annual (seasonal) variations occur both in trapped foraminiferal Mg/Ca and d 18 O c for the two planktonic species. In general, high Mg/Ca coupled with light d 18 O c are observed in summer and fall months (May to October). Low Mg/Ca and heavy d 18 O c, on the other hand, are characterized in winter season (December to March) (Figure 3). (2) Deep-dwelling species showed a relatively small variation in Mg/Ca, in an order of N. dutertrei < G. sacculifer < G. ruber (Figure 3). (3) Mg/Ca is positively correlated with estimated T H (Figure 4a) 5. Sediment Trap Results Table 2. Species of Planktonic Foraminifera in the SCS Used in This Study and Estimations of Calcification Depth for Individual Species Based on the Comparison of Foraminiferal d 18 O With Calcite in Equilibrium With Seawater Planktonic Species Calcification Depth, m Globigerinoides rubber white 0 20 Globigerinoides sacculifer 0 50 Neogloboquadrina dutertrei of20

7 Figure 3. Intra-annual variation in Mg/Ca (mmol mol 1 ) and salinity-corrected d 18 O(%) forg. ruber (white), G. sacculifer (without sac), and N. dutertrei. Trap depths: (a, d, and g) <1000 m; (b, e, and h) m; and (c, f, and i) >2000 m. Note that the correlation between shell Mg/Ca and salinity-corrected d 18 O becomes weak at depth and only Mg/Ca data in N. dutertrei are available. and shows negative relationships with d 18 O derived calcification temperature (Figure 4b). (4) An exponential equation fits Mg/Ca and mean habitat temperature, but shows a distinguishable difference with trap data from the Sargasso Sea (Figure 5). (5) Partial dissolution artifacts at various depths play a significant role to alter Mg/Ca in foraminiferal 7of20

8 Figure 4. (a) Mg/Ca ratios and the mean habitat temperature correlation for the three planktonic foraminifers in the SCS. (b) Plot of Mg/Ca versus d 18 O calcification temperature. Red and blue rectangles: data points do not follow the general trend (i.e., high Mg/Ca at shallow depth), most likely due to lateral advection. calcite, even at water depths shallower than lysocline or CCD (Figures 6 and 7). 6. Discussion 6.1. Interspecies Foraminiferal Mg/Ca [20] The three most abundant planktonic species of G. ruber white, G. sacculifer and N. dutertrei separated from the trap sediments in the SCS show an interspecies Mg/Ca ratio variability, possibly related to variation in calcification depths. Average Mg/Ca of G. ruber (white), G. sacculifer (w/o sac) and N. dutertrei is mmol mol 1, mmol mol 1, and mmol mol 1, respectively (Table S1). The apparent Mg/Ca difference among species is rather large compared with the potential analytical error (2s) of ±0.005 mmol mol 1. The seawater Mg/Ca ratio does not vary under modern hydrological conditions due to the long residence time of Mg and Ca [Broecker and Peng, 1982]. Thus, it is reasonably to infer that the observed Figure 5. (a) Plot of Mg/Ca versus mean habitat temperature for all species in the SCS (blue line) compared with the Sargasso Sea sediment trap (purple line) [Anand et al., 2003]. All regression lines assume a constant exponential constant (=0.090). (b) Comparison of seasonal variability in Mg/Ca during the two deployment periods (M2, fallwinter; M3, spring-summer). Two dashed lines represent uncertainties proposed by Anand et al. [2003]. 8of20

9 Figure 6. The regression of Mg/Ca and mean habitat temperature for various trap depths in the SCS during (a) fallwinter period and (b) spring-summer. The pre-exponential constants for all regression lines are also shown in cold season. Mg/Ca differences between species were affected by Mg partition coefficient (D Mg ). [21] Inorganic carbonate precipitation and foraminiferal culture experiment indicate that incorporation of Mg into calcite depends remarkably upon the temperature of the ambient solution [Delaney et al., 1985; Oomori et al., 1987; Lea et al., 1999], and can be partly explained by thermodynamic Mg 2+ substitution of Ca 2+ into the calcite crystal lattice [Katz, 1973; Oomori et al., 1987]. Postdepositional dissolution could play a significant role to alter foraminiferal Mg/Ca by preferential