Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
 An 8 million year record of subtropical and midlatitude shelf-sea temperatures, derived from oxygen isotopes of well-preserved brachiopods from a variety of European sections, demonstrates a long-term Cenomanian temperature rise (16–20°C, midlatitudes) that reached its maximum early in the late Turonian (23°C, midlatitudes). Superimposed on the long-term trend, shelf-sea temperatures vary at shorter timescales in relation to global carbon cycle perturbations. In the mid-Cenomanian and the late Turonian, two minor shelf-sea cooling events (2–3°C) coincide with carbon cycle perturbations and times of high-amplitude sea level falls. Although this evidence supports the hypothesis of potential glacioeustatic effects on Cretaceous sea level, the occurrence of minimum shelf-sea temperatures within transgressive beds argues for regional changes in shelf-sea circulation as the most plausible mechanism. The major carbon cycle event in the latest Cenomanian (oceanic anoxic event 2) is accompanied by a substantial increase in shelf-sea temperatures (4–5°C) that occurred ∼150 kyr after the commencement of the δ13C excursion and is related to the spread of oceanic conditions in western European shelf-sea basins. Our oxygen isotope record and published δ18O data of pristinely preserved foraminifera allow the consideration of North Atlantic surface water properties in the Cenomanian along a transect from the tropics to the midlatitudes. On the basis of fossil-derived δ18O, estimated δw ranges, and modeled salinities, temperature-salinity-density ranges were estimated for tropical, subtropical, and midlatitude surface waters. Accordingly, the Cenomanian temperate shelf-seas waters have potentially the highest surface water density and could have contributed to North Atlantic intermediate to deep waters in the preopening stage of the equatorial Atlantic gateway.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The Mesozoic-Cenozoic greenhouse climate peaked during Cenomanian-Turonian times and reached its thermal maximum during the late Turonian [Clarke and Jenkyns, 1999; Wilson et al., 2002]. Unusually warm sea surface temperatures (SST) were reported for the tropical (33–34°C) [Norris et al., 2002; Wilson et al., 2002; Schouten et al., 2003], and southern subpolar Atlantic Ocean (30–32°C) [Bice et al., 2003], based on oxygen isotope data from glassy foraminifera and archaeal membrane lipids, and are interpreted to be the consequence of intensified meridional heat transport caused by increased greenhouse gas forcing and accelerated ocean crust production [Larson, 1991]. However, evidence of SSTs warmer than 32°C are only snapshots of Cenomanian-Turonian time. A first paleotemperature record with a high temporal resolution from the late Albian to the earliest Cenomanian shows pronounced short-term variability (10–100 kyr) in both SSTs and the thermal structure of the surface ocean, suggesting a highly dynamic ocean hydrology during greenhouse climate conditions [Wilson and Norris, 2001].
 Opposing the evidence for extraordinarily warm tropics, a variety of authors have repeatedly suggested that short glaciations (<100 kyr) may have occurred within the ice-free Cenomanian-Maastrichtian greenhouse world, based on correlation between shallow water sequences and δ18O-derived paleotemperature records [Miller et al., 1999, 2003; Stoll and Schrag, 2000]. In Cenomanian time, third- and fourth-order sea level changes were eustatic and occurred synchronously within the resolution of ammonite biozones [Gale et al., 2002]. Glacioeustasy is the only known mechanism that can drive large (amplitudes >10 m) and rapid (<100 kyr) sea level changes [Pitman and Golovchenko, 1983]. However, evidence for glacioeustasy in Cenomanian-Turonian times is complicated either by low temporal resolution of deep-sea δ18O records, which do not resolve variations at timescales less than 100 kyr [e.g., Huber et al., 2002], or by diagenetic alteration of bulk carbonate δ18O data [e.g., Clarke and Jenkyns, 1999; Stoll and Schrag, 2000]. Additionally, uncertainties in correlations along coast-shelf-ocean transects at timescales of less than 100 kyr have so far inhibited the direct comparison of sea level and paleotemperature records.
 Here we present an 8 million year record of mid-Cretaceous shelf-sea temperatures based on new and previously published oxygen isotopes of well-preserved brachiopod calcite from subtropical and temperate European basins [Voigt, 2000; Voigt et al., 2003; S. Voigt et al., Chronology of short-termed sea-level, carbon cycle, and climate variations during the Cenomanian-Turonian boundary event in NW Europe, submitted to Cretaceous Research, 2004]. Low-magnesium calcite of brachiopod shells is the most suitable substrate to reconstruct the paleoceanographic history in shelf sea environments because of the complex history of early, burial and meteoric diagenesis of land sections. The record resolves pronounced short-term temperature variations (in the order of 100 kyr) through Cenomanian-Turonian carbon cycle events. In combination with δ18O-derived surface, thermocline and deep-sea temperatures from ODP sites, we discuss tropical, subtropical and midlatitude surface water properties. The record suggests that (1) the temperate midlatitudes shelf-seawaters could have contributed to Cenomanian intermediate to deep waters, and (2) Cenomanian-Turonian carbon cycle perturbations occurred concomitant to high-amplitude sea level changes, whereby falling and rising sea level is related to decreasing and rising shelf-sea temperatures.
2. Material and Methods
 A composite of 24 different Cenomanian and Turonian sections from midlatitude (∼35–36°N paleolatitude) and subtropical (26–28°N) European shelf-sea basins were sampled for specimens of articulate brachiopods (Figure 1; see auxiliary material). Localities at midlatitudes are situated in southern England and northern Germany, with Cenomanian sample sites at Southerham, Eastbourne, Lydden Spout (England), and Quedlinburg (Germany), and Turonian sections at Dover, Kensworth, Lewes (southern England), Söhlde and Hoppenstedt (northern Germany). All successions expose a pelagic carbonate facies consisting of marly chalks, chalks and pelagic limestones. Subtropical localities are situated at the northern Tethys margin in northern Spain (North Cantabrian Basin) and southeastern France (Vocontian Basin). Sections in northern Spain expose a carbonate shelf in the Cenomanian and mixed hemipelagic silicilastics and carbonates in the Turonian [Wilmsen et al., 1996]. Two localities in the Vocontian Basin expose a basinal marl-limestone facies (St. Lions) and a shallower transitional facies between platform carbonates and basinal marls (Col des Abesses).
 A total of 417 stratigraphically horizoned brachiopod shells were studied for their preservation and stable isotopic composition (for an overview of new and already published data see auxiliary material). The occurrence of distinct brachiopod species is frequently limited to certain stratigraphic horizons (e.g., acme occurrences of Orbirhynchia mantelliana in the early and middle Cenomanian [Paul et al., 1994]). Therefore the composite Cenomanian-Turonian δ18O record is based on different brachiopod taxa consisting of rhynchonellid (Orbirhyncia, Grasirhynchia, Modestella, Gemmarcula) and terebratulid (Concinnthyris, Capillithyris, Kingena, Terebriostra, Terebratulina, Gibbithyris) genera.
 The samples were mechanically scraped and cleaned in an ultrasonic bath to separate shell calcite from the adherent matrix. Polished thin sections were prepared from one half of the samples for conventional and cathodoluminescence microscopy (CL) carried out with a hot cathode luminescense microscope (HC1-LM). Selected shell samples were broken into small pieces, and their textural preservation was examined with a scanning electron microscope (SEM, CamScan CS44 ED). In order to perform element geochemistry and stable isotope analysis, about 1.5 mg of sample material was drilled from the best preserved domain of the fibrous secondary (all rhynchonellids and most terebratulids) or the prismatic tertiary (terebratulid genera Concinnthyris and Gibbithyris) shell layer using a binocular microscope. If both valves were preserved, the posterior part of the dorsal valve was used preferentially to avoid kinetic fractionation effects as described by Curry and Fallick . Element concentrations were measured with ICP-AES (Oldenburg) or ICP-MS (Cologne) by the reaction of 1 mg powder with 5% HNO3, and determination of element concentrations for Mg, Fe, Mn, and Sr in μg/g. Stable isotope ratios were measured with a mass spectrometer of Analytical Precision (Jülich) coupled to an automatic preparation system. Values are given against the VPDB standard, and reproducibility of repeated standard measurements was better than 0.1‰ for carbon and 0.15‰ for oxygen.
 Temporal resolution of the composite brachiopod record depends on the abundance of brachiopods and the quality of interbasinal correlations, and ranges from the level of ammonite zones (northern Spain) to the level of orbital timescales (<100 kyr) in the middle and late Cenomanian and late Turonian (midlatitude localities, Figure 2). Cenomanian samples from midlatitude localities are calibrated against the orbitally tuned cyclostratigraphy of Gale et al. , which has been fixed to the chronostratigraphic age of the C-T Boundary (93.5 Ma; Gradstein et al. , FO W. devonense, Q. gartneri). The late Turonian age model is based on the highly coherent correlation of two continuous δ13C records from a boreal (Salzgitter-Salder; Voigt and Hilbrecht ) and a Tethyan (Contessa; Stoll and Schrag ) section (Figure 2). The mean sedimentation rate is two times higher at Salzgitter-Salder (99.3 m/myr) than at Contessa (9.86 m/myr). Both records were interpolated between the ages for the middle-upper Turonian (FO S. neptuni, ∼90.4 Ma) and the Turonian-Coniacian boundary (89.0 Ma) [Gradstein et al., 1995]. Numerical ages were assigned to late Turonian brachiopods based on interpolation of carbon isotope stratigraphy and five bentonite layers, which occur in both southern England and northern Germany [Wray, 1999]. In the early to middle Turonian interval, strong biogeographic differentiation and a poor calibration of microfossil and macrofossil biozones with carbon isotope stratigraphy results in a higher degree of uncertainty of chronological ages. The estimated error is in the order of 100 to 500 kyr. Here the Contessa-δ13C record has been interpolated between the ages of the Cenomanian-Turonian and middle-upper Turonian boundaries.
 Macroscopic shell preservation ranges from moderate to very good, and is better in sections with chalks or platform carbonates than in hemipelagic chalks and calcareous siltstones. In order to evaluate the preservation state of the brachiopod shell calcite, an integrated approach of trace element geochemistry, SEM and CL microcopy is used. CL microscopy allows the identification of the best preserved domains of the brachiopod shells. Diagenetically unaltered low-Mg calcite appears nonluminescent, whereas early diagenetic incorporation of Mn2+ in the calcite lattice acts as the most prominent activator of an orange-colored luminescence, which can be quenched by additional incorporation of Fe2+ [Machel et al., 1991]. Both elements are enriched in fluids under reducing conditions, are common in early diagenetic environments, and may serve as indicator of diagenetic recrystallization of low-Mg calcite [Veizer, 1983]. In this study, brachiopod shells with less than 100 μg/g Mn and 500 μg/g Fe are considered to be very good preserved, whereas shells with Mn and Fe concentrations between 100 and 200 μg/g and 500 and 1500 μg/g are classified as good to moderately preserved (Figure 3).
