The Mainz Basin is located near the northwestern margin of the Upper Rhine Graben. Its formation dates back to the middle Eocene and is related to the taphrogenesis of the European Cenozoic rift system (Dèzes et al., 2004; Ritzkowski, 2005; Grimm et al., 2011; Fig. 1). Its sedimentary succession, which was mainly deposited in shallow marine and brackish settings, covers a time interval of ca. 20 My, from the Lutetian (∼ 49.5 Ma) to the early Burdigalian (∼ 18.5 Ma). During the late Rupelian (i.e., late Early Oligocene; 34.9–28.4 Ma; Gradstein et al., 2004), central Europe experienced a period of extended marine transgressions due to local tectonic subsidence and eustatic sea level rise. As a consequence, the Mainz Basin became part of a marine strait that extended from the paleo-North Sea basin to the southernmost Upper Rhine Graben (Picot, 2002; Sissingh, 2003; Berger et al., 2005a). A possible southern connection with the western Molasse Basin has been controversially debated (Martini, 1982; Picot, 2002; Berger et al., 2005a, b; Grimm, 2006).

According to Berger et al. (2005b), marine conditions in the Mainz Basin lasted for about 2.5 My from the sea-level highstands Ru2/Ru3 (∼ 32 Ma) to Ru3/Ru4 (∼ 29.5 Ma) of Haq et al. (1988). Nearshore deposits representing that time comprise coarse-grained (sand to gravel) siliciclastics of the Alzey formation (from which the studied fossil material was collected – see below) and the overlaying Stadecken formation. Contemporaneously deposited basinal sediments (pelites) belong to the Bodenheim formation (Grimm et al., 2000; Sissingh, 2003; Berger et al., 2005b). Age control for the marine strata of the Mainz Basin is mainly based on calcareous nanoplankton (Martini and Müller, 1971; Martini, 1982), dinoflagellate cysts (e.g., Pross, 1997) and, to a lesser extent, benthic foraminifera (Grimm, 1998, 2002). With regard to the nanoplankton zonation, the Alzey and Bodenheim formations comprise the upper part of nanoplankton zone NP23 and the lower part of nanoplankton zone NP24 (Grimm, 1994; Pross and Schmiedl, 2002; Berger et al., 2005b).

Paleoenvironmental reconstructions of the Alzey formation are based on palynological and faunal data indicating an overall warm climate comparable to modern subtropical climate zones of the Mediterranean (Grimm et al., 2011). Reconstructed mean annual air temperature in the hinterland fluctuated between ∼ 16 and ∼ 17 ∘C, and mean annual precipitation was between 1000 and 1250 mm per year (Pross et al., 1998, 2000). The Mainz Basin experienced repeated alternations between drier and wetter conditions during the deposition of the Alzey formation, which caused fluctuations in surface water salinity levels (remaining within the range of normal marine conditions) and the episodic formation of water-mass stratifications (Pross, 2001; Pross and Schmiedl, 2002). Temperature estimates of the seawater have been derived from the δ18O values of shark teeth and foraminiferal tests. These estimates yielded values of 6.9 to 23.3 ∘C for shallow-water settings (Tütken, 2003) and 5.9 to 14.9 ∘C for bottom waters (maximum depth: 150 m; Grimm, 1994; Grimm et al., 2011).

The studied shell material was collected by Jürgen Boy during the 1970s and 1980s and has been stored at the Paleontology Collection of the Institute of Geosciences in Mainz. Samples originate from the “Trift” outcrop near Weinheim, the stratotype of the Alzey formation (Grimm et al., 2000; Fig. 1). Additional information about the precise stratigraphic position of the sampled layer is not available. The outcrop is ∼ 8 m thick and ∼ 40 m wide; it consists of fossiliferous middle-to-coarse sands and fine gravels. Numerical dating with strontium isotope (87Sr/86Sr) stratigraphy of a well-preserved bivalve shell from the outcrop yielded an age of 30.1 ± 0.1 Ma (Grimm et al., 2003). The outcrop exhibits a highly diverse and fully marine benthic fauna dominated by bivalves, gastropods and scaphopods that dwelled in shallow subtidal waters. Furthermore, corals were found, suggesting limited seasonal salinity changes. Water-depth estimates, which are based on sedimentological features (Grimm et al., 2003) and ichnofossils (Schindler et al., 2005), range from ∼ 30 to ∼ 40 m. The sea cow teeth originated from four localities located along the southwestern paleo-coastline of the Mainz Basin (Fig. 1; Table 1), and were stored at the Paleontology Collection of the Institute of Geosciences and the collection of the Museum of Natural History Mainz/Landessamlung für Naturkunde Rheinland Pfalz.

From the Glycymeris planicostalis specimens collected at the Trift outcrop, three large valves (∼ 8 cm in height) that visually appeared well-preserved were selected for further investigations. These valves were labeled (MB-Wht-2, MB-Wht-4 and MB-Wht-7), mounted on Plexiglas cubes with GlueTec Multipower plastic welder and coated with WIKO metal epoxy resin to avoid fracture during cross sectioning. From each valve, two ca. 3 mm thick slabs were cut perpendicular to the growth lines and along the axis of maximum growth from the umbo to the commissure using a low-speed precision saw (Buehler Isomet 1000; at 200 rpm) equipped with a wafer-thin (0.4 mm), diamond-coated blade. Both shell slabs were glued to glass slides with the mirroring sides facing up, ground on glass plates (320, 800, 1200 grit SiC powder) and polished with 1 µm Al2O3 powder. After each preparation step, the samples were ultrasonically rinsed in deionized water.

One polished slab of each specimen was firstly used for diagenetic screening. For this purpose, a set of different methods was employed including cathodoluminescence petrography, Raman spectroscopy and immersion of the shell slabs in Feigl solution. The presence of Mn2+ (> 10–20 ppm) in calcium carbonates produces an orange cathodoluminescence (Machel et al., 1991) and is typically regarded as an indicator of diagenetic neomorphism (Grossman et al., 1996; Flügel, 2004) because biogenic aragonite is non-cathodoluminescent (Major, 1991). Like modern Glycymeris spp., shells of fossil representatives of this genus consisted of aragonite, which is prone to change to calcite during diagenesis. Raman spectroscopy can yield detailed and spatially highly resolved information on the type of polymorphs of CaCO3. The Feigl test can likewise distinguish between aragonite and calcite (Feigl, 1958). Feigl solution stains aragonite black and calcite pale grey. After diagenesis screening, the shell slabs were ground and polished again, and prepared for sclerochronological studies and subsequent scanning electron microscopic (SEM) analyses. For this purpose, polished cross sections were immersed in Mutvei's solution for 40 min under constant stirring at 37–40 ∘C (Schöne et al., 2005a). After the staining process, the samples were gently rinsed in deionized water, air dried and then photographed with a digital camera (Canon EOS 600D) mounted to a binocular microscope (Wild Heerbrugg M8). Growth increments were counted and their width measured with the image processing software Panopea (^©Peinl and Schöne). Subsequently, samples were sputter-coated with a 2 nm thick gold layer and viewed under a scanning electron microscope (LOT Quantum Design Phenom Pro, second generation) in order to describe the prevailing microstructures and identify possible neomorphic minerals that may have formed during diagenesis (Fig. 2).

The other polished slab of each specimen was used for the analysis of oxygen isotope values (δ18Oshell). Prior to the analysis, the outer ca. 0.5 to 1 mm thick chalky rim of the shell surfaces was physically removed (Fig. 2a–c). Then, 675 individual carbonate powder samples were micro-milled from the outer shell layer (ventral margin) of the three specimens (316, 193 and 166 samples from specimens MB-Wht-2, MB-Wht-4 and MB-Wht-7, respectively) using a Rexim Minimo dental drill mounted to a stereo microscope and equipped with a cylindrical, diamond-coated bit (1 mm diameter; Komet/Gebr. Brasseler GmbH & Co. KG, model no. 835 104 010). Sampling was performed in the ontogenetically youngest part of the shells. Individual milling steps contoured the shell growth patterns and measured between 100 and 200 µm in width. Carbonate powder samples weighing between 50 and 120 µg were reacted with 100 % phosphoric acid in He-flushed borosilicate exetainers at 72 ∘C. The resulting CO2 was measured with a GasBench II-coupled Thermo Finnigan MAT 253 gas source isotope ratio mass spectrometer in continuous flow mode at the Institute of Geosciences of the University of Mainz. Oxygen isotope values are reported in δ notation and given as parts per mill (‰). Data were calibrated against a NBS-19 calibrated IVA Carrara marble (δ18O = -1.91 ‰). On average, replicated internal precision (1σ) and accuracy (1σ) were better than 0.05 ‰, respectively.

