Paleoceanography of the Late Cretaceous northwestern Tethys Ocean: Seasonal upwelling or steady thermocline?

In this study we attempted to assess whether seasonal upwelling or a steady thermocline persisted at the western margin of the Tethys Ocean during the late Turonian–early Coniacian interval. For this scope, we employed novel and published stable oxygen isotope (δ18O) data of various organisms (bivalves, bivalves, brachiopods, fish and belemnites). New seasonally resolved temperature estimates were based on the δ18O record of sequentially sampled inoceramid (Inoceramus sp.) and rudist (Hippurites resectus) shells from the Scaglia Rossa and Gosau deposits of northern Italy and western Austria, respectively. Diagenetic screening was performed using reflected light, cathodoluminescence (CL), scanning electron microscopy (SEM) and stable isotope analysis. Originally preserved δ13C and δ18O values were used to characterize the lifestyle of the bivalves and detect vital effects that could have biased oxygen isotope-based temperature reconstructions. Inoceramid δ18O values provide–for the first time–information on temperatures of Tethyan benthic waters, which were, on average, 14.4 ± 0.6 °C and fluctuated seasonally within a range of less than 2 °C. Such a thermal regime is in line with the temperatures postulated for late Turonian boreal water masses and support the existence of a cold water supply from the North Atlantic to the Tethyan bottom. Bottom cooling, however, did not affect the shallow water environment. In fact, the rudist-based temperature estimates for shallow water environment revealed a mean annual range of 11 °C, between 24 and 35 °C (assuming a seasonally constant δ18Ow = 1.0 ‰), which are among the warmest temperatures recorded over the entire Late Cretaceous. Our findings, thus, suggest a strong thermal and food web decoupling between the two environments. The absence of a seasonal vertical homogenization of different water bodies suggests the existence of a steady thermocline and, therefore, contrasts with the presence of an active coastal upwelling in the region as hypothesized by previous authors.

In the present study, we sclerochronologically sampled δ 13 C and δ 18 O values of well-preserved rudist and inoceramid shells to estimate the sub-annual temperature variability that prevailed in the shallow and deeper water environments of the northwestern Tethyan Ocean during the Late Turonian-Early Coniacian interval (Fig 1). Shallow water seasonal variability was reconstructed by combining novel and published sclerochronological stable isotope data of rudist material from shallow water deposits of the Gosau Formation of Austria. We also present the first temperature reconstruction of the seafloor water temperature that prevailed in the hemipelagic environment of the Tethys. Seasonal temperatures were estimated from sequentially sampled δ 18 O values of Inoceramus sp. shells from the 'lastame' deposits of the Scaglia Rossa Formation of northern Italy. Prior to using the shells for environmental reconstructions, we performed a detailed diagenetic screening of the shell material by employing a combination of cathodoluminescence (CL), scanning electron microscopy (SEM) and stable isotope analysis of different carbonate phases. Well-preserved δ 13 C and δ 18 O values were used to discuss the autoecology of the bivalves and the impact of possible vital effects on the paleowater temperature estimates. Additionally, we combined our results with previously published temperature reconstructions of coeval seawaters of the Central European basins to gain information on the vertical temperature gradient of the studied water body. Finally, temperature estimates were used to test the hypothesis of a thermal decoupling between hemipelagic and shallow water environments in study region. The present study highlights the potential of bivalve sