Inter- and intra-specimen variability masks reliable temperature control on shell Mg/Ca ratios in laboratory- and field-cultured Mytilus edulis and Pecten maximus (bivalvia)

Abstract. The Mg/Ca ratios of biogenic calcite is commonly seen as a valuable palaeo-proxy for reconstructing past ocean temperatures. The temperature dependence of Mg/Ca ratios in bivalve calcite has been the subject of contradictory observations. The palaeoceanographic use of a geochemical proxy is dependent on initial, rigorous calibration and validation of relationships between the proxy and the ambient environmental variable to be reconstructed. Shell Mg/Ca ratio data are reported for the calcite of two bivalve species, Mytilus edulis (common mussel) and Pecten maximus (king scallop), which were grown in laboratory culturing experiments at controlled and constant aquarium seawater temperatures over a range from ~10 to ~20°C. Furthermore, Mg/Ca ratio data of laboratory- and field-grown M. edulis specimens were compared. Only a weak, albeit significant, shell Mg/Ca ratio–temperature relationship was observed in the two bivalve species: M. edulis (r2=0.37, p


Inter-and intra-specimen variability masks reliable
record high-resolution time-series of those environmental conditions in which the organism grew. Furthermore, marine bivalves occupy widely distributed habitats in the modern-day oceans, as well as being relatively common throughout the fossil record since the Cretaceous. Information on past environmental conditions that are preserved in carbonates can be obtained through the use of proxies, i.e. physical and chemical 10 signals that provide information on sought after variables that cannot be measured directly, such as seawater temperature or salinity. However, a proxy is rarely dependent on a single variable, and the influence of other secondary independent variables complicates, to a lesser or greater extent, proxy use in palaeo-studies; such factors must be assessed rigorously via calibration and validation studies prior to successful 15 application (for reviews, see e.g. Wefer et al., 1999;Lea, 2003). The use of the oxygen-isotope composition ( 18 O/ 16 O ratios expressed as δ 18 O values) of biogenic carbonate archives as a proxy for seawater temperature (for reviews, see e.g. Emiliani, 1966;Wefer and Berger, 1991) is one of the most powerful tools in palaeoceanographic studies (e.g. Shackleton, 1967;Shackleton and Opdyke, 1973; fluence the Mg/Ca ratios of biogenic calcites. For example, biological influences such as gametogenesis, ontogeny, growth rate and size, as well as environmental and physical parameters such as salinity, pH and post-depositional dissolution, have all been proposed to significantly influence foraminiferal Mg/Ca ratios (Delaney et al., 1985;Lea et al., 1999;Elderfield et al., 2001;Bentov and Erez, 2005). Furthermore, observations of significant small-scale intra-shell heterogeneity in Mg contents indicates a strong physiological control on the Mg/Ca ratio of biogenic calcites, such as observed in foraminifera (Rio et al., 1997;Hathorne et al., 2003;Eggins et al., 2004;Bentov and Erez, 2005;Sadekov et al., 2005), ostracodes (Rio et al., 1997) and bivalves (Lorens and Bender, 1980;Rosenberg et al., 2001). 15 In calcitic bivalve molluscs the occurrence of a temperature control on shell Mg/Ca ratios has been the subject of several studies that have returned contrasting results, but nevertheless shell Mg/Ca ratios have been used to reconstruct palaeotemperatures from fossil bivalves (Klein et al., 1997;Immenhauser et al., 2005). In an early study, a weak positive correlation between shell calcite Mg concentration with temperature Introduction  (Freitas et al., 2005). For other bivalve species, such as Pecten maximus (king scallop), there also exists no clear temperature relationship; Lorrain et al. (2005) reported an absence of a significant correlation between Mg/Ca ratios and temperature for this species while a weak, albeit significant, Mg/Ca ratio to temperature relationship was also observed (Freitas et al., 2006), with the relationship breaking down during winter 5 months. Furthermore, several studies report, or suggest, the occurrence of significant non-thermodynamic controls on the Mg content of bivalve mollusc calcite, such as salinity (Dodd, 1965), solution Mg/Ca ratios (Lorens and Bender, 1980) or the animal's metabolism (Lorens andBender, 1977, 1980;Vander Putten et al., 2000). Significant small-scale heterogeneity in Mg content also has been described for bivalve shell 10 calcite. Such variability has been associated with stress (Lorens and Bender, 1980), metabolic activity (Rosenberg and Hughes, 1991) and control of shell crystal elongation (Rosenberg et al., 2001). The purpose of this study was to advance an understanding of the degree of variability of Mg/Ca ratios in calcite bivalve shells using a controlled laboratory aquarium cul- 15 turing approach. Specifically, no laboratory calibration of the Mg/Ca ratio-temperature relationship in bivalve