removal of Mg-enriched calcites [Rosenthal and Boyle, 1993; Brown and Elderfield, 1996]. Other environmental factors, such as disequilibrium (or kinetic) effects, may also affect Mg uptake from the ambient seawater based on the observed relationships between shell size and Mg/Ca ratios of planktonic foraminifers [Elderfield et al., 2002]. Moreover, there is evidence for Mg/Ca variability among different morphotypes of the same species, which might be related to differences in either growth season (e.g., G. ruber white versus pink [Anand et al., 2003]) or dwelling habitat (e.g., G. ruber s.s. and G. ruber s.l. [Steinke et al., 2005]). In this study, we use mixed morphotypes of G. ruber Figure 7. The Mg/Ca data for the three species of planktonic foraminifer at various trap depths during the two deployment periods. (a) G. ruber, (b) G. sacculifer, and (c) N. dutertrei. Also shown are the calcite lysocline (3000 m, dashed purple line) and calcite saturation depth (2500 m, dashed red line) [Chou et al., 2007]. 9of20

10 white (i.e., s.s. and s.l.) due to size limitation required for precise TE/Ca determination. [22] To minimize possible influences of partial dissolution and disequilibrium processes, only foraminiferal Mg/Ca taken from the shallowest trap were used to assess correlations between interspecies shell Mg/Ca and depth habitat. The increases in interspecies Mg/Ca are accompanied by decreasing d 18 O c and reflect an increased calcification temperature in the sequence of G. ruber > G. sacculifer > N. dutertrei. This is consistent with previous studies, low Mg/Ca and enriched d 18 O c for deep dwellers (e.g., N. dutertrei) compared to the mixed-layer dwelling species (i.e., G. ruber and G. sacculifer) [Fairbanks et al., 1982; Rosenthal and Boyle, 1993]. Alternatively, the offset between G. ruber and G. sacculifer is likely related to the addition of gametogenic calcite (28%) in G. sacculifer [Bé, 1980]. Gametogenic calcite of G. sacculifer is encrusted toward to its late life and sinks to the seafloor [Bé, 1977], the low temperature in deeper water might be recorded during their calcification and, therefore, would decrease the average Mg/Ca of G. sacculifer. Although the interspecies differences in Mg/Ca are commonly linked to specific calcification depths of planktonic species, more recent studies have shown possible biological influence on Mg uptakes during biomineralization of foraminiferal shells [Bentov and Erez, 2005, 2006] Partition Coefficient for Mg [23] The element-to-calcium ratios in biogenic calcites depend on corresponding ratios of ambient ocean water and the partition coefficient of elements between the solid carbonate and the seawater. These relationships are expressed as (Mg/Ca) foram = D Mg (Mg/Ca) seawater, where (Mg/Ca) foram is a molar ratio of Mg to Ca in the foraminifer, (Mg/ Ca) seawater is a molar ratio of Mg to Ca in seawater (assuming seawater Mg/Ca = 5.15 mol mol 1 [Broecker and Peng, 1982]), and D Mg is the partition coefficient of Mg in shell calcite. Using this relationship, we obtain D Mg for all three planktonic species in the SCS. The apparent D Mg values fall in a range of (ave ), (ave ), and (ave ) for G. ruber, G. sacculifer and N. dutertrei, respectively. This can be linked to depth habitats of specific life cycles for planktonic species (i.e., surface dwelling species have high D Mg, related to high calcification temperature or shallow habitat depth), suggesting that D Mg in foraminiferal calcite is mainly a function of seawater temperature. These calculated values are consistent with estimates (D Mg ) from core top sediments [Elderfield et al., 2002] and culture experiments [Nüernberg et al., 1996]. However, these D Mg values are in strong contradiction to the inorganic experimental determination [Katz, 1973; Morse and Bender, 1990], about two orders of magnitude higher (D Mg = ), implying biological processes exert an important impact on Mg coprecipitation in foraminifers Intra-annual Mg/Ca-d 18 O c Relationship [24] Mg/Ca and salinity-corrected shell d 18 O c data for two mixed-layer dwelling species show