 Independent of the redox state, postdepositional recrystallization causes a loss of strontium in skeletal calcite [Veizer et al., 1986]. Primarily, incorporation of strontium into biogenic calcites is controlled by metabolic processes governing the rate of calcite precipitation, and by environmental conditions such as variations in seawater salinity, temperature and Sr/Ca ratio [Klein et al., 1996; Stanley and Hardie, 1998; Steuber, 2002]. Modern brachiopods have a slow metabolism and a low rate of strontium uptake in the lattice of their secondary layer [Curry et al., 1989; Brand and Brenckle, 2001; Curry and Fallick, 2002]. Strontium concentrations of recent brachiopods are between 450 and 1550 μg/g with a mean value of ∼1000 μg/g [Morrison and Brand, 1986; Brand et al., 2003]. The mean strontium concentration of Cenomanian-Turonian brachiopod shells is lower (700 μg/g) in comparison to modern values, but shows a large range of values from 200 to 1700 μg/g (Figure 3). The strontium concentration of brachiopod shells is also lower than strontium concentrations reported for mid-Cretaceous rudists and belemnites (1100–1500 μg/g; Steuber and Veizer ). Low Sr concentrations in brachiopod shells can be either attributed to an early diagenetic loss or to a taxonomically mediated lower rate of strontium uptake (lower distribution coefficient). The large range of strontium concentrations in Cenomanian-Turonian brachiopod shells is related to a pronounced variability through time (Figure 4). The majority of Cenomanian-Turonian brachiopod shells have strontium concentrations between 300 and 700 μg/g, however, brachiopod specimens of early middle Cenomanian age have strontium concentrations, which are about three times higher (1200–1700 μg/g). The mid-Cenomanian brachiopod specimens do not differ in their structural and chemical shell preservation from other specimens of Cenomanian-Turonian age. Therefore it is not very reasonable to attribute the variability in strontium concentrations to different degrees of early diagenetic alteration. The substantial increase of strontium concentrations in the early middle Cenomanian, seems more to be related to environmental conditions, such as fast changes in the Sr/Ca ratio of seawater. It remains to be tested, if other taxonomical groups show similar temporal variations in their Sr/Ca ratios and if the low brachiopod strontium concentrations are biologically controlled. In order to take an early diagenetical loss of Sr into account, we used the lower value (400 μg/g) of the modern range to discriminate between good and moderately (300–400 μg/g) preserved brachiopod specimens.
 The majority of studied brachiopod samples have nonluminescent secondary and tertiary shell layers (Figures 5a, 5b, 5c, 5d, and 5f). Some rhynchonellids show signs of secondary layer cementation in the space between calcitic fibres (Figure 5e). The hollows of terebratulid punctae are commonly filled with secondary cements, which is visible in the texture (Figures 5g and 5i) and by the orange luminescence (Figure 5c). To minimize the influence of cemented punctae, the posterior part of the brachiopod shell was chosen for sampling. Punctae are small in this shell domain and hold only a minor portion of the shell calcite (< 5%). Rhynchonellid brachiopod shells are best preserved in the posterior shell portion as well.
 About 70% of all studied brachiopod shells are classified as very good to good preserved and their isotopic composition as primary. Moderately preserved shells do not match all geochemical screening criteria, but are usually not significantly offset in their stable isotopic composition from the good and very good preserved shells (Figure 4). Oxygen and carbon isotope values of moderately preserved shells show a broader scatter during certain time intervals, but are not systematically enriched or depleted. In oder to enhance the stratigraphic resolution, we do not exclude moderate shells from the data base, but discuss their isotopic signature only in terms of mean values.
4. Equilibrium Precipitation
 The first systematic isotopic study of brachiopod shells goes back to the work of Lowenstam , who measured the oxygen isotopic composition of different species of all modern superfamilies at different latitudes and corresponding water samples, and found the 18O/16O ratio of brachiopod shells strongly to be related to seawater temperature but not to species. Later, Carpenter and Lohmann  demonstrated substantial deviations from seawater isotopic equilibrium for the primary shell layer and specialized portions of the brachiopod shell as brachidium, foramen, and muscle scars. However, the isotopic composition of the secondary shell layer were found to be close to the isotopic equilibrium, a result that was reproduced for the species Laqueus californianus [Buening and Spero, 1996] and for four terebratulid species from the Lacepede shelf in Australia [James et al., 1997]. More recently, isotopic studies on specimens of single species have shown substantial offsets from the isotopic equilibrium. Curry and Fallick  found systematic differences in the oxygen isotope ratio of matching dorsal and ventral valves for the terebratulid species Calloria inconspicua (Sowerby), with the thinner dorsal valve yielding the higher and less variable δ18O values. Auclair et al.  demonstrated seasonal variable kinetic fractionation effects for the primary and the upper secondary layer of Terebratalia transversa (Sowerby), with deviations from isotopic equilibrium of up to 6‰. Owing to the limited numer of recent studies so far, it remains to be tested whether kinetic disequilibrium precipitation of brachiopod shells is restricted to certain taxonomic groups or to environmental conditions, and how the isotopic composition of fossil brachiopods is affected.