calcite has previously been performed under constrained and constant seawater temperatures. This approach is a significant advancement on previous studies, since it enables manipulation of specimens, control of environmental variables, and measurement of other parameters, such as size and growth rate. It 20 must be acknowledged, however, that laboratory aquaria are not a true representation of the animal's natural habitat. However, the outcomes of laboratory culturing studies are only of value when validated by field-based studies, albeit with the latter suffering from a lesser degree of constraint of environmental variables. In summary, the ultimate goal of this investigation was to determine whether a reliable calibration of the Mg/Ca Introduction individual control systems, with a maximum resolution achievable by these controllers of 1 • C. For improved constraint, seawater temperature also was monitored in each aquarium every 15 min using submerged temperature loggers (Gemini Data Loggers TinyTag -TGI 3080; accuracy of ±0.2 • C). The intermittent lack of temperature control in some aquaria is a limitation of the aquarium system used and most manifest at the 10 lowest nominal temperature of 10 • C, when the cooling system sometimes struggled to compensate for fluctuations in the temperature of the external seawater supply. Natural seawater is pumped from the proximal Menai Strait into settling tanks before being introduced as a common supply into the laboratory aquaria.

EGU
Once the animals had acclimatised, individual specimens were removed at weekly 15 intervals (with the exception of the last growth interval in experiment two, which was longer than a week for both the 15 • C and 20 • C aquaria) to be processed. Each time the M. edulis specimens were removed from the aquaria they were exposed to the air for 5 to 6 h, while P. maximus specimens were kept in small holding tanks for periods of 30 to 45 min. Both methods resulted in emplacement of a disturbance mark on the surface of 20 the shells. The shells then were photographed and digitally imaged using the AnalySIS software package. The combination of disturbance marks and photographs was used to identify and measure all shell growth between emersions and provided a time control of the new shell growth laid down throughout the experiments. Subsequently the term "growth interval" has been used to describe the time intervals between emersions of 25 animals (Table 1). The duration of the experiments, and hence the number of growth intervals, varied with species and aquarium temperature (  The animals were all less than 1 y old when deployed, obtained from one spat cohort and initially ranged from 2.0 to 2.7 cm in shell length. This raft is moored in 20 the close vicinity (ca. 500 m) of the School of Ocean Sciences, Bangor University in a section of the Menai Strait where the water column is completely mixed, due to strong turbulent tidal mixing (Harvey, 1968). Animals were deployed in mesh cages and each shell was identified by a mark hand drilled on its surface. Two different, but parallel, experimental approaches were taken: 1) "short" deployment specimens were placed into 25 cages for 16 short, well-defined and consecutive growth intervals that together covered the duration of the entire field experiment. The duration of each growth interval varied during the experiment according to expected seasonal changes in shell growth rate; 2) 538 Introduction

EGU
In contrast to the short specimens, "annual" deployment specimens were placed in the field for the entire duration of the experiment. To ensure that short-deployment specimens were in the same physiological condition as their annual counterparts, and to avoid the inevitable period of acclimation if animals were deployed directly from laboratory conditions, the former were taken at the start of each growth interval from a stock 5 of animals deployed at the beginning of the experiment and kept in a separate cage. At the end of each growth interval (which also was the beginning of the next growth interval) all short-deployment specimens from the preceding deployment and all annual specimens were removed from the raft, together with a new set of short-deployment specimens taken from the stock that were to be deployed during the next growth inter-10 val. All of these shells then were photographed and digitally imaged using the Anal-ySIS software package. Between each deployment growth interval both short-and annual-deployment M. edulis specimens were exposed to the air for 5 to 6 h resulting in emplacement of a disturbance mark on the surface of the shells. The combination of disturbance marks and photographs was used to identify and measure all shell growth 15 for each growth interval, as well as shell height (i.e. the distance from the umbo to the shell margin along the main axis of growth), and thus provide a time control of the new shell growth laid down throughout the field experiment by assuming shell growth rate to be constant during each growth interval.