a strong, but analogous intra-annual variation (Figure 3) over a 6 C change in estimated isotopic temperature. The amplitude of Mg/Ca and d 18 O c variability in G. ruber is significantly larger than G. sacculifer and N. dutertrei, implying that G. ruber grew and calcified at relatively shallower depths. This is supported by the estimated habitat depth from plankton tows and sediment traps in this area [Lin et al., 2004]. In general, for two mixed-layer dwelling species, high Mg/Ca values and light d 18 O c occur in warm periods, and low Mg/Ca and heavy d 18 O c during cold seasons. In contrast, the thermocline dweller N. dutertrei shows a small, but complicated Mg/Ca pattern (see Figures 3g 3i), possibly reflecting relative stable ambient conditions, changes in thermocline depth or hydrographic variations by regional monsoon upwelling [Liu et al., 2002]. [25] It is interesting to note that Mg/Ca and d 18 O c becomes weaker correlated and foraminiferal Mg/ Ca ratio seems to decrease gradually with trap depth for the mixed-layer dwellers (Figures 3a 3f). As d 18 O c do not change significantly with depth, other factors should have changed the primary shell Mg/Ca. Calcification of the planktonic foraminifer occurs in the upper 100 m and descend to trap collectors after death. It takes 3 days to 2 weeks to reach collecting cups depending on its mooring depth, shell density, and sinking process in the water column. These tests are subject to chemical and physical alteration in the water column. Consequently, dissolution and/or secondary overgrowth associated with diagenetic phases cause alteration of Mg/Ca in the foraminiferal calcite [Lohmann, 1995]. Furthermore, foraminiferal Mg/Ca could be affected by lateral advection 10 of 20

11 from the calcification area with different temperature before settling [Boltovskoy et al., 1996]. However, more detailed information on oceanic current system in the SCS is necessary for further understanding of the lateral transportation effect on Mg/Ca ratios. [26] Although foraminiferal d 18 O c has been widely used as a temperature-dependant proxy, an ontogenetic artifact can contribute to the discrepancy due to differences in size used for d 18 O c and Mg/Ca measurements. In addition, S-d 18 O sw correlations applied for salinity correction is not contemporaneous with actual sampling periods. The similar annual d 18 O c cycles observed at various trap depths suggest that main discrepancy between Mg/Ca and d 18 O c is likely caused by partial dissolution artifacts [Brown and Elderfield, 1996; Rosenthal et al., 2000]. However, there is a small d 18 O c variation at some depths (e.g., Figures 3a 3c) and implies that shell d 18 O c may have been slightly altered by partial dissolution [Wu and Berger, 1989; Rosenthal et al., 2000] or other complicated factors In Situ Mg/Ca-SST Calibration [27] Mg/Ca in planktonic foraminiferal shells are primarily controlled by seawater temperature and cause significant correlation (r 2 = 0.64) between Mg/ Ca and estimated habitat temperature (Figure 4a). A further comparison of foraminiferal Mg/Ca and d 18 O calcification temperature calculated using paleothermometer equation by Bemis et al. [1998] (Figure 4b) shows that higher Mg/Ca is not always related to shallow depth, indicating that in addition to seawater temperature and partial dissolution, other environmental variables, such as lateral transport, could also significantly modify the primary foraminiferal Mg/Ca ratio. [28] To compare with data of Anand et al. [2003] using d 18 O calcification temperature, we assumed a temperature sensitivity (exponential constant = 0.090) for multiple trap species in the Sargasso Sea. A SCS temperature calibration can be described for all three species: Mg/Ca = 0.32 (±0.05) exp (0.090*T), r 2 = This calibration curve (see Figure 5a) deviates slightly from Anand et al. [2003] in the Sargasso Sea, Mg/Ca = 0.38 (±0.02) exp [0.090 (±0.03)*T]. A relative large scatter of the Mg/Ca ratio for all species was apparent in this study and can be understand in terms of regional hydrological differences caused by