If the bivalves formed their shell in oxygen isotopic equilibrium with the ambient water, the δ18Oshell values could provide information on water temperature during growth (Epstein et al., 1953). For aragonitic shells, the paleothermometry equation of Grossman and Ku (1986) with a scale correction of -0.27 ‰ (see Dettman et al., 1999) is typically employed: Tδ18O∘C=20.60-4.34⋅δ18Oshell-δ18Owater-0.27, where δ18Oshell is measured relative to VPDB and δ18Owater relative to VSMOW. Computing reliable temperatures from δ18Oshell values also requires knowledge of the δ18Owater value during shell formation. This value was reconstructed from δ18OPO4 values of tooth enamel of sea cows, i.e., homoeothermic marine mammals, from the same stratigraphic level. The average Tδ18O error was calculated by combining the average precision errors of the mass spectrometric analyses of bivalve shells and sirenian teeth (error propagation method). This resulted in an average Tδ18O error of ±0.4 ∘C. An additional source of uncertainty is represented by the standard deviation of the reconstructed average δ18Owater values (±0.3 ‰). The integration of both values results in a combined temperature error of ±1.3 ∘C.

The oxygen isotope composition of tooth enamel from marine vertebrates can provide information on the δ18Owater value of ambient seawater (e.g., Lécuyer et al., 1996; Clementz and Koch, 2001; Clementz et al., 2006; Clementz and Sewall, 2009). Therefore, we measured the phosphate oxygen isotope composition (δ18OPO4) of the enamel from seven sirenian teeth of Halitherium schinzii recovered from the deposits of the Alzey formation.

The surface of the teeth was physically cleaned and then sampled with a handheld dental drill. Five teeth which were large enough were sampled twice: once from the top and once from the bottom of the crown. A fraction of each enamel powder sample was then treated with 2 % NaOCl and 0.18 mL of 0.1 molar acetic acid to remove organics and potential diagenetic carbonates, respectively. Subsequently, ca. 4 mg of each pretreated sample was converted into silver phosphate (Ag3PO4) following the method of O'Neil et al. (1994) with modifications of Dettman et al. (2001) and Tütken et al. (2006). Triplicates of 500 µg of each Ag3PO4 sample were analyzed with a Thermo Fisher Delta Plus XL mass spectrometer coupled to a thermal conversion elemental analyzer (TC/EA) at the University of Tübingen, Germany. Measured values were normalized to calibrated in-house standards, Tu-1 and Tu-2 (Vennemann et al., 2002) and reported in δ notation versus VSMOW. Replicate (n= 6) analyses of NBS 120c (pretreated as the samples) yielded a δ18OPO4 value of 21.6 ± 0.13 ‰, which agrees well with the value of 21.7 ‰ reported by Lécuyer et al. (1993) and subsequently confirmed by many other laboratories (summarized in the appendix of Chenery et al., 2010). The δ18OPO4 values of the H. schinzii teeth of the Mainz Basin were converted into δ18Owater values using the equation determined for modern sirenians by Tütken (2003): δ18Owater=δ18OPO4-20.230.86, where δ18Owater and δ18OPO4 are given relative to VSMOW. We used the equation of Tütken (2003) instead of the one proposed by Lécuyer et al. (1996) because the calibration of Eq. (2) is based on more sea cow specimens and covers a range of measured ambient δ18Owater values 3 times larger. However, δ18Owater values reconstructed using both equations yielded similar values that are statistically invariant (Tütken, 2003: -0.9 ± 0.3 ‰; Lécuyer et al., 1996: -0.6± ca. 0.8 ‰).