Seawater temperature was monitored every two hours throughout the experimen-20 tal deployment period using submerged temperature loggers placed in the mesh cages containing the animals (Gemini Data Loggers TinyTag -TGI 3080; accuracy of ±0. were cleaned in a similar manner to the Pecten shells but, in addition, the outer organic periostracum was milled away with the drill until periostracum-free shell was visible in the entire sampling area. Shell powder samples subsequently were taken from the new shell growth by milling to a depth of ca. 200µm. Accurate milling was completed under a binocular microscope fitted with an eyepiece graticule, and depth and width 10 of milling were controlled carefully. Each milled powder sample was taken from the main axis of shell growth: in P. maximus from the mid 2-3 axial ridges (ribs), and in M. edulis from the middle section, to avoid the increase in shell curvature that occurs away from the main growth axis (Fig. 2). Only one powder sample was milled from each individual growth interval and, particularly at the lower temperatures, the milled powder 15 from one or more growth intervals had to be combined to provide enough shell material for analysis. Whenever the amount of sample permitted, single milled powder samples were split into separate aliquots for Mg/Ca and stable-isotope ratio determinations, otherwise only Mg/Ca ratios were measured. Lorens and Bender (1980) have described that the stress of capture and adaptation 20 to a new laboratory environment induced the deposition of a region of shell (termed "transition zone calcite" by those authors) with higher Mg/Ca ratios. Therefore, it is possible that the regular handling disturbance (for measurement purposes) imposed on the animals of both species during the present study may explain some of the variability of shell calcite Mg/Ca ratios that have been observed. However, Lorens and 25 Bender (1980)  EGU sequential shell growth occurs at an angle to the shell surface. Also, no spatial scale was reported in the Lorens and Bender (1980) study making it difficult to determine the size of the region of "transition zone calcite". Therefore, in the current study the sampling of individual growth intervals by hand-milling of powder samples between the disturbance marks on the surface of the shell, that represent the times of immersion 5 during the experimental period, has minimized the influence of handling disturbances on measured Mg/Ca ratios.

Field culturing experiment
The left hand valve of two short-deployment M. edulis specimens were sampled for each growth interval, while three annual M. edulis specimens (A2, A6, A20) were se- 10 quentially sampled for all growth intervals. The milling of shell powder samples was as described for the laboratory culturing experiment. Whenever the amount of shell growth permitted more than one sample was collected from a single growth interval. On such occasions the new shell growth was equally divided between the number of samples collected (2≤N≤4).

Shell stable-isotope and elemental ratio analyses
The shell milled powder sample preparation and analytical methodologies used in this study are as described in detail in Freitas et al. (2005;2006). Shell and water oxygen stable-isotope data are reported in per mil (‰) deviations relative to VPDB and VSMOW, respectively. The overall analytical precision for shell δ 18 O measurements 20 based on analyses of an internal laboratory standard run concurrently with all M. edulis and P. maximus samples analysed in this study is 0.08‰ (1 σ; N=32). Sufficient material was not available from any one growth interval to enable replicate stable-isotope analyses for an assessment of true sample precision; however, Freitas et al. EGU maximus specimen. Calibration for Mg/Ca ratio determinations was performed via an established ICP-AES intensity-ratio method (de Villiers et al., 2002), using synthetic standard solutions in the range 0t25 mmol/mol for Mg/Ca, and most at Ca concentrations of 50 (N=304) and 60µg/ml (N=161). The smallest milled powder samples were analysed at 30µg/ml 5 (N=102). Measurements were made using the Perkin Elmer Optima 3300RL ICP-AES instrument housed at the NERC ICP Facility, Royal Holloway University of London. Instrumental drift was monitored by running an intermediate (16 mmol/mol) calibration standard every 5 to 10 samples and data then were corrected accordingly. Analytical precision (expressed as relative standard deviation or RSD) was 0.5% for the labora-10 tory cultured specimens (N=86) and 1.3% for the field cultured specimens (N=29). In the laboratory culturing experiments, sufficient material was not available from any one growth interval to enable replicate analyses for an assessment of true sample precision; in the field experiment, however, sample precision was better than 6.2% RSD for replicate measurements (N=3) of the same milled powder samples obtained from 15 five M. edulis specimens. Furthermore, Freitas et al. (2006) used the same method as reported here and obtained a Mg/Ca ratio precision of 3.5% RSD for five replicate measurements of the same milled powder sample obtained from one P. maximus specimen. For comparison with past and future datasets, Mg/Ca ratio measurements also are reported for a set of solutions prepared by the Elderfield group at the University of 20 Cambridge, UK (Greaves, personal communication, 2003;cf. de Villiers et al., 2002), as well as for three solutions (BAM-RS3, ECRM-752 and CMSI-1767) that have been proposed as certified reference materials (CRMs) for Mg/Ca ratio measurements in carbonates (Greaves et al., 2005) and that are subject to an ongoing international interlaboratory calibration study (Table 2). For each CRM, approximately 50 mg of powder 25 was dissolved in 50 g of 0.075 M HNO 3 (Merck Ultrapur), resulting in Ca concentrations in solution of ca. 400 µg/ml. Subsequently, 1.5 ml of each solution was centrifuged for 10 min and an aliquot then was pipetted into clean 12 ml centrifuge tubes and diluted to 10 ml to give final Ca concentrations of 50 and 30 µg/ml in order to match the sample EGU and standard solutions. The linearity of the intensity-ratio calibration lines, combined with the independent assessment of the accuracy of the analytical procedure (Freitas et al., 2005(Freitas et al., , 2006, confirms the veracity of the M. edulis and P. maximus Mg/Ca ratios obtained in this study.