seasonal monsoon and temporal water mass variation in marginal sea. Hydrographic forcing could result in large differences in chemical properties at surface layers. Considering the uncertainty, there is no large difference between the equations obtained from the SCS and the Sargasso Sea (Figure 5a). However, the Mg/Ca data in the SCS obviously show a much larger scatter and fall below the Anand et al. s [2003] calibration curve, especially for those data collected from the spring-summer season. The calibration curve obtained for SCS during the warm period deviates significantly from Anand et al. [2003] in the Sargasso Sea (see Figure 5b). [29] Previous studies have shown that the reductive steps in cleaning planktonic foraminifers caused slight dissolution in the tests and biased the shell Mg/Ca [Barker et al., 2003; Rosenthal et al., 2004]. This effect, however, cause less than 10% Mg/Ca variation, agrees with our own experimental evaluations (see auxiliary material Figure S2), and is much smaller than the Mg/Ca variation (>30%) observed in the SCS trap samples. Other natural variables (e.g., partial dissolution and lateral transport), therefore, may be responsible for the detected Mg/Ca variations. The degree of partial dissolution is an important factor to consider for causing scatter in Mg/Ca ratio for the studied foraminiferal species. Dissolution due to degradation or growth of organic matters inside the traps was minimized by proper preservative during the trap deployment (see section 4.1 for details). [30] A plot of Mg/Ca and estimated temperature shows distinct trends between two collecting periods (Figure 5b). The fall-winter trend (Mg/Ca = 0.36 (±0.08) exp (0.090*T), r 2 = 0.74, from Oct. to Jan.) is more close to the Mg/Ca-SST equation of Anand et al. [2003] than that of the spring-summer (from Mar. to Aug.) trend. The spring-summer trend (Mg/Ca = 0.30 (±0.04) exp (0.090*T), r 2 = 0.83) is clearly biased toward the low Mg/Ca ratio or cold temperature. This most likely is a consequence of the partial dissolution artifact in the SCS (see section 6.5 for a more detailed discussion). However, the species-specific calibration at SCS could not be established statistically because sample collections were conducted at different water depths and seasons, as well as the small changes in habitat temperatures for each species. Future study focused on the species-specific calibration would provide valuable information on temporal variability of thermal structures in water columns in the climate-sensitive at low latitude region. [31] Another approach for the Mg/Ca empirical equation can be theoretically done by incorporating the dissolution term (e.g., depth or [CO 2 3 ][Dekens et al., 2002]). Since there is no detectable change 11 of 20

12 in Mg/Ca-derived equations from different depths during warm periods (Figure 6b), we can utilize the calibration curves without any dissolution correction. In contrast, in the fall-winter period, the Mg/ Ca-SST equations are correlated with depths (Figure 6a) where pre-exponent constant is a function of depth, *depth (km), r 2 = In principle, one can describe a statistical correlation between pre-exponential constant and depth for all three species as: Mg=Ca ¼ ð0:38 0:02* water depth ðkmþþ expð0:090*tþfor fall winter ð2þ Mg=Ca ¼ 0:30 expð0:090*tþfor spring summer ð3þ Although the dissolution-corrected term for preexponential constant is derived from limited data points (i.e., 4 trap depths), this calibration is rather reliable as it faithfully reflected water column conditions in the SCS, and high correlation coefficients (r 2 >0.71) of the best-fit equations were obtained at various depths. One should be able to estimate annual SST records if average warm and cold SSTs were obtained using dissolution-corrected calibration curves established here Effect of Partial Dissolution on Shell Chemistry [32] Previous studies have shown that shell chemistry systematically decreases through post-depositional dissolution in un-saturated deep-water or pore fluids [Lorens et al., 1977; Rosenthal et al., 2000] and is thought to be influenced by preferential dissolution of high Mg calcite of the foraminiferal test. These high Mg calcites were formed at warmer temperature and partial dissolution therefore plays a major factor in shell chemistry [Rosenthal and Boyle, 1993; Brown and Elderfield, 1996]. [33] Considerable calcium carbonate dissolution (up to 60 80%) might take place in the upper m water column based on excess alkalinity signals [Milliman et al., 1999]. Although estuarine and/or coastal benthic processes [Chen, 2002], or mixing of deeper waters with shallower waters [Friis et al., 2006] may contribute to excess alkalinity in the upper ocean, several hypothesis, including biological controls (e.g., a corrosive micro-environment due to microbial oxidation of organic matter [Milliman et al., 1999]) or preferential dissolution of more soluble carbonate phases (e.g., high-mg calcite [Feely et al., 2002]), have been proposed to explain these dissolution artifacts. Consequently, similar processes may result in dissolution of foraminiferal shells and modify Mg/ Ca at depths well above lysocline or CCD. [34] The partial dissolution problem has been tackled successfully by incorporating a dissolution term using information of water depth, seawater [CO 2 3 ], foraminiferal test weights [Dekens et al., 2002; Rosenthal and Lohmann, 2002]) into calibration equations, or applying a calcite dissolution proxy (e.g., Globorotalia menardii fragmentation index, MFI, or Mg/Ca and Mg/Sr in deep dwelling planktonic species [Mekik and François, 2006]). These methods, however, have unavoidable assumption on the extent of dissolution. For example, the initial weight of planktonic foraminiferal species needed to be assumed for dissolution weight loss calculation. This assumption has been challenged as the initial weight could vary regionally and temporally [Barker and Elderfield, 2002]. In this study, we use foraminiferal shells from different trap depths to evaluate directly of regional dissolution artifacts on shell Mg/Ca. It provides not only the dissolution-corrected temperature equation, but also an opportunity to evaluate variability of shell chemistry with depth, which is a function of partial dissolution. [35] The correlation coefficients (r 2 ) of foraminiferal Mg/Ca and T H derived from different depths are always better than 0.64 (Figure 6). The best-fit lines show effects of partial dissolution in the SCS and are characterized by a gradual decrease of the pre-exponential constant from surface to deep collecting cups (Figure 6). Another feature is that the shallowest curve at 240 m showed a similarity with species established in the Sargasso Sea [Anand et al., 2003]. The shallowest and the deepest (3240 m, below lysocline) traps in cold months shows 20% difference in Mg/Ca at the same temperature (Figure 6a), equivalent to 2.2 C assuming of 9% sensitivity. For the warm months, because the shallowest trap depth (600 m) is much deeper than that in the cold months, a less Mg/Ca variation is found among different deployed depths (Figure 6b). If assumed the same empirical Mg/Ca-SST equation for the shallowest depth (240 m), the dissolution effect in warm months seems to be more intense (10% and 25% decreases in Mg/ Ca for 1240 m and 1126 m, respectively). The observed large differences in Mg/Ca calibration equations at different localities and seasons emphasize the importance of in situ calibration for dissolution artifact in various ocean basins [Rosenthal 12 of 20

13 Table 3. Comparison of Dissolution Effects on Foraminiferal Shell Mg/Ca as Observed in Specimens From Core Top Sediments and the Sediment Trap a Species Sample Location Mg/Ca, %/km Water Depth, m DCO 3 2, mmol kg 1 Reference G. ruber Core-top sediment Ontong Java Plateau to 9 Lea et al. [2000] Ontong Java Plateau to 15 Deken et al. [2002] Ceaea Rise to 3 Deken et al. [2002] Sierra Leone Rise to 7 Deken et al. [2002] South China Sea to 7 Huang et al. (in preparation, 2008) Sediment Trap South China Sea to 16 this study G. sacculifer Core-top sediment Ceaea Rise to 3 Russell et al. [2004] Ontong Java Plateau to 15 Deken et