 In order to test the possibility of metabolic fractionation effects for Cenomanian-Turonian brachiopods, we compared the δ13C and δ18O values of different genera from middle to late Cenomanian localities in northern Spain and southern England (Figure 6). Middle to late Cenomanian terebratulid brachiopods from northern Spain are depleted in 13C by 1.0‰ (Dereta, Gemmarcula) to 2.0‰ (Terebriostra, Sellithyris) in comparison to the rhynchonellid genus Grasirhynchia. However, the oxygen isotopic composition of these genera shows only a small offset of less than 0.5‰ (Figure 6a). The abundant occurrence of a diverse mid-Cenomanian brachiopod fauna in southern England allows the comparison of eight genera within a relatively short period of time (∼500 kyr). We used specimens from the interval before (couplets B35–B40 sensu; Gale ) and after (couplets C2–C10) the MCE carbon isotope excursion (Figure 6b). Accordingly, the mean δ13C and δ18O values of five rhynchonellid and terebratulid genera plot within a narrow range of 2.8 to 3.2‰ and −0.8 to −0.6‰, respectively, whereas the δ13C and δ18O values of the terebratulid genera Kingena, Terebratulina and Gemmarcula are depleted by 0.5 to 1.0‰. The isotopic offset for the genus Gemmarcula is similar to that observed for the brachiopods in northern Spain. The shells of modern brachiopods contain a high proportion of total organic mass, and terebratulid brachiopods are capable of storing metabolites within extensions of the mantle tissue in their hollow endopunctae [Curry et al., 1989]. It might be possible that genera-dependent incorporation of metabolic CO2 or early diagenetic alteration due to degradation of organic tissue could have influenced shell carbon isotopic composition of some terebratulid genera. However, high absolute values (∼1.0‰ enrichment in comparison to the bulk-carbonate signal for carbon) and low variability in the mean carbon and oxygen isotopic composition of shells from the genera Orbirhynchia, Grasirhynchia, Capillithyris, and Modestella argue for a precipitation close to the isotopic equilibrium. We cannot exclude the possibility of metabolic effects on the oxygen isotopic composition of Cenomanian-Turonian brachiopods, but estimate the magnitude of these effects <0.5‰ for most of the studied genera. Systematic studies on modern brachiopods are necessary to evaluate taxon-related nonequilibrium isotopic fractionation effects.
5. Results and Discussion
 Brachiopods are epifaunal organisms and thrived on the seafloor at shallow water depths. All chalk-sea localities represent a distal shelf environment with water depth between 30 and 100 m, but the occurrence of brachiopods is mostly restricted to the shallower part of sedimentary sequences. Their estimated paleohabitat was in depths between 30 and 50 m. Specimens from northern Spain are associated with shallow water limestones (0–30 m), whereas the localities in the Vocontian basin reflect a deeper environment with an estimated depth range of 30 to 100 m.
 Modern articulate brachiopods in temperate regions show slow and continuous shell growth for 8–12 years [Peck, 2001]. Assuming a similar age for Cretaceous brachiopods, which usually have sizes between 0.5 and 2 cm, our drilled shell samples most probably reflect an average of several months. Some of the studied brachiopods genera show distinct growth lines, which could reflect seasonal interruptions of shell growth. The relatively small size and the poorer preservation of the anterior shell inhibit an ontogenetic monitoring of shell oxygen isotopic composition within a single specimen, thus we cannot determine the season for which our drilled shell sample is representative. However, seasonality in midlatitude shelf-seas should have an influence on the oxygen isotopic composition of brachiopod shells, and possibly, most of our oxygen isotope values reflect summer conditions. A raw estimation of seasonality could be provided by individuals of the very well preserved Cenomanian species Orbirhynchia mantelliana, a species that has no growth lines. The range of δ18O values in the same interval is about 1‰ (−1.54 to −0.46‰ in couplets B39–40). This quantification is compromised by our uncertainty about possible nonequilibrium isotopic fractionation effects and can not be accepted to be valid for the overall scatter of data, which is probably more related to shell preservation than to seasonality. In order to avoid overinterpretation of single shell data, we applied a nonlinear regression (250 kyr filter) to the composite brachiopod-δ18O data set (Figure 2). The filtering process averages possible effects of preservation, nonequililibrium precipitation and seasonality, thus we consider the filtered mean δ18O values (1σ standard deviation) to resemble a mean annual temperature signal, which can be biased somewhat toward summer values (Table 1).
Table 1. Brachiopod-Derived Paleotemperature Estimates for Midlatitude and Subtropical Shelf Seas in Comparison to Temperatures Derived From Paleobotany (Leaf Physiognomy), AGCM, and OGCM Simulations
 Conservative temperature estimates derived from brachiopod oxygen isotopes were calculated using the equation for shell calcite by [Anderson and Arthur, 1983] and a seawater oxygen isotopic composition (δw) of −1‰ SMOW for an ice-free world [Shackleton and Kennett, 1975]. Accordingly, mean background chalk-sea temperatures vary between 15 and 16°C in the early middle Cenomanian and display a long-term warming to 16–17°C in the late Cenomanian and to 19–21°C in the early and late Turonian (Figure 2; Table 1). Superimposed on this long-term trend, two periods of decreasing shelf-sea temperatures occurred during the MCE (∼2°C) and the LTE (∼3°C), whereas substantial warming (4–5°C) occurred during the CTBE. In northern Cantabria, mean shelf-sea temperatures range from 22–27°C in the Cenomanian and 20–23°C in the Turonian, whereas in the Vocontian Basin, temperature estimates for the MCE (13–14°C) and the CTBE (18–22°C) are similar to those of midlatitude chalk seas.