Statistical analyses
5 Two-sample t-tests were used to determine statistically whether significant differences existed between measured shell Mg/Ca ratios precipitated at different seawater temperatures in pairs of constant-temperature aquaria. Herein, probability levels less than 5% (p<0.05) are considered significantly different. Linear regressions and ANOVA analyses of shell Mg/Ca ratios and seawater temperature were performed using the software 10 package MINITAB. Regressions were compared by testing the equality of variance in the regression residuals, since unequal variance in the regression residuals (F -test, p<0.05) indicates significantly different regressions. GLM ANOVA was used to test for differences in the slope and intercepts of the regressions. The variability in shell Mg/Ca ratios attributable to different factors was determined using fully nested ANOVA.

Culture conditions and confirmation of shell precipitation in thermodynamic equilibrium
Seawater temperature was stable during laboratory culturing experiment one, but more variable during experiment two, especially in the lower (10 • C) and 20 mid (15 • C) temperature aquaria (Fig. 3, http://www.biogeosciences-discuss.net/5/ 531/2008/bgd-5-531-2008-supplement.pdf). Nevertheless, clear temperature differences were maintained in the three different aquaria in each of the experiments (Fig. 3). Aquaria mean seawater temperatures were 11.96±0.12 • C, 15 was compared to previously derived data for the precipitation of inorganic calcite from seawater in oxygen-isotope thermodynamic equilibrium (Kim and O'Neil, 1997), but also to species-specific palaeotemperature equations obtained for P. maximus (Chauvaud et al., 2005) and M. edulis (Wanamaker et al., 2007 15 edulis that was cultured in both experiment one and two there is a significant difference in the Mg/Ca ratio to temperature relationship between experiments and shell Mg/Ca ratios were higher in experiment two than in experiment one (Fig. 6a).
Irrespective of the observation that shell Mg/Ca ratios are significantly, albeit weakly, correlated with temperature in both bivalve species, the significant differences evident 20 in the absolute shell Mg/Ca ratios of the two species, evident in the laboratory-culturing experiment, indicate a clear species-specific Mg/Ca ratio-temperature relationship for the two bivalve species investigated in this study. Furthermore, the degree of variability of shell Mg/Ca ratios at each aquaria temperature also is higher in P. maximus than in M. edulis. Unequal variance in the residuals indicates significantly different regressions of Mg/Ca ratios with temperature between M. edulis and P. maximus from laboratoryculturing experiment two (F -test, p<0.05 Mytilus edulis shell Mg/Ca ratios also are statistically different between the two laboratory-culturing experiments, with higher values during experiment two (Fig. 6). The correlation between M. edulis shell Mg/Ca ratios and temperature also is stronger in experiment two than in experiment one (r 2 =0.38 and 0.57 in experiment one and 10 two, respectively, p<0.001 in both experiments), although this may be due solely to the smaller number of individuals analysed in experiment two compared to experiment one, i.e. the capture of a smaller degree of Mg/Ca ratio variability, as well the greater temperature range for the experiment two regression. Unequal variance in the residuals confirms significantly different regressions of Mg/Ca ratios with temperature in M. 15 edulis between experiments one and two (F -test, p<0.05). Further analysis of variance of the regressions of M. edulis Mg/Ca ratios and temperature shows that the slope of the regressions is not significantly different (F =2.50, p=0.116), but that the intercept (F =127.92, p<0.001) is different in the two experiments.