al. [2002] Ceaea Rise to 3 Deken et al. [2002] Sierra Leone Rise to 7 Deken et al. [2002] Ontong Java Plateau to 3 Rosenthal et al. [2000] Ceaea Rise to 21 Rosenthal et al. [2000] Sierra Leone Rise to 7 Rosenthal et al. [2000] South China Sea to 7 Huang et al. (in preparation, 2008) Sediment Trap South China Sea to 16 this study N. dutertrei Core-top sediment Ontong Java Plateau to 15 Deken et al. [2002] Ceaea Rise to 3 Deken et al. [2002] Sierra Leone Rise to 7 Deken et al. [2002] South China Sea to 7 Huang et al. (in preparation, 2008) Sediment Trap South China Sea to 16 this study a Changes in Mg/Ca ratios of core top sediments were calculated and expressed as percentage change per km water depth, relative to the dissolution onset in a water depth profile. For the sediment trap samples, the percentage changes were calculated using average Mg/Ca at the shallowest and deepest water depths. Values for DCO 2 3 (= [CO 2 3 ] in situ [CO 2 3 ] saturation ; CO 2 3 ] saturation were obtained on the basis of [CO 2 3 ] saturation =90exp [0.16(Z-4)]) were calculated using hydrographic data from nearby time series station (W.-C. Chou et al., unpublished data, 2007). Observed trends vary intraspecies and interspecies and show a complicated relationship between changes in Mg/Ca and deep ocean DCO 2 3. Note that decreases in Mg/Ca from core top sediments in the South China Sea are much larger than sediment trap results. and Lohmann, 2002; Dekens et al., 2002], especially for those with distinctive water chemistry. [36] Particle sinking rate estimated in the SCS is 300 m/day (H.-L. Lin et al., unpublished data, 2007). The observed of Mg/Ca trend thus should reflect the special variations of dissolution at various depths as these samples were collected at the same time. Among the three planktonic species, N. dutertrei appears to be more susceptible to dissolution and shows the strongest decrease in Mg/Ca of 12%/km, compared to 4%/km and 5%/ km for G. ruber and G. sacculifer, respectively (see Table 3). These percentage changes were calculated using average Mg/Ca at the shallowest and the deepest depths, and assumed linear variation at other depth. In summary, N. dutertrei, the nonspinose thermocline dweller, is the most sensitive species to dissolution and the spinose, mixed-layer dwelling species (G. ruber and G. sacculifer) are relatively resistant. This agrees well with previous observations of selective dissolution susceptibility for planktonic species in core top samples from several transects in tropical Pacific and subtropical Atlantic [Dekens et al., 2002] (Table 3). [37] The foraminiferal Mg/Ca in the SCS shows a relatively strong dissolution compared with those in open oceans (Figures 5 and 6). It is possible that upwelling may play a significant role in changing the chemical composition of surface seawater through uplift of the deep subsurface, nutrientenriched water. Several evidences show that regional upwelling forced by wind-driven circulation have occurred in the northwest Luzon and offshore Vietnam during prevailing winter and summer monsoons, respectively [Liu et al., 2002]. Deep water with relatively low ph is admixed upwardly 13 of 20

14 with surface water and becomes more corrosive, which in turn results in more intense carbonate dissolution than other marginal seas. In summertime, more corrosive surface seawaters caused by upwelling offshore Vietnam could be delivered to the northern SCS through a clockwise gyre, which may partly explain the cause of strong dissolution in warm seasons [Wyrtki, 1961]. Alternatively, a recent study at the SEATS time series site in the northern SCS indicates slightly high pco 2 (i.e., lower ph) during the warm seasons (W.-C. Chou et al., unpublished data, 2007), and provides another example for more corrosive waters in the summertime. More detailed hydrographic studies are required to evaluate further of this contribution. [38] This study found that Mg/Ca is strongly affected by selective partial dissolution in the SCS, even at depths well above the lysocline and calcite saturation depth (Figure 7 and Table 3). This argument is supported by weight data of shell separated from core top sediments showed a depth-related trend for selected species in the SCS (K.