 The assumption of a temporally and spatially constant δw value is not very reasonable. The validity of our conservative temperature estimates is complicated by the local variability of shelf-sea δw values due to regional variations in evaporation, precipitation, and runoff, which are difficult to estimate. A prominent facies shift from marly to pure chalk occurred in the late Cenomanian (guerangeri zone in NW Germany, uppermost geslinianum zone in southern England) in midlatitude shelf sea basins, indicating a diminished terrestrial and fresh-water influx as open oceanic conditions spread [Hay, 1995; Gale et al., 2000]. Therefore the effects of increased continental runoff on shelf-sea δw should have been subordinate later than late Cenomanian. In the early to middle Cenomanian, a higher freshwater influx due to increased runoff is a possible scenario. In order to provide a range of possible paleotemperature estimates, two regional δw values are considered as possible end-members for midlatitude localities (Table 1). The lower δw value of −1.5‰ takes into account that continental runoff has effected the oxygen isotopic composition of shelf-seawater, whereas the higher δw value (−0.5‰) considers the effect of the net-precipitation-evaporation balance on surface salinities and seawater δw at paleolatitudes around 35°N [Zachos et al., 1994]. At subtropical shelf localities, evaporation was the main influence on the oxygen isotopic composition of seawater, because of the latitudinal position close to the subtropical high and the reduced terrestrial influx in carbonate platform systems. δw values between −0.5‰ and −0.1‰ are considered as a reasonable range. Consequently, shelf-sea temperature estimates would increase by 2–3°C at midlatitudes, and by 2–5°C in the subtropics for the case of increased evaporation. In the case of increased continental runoff, midlatitude temperature estimates would be lowered by 2–3°C. The imbalances between sinks and sources for 18O within the hydrological cycle are modeled to result in a −0.2‰ deviation of δw from the present-day value in the Cretaceous [Wallmann, 2001]. Accordingly, our estimated temperatures would be lowered by 1–2°C. However, effects of changes in seawater pH on δw underestimate shelf-sea temperatures for higher seawater acidity as a consequence of elevated atmospheric CO2 levels and would, therefore, increase our temperatures estimates by 2–3.5°C [Zeebe, 2001].
 Several independent paleotemperature estimates derived from climate and ocean modeling and paleobotany can be compared with the different scenarios for δ18O-derived temperatures (Table 1). Results of a Cenomanian-Turonian Earth system model with 6 × atmospheric CO2 (GENESIS, version 2.0) predict mean annual temperatures of 21–22°C for 35°N, and 25–26°C for 27°N [Flögel, 2002]. According to the results of a Turonian ocean circulation model (OGCM) with 4 times atmospheric CO2 (parallel ocean climate model (POCM)), mean SSTs are between 18–20°C at midlatitudes and 22–24°C in the subtropics [Poulsen et al., 2001]. Analysis of the leaf physiognomy of mid-Cenomanian flora from Bohemia using the CLAMP technique shows mean annual temperatures between 17 and 20°C for the mid-European Island (∼33°N) [Herman et al., 2002].
 In the Cenomanian, model-predicted and floral-derived temperatures agree best with midlatitude δ18O-derived temperatures, which use a higher δw value for effects of increased evaporation. The assumption of a reduced δw value due to enhanced continental runoff results in relatively low temperatures, which underestimate ocean-model-predicted temperatures by 3–5°C. In the Turonian, the best agreement between δ18O- and model-derived temperatures is obtained for the scenario with low δw values. A shift from evaporation-dominated to runoff dominated seawater oxygen isotopic composition would explain much of the observed long-term Cenomanian-Turonian δ18O decrease in brachiopod shell calcite. However, such a temporal trend is opposite to the observed facies change from marly to pure chalk, which indicates a reduction of terrestrial and freshwater influx in the Turonian. We, therefore, assume that the long-term decrease in brachiopod δ18O values is probably related more to temperature than to changes in the oxygen isotopic composition of seawater. The estimated temperature rise of 6°C for stable δw values in the Cenomanian-Turonian would be amplified if a reduction of continental runoff (increase of δw) is taken into account, and thus represents a conservative minimum estimate.
 At subtropical localities in northern Spain, the lower δw estimate (−0.5‰) shows the best agreement to model-predicted temperatures, whereas in southern France, the lower δw estimate results in temperatures, which underestimate model temperatures by 2–7°C. The relatively low temperature estimates are explained by the deeper depth habitat of brachiopods in the Vocontian Basin.