EGU
In addition to inter-individual shell variability, there is also a degree of intra-individual shell variability in Mg/Ca ratios within the dataset, i.e. between milled samples taken from different growth intervals (Fig. 7). For either species, the proportion of individual shells that produced Mg/Ca ratios significantly different among samples milled from the same specimen (i.e. the difference between any two Mg/Ca measurements was larger than twice the analytical error) was similar at each temperature. However, P. maximus showed a higher frequency of milled samples with different Mg/Ca ratios within an individual shell (>97% in all aquaria) than did M. edulis, (68%< experiment one <73%, and 72%< experiment two <83%). For the field-cultured M. edulis, measured shell Mg/Ca ratios range from 2.96 to 9.16 mmol/mol in the short-deployment specimens; from 2.86 to 8.34 mmol/mol in the A2 specimen; from 2.78 to 5.97 mmol/mol in the A6 specimen and from 2.75 to 6.11 mmol/mol in the A20 specimen (Figs. 4b and 6, http://www. 15 biogeosciences-discuss.net/5/531/2008/bgd-5-531-2008-supplement.pdf). During the field-culturing experiment, shell Mg/Ca ratios from short-and annual-deployment M. edulis specimens showed a clear seasonal pattern (Fig. 4b). Shell Mg/Ca ratios were low at ca. 4 mmol/mol during the winter and spring months up to May, and then increased to maxima higher than 7 mmol/mol during June-July. From that time to Octo-20 ber shell Mg/Ca ratios remained high at ca. 7 mmol/mol, and then decreased to ca. 3.5 mmol/mol in December 2005 (Fig. 4b).

Shell
In field-cultured M. edulis, a significant (p<0.001 for all specimens) correlation exists between seawater temperature and shell Mg/Ca ratios (Fig. 6b): r 2 =0.54 for the short-deployment specimens (N=62); r 2 =0.77 for the A2 specimen (N=28); r 2 =0.72 25 for the A6 specimen (N=34) and r 2 =0.81 for the A20 specimen (N=30 from M. edulis specimens grown under the same field-culturing conditions. Maximum shell Mg/Ca ratios, in particular, are markedly different between individual specimens and range from 5.97 to 9.16 mmol/mol. For the same range of temperature, shell Mg/Ca ratios of M. edulis grown in the laboratory-and field-culturing experiments showed a similar range (Fig. 6). However, 15 the correlation between Mg/Ca ratios and temperature was stronger in field-cultured (0.54<r 2 <0.81) than in laboratory-cultured M. edulis specimens (0.38<r 2 <0.57). Furthermore, ANOVA analysis of the regressions of Mg/Ca with temperature between laboratory-cultured and field-cultured M.edulis specimens shows that the slope of the linear regressions does not differ significantly (F =0.70, p=0.799), but that the intercept 20 does (F =224.68, p<0.001).

Inter-species, inter-individual and intra-individual variability in shell Mg/Ca ratios
In addition to the weak, but significant, relationships with seawater temperature, the shell Mg/Ca ratio data obtained in this study also clearly show a large degree of vari- EGU 6 and 7). Like with other bivalve geochemical and physical proxies (for review see e.g. Richardson, 2001) variability of shell Mg/Ca ratios occurs at different levels, requiring consideration of differences between the two bivalve species cultured (inter-species level), between shells of different individuals grown simultaneously in the same aquarium or in the same cages and under the same laboratory-and field-culturing conditions 5 (inter-individual shell level) and within individual shells, i.e. between milled samples taken from one individual shell that correspond to different growth intervals during the experimental period (intra-individual shell level). Differences in shell Mg/Ca ratios of the same species have been observed in previous field-based studies at levels of both inter-and intra-individual shell variability 10 (Rosenberg and Hughes, 1991;Klein et al., 1996;Vander Putten et al., 2000;Freitas et al., 2005;Lorrain et al., 2005;Freitas et al., 2006). For example, Klein et al. (1996) presented data from two field-collected shells (British Columbia, Canada) of the mussel M. trossulus which clearly show large Mg/Ca ratio differences at inter-(up to 2.5 mmol/mol) and intra-(up to 1.5 mmol/mol) individual shell levels, in addi- 15 tion to a temperature relationship (r 2 =0.74, p<0.001) over a range from 5.5 to 22.7 • C. By comparison, Vander Putten et al. (2000) reported inter-individual differences in Mg/Ca ratios between four M. edulis field-grown shells (Schelde Estuary, Netherlands) as high as ∼7 mmol/mol. Similarly, Lorrain et al. (2005) presented