Evaluating Mg/Ca in belemnite calcite as a palaeo-proxy

The Mg/Ca, Sr/Ca, δ 18 O and δ 13 C compositions are given for well-preserved specimens of ten belemnite species/genera from three stratigraphic intervals. The data help assess the use of these proxies for palaeo-oceanography. Samples are from Dorset, UK (Pliensbachian; 5 species); Cambridgeshire, UK (Callovian; 1 species); and the Vocontian Basin, SE France (Valanginian; 4 genera). In none of these belemnite populations (at species or genera level) does Mg/Ca correlate with δ 18 O. Neither do values of δ 18 O correlate with Mg/Ca along a microsampled radial profile across a single specimen of Cylindroteuthis puzosiana (Callovian). The use of Mg/Ca is therefore considered to be unreliable as a palaeotemperature indicator for these belemnite species and genera.


Introduction
The oxygen isotopic composition (δ 18 O) and Mg/Ca ratio (and later Sr/Ca) of belemnite calcite have been used to determine palaeo-temperatures and ocean salinity in Jurassic-Cretaceous time. This use is based on two assumptions: 1) that δ 18 O of belemnite calcite reflects calcification temperature and the isotopic composition of the ambient seawater (Urey et al., 1951;Saelen et al., 1996;Podlaha et al., 1998 and many others); and 2) the Mg/Ca, and possibly Sr/Ca, of belemnite calcite is temperature dependent (Berlin et al., 1967;Yasamanov, 1981;Bailey et al., 2003;McArthur et al., 2007a, b;Wierzbowski and Joachimski, 2009;Nunn and Price, 2010). Some credence is given to such assumptions by the demonstration of an inverse covariance between δ 18 O and Sr/Ca, or Mg/Ca, or both, in some Jurassic-Cretaceous belemnites (McArthur et al., 2000(McArthur et al., , 2007aBailey et al., 2003;Rosales et al., 2004a, b). The covariance seen in Fig. 1 is for undifferentiated belemnites. By analogy with other calcitic organisms, such as foraminifera, the response of these putative palaeo-proxies in belemnites may be species-dependent, so interpreting undifferentiated belemnites, at genera or family level, may lead to misinterpretation. No correlation between δ 18 O and Mg/Ca was seen for Jurassic-Cretaceous belemnite species by Wierzbowski and Rogov (2011), Li et al. (2012), or Benito and Reolid (2012). Previous studies of Mg/Ca, Sr/Ca, and Na/Ca in belemnite species or genera noted intra-rostral variations and inter-species/genus differences (McArthur

ACCEPTED MANUSCRIPT
4 that is exposed in a quarry at Kings Dyke, Whittlesey, near Peterborough, UK. This 2 cmthick bed is the uppermost layer of a condensed Jason Subzone. The biostratigraphy of the section is given in Callomon (1968), Duff (1975), and Martill (1991, 1994).
Valanginian specimens were from the section exposed at Vergol, in the Vocontian Basin, SE France, the stratigraphy and biozonation of which are detailed in Reboulet et al., (2006) and McArthur et al., (2007a). Specimens from Vergol are new to this paper and are not those of McArthur et al. (2007a).

Analytical methods
In order to approach statistical significance, data are reported only where five or more specimens of a species/genus were analysed. Prior to analysis, belemnite specimens were cut at the base of the alveolus, perpendicularly to the apical line, yielding front and back slabs ( Fig. 2); the better preserved was used for analysis. Altered areas of the slab (typically the exterior and the apical line) were removed using diamond cutting tools, or dissolved away using 10% HCl for very small samples. The reminder was fragmented to sub-mm size and briefly immersed in 1.2 M HCl to remove surface dust. The best preserved fragments were then picked from a bath of ethanol under the binocular microscope, washed in ultrapure water, and dried in a clean-hood prior to analysis. Fragments that were cloudy were avoided as cloudiness indicates alteration (McArthur et al., 2007b;Rosales et al. 2004a;Benito & Reolid, 2012). Cathodoluminescence was not used to screen for alteration as the methods used proved adequate for that task; visual inspection through the microscope remains one of the best methods of assessing alteration in belemnite calcite.
Stable isotopic analysis of picked fragments was performed either in UCL, using a ThermoFinnegan Delta mass spectrometer and Gasbench preparation system, or in the NERC Isotope Geosciences Laboratory (NIGL), UK, using a VG Optima mass spectrometer. All