6. Paleoclimatic and Paleoceanographic Implications
 The middle to early late Cenomanian part of our two brachiopod shelf-sea temperature records supplement δ18O-derived temperatures of tropical, subtropical and temperate oceanic surface, thermocline and deeper waters (Sites 144, 1050, 551) [Huber et al., 2002; Norris et al., 2002; Gustafsson et al., 2003], forming a transect across the central northern Atlantic ocean from the tropics to the midlatitudes (Figure 1). The mean oxygen isotopic composition of calcitic shells at each locality in combination with modeled salinities [Poulsen et al., 2001] allow the estimation of possible salinity-temperature ranges for different water masses, thus providing a latitudinal salinity-temperature-density profile (Table 2; Figure 7). In order to provide a reasonable paleotemperature range for each locality and water depth, a range of regional δw values is taken into account (Table 2). Surface water δw values at midlatitudes (European shelf sea and Site 551, Goban Spur) and at subtropical localities (northern Spain and Site 1050, Blake Nose) are considered within the ranges discussed above. The tropical surface water δw range is discussed by Norris et al. , where the authors used the present-day range of δw values at Demerara rise corrected for an ice-free world. In the deeper ocean (intermediate to deep waters), the conservative δw value of −1.0 0.2 l SMOW for an ice-free world is a reasonable estimate. Paleotemperature ranges are calculated by using the equations of Anderson and Arthur  for brachiopods, of Bemis et al.  for nonsymbiontic planktic foraminifera and of Shackleton  for benthic foraminifera. Application of these different paleotemperature equations implies an error of ∼1°C, but considers differences in biogenic calcite precipitation. Regional surface water salinities, predicted by Albian and Turonian ocean circulation simulations (POCM) [Poulsen et al., 2001], are attributed to the calculated temperature ranges. The Turonian simulation of Poulsen et al.  provides the best scenario for surface waters exchange across shallow seas after the Cenomanian first-order sea level rise. The Cenomanian deeper proto-North Atlantic probably had no deep-water exchange with the Southern Ocean [Summerhayes, 1987; Pletsch et al., 2001], thus calculated temperature ranges of deeper water masses are assigned to model-predicted salinities of the Albian simulation, which consider the North Atlantic ocean as an isolated basin. Model-predicted surface water salinities at midlatitude shelf seas cover a broad range from low boreal (28 psu) to high Tethyan (36 psu) salinities [Poulsen et al., 2001]. The low boreal model salinities result from geographic isolation of boreal seas as a consequence of low-resolution paleogeography. We, therefore, used an intermediate range of 32–35 psu for the European shelf-sea, in order to account for the open exchange of surface waters between the Boreal and Tethyan seas in the Cenomanian-Turonian, which is evident from the geological record but not resolved by the ocean model. In order to test the validity of our approach of relating δ18O-derived temperatures to modeled salinities, we examined the agreement of δ18O-derived and ocean model-predicted temperatures (Table 2). The δ18O-derived temperatures match very well for intermediate to deep waters and midlatitude shelf-seawaters. For tropical to subtropical surface and subtropical to temperate shelf-seawaters, the lower value of the estimated δ18O-derived temperatures range agrees with model-predicted temperatures.
Table 2. Ranges of Calcite δ18O Values, Estimated δw Values and Temperatures, and Modeled Salinities at Different Localities in the Late Cenomanian Proto-North Atlantic Ocean
 The corresponding temperature-salinity pairs of variables for each water mass describe distinct arrays within the salinity-temperature-density diagram [Millero and Poisson, 1981] (Figure 7). Surface waters with the lowest density occurred in the tropics (Site 144), where sea-surface temperatures were extraordinarily warm and salinities were partly influenced by the Intertropical Convergence Zone (see discussion in the work of Wilson et al. ). Toward the subtropics (Site 1050 and northern Spain), surface water density increased as a consequence of increasing salinity and decreasing temperature beneath the subtropical high and within the subtropical gyre. At midlatitudes, temperate surface and shelf-seawaters (Site 551 and European shelf sea) cover a broad array of possible temperature-salinity ranges due to uncertainties about the regional effects of continental runoff. In the case of lower surface salinity (lower δw estimate due to enhanced runoff and/or precipitation), corresponding temperatures are relatively low (15–19°C), and the density resembles that of the warmer and more saline subtropical surface waters. In the case of higher salinities and temperatures (higher δw estimate), the temperate surface water at Goban Spur and the European shelf-seas form the highest density surface water in the Cenomanian proto-North Atlantic ocean (Figure 7). The oxygen isotopic composition of benthic foraminifera at Site 1050 [Huber et al., 20002] and Site 551 [Gustafsson et al., 2003] is similar at both sites and reflects the conditions of intermediate to deep waters at 1500 to 2500 m water depth. The estimated temperature range is 15–18°C at both localities.
 Results of ocean and coupled atmosphere-ocean simulations for the Campanian and Cenomanian-Turonian indicate that oceanic deep waters can be formed by cooling warm and salty water in the high latitudes of the northern Pacific and southern hemisphere [Brady et al., 1998; Poulsen et al., 2001; Bice and Norris, 2002], which is contrary to the older theory of warm saline deep water formation at low latitudes [Brass et al., 1982]. During the Cenomanian, the North Atlantic ocean was an isolated basin with either no or only limited global deep water exchange. Deep water connections toward the Arctic ocean were blocked until the opening of the Greenland-Iceland-Norwegian Sea in the Oligocene [Wold, 1995]. The isthmus between North and South America was only open to depths of 400–700 m [Otto-Bliesner et al., 2002], and toward Tethys, deep water connections were inhibited or restricted by the continental terranes of Apulia and the Taurides [Dercourt et al., 1992]. The deep gateway between the North and South Atlantic oceans is supposed to have opened in the latest Cenomanian to early Turonian [Pletsch et al., 2001]. Hence during most of the Cenomanian time, proto-North Atlantic deep waters originated from a regional source. The oxygen isotopic composition of midlatitude brachiopods, which reflect mean annual or summer conditions, is very similar to those of benthic foraminifera at sites 1050 and 551. The temperate shelf-seawaters of northwestern Europe had the highest surface water density, if higher rates of evaporation and low continental runoff (higher δw estimate) are assumed, which describes a situation similar to the Mediterranean Sea today. The density of temperate surface waters would be even higher, if the range of brachiopod-δ18O values during a certain time interval is a measure of seasonality, as discussed for the middle Cenomanian species O. mantelliana (see above), and the higher δ18O value (−0.5‰) would reflect winter conditions. We, therefore speculate, that the dense temperate shelf-seawater could have contributed either seasonally (during winter times) or episodically to intermediate and/or deeper waters in the Cenomanian isolated proto-North Atlantic basin. This idea is supported by a north-south gradient in preservation of organic matter. Highest rates of organic carbon accumulation occurred at the southern edge of the North Atlantic basin (DSDP Sites 144, 367, Tarfaya Basin) [Kuhnt et al., 1990; Kuypers et al., 2002], indicating the occurrence of the oldest and less oxygenated deep waters there.