In order to assess the possibility of diagenetic alteration of the enamel oxygen isotope composition, the carbonate (δ18OCO3) and the phosphate (δ18OPO4) group of the enamel were plotted against each other and compared to a compilation of δ18OCO3 vs. δ18OPO4 pairs from extant mammals published by Pellegrini et al. (2011). The δ18OCO3 values were determined in the remaining fraction of the pretreated H. schinzii enamel powders. About 800 µg of each enamel powder sample was analyzed with a Thermo Finnigan MAT 253 gas source isotope ratio mass spectrometer in continuous flow mode equipped with a GasBench II at the University of Mainz. The δ18OCO3 values were measured against VPDB and normalized to a NBS-18 and NBS-19 calibrated Laas marble (-5.21  ‰; replicated precision, 1σ, better than 0.1 ‰). Afterward, the results were converted to the SMOW scale using the equation of Coplen et al. (1983): δ18OSMOW=1.03091⋅δ18OPDB+30.91.

In bivalves, shell growth rate declines during ontogeny resulting in increasingly narrow annual growth increments with increasing lifespan (Jones and Quitmyer, 1996). Since the isotope samples were taken at approximately equidistant intervals (100 to 200 µm), the number of samples per year decreases towards the ventral margin, and the time represented by each carbonate sample (time-averaging) increases in ontogenetically older shell portions. To compensate for that bias and make the isotope samples from different ontogenetic years comparable to each other, the number of δ18Oshell values per year was mathematically equalized by a re-sampling technique similar to that described in Schöne et al. (2004) and Hallmann et al. (2011). Following previous work (e.g., Schöne and Fiebig, 2008; Wanamaker et al., 2011), we fitted the isotope data of each annual increment with a seven-point cubic spline using Analyseries 1.1 software (Paillard et al., 1996) and re-sampled each intra-annual curve so that the same number of isotope values were available for each annual increment, i.e., seven δ18Oshell' values. This re-sampling method slightly deviated from previous approaches (Schöne et al., 2004; Hallmann et al., 2011) because it was impossible to determine seasonal growth curves from microgrowth patterns. Hence, the δ18Oshell' values within a given year most likely represented different amounts of time, but the first, second, third, etc. δ18Oshell' values of different years represented the same amounts of time.

According to a set of different diagenesis screening tests outlined above, the studied shells of Glycymeris planicostalis from the Rupelian of the Mainz Basin consist of aragonite and were remarkably well preserved. This even applies to the chalky rims of the shells, i.e., the shell portions that were only pale blue stained by Mutvei's solution and lost almost all organics during taphonomy. Orange cathodoluminescence was only emitted from a very few isolated spots, i.e., cracks containing neomorphic mineral phases. Most other portions of the shells were dark blue to non-luminescent. Moreover, both reflected light microscope and electron microscope analyses revealed the same shell microstructures that occur in modern representatives of this genus, i.e., crossed-lamellar structures (Fig. 2d–f). Alternately, the first-order lamellae appear dark and bright because the higher-order lamellae are arranged in a fence-like manner and stand perpendicular to each other (compare Füllenbach et al., 2014). Furthermore, both shell layers are perforated by numerous hollow microtubuli (Fig. 2g), especially in the juvenile portion of the shells. On rare occasions, these tubuli (ca. 10 µm in diameter) are filled with pyrite crystals or iron oxides.

Like the bivalves, the studied sirenian teeth are well preserved (Fig. 3). The δ18OCO3 vs. δ18OPO4 pairs of the seven specimens plot well within the 95 % prediction intervals of modern and other well-preserved fossil mammal bioapatite data compiled by Pellegrini et al. (2011; Table 1; Fig. 3). Digenetic alteration of tooth enamel would in the first place have affected the carbonate-bound oxygen (Iacumin et al., 1996) and resulted in δ18OCO3 vs. δ18OPO4 pairs plotted farther away from the regression line depicted in Pellegrini et al. (2011). Given the excellent preservation, δ18Owater values were computed from δ18OPO4 values of the enamel using Eq. (2). On average, the δ18Owater value of the ambient seawater was -0.9 ± 0.3 ‰ (1σ).