ACCEPTED MANUSCRIPT
5 isotopic values are reported as permil relative to the V-PDB scale. Analytical precision, estimated within each run using an in-house standard and repeat analyses of NBS19, was better than 0.1‰ (2sd) for δ 13 C and δ 18 O.
Elemental analysis of picked, fragmented, belemnites for Na, Mg, S, Ca, Fe, Mn, Sr and Ba, was done using a Perkin-Elmer ICP-AES at Royal Holloway University of London using 20 mg samples dissolved in 10 ml of 2% Analar ® HNO 3 . Reproducibility, estimated from duplicates, was better than 4% (2sd) for Ca, Mg and Sr, but more for other elements present as impurities.
One specimen of C. puzosiana (sample 13-49; Fig. 3a) was profiled radially for  13 C,  18 O, and elemental composition. This species has an asymmetrical rostrum and the specimen was profiled along the longer radius of the rostrum (Fig. 3b). The isotopic profile was 15.5 mm in length and the El/Ca profiles was 12.5mm. The two profiles were correlated by matching the prominent concentric alteration bands (usually termed growth rings). This specimen, however, was not particularly well preserved. It showed many concentric rings of cloudy calcite (Fig. 3b) which are usually indicative of alteration. It was used, nevertheless, because better preserved specimens were all smaller and failed to yield sufficient sample for isotopic analysis.
For isotopic profiling of C. puzosiana, a New Wave Research micromill was used to micro-sample parallel to growth boundaries at intervals of 250 μm using a longitudinal section of the front slab (Fig. 3c). Samples were measured for δ 13 C and δ 18 O using a ThermoFinnegan Delta mass spectrometer. For elemental profiling, LA-ICPMS was used (Muller et al., 2009) on the back slab of this specimen (Fig. 3b). An ArF excimer (193 nm) laser-ablation system (RESOlution M-50 prototype, Resonectics LLC, USA) with a twovolume laser ablation cell (Laurin Technic, Australia) was connected to an Agilent 7500ce A C C E P T E D M A N U S C R I P T

ACCEPTED MANUSCRIPT
6 quadruple ICP-MS operated at a RF power of 1250W. The spot diameter was 57 μm, the laser repetition rate 10 Hz, and scan speed 2 mm/min. Analysis was preceded by pre-ablation cleaning using 74 µm spot, 50 Hz repetition rate and 10 mm/min scan speed. Data reduction to elemental composition followed Longerich et al. (1996). NIST612 was used as external standard and pressed powder carbonate standard pellet USGS MACS-3 as secondary standard.
Ca was used as internal standard, and a Ca concentration of 40.08% was assumed in belemnite calcite for calculation, although for element to Ca ratios (El/Ca) this value is cancelled out. Accuracy, estimated using standard MACS-3 was 1.5% -4.5% (2sd) for Na/Ca, Mg/Ca and Mn/Ca, and 8-11% (2sd) for Zn/Ca and Sr/Ca, comparable to published accuracy of MACS-3 pellet (Jochum et al., 2012;Griffiths et al., 2013).

Results
The isotopic and elemental data of belemnite specimens are given in Figs

Isotopic compositions
Within belemnites from each locality and time interval, isotopic compositions differ between species/genera. Differences of δ 13 C are larger than are differences in δ 18 O  Fig. 6).
There is no statistically significant correlation between δ 13 C and δ 18 O for Pliensbachian belemnites (

Relationship between elemental and isotopic compositions.
In none of the analyzed belemnite species/genera, does Mg/Ca correlate well with δ 18 O at species/genus level, nor does it do so for the lumped belemnites from each locality ( Fig. 4-6).