 Shelf-sea temperatures varied considerably during the Cenomanian-Turonian time interval, revealing changes in the order of 100 kyr during the three carbon cycle events MCE, CTBE and LTE (Figures 2 and 8). The minor δ13C events, MCE and LTE, have similar temperature trends, showing an initial cooling (2–3°C) that peaks close to the δ13C maximum, and a subsequent warming interval. This overall trend is closely related to large-amplitude sea level changes indicated by incision, condensation, lowstand sediments and subsequent transgressive sequences at shallow shelf settings.
 The commencement of the MCE-δ13C excursion falls in the C.inerme ammonite zone, which was regressive on a global scale (India, Boreal and Tethyan European Basins, New Jersey margin) [Gale et al., 2002; Miller et al., 2003; Wilmsen, 2003]. The subsequent transgression coincides with the δ13C maximum and is related to the basal A. rhotomagense ammonite zone. The sudden increase of Sr concentrations in brachiopod shells is also of C. inerme zone age (Figure 4). The large magnitude of changes in the Sr concentration exceeds any variability, which is known from modern calcite precipitating organisms due to the influence of changes in seawater temperature or rates of growth and calcification [e.g., Lea et al., 1999]. The most probable explanation for the fast and large strontium increase is a variation in the Sr/Ca ratio of seawater as a consequence of the rapid, short-term regression [Stoll and Schrag, 1998; Stoll and Schrag, 2001]. The large amount of Sr is recycled to seawater during subaerial exposure and meteoric diagenesis of previously deposited shallow water carbonates, which were dominated by aragonitic communities in the Albian-Cenomanian time [Stanley and Hardie, 1998; Steuber and Veizer, 2002].
 In western Europe, the southward spread of boreal invertebrates (bivalves, belemnites) indicates the occurrence of two MCE cooling pulses, the first related to the initial δ13C bulk-carbonate increase (arlesiensis bed) and the second to the δ13C maximum (primus bed; Figure 8a) [Paul et al., 1994]. Both bioevents are sandwiched between a lower and an upper Orbirhynchia event forming a bundle of bioevents across the lower-middle Cenomanian transition. δ18O-derived shelf-sea temperatures indicate that the temperature minimum corresponds to the occurrence of the boreal belemnite Praeactinocamax primus. The main temperature decrease correlates with the sea level lowstand and is followed by subsequent warming during transgression. The overall duration of the MCE is about 200 kyr.
 Cenomanian carbon isotopic data from benthic and planktonic foraminifera at Site 1050 show a distinct positive excursion, thus providing the first evidence for the presence of the MCE in a deep-sea section [Huber et al., 2002] (Figure 2). During the MCE, thermocline and deep water temperatures decrease by ∼3°C, thus showing the same trend as our midlatitude shelf-sea temperatures. Overall cooling in deeper, thermocline and shelf waters indicates a general change in North Atlantic ocean circulation that probably favored increased oceanic ventilation and upwelling of nutrient-rich water masses, and could drive the positive carbon isotope excursion.
 The relation between eustasy and the LTE is more difficult to reconstruct due to the lack of overregional correlations of sedimentary sequences. Biostratigraphically, the LTE is placed in the mid-S.neptuni ammonite zone of the European zonation and refers to the Scaphites nigricollensis to Prionocyclus quadratus ammonite zones in North America [Walaszczyk and Cobban, 2000], and to the Inoceramus costellatus and I. inaequivalvis inoceramid zones in Russia and northern Siberia [Zakharov et al., 1991; Sahagian et al., 1994]. In western Europe, the LTE is characterized by a distinct cooling trend (3–4°C) associated with a positive δ13C excursion (Figure 8c) [Voigt, 2000; Voigt and Wiese, 2000]. According to the age model, the duration for the LTE can be estimated to be in the order of 200 to 400 kyr. A major late Turonian third-order sea level drop is indicated by the Exxon sea level curve [Hardenbol et al., 1998] (Figure 8a). This large regression lasted about 1 million years, caused a widespread hiatus in the Western Interior Basin [Cobban and Scott, 1972; Leckie et al., 1997; Sageman et al., 1997], in western Europe [e.g., Hancock, 1989] on the Russian Platform and in northern Siberia [Sahagian and Jones, 1993; Sahagian et al., 1994]. The LTE has so far only been recorded from different European basinal sections, where several sedimentary sequences are superimposed on the late Turonian long-term sea level fall and rise (Figure 8c) [Wiese et al., 2004]. The occurrence of temporal expanded stratigraphical gaps at marginal settings and the lack of δ13C stratigraphies from other continental cratons have inhibited so far a precise correlation of high-order sequences with eustatic sea level changes. However, the late third-order sea level lowstand can be correlated with the initial LTE-δ13C increase. The subsequent transgression caused the development of sedimentary onlap surfaces in all basins of western Europe, North America, Russia and northern Siberia, and corresponds to the interval yielding the δ13C decrease in the Prionocyclus germari ammonite zone. The decrease of shelf-sea temperatures during the LTE refers only to a short interval of the global sea level lowstand and represents one of the European fourth-order sequences (Tu 6 in Figure 8a) [Wiese et al., 2004]. The subsequent rise of shelf-sea temperatures corresponds to the third-order transgression.