The studied fossil G. planicostalis specimens show distinct growth lines in the ventral margin and the hinge plate of Mutvei-stained cross sections (Fig. 2h). These lines were previously identified as periodic annual features (Berthou et al., 1986; Royer et al., 2013; Bušelić et al., 2014) separating the growth pattern in annual time slices, i.e., annual growth increments. The annual growth lines are more distinctly developed and hence easier to discern in the hinge plate than in the ventral margin. Based on annual increment counts, it was possible to determine the ontogenetic ages of the specimens. Specimens MB-Wht-2, MB-Wht-4 and MB-Wht-7 reached life spans of 77, 84 and 67 years, respectively.

Oxygen isotope curves of all three specimens exhibit distinct seasonal oscillations (16, 14 and 10 cycles in specimens MB-Wht-2, 4, and 7, respectively) with the annual growth lines occurring shortly after the most negative δ18Oshell values of each cycle (Fig. 4; see Supplement). In other words, the full seasonal amplitudes are preserved in the shells including winter and summer values. The annual growth line formation occurred in late summer/early fall.

The shells grew faster during the first half of the year than after summer. This is well reflected in the seasonal temperature curve based on the averaged δ18Oshell' values of all 40 measured annual increments (Fig. 5). There are more data points in shell portions formed during spring than in shell portions formed during fall (Fig. 5). Accordingly, the reconstructed temperature curve is right skewed.

The average annual δ18Oshell values and seasonal δ18Oshell ranges are fairly similar among the three studied specimens (Table 2). Seasonal extremes fluctuate between -1.48 ‰ (summer value) and 0.75 ‰ (winter value) in specimen MB-Wht-2, between -1.16 and 0.67 ‰ in specimen MB-Wht-4 and between -1.19 and 0.60 ‰ in specimen MB-Wht-7. Using the reconstructed δ18Owater value, this translates into total temperature (Tδ18O) ranges of 9.7, 7.6 and 7.8 ∘C in specimens MB-Wht-2, MB-Wht-4 and MB-Wht-7, respectively. Taking the resampled values of the 40 seasonal cycles of all three specimens, the average annual temperature is 15.4 ± 0.7 ∘C (1σ), and the seasonal temperature range equals 3.7 ∘C with average minimum (winter) values of 13.6 ± 0.8 ∘C (1σ) and average maximum (summer) values of 17.3 ± 1.2 ∘C (1σ). It is worth noting that the seasonal amplitudes vary through time. In some years, the seasonal Tδ18O range was less than 2 ∘C (Fig. 4).

According to diagenetic screening, the studied G. planicostalis shells are well preserved. The shells consist of pristine aragonite. Furthermore, SEM analysis revealed original delicate shell microstructures including the typical skeletal feature of Glycymeris spp., i.e., microtubuli. These cylindrical cavities perforate the inner and outer shell layers and are filled with organics during the lifetime of the animal (e.g., Waller, 1980; Crippa, 2013). The diagenetic loss of organic material leaves behind hollow cavities that could potentially be filled with neomorphic mineral phases. However, the microtubuli of the studied specimens were typically hollow and only rarely contained pyrite. Pyrite crystals can even occur in shells of living bivalves and are possibly related to the bacterial degradation of organic matter (Clark and Lutz, 1980).

In fact, the recovery of pristinely preserved fossil Glycymeris spp. has been reported from many other localities and geological time intervals (e.g., Tsuboi and Hirata, 1935; Crippa, 2013). Since Glycymeris spp. dwell in sandy to fine gravelly habitats, shells of this genus are usually embedded in coarse grained and highly porous sediments. In such types of host rock and stratigraphic age, one would not expect aragonitic shell preservation, particularly if the burial depth is shallow (few tens of meters) and the sediment is still unconsolidated as is the case with the weakly cemented sandstones of the Alzey formation. Under surface conditions, aragonite is metastable and slowly turns into the more stable polymorph of CaCO3, i.e., calcite (Boettcher and Wyllie, 1967). This conversion into calcite can be expedited when a fluid is present (Putnis and Putnis, 2007) and when temperature is increased (Dasgupta, 1963). The resistance of Glycymeris shells against diagenetic alteration likely resulted from low organic content and the dense crossed-lamellar microstructure (Taylor and Layman, 1972).