ACCEPTED MANUSCRIPT
8 Only the small number of Castellanibelus specimens appears to show a weak negative correlation between Mg/Ca and δ 18 O (r = -0.51, n = 8 and P < 0.1, Fig. 7), but the correlation degrades after including data for 6 more Castellanibelus specimens from McArthur et al.
(2007a) that were from similar stratigraphic levels in the Vocontian Basin to those used here.

Radial profiles of El/Ca and stable isotopes in C. puzosiana
Along the centre-to-edge profile, δ 18 O ranges from +0.01 to -0.91‰ (Table 2)

El/Ca ratios
The absence of a correlation between Mg/Ca and δ 18 O in our belemnite species questions the assumption that Mg/Ca in belemnite calcite reflects mainly calcification temperature Berlin et al., 1967;Yasamanov, 1981;Bailey et al., 2003;McArthur et al., 2007a). That assumption was, in itself, based on the view that  18 O in belemnite calcite reflects the temperature and δ 18 O of ambient water (Urey et al. 1951 and many since) and so good correlations between Mg/Ca and δ 18 O meant Mg/Ca also reflects temperature of calcification.
Factors other than temperature that may affect either El/Ca, or δ 18 O, or both, are El/Ca of the seawater, salinity, and vital effects (i.e. growth rate, ontogeny, and species-specific fractionation). Firstly, the El/Ca of the seawater in each time interval studied here is assumed to be invariant. The durations of these intervals can be determined with reasonable accuracy.
Pliensbachian samples were collected over two ammonite subzones (Appendix 1) each around 0.3 myrs in duration (Gradstein et al., 2012) and so represent a period of around 0.6 myrs at most. Callovian belemnites are from a single condensed horizon in one subzone (Appendix 2) and may represent spawning accumulations that accumulated in perhaps as little as a few tens to hundreds of years (Doyle and Macdonald, 1993 (Horita et al., 2002;Dickson, 2004).
Secondly, the effects of changes in salinity on δ 18 O cannot be reliably estimated in deep time and are considered by some to be subordinate to the effects of changes in temperature

ACCEPTED MANUSCRIPT
10 ( Price and Mutterlose, 2004;Li et al., 2012). Salinity variation may have introduced scatter into the correlations of δ 18 O with Mg/Ca and Sr/Ca, but it is assumed here that salinity affected δ 18 O less than did temperature (Li et al., 2012). Modeling of end-member mixing of river and seawater as a function of salinity and δ 18 O of ambient water (Fig. 9), shows that Mg/Ca and Sr/Ca in modern seawater are insensitive to salinity in the range 35 to 28 psu. It follows that changes in salinity would not much alter marine Sr/Ca and Mg/Ca ratios except at salinities too low for belemnites to flourish (Fig. 9). Furthermore, as belemnites were probably nekto-benthic in habitat (Dutton et al., 2007) the influence of salinity on their composition was probably small.
Finally, although biological fractionation of oxygen isotopes is generally small (Rexfort and Mutterlose, 2006)  Notwithstanding that, the line of correlation in Fig. 10 shows increasing El/Ca with increasingly negative  18 O, which is what would be expected from operation of a temperature effect on composition. The possibility remains, therefore, that large differences in habitat temperature between belemnite species and genera may be reflected in El/Ca, whilst biofractionation effects work to obscure such a signal within each species and genera, not least because the habitat range of a species may be small e.g. the Toarcian species