 At a first glance, our observed shelf-sea temperature variations correlate with falling and rising sea level during the MCE and LTE. The amplitude of the eustatic sea level change is larger than 10 m (20–30 m) in both cases, and the duration of the δ13C event approaches frequencies of orbital forcing. Thus our data would support the hypothesis of short-term growth of glaciers as a glacioeustatic mechanism for fast and high-amplitude sea level changes during greenhouse climate conditions [Miller et al., 2003]. However, the linkage between sea level and temperature is not straightforward in the shelf-sea settings. During the MCE, the coolest temperature estimate is related to the bed identified as a transgressive surface. During the LTE, the cooling of shelf-sea temperatures occurred only within a short interval during the major third-order sea level lowstand. If there was a linkage between waxing and waning of polar ice and sea level, the cooling of shelf-sea temperatures underwent probably not a direct forcing. Regional changes of shelf-sea circulation as a consequence of changes in sea level and North Atlantic circulation are a more realistic mechanism to explain the observed shelf-sea temperature variation.
 Shelf-sea temperature evolution throughout the CTBE shows a prominent increase of about 4°C during the late δ13C excursion, when a second δ13C increase in the inorganic carbon reservoir occurred (∼150 kyr after initiation of OAE 2, Figure 8b). The temperature rise is preceded by a brief cooling interval (10–20 kyr duration), which is not resolved by our δ18O record, but evident from the southward spread of boreal taxa [Jefferies, 1962; Gale and Christensen, 1996]. The short cool-pulse in the middle of OAE 2 occurred shortly after the main δ13C excursion (10–20 kyr), and can be seen as a first climate feedback related to increased rates of global organic carbon burial and reduction of atmospheric CO2 [Arthur et al., 1987; Kuypers et al., 1999]. Instead, the late and longer warming trend seems more to be related to the causes of the second δ13C rise and ongoing carbon cycle perturbation. Increased rates of volcanism and hydrothermal activity, as a consequence of the emplacement of large oceanic plateau basalts, were suggested as the cause of increased greenhouse gas forcing and enhanced nutrient availability, considered to be crucial for both the initiation and the response of the Cenomanian-Turonian carbon cycle perturbation [Larson, 1991; Orth et al., 1993; Sinton and Duncan, 1997; Kerr, 1998]. An additionally proposed mechanism for OAE 2 is the opening of a deep water connection between the North and South Atlantic oceans [Tucholke and Vogt, 1979; Summerhayes, 1987; Pletsch et al., 2001; Kuypers et al., 2002]. Poulsen et al.  performed coupled atmosphere-ocean model simulations to study the climatic effect of gateway opening, and found substantial oceanographic changes in their postrifting experiment. Accordingly, North Atlantic surface waters experienced both freshening (0.5–1‰) and warming (0.5°C), which would partly account for the observed 1.0‰ δ18O decrease in brachiopod calcite. A substantial change in ocean and shelf-sea circulation is also indicated by the spread of oligotrophic conditions in western European shelf seas in relation to the sea level rise contemporaneous with the second δ13C rise [Gale et al., 2000]. A further implication of warming and freshening of European shelf-seawater would be the lowering of its density, which could have forced a cessation of deep water formation. Interesting in this context is, that the evidence of photic zone anoxia in the southern North Atlantic ocean is also related to the late OAE 2 carbon isotope excursion [Kuypers et al., 2002], supporting the idea of a diminished or collapsed oceanic ventilation within the North Atlantic as a late feedback. On the other side, enhanced production of deepwater on shelves, which drives oceanic overturn, productivity, and the positive δ13C excursion is a long-time proposed mechanism for OAE 2 [Arthur et al., 1988]. Here we can only speculate that increased oceanic overturn and ventilation were probably related to the early phase of OAE 2. Oceanic surface-to-deep water temperature gradients with a sufficient temporal resolution are needed to address the question whether diminished and/or enhanced oceanic circulation are accompanied with the Cenomanian-Turonian carbon cycle perturbation.
 An 8 million year midlatitude δ18O-derived shelf-sea temperature record, based on well-preserved brachiopod shells, reveals a long-term climate warming through the Cenomanian that reached a maximum (23°C) in the early to early late Turonian. The most prominent temperature rise of 4–5°C occurred during the CTBE, indicating a major turnover of thermohaline circulation in the North Atlantic basin. A comparison of open oceanic and shelf sea δ18O data from tropical and subtropical sites with those from the temperate midlatitudes demonstrates the opportunity that the temperate shelf-seawaters could have had the highest density surface waters in the Cenomanian North Atlantic ocean if the continental runoff was low. Owing to the restriction of deep water connections by geographic barriers, waters of the temperate to boreal shelf seas could have contributed to the formation of intermediate and/or deeper waters. Two minor shelf sea cooling events in the mid-Cenomanian and late Turonian coincide with carbon cycle perturbations and large amplitude sea level falls, and thus provide supporting evidence for potential glacioeustatic effects. However, the linkage between sea level and temperature is not straightforward in shelf-sea settings. During the MCE, the temperature minimum correlates with a transgressive bed, and during the LTE, shelf-sea cooling represents only a brief interval within the major third-order sea level lowstand. Regional changes in shelf-sea circulation as a consequence of changes in sea level and North Atlantic circulation are the most plausible mechanism to explain the observed shelf-sea temperature variations.
 We thank Frank Wiese, Markus Wilmsen and Rory Mortimore for providing brachiopod samples and assistance during field work. The insightful reviews by Karen Bice, Paul Wilson and Mike Arthur are gratefully acknowledged. Kay Stone is thanked for her grammatical corrections. This work was supported by grants of the Deutsche Forschungsgemeinschaft (VO 387/3 and Ha 2891/3).