Pristine preservation is a major prerequisite for the reconstruction of environmental variables from geochemical properties of the shells including ambient water temperature from δ18Oshell values. According to shell oxygen isotope data, the studied shells grew during winter and summer. Similar findings on shell growth during seasonal extremes were recently reported for modern Glycymeris bimaculata from Croatia (Bušelić et al., 2014). The only difference is that the Oligocene shells formed annual growth lines in late summer/early fall, whereas the period of extremely slow or no shell growth in specimens from Croatia occurs during spring. On the contrary, modern G. glycymeris from the North Atlantic form annual growth breaks in winter (Berthou et al., 1986; Royer et al., 2013). These findings suggest that the timing and rate of shell growth can vary greatly among different species of the same genus and most likely even among specimens of the same species living in different localities (e.g., Ansell, 1968; Jones and Quitmyer, 1996).

A number of explanations have been proposed to explain the reason for periodic cessation of shell growth. Temperature stress seems to limit shell growth in many bivalves. Above and below a taxon-specific water temperature range, biomineralization ceases and results in the formation of winter or summer growth lines (Jones and Quitmyer, 1996) or even both (Schöne et al., 2002). For example, Mercenaria mercenaria stops shell growth above 31 and below 9 ∘C (Ansell, 1968). Following this explanation, modern G. glycymeris from the North Atlantic forms winter lines when temperatures fall below the tolerance limit of this species, as suggested by Royer et al. (2013). In some taxa, annual growth line formation can also be linked to the reproductive cycle. For example, Phacosoma japonicum from Japan not only forms winter lines but also slows down shell growth regularly during June and July, i.e., during the peak spawning phase (Sato, 1995). Instead of biomineralizing shell, the energy is then allocated to the formation of eggs and sperm. Spawning breaks may be limited to species lacking specific tissues for energy (lipids) storage. Modern G. glycymeris from the North Atlantic, for example, belongs to this group of bivalves and reportedly spawns once or twice per year between spring and fall (Galap et al., 1997). In the Mediterranean Sea, G. nummaria spawns in July and August, followed by a sudden decrease of the condition index in late summer/fall (Crnčević et al., 2013). However, it has remained unresolved whether modern Glycymeris spp. can only form shell during periods of sexual inactivity. Existing studies on seasonal shell growth of this genus were based on juvenile specimens or juvenile portions of adult specimens (Peharda et al., 2012; Bušelić et al., 2014; Royer et al., 2013). Therefore, it is difficult to draw conclusions on how the reproductive cycle affects seasonal growth in (modern) Glycymeris spp. It is not possible to determine whether growth line formation of G. planicostalis was governed by reproduction or other environmental factors. At the least, the annual growth lines in the studied specimens from the Oligocene are unrelated to seasonal temperature extremes because the lines do not fall together with the most negative and positive oxygen-isotope-derived water temperatures (Fig. 4).

Shell growth rates of the studied bivalves from the Early Oligocene of the Mainz Basin also varied during the main growing season. For example, shell production was faster during spring and summer than during fall and winter. This finding has implications for geochemical sampling strategies. In order to obtain reliable information on the actual seasonal temperature spread, a higher sampling resolution has to be applied in slow-growing shell portions.

Only few temperature estimates of the Mainz Basin and adjacent regions during the Rupelian are currently available. For example, sediments of the Alzey formation contain diverse warm-water fauna including marine fish, mammals and crocodiles as well as terrestrial turtles. Based on this fossil assemblage, subtropical climate conditions – similar to the modern southeastern Mediterranean – were inferred for the Mainz Basin (Grimm et al., 2003, 2011). Furthermore, macroflora and palynological data from the Bodenheim formation yielded winter and summer air temperatures of 7.1–10.2 and 25.7–28.1 ∘C, respectively (Pross et al., 1998; Pross et al., 2000). These estimates compare well with those obtained from fossil floras of other contemporaneous localities in central Europe (Mosbrugger et al., 2005; Erdei et al., 2012).