Intra-rostral profiles of belemnite C. puzosiana 4.3.1 δ 18 O and El/Ca profiles
The δ 18 O variation within the specimen of C. puzosiana analyzed here is 0.92‰, which is less than the 1.6‰ reported for late Cretaceous belemnites Dimitobelus seymouriensis (Dutton et al., 2007), or the range of 2‰ reported by Wierzbowski and Joachimski (2009) for Middle Jurassic (Bathonian) Hibolithes. If attributed solely to a change in temperature, the 0.92‰ shift in δ 18 O represents ~4°C, using the palaeo-temperature equation of Anderson & Arthur (1983). If attributed to a change of salinity, the shift represents a change of only 2 psu, assuming a  18 O/Salinity of 0.5‰ per 1 psu, based on the modern subtropical Atlantic Ocean (Broecker, 1989).
Whether reflecting changes in salinity, or temperature, or both, the three-stage radial profile of δ 18 O in the C. puzosiana studied here is interpreted to represent three stages of life (Fig. 8). The δ 18 O suggests that the specimen spent the time represented by the first and third stages in water that was either warmer, or less saline, or both, than the water signature captured in stage 2. Stages 1 and 3 are thought to represent periods of shallower water than stage 2, which is interpreted to represent deeper, more offshore, environment. Such an interpretation matches a lifestyle in which inshore hatching and growth is followed by

ACCEPTED MANUSCRIPT
13 migration offshore and an eventual return to inshore shallow waters. Assuming linear growth, each stage represents around one third of a belemnite lifetime. Migration, if not on this timescale, has been noted as a lifestyle in some cuttlefishes (Rexfort and Mutterlose, 2006), a presumed modern analogue of belemnites, and has also been invoked to by Lukeneder et al. (2010) (Fig. 5). This alternative also borrows support from the proposition that a change of salinity of 2 psu would likely be associated with a change in temperature, especially if surface/deep migration was the driver of change.

δ 13 C profile
The δ 13 C profile reflects sharper fluctuations than that of δ 18 O, and does not correlate with δ 18 O profile (Fig. 8). Carbon isotopes in marine biogenic carbonates are often more difficult to interpret than oxygen isotopes. This is because firstly carbon can have multiple sources during carbon fixation. The main source of carbon for the marine calcareous skeleton is dissolved inorganic carbon (DIC), but metabolic carbon, food and fresh water can also contribute carbon (McConnaughey et al., 1997;McConnaughey and Gillikin, 2008). The

ACCEPTED MANUSCRIPT
14 second reason is the occurrence of disequilibrium, or biological fractionation, of carbon isotopes in biogenic carbonates (McConnaughey, 1989).
For the Callovian belemnite specimen C. puzosiana, values of δ 13 C are mostly between +2.0 and +2.5 ‰. Excursions at 5.0, 9.5, and 11.0 mm depart the main trend but show no relation to alteration as evidenced by spikes in Mg/Ca, Mn/Ca, and Zn/Ca (Fig. 8). Similar variations of δ 13 C have been reported in the cuttlebone of a wild cuttlefish Sepia by Rexfort and Mutterlose, (2006), and are interpreted as the metabolically controlled biofractionation of carbon isotopes (see also (Rexfort and Mutterlose, 2006;Wierzbowski and Joachimski, 2009).
Vertical migration of belemnites in the water column would change their δ 13 C values because increasing mineralization of organic carbon with depth drives  13 C to lower values. The trend of the δ 13 C profile of C. puzosiana, however, doesn't fit this explanation if we have correctly interpreted the temperature/depth changes implied by the δ 18 O profile.