data from four P. maximus specimens collected from the Bay of Brest, France, where differences of up 20 to 6 mmol/mol in Mg/Ca ratios were observed between individual specimens for shell samples that corresponded to the same time of calcification. Most recently, in three P. maximus specimens grown in a field-based experiment, and for a similar temperature range to that used in the present laboratory culturing study (10 to 20 • C), differences were observed in Mg/Ca ratios of up to 7.5 mmol/mol between shells (Freitas et al., EGU ies thus are of similar magnitude. Significant differences in absolute shell Mg/Ca ratios can be observed between the two cultured bivalve species (Fig. 6); P. maximus shell Mg/Ca ratios being approximately three times higher than those in M. edulis. Large variations in the Mg content of biogenic calcite from different species has been observed previously in bivalves 5 (Lorens and Bender, 1980;Klein et al., 1996;Vander Putten et al., 2000;Lorrain et al., 2005). The Mg/Ca ratio data obtained in this study for laboratory cultured M. edulis and P. maximus have been compared to previously published data for other marine bivalve species investigated in field-based studies (Fig. 8). On the whole, a large degree of overlap can be observed between the Mg/Ca ratio data derived from the laboratory-10 and field-cultured M. edulis and P. maximus specimens. Nevertheless, laboratory-and field-cultured M. edulis show lower shell Mg/Ca ratios than data reported from field experiments for M. edulis by Vander Putten et al. (2000), although the latter data were obtained by laser ablation ICP-MS and there is the potential for calibration issues between datasets. The Mg/Ca ratios for M. edulis cultured in this study are, however, 15 similar to Mg/Ca ratios reported for M. trossulus (Klein et al., 1996), a close relative of M. edulis. Shell Mg/Ca ratios in the P. maximus animals cultured in this study are similar to Mg/Ca ratios reported for specimens of the same species grown or collected in field studies (Lorrain et al., 2005;Freitas et al., 2006), but extend to higher values and also show a larger variability than in specimens grown at a field location adjacent 20 to the present aquarium based study (Freitas et al., 2006). This latter observation suggests that the influence of any non-temperature control (i.e. a physiological control) on P. maximus shell Mg/Ca ratios may well be stronger under laboratory culture conditions than in field-based situations that more closely mimic the conditions best suited for optimal growth of natural populations. 4.2 Imprecise temperature control on shell Mg/Ca ratios Despite the observation that the shell oxygen-isotope composition was deposited in or near to oxygen-isotope thermodynamic equilibrium (Fig. 5) EGU the measured shell Mg/Ca ratios obtained from specimens of M. edulis and P. maximus cultured in the constant-temperature aquaria in this study is that there is only a weak dependence on temperature ( Fig. 6a and Table 3). Nevertheless, in M. edulis specimens cultured in the field, shell Mg/Ca ratios were significantly correlated with temperature (0.54<r 2 <0.81). However, the inter-individual variability of Mg/Ca ratios is 5 large (Fig. 6b) and results in a weaker correlation (r 2 =0.50, p<0.001) when data from all field cultured specimens are pooled together. Furthermore, linear regressions of Mg/Ca ratios with temperature are different: between laboratory cultured P. maximus and M. edulis, between M. edulis specimens grown in the two laboratory culturing experiments, between laboratory-and field-cultured M. edulis specimens, and between individual field-cultured M. edulis specimens. Therefore, the establishment of even a species-specific valid regression between Mg/Ca ratios and temperature has not been possible for the two species studied. The temperature dependent incorporation of Mg into inorganic calcite experimentally precipitated from seawater (Chilingar, 1962;Katz, 1973;Burton and Walter, 1987;15 Mucci, 1987;Oomori et al., 1987) and in other biogenic calcites (e.g. Dwyer et al., 1995;Nurnberg et al., 1996;Stoll et al., 2001) has been well-documented. Clearly, as has been observed previously in some field studies, the weak temperature dependence of shell Mg/Ca ratios in the two marine bivalve species that were investigated in this study is a feature specific to the incorporation of Mg into the calcite shells of these organisms 20 and must therefore relate to their specific biomineralization processes, including any secondary physiological influences.