Knowledge on water temperatures of the Mainz Basin comes from oxygen isotope compositions of biogenic skeletons. Tütken (2003) reported δ18OPO4 values of shark teeth that correspond to absolute temperatures between 6.9 and 23.3 ∘C (temperatures recalculated assuming a δ18Owater value of -0.9 ‰; Table 3), using the thermometry equation of Longinelli and Nuti (1973). Grimm (1994) reported oxygen isotope data of planktonic and benthic foraminifera that can be converted into absolute temperatures using the paleothermometry equation by Anderson and Arthur (1983) and a δ18Owater value of -0.9 ‰. Based on this calculation, sea surface temperatures of the Mainz Basin fluctuated between 11.7 and 21.3 ∘C (Table 4), whereas bottom water (up to 150 m depth; Grimm et al., 2011) temperatures were as cold as 5.9 to 14.9 ∘C during the Rupelian. Similar surface water temperatures were reconstructed from bivalve shells in the present study (12.3 and 22.0 ∘C), although the lowest temperatures are ∼ 5 ∘C higher than those obtained from shark teeth (Fig. 6). Leaving aside the fact that it is rather unlikely that the studied bivalves, sharks and foraminifera lived during the exact same time interval, a direct comparison of temperature extremes derived from the different marine archives seems problematic for a variety of reasons:

(i) The temporal resolution provided by foraminifera is much higher than that of bivalves. Foraminiferan tests can grow within a few weeks (Bé et al., 1981). Thus, each specimen recorded environmental conditions during a very short time interval of the year. In contrast, each sample taken from the bivalve shells represents, on average, about 2 weeks to 1 month worth of growth. Foraminifera samples analyzed so far may not necessarily have grown when the most extreme seasonal temperatures occurred. Accordingly, actual winter temperatures may have been slightly colder and summers slightly warmer than suggested by the δ18O values of foraminifera.

(ii) Seasonal temperature extremes given by sharks may not represent the actual temperatures of where the bivalves lived. Being highly mobile, nektonic organisms, the sharks may have foraged temporarily in the shallowest waters near the coast and at other times dived to the very bottom of the sea. In addition to vertical movements, they may have traveled large distances such as modern sharks (e.g., Domeier and Nasby-Lucas, 2008). The lowest temperatures recorded by sharks may thus represent conditions below the thermocline or settings much further north instead of winter temperatures in the Mainz Basin. In turn, those teeth that provided temperature estimates of 22 ∘C may actually have been formed while the animals lived in warmer waters farther south or near the coast.

(iii) Actual sea surface temperatures during both winter and summer may have been underestimated by the planktonic foraminifera (and sharks while they resided in shallowest, coastal waters) because the assumed average δ18Owater value did not reflect the actual isotope signature of the water. Planktonic foraminifera lived in the upper few meters of the water column in a narrow, shallow epicontinental sea. In such a habitat, seasonally varying riverine freshwater influx, precipitation and evaporation rates likely resulted in seasonal changes of the δ18Owater value. Increased evaporation during summer may have shifted the δ18Owater value toward more positive values resulting in lower-than-actual reconstructed water temperatures near the sea surface. In contrast, higher precipitation rates during winter may have shifted the δ18Owater value toward more negative values so that the oxygen-isotope-derived temperatures appeared colder than they actually were. Indications for a seasonally varying δ18Owater value come, to some extent, from sirenian teeth. Like their extant relatives, sea cows lived in the upper 10 meters of the ocean and near the coast (Louise Chilvers et al., 2004). Thus, they have recorded the isotope signature of the near-coastal surface water in which the planktonic foraminifera (Globigerina sp.) lived. Reconstructed δ18Owater values fluctuated between -0.2 and -1.4 ‰. If the latter value reflects conditions during winter and the former during summer, actual seasonal temperatures in the upper few meters of the Mainz Basin reconstructed from foraminifera would range between ca. 11 and 27 ∘C.

(iv) Bivalve shell-based temperature estimates cannot be compared directly to those of planktonic foraminifera even if the fluctuating δ18Owater values in the upper few meters of the ocean are precisely known. At ca. 30–40 m water depth, bivalves likely experienced a smaller seasonal temperature range than organisms in the upper few meters of the sea.