Comparison to modern marine calcareous organisms: e.g. foraminifera
The biological control on the uptake of trace elements from the ambient seawater during biomineralization has been documented in modern marine biogenic calcite, such as foraminifera, corals, molluscs and bivalves (Purton et al., 1999;Erez, 2003;Mitsuguchi et al., 2003;Weiner and Dove, 2003;Sinclair et al., 2005;Elliot et al., 2009). Here, foraminifera are used as an example because the Mg/Ca ratio of their tests is a proven palaeo-thermometer and the incorporation of trace elements including Mg in foraminifera has been intensively studied (Lea et al., 1999;Eggins et al., 2003;Erez, 2003;Eggins et al., 2004). In foraminiferal species Orbulina universa, detailed profiles of Mg/Ca indicate that temperature ranges during growth deduced from Mg/Ca exceed the maximum temperature range within the organism's habitat (Eggins et al., 2004). Factors such as biological activities may therefore in part control the Mg/Ca ratios in O. universa (Spero, 1988;Rink et al., 1998;Wolf-Gladrow et al., 1999; A C C E P T E D M A N U S C R I P T  (Bé et al., 1983;Hamilton et al., 2008), and this gametogenic calcite has a different Mg/Ca to the inner ontogenetic calcite of the test (Hathorne et al., 2003;Eggins et al., 2004;Sadekov et al., 2005). Ontogenetic calcification occurs throughout the lifespan of a species, whereas gametogenic calcification occurs over a short period of time (Bé et al., 1983;Hamilton et al., 2008). The difference in Mg/Ca between the two calcite phases may reflect calcification at different water temperature and chemistry and different calcification rates influenced by the differing biogenic precipitation regimes (Hathorne et al., 2003), or fundamentally different styles of calcification driven by cellular processes (Erez, 2003).
In comparison to foraminifera, Mg/Ca composition in belemnites is also likely to be affected by factors other than temperature and Mg/Ca in ambient seawater. The Mg/Ca ratios in belemnites are certainly species-dependent and the evidence presented here and elsewhere suggests that temperature may not be a strong control on El/Ca values.

Does Mg/Ca work as a palaeo-proxy in belemnites?
The use of belemnite δ 18 O as a proxy for palaeo-temperature has always been compromised by uncertainty over what proportion of a change in  18 O recorded in belemnite calcite relates to change in temperature as opposed to change in salinity (e.g. Spaeth et al., 1971;Stevens and Clayton, 1971;Podlaha et al., 1998;Rosales et al., 2004a). Recently, isotopic temperatures for belemnites have been reported to be several degrees cooler than those derived from TEX 86 (Mutterlose et al., 2010). The difference is consistent with temperature being the major control on the belemnite calcite, with TEX 86 temperature

ACCEPTED MANUSCRIPT
16 reflecting the surface mixed layer and belemnites recording deeper, cooler, temperatures in coastal waters (Wierzbowski, 2002;Mutterlose et al., 2010). This interpretation is also consistent with the finding of Dutton et al. (2007)  These studies highlight the importance of the fact that, in the belemnite species and genera presented here, no correlation is seen between Mg/Ca and  18 O. This present work suggests that unless the effects of temperature and species-specific fractionation on belemnite El/Ca ratios can be separately quantified, El/Ca ratios of belemnites may be of restricted use as palaeo-temperature indicators and need further evaluation.

Conclusions
1) In the 10 species/genera of Jurassic-Cretaceous belemnites studied here, Mg/Ca ratios and δ 18 O in bulk analysis do not correlate, either within a single belemnite species or genus, nor within a belemnite population of mixed species or genera. It is concluded that Mg/Ca ratios do not reflect calcification temperature in these belemnites.
2) A radial profile of δ 18 O across a specimen of Cylindroteuthis puzosiana suggests an early and late life-stages spent in shallow inshore, waters, with an intervening period spent in deeper water that was either more saline or as much as 4°C cooler.
3) The profiles of Mg/Ca and δ 18 O across the sections of the same C. puzosiana specimen show little comparability between the two, suggesting that intra-rostral Mg/Ca profile is unlikely to be reflecting temperature for this species.
grateful to Finn Surlyk and two anonymous reviewers for their detailed and valuable comments, which greatly improved the final manuscript.
List of Figures   Fig. 1. Cross-plots of El/Ca and δ 18 O for Pliensbachian-Toarcian belemnites from Yorkshire of the UK (data of McArthur et al. 2000), southern Germany (data from Bailey et al. 2003) and northern Spain (data from Rosales et al. 2004a). Correlation lines derived by reducedmajor-axis regression.      Elemental data from a radial profile of a back slab (Fig 2) of the specimen, stable isotope data from the front slab after sectioning (Fig. 2). The profiles were matched using concentric alteration bands, usually termed 'growth rings'.  belemnires from Dubki section, near Saratov, Russian Platform: it is emphasized that these authors did not claim that Mg/Ca or Sr/Ca was correlated to δ 18 O. is Olcostephanus.  Fig. 3c).