The suggestion of additional physiological controls on Mg/Ca ratios in bivalve calcite (i.e. metabolic or kinetic controls) is supported further by the significant difference in the absolute shell Mg/Ca ratios in M. edulis grown at 15 • C in two aquaria in experi-25 ments one and two of ∼1.1 mmol/mol (t-test, p=0.004, DF=24), but also by the large inter-and intra-individual variability of Mg/Ca ratios observed in both the laboratoryand field-cultured specimens (Fig. 6). These observations clearly indicate that specimens from the same species cultured at different times at the same seawater temper- EGU ature can have different shell Mg/Ca ratios. It is therefore important to recognise that other non-thermodynamic factors in the two experiments must also have influenced shell Mg/Ca ratios. Furthermore, it is not possible to discount the possibility that the M. edulis animals cultured in experiment two were better conditioned for the laboratory environment than those in experiment one, due to their longer acclimation in the exper-5 imental aquaria prior to commencement of the experimental period. Consequently, the outcomes of this laboratory culturing experiment must be compared to other laboratory culturing and field-based studies in order to obtain the most robust interpretations. Given the experimental design in this study, only factors that were entirely independent of seawater temperature can be discussed as additional potential controls on shell 10 Mg/Ca ratios. This consideration thus prohibits a detailed discussion of the influence of shell growth rate on shell Mg/Ca ratios, since growth rates co-vary significantly with temperature in both the laboratory-(for P. maximus, r 2 =0.62, p<0.001; for M. edulis, r 2 =0.23, p=0.001 and r 2 =0.15, p=0.032 in experiment one and two, respectively) and field-culturing experiments (for M. edulis, 0.26<r 2 <0.43, p<0.002). Nevertheless, as 15 reported by Lorens and Bender (1980) for laboratory-cultured M. edulis specimens, shell Mg/Ca ratios also were only weakly correlated to shell growth rates in both the laboratory-and field-culturing experiments (Table 3). A metabolic control, i.e. the physiological exclusion of Mg from its shell-forming fluid (the extra-pallial fluid or EPF), on calcite Mg content has been proposed previously 20 for M. edulis (Lorens andBender, 1977, 1980). Metabolic control also was suggested as a possible way of explaining an observed seasonal breakdown in the relationship between Mg/Ca and temperature reported for M. edulis (Vander Putten et al., 2000). An apparent ontogenetic control of Mg/Ca ratios has been described in the fan mussel Pinna nobilis, although a temperature control on shell Mg/Ca ratios also was present 25 in this species (Freitas et al., 2005). For P. maximus, recent field-based studies have shown the absence of a significant correlation between shell Mg/Ca ratios and seawater temperature (Lorrain et al., 2005) or a strong seasonal variation in the strength of the correlation between shell Mg/Ca ratios and seawater temperatures, again suggest-

EGU
ing that other factors must influence Mg/Ca ratios in P. maximus shell calcite (Freitas et al., 2006). Seawater salinity is a truly independent variable in the laboratory culture experiment, but not in the field-culturing experiment where it co-varies with seawater temperature (r 2 =0.50, p<0.001). Any differences in seawater salinity between the two laboratory-5 culturing experiments thus could influence the amount of magnesium available for incorporation, assuming that shell Mg/Ca ratios are not solely related to seawater Mg/Ca ratios. Indeed, Lorens and Bender (1980) have shown that shell Mg/Ca ratios increase with increasing solution Mg concentrations, albeit at much higher concentrations than would be expected from natural changes in seawater salinity. By comparison, an ear-10 lier study by Dodd (1965) observed the opposite trend of increasing Mg concentrations in M. edulis shell calcite with decreasing salinity. In addition, salinity has been reported to significantly influence the Mg/Ca ratios of foraminifera calcite (Lea et al., 1999). In M. edulis, salinity was not significantly correlated with shell Mg/Ca ratios in the two laboratory experiments in this study (p>0.05), with only a weak correlation for P. maximus 15 (r 2 =0.21, p<0.001). The strength of this correlation between shell Mg/Ca ratios and salinity is, however, of comparable magnitude to that observed between temperature and shell Mg/Ca ratios (r 2 =0.21, p<0.001). Nevertheless, temperature and salinity together (r 2 =0.37, p<0.001) still do not explain much more of the observed shell Mg/Ca variability in P. maximus than just temperature alone. Bivalves, like other calcifying organisms, are capable of regulating, or at least influencing to variable extents, the Mg content of their calcium carbonate skeletons (Dodd, 1965;Lorens