The benthic faunal associations of the Mainz Basin have been interpreted as reflecting paleoclimate conditions similar to those of the modern southeastern Mediterranean Sea (Grimm et al., 2003, 2011). However, temperature estimates derived from G. planicostalis shells suggest lower water temperatures. According to hydrographical studies from coastal regions in the northwestern Mediterranean (France) and southeastern Mediterranean (Lebanon), the temperature in 30 to 40 m water depth is still influenced by surface conditions (Abboud-Abi Saab et al., 2004). At 35 m water depth, the temperatures off Lebanon ranged between 16.9 and 29.1 ∘C with an annual average of 22.5 ± 4.1 ∘C (1σ), whereas the coastal waters off France ranged between 12.3 and 21.8 ∘C with an annual average of 15.2 ± 2.2 ∘C (1σ). Temperatures recorded by G. planicostalis lie well within these ranges, which would suggest that water temperatures were more similar to regions in the northwestern Mediterranean than those in subtropical areas. However, the mean annual precipitation in the area of Marseille (Mediterranean coast of southern France) equals 751 ± 172 mm (Harris et al., 2014), which is considerably lower than 1000–1250 mm a-1 reconstructed precipitation rates for the hinterland of the Mainz Basin (Pross et al., 1998, 2000). A possible explanation for the high precipitation rates in central Europe during the Oligocene has been provided by Pross and Schmiedl (2002). The deposition of the Alzey formation and its basinal counterpart, the Bodenheim formation, took place during sea-level highstands, which could have increased the moisture concentration in the atmosphere and thus intensified rainfalls. Such linkage between sea level rise and precipitation has recently been postulated for the early Holocene intensification of Australian–Indonesian monsoon rainfall (Griffiths et al., 2009).

The studied specimens of G. planicostalis offer a number of advantages over existing marine paleoclimate archives. As sessile organisms, bivalves record the water properties at a specific locality and depth throughout their lifetimes. Since their shells grew almost year-round, each isotope sample can be assigned to a particular season. If preservation permits, daily microgrowth increments can be employed to temporally contextualize the seasonal shell growth to the nearest week or so (e.g., Schöne et al., 2005b). Such an internal calendar is missing in foraminifera.

The studied G. planicostalis specimens lived for several decades and recorded seasonal temperature changes over the course of many consecutive years. This is a clear advantage over other climate archives that only provide very short temporal snapshots of unknown timing within the year such as foraminifera or shark teeth, or over a few consecutive years, such as fish otoliths. Analogously to bivalve shells, they form growth lines, and their oxygen isotope composition can be used for seasonal paleotemperature reconstructions, for example, at the Eocene–Oligocene boundary (Ivany et al., 2000). Although fish otoliths are very common components of marine nektonic microfossil assemblages, their reduced size makes them difficult to sample, and analyses of their chemical composition usually cover only short time intervals.

Long proxy records offer the possibility to examine the variability of winter and summer temperatures over the course of several years. Future studies should generate δ18Oshell time series of G. planicostalis that are long enough to permit spectral analyses. These data should then be combined with numerical climate models for that time. Furthermore, δ18Oshell chronologies should also be compared to increment widths in order to identify potential influences of temperature on shell growth rates.

The δ18Owater value at 30–40 m water depth was most probably much less variable than near the sea surface. Seasonal changes in freshwater influx into the Mainz Basin likely did not have any significant effect on the isotope signature of the water in which the bivalves lived. In fact, modern G. glycymeris from the North Atlantic is most prolific in water with stable salinity of 34–35 PSU (Rombouts et al., 2012). If the same preference is true for the Oligocene relatives of this genus, water temperatures can be reconstructed with smaller error bars from oxygen isotope values of the bivalve shells than from skeletal hard parts of nektonic and planktonic organisms. Evidently, absolute temperature estimates from δ18O values require knowledge of the oxygen isotope composition of the ambient water, which is rarely available for fossil environments. In the present study, the δ18Owater value was reconstructed from the tooth enamel of sea cows from the same stratigraphic level. Although the bivalves and the sea cows did most certainly not live during the exact same time and the sphere of action of the sea cows was the upper 10 meters of the ocean, the average δ18OPO4 value of the sirenian teeth serves as a reasonable estimate of the Rupelian δ18Owater value (-0.9 ‰) of the Mainz Basin. A similar value (-1 ‰) was also assumed by Grimm (1994). For comparison, the δ18Owater of the open ocean was -0.5 ‰ at that time (Lear et al., 2000). To test the temperature estimates obtained from δ18Oshell values and circumvent uncertainties related to the precise δ18Owater signature during shell formation, future studies should explore other potential temperature proxies such as Sr / Ca and Δ47 values (Eagle et al., 2013).