-F. Huang et al., manuscript in preparation, 2008) (see auxiliary material Figure S3). Because shell weights may also vary with growth condition [de Villiers, 2004], it is necessary to evaluate the carbonate ion concentration in the surface SCS. On the basis of carbonate data from time series SEATs station (W.-C. Chou et al., unpublished data, 2007), there is no seasonal variability in the study area. Therefore, the observed shell weight loss should be mainly affected by partial dissolution Other Environmental Factors on Foraminiferal Mg/Ca [39] Although work to date has led to an important advance in the use of foraminiferal Mg/Ca as a reliable paleothermometry, uncertainties still exist and need to be evaluated for other secondary environmental factors. These factors include shell growth rate, ph, salinity and other biological effects [Delaney et al., 1985; Elderfield et al., 1996; Rosenthal et al., 1997; Lea et al., 1999]. [40] The d 13 C in foraminiferal shell is commonly in disequilibrium with ambient seawater. A considerable body of work regarding the foraminiferal d 13 C has interpreted in terms of kinetic isotopic fractionation and incorporation of metabolic or photosynthetic CO 2 [Spero and Williams, 1998]. Core top studies have demonstrated that Mg/Ca was affected by non-equilibrium process to form correlation between d 13 C and size fraction [Elderfield et al., 2002]. Lin et al. [2004] reported a tight correlation between d 13 C(G. ruber and G. sacculifer) and integrated primary production in the northern SCS, where low d 13 C is associated with high nutrient level in surface water, possibly related to a prevailing winter monsoon. Consequently, d 13 C in planktonic foraminifers may serve as an indicator for nutrient levels in surface water. [41] Lea et al. [1999] and Russell et al. [2004] indicate that shell Mg/Ca decreases with ph or [CO 2 3 ]ioninorbulina universa and Globigerina bulloides and explain this in terms of the secondary kinetic influence, high [CO 2 3 ] leading to fast shell growth. It is evident also that high [CO 2 3 ] leads to an increase in shell thickness and weights [Barker and Elderfield, 2002]. Our Mg/Ca analyses were restricted to shells within a narrow size range, mm, to eliminate ontogenetic effect and to minimize artifacts caused by non-equilibrium processes. The changes in foraminiferal d 13 C, therefore, are primarily controlled by fertilization of surface water [Lin et al., 2004]. There is no appreciable correlation between sphere Mg/Ca and d 13 C (Figure 8a), indicating minor kinetic and nutrient level artifacts. [42] Temperature is a conservative variable that is highly correlated with other oceanographic parameters. For instance, it correlates positively with salinity and seawater [CO 2 3 ] in the open ocean. Yet separation of these temperature-induced artifacts in modern empirical sediment trap is not straightforward. Salinity is another factor that may bias foraminiferal Mg/Ca ratios. Living laboratory culture experiments showed about +4(±3)% Mg/Ca per psu for O. universa at 22 C [Lea et al., 1999], suggesting potential salinity influence on Mg/Ca in shell. However, salinity changes are relatively small in natural systems, compared with potential errors derived from the culture experiments. Therefore, it is difficult to evaluate the salinity effect using culture samples. Besides, species difference may generate uncertainties in estimating biological response under various environmental conditions. [43] Modern hydrological data have shown that continental runoffs and precipitation were modulated by the EAM, SSS seasonal variation is psu. Applying an empirical relationship derived from the culture experiments, it is expected a 6% or maximal to 10% variation in shell Mg/Ca between wet and dry periods. Several observations lead to a similar conclusion that salinity is not a predominant factor influencing foraminiferal Mg/ Ca in the SCS. First of all, there is a negative 14 of 20