and Bender, 1977;Neri et al., 1979;Onuma et al., 1979;Lorens and Bender, 1980;Rosenberg and Hughes, 1991;Rosenberg et al., 2001). This phenomenon EGU ferences in Mg/Ca ratios between and within individuals of a single species, suggest a strong physiological control of the incorporation of Mg into biogenic calcites. Examples of such physiological effects that may influence, either directly or indirectly, the Mg content of bivalve shell calcite are: variable chemical composition of the precipitating fluid, i.e. the EPF, resulting from biological control on differential transport of ions into and 5 out of the EPF; variable calcification rates; the transport and diffusion conditions of the local precipitation microenvironment (Wasylenki et al., 2005); and differences in crystal growth orientation and morphology (Mucci and Morse, 1983;Reeder and Grams, 1987;Debeney et al., 2000;Erez, 2003 have been shown to derive from Mg being concentrated along the margins of calcite prisms, especially along the terminations of the crystals, with the alignment of adjacent crystals then producing compositional growth bands within the shell (Rosenberg et al., 2001). The latter observation lead to the suggestion that in M. edulis Mg and also sulphur in the shell could control rates of shell crystal elongation, shell curvature along 20 different axes and ultimately the Mg distribution throughout the shell (Rosenberg et al., 2001). The use of Mg/Ca ratios from bivalve calcite shells as a reliable and accurate temperature proxy thus remains unlikely at present, at least in the species studied to date. The now well-documented variation of Mg/Ca ratios in bivalve calcite at species-specific, 25 inter-and intra-individual shell levels prevents the establishment of valid Mg/Ca ratiotemperature relationships, even for individual species. Furthermore, there exists support for a strong metabolic control of Mg/Ca ratios in bivalve shells (Lorens andBender, 1977, 1980;Rosenberg and Hughes, 1991;Vander Putten et al., 2000; al., 2001), although the mechanisms by which such a control acts are still not fully clear, as well as for extensive small-scale heterogeneity in shell Mg contents (Lorens and Bender, 1980;Rosenberg et al., 2001). Future research should address these issues in greater detail, if ever this geochemical proxy is to be used as a reliable and accurate temperature proxy in bivalve calcite.

Conclusions
In laboratory-and field-culturing experiments only a weak dependence on temperature, as well as a large degree of variability, has been observed for shell Mg/Ca ratios in calcite sampled from two marine bivalve species, M. edulis and P. maximus. Such variability is significant at the species, inter-and intra-individual shell levels, and most 10 likely reflects the influence of additional secondary physiological factors influencing shell biomineralisation and Mg content. Shell Mg/Ca ratios were different between M. edulis and P. maximus, being three to five times greater in the latter species. The variability of shell Mg/Ca ratios for laboratory-and field-cultured M. edulis in the present study was similar to the variability observed in previous field-grown specimens. Labo-15 ratory cultured P. maximus specimens, however, showed approximately twice the variability of shell Mg/Ca ratios than has been reported previously for field-grown specimens. In the two species, shell Mg/Ca ratios were not found to be controlled by shell growth rate or salinity.  Geochim. Cosmochim. Ac., 44, 1265-1278, 1980 Strong kinetic effects on Sr/Ca ratios in the calcitic bivalve Pecten maximus, Geology, 33, 965-968, 2005 EGU chemistry and mantle metabolism, Am. Malacol. Bull., 16, 251-261, 2001. Rosenthal, Y., Boyle, E., and Slowey, N.:     ---- ( Kim and O'Neil (1997)  on the VPDB and VSMOW scale, respectively) for laboratory-cultured M. edulis (• -experiment one and • -experiment two) and P. maximus (+ experiment two only), and field-cultured M. edulis (∆). Plotted also are the data ( ✷ -solid black line) for inorganic calcite deposited from seawater in oxygen-isotope thermodynamic equilibrium from Kim and O'Neil (1997), but also species-specific palaeotemperature equations obtained for P. maximus (solid grey line) by Chauvaud et al. (2005) and M. edulis (dashed black line) by Wannamaker et al. (2007). Due to the use of different acid fractionation factors between the present study and Kim and O'Neil (1997) (Klein et al. 1996) ab δ Fig. 8. Comparison of bivalve calcite shell Mg/Ca ratios, plotted against temperature, from: (a) laboratory aquaria culturing completed in this study for Pecten maximus 1 and Mytilus edulis 1 ; and (b) field culturing completed in this study for Mytilus edulis 1 and other field-based studies, for the species: Mytilus edulis (Vander Putten et al., 2000) 1 , Pecten maximus (Freitas et al., 2006) 2 , Mytilus trossulus (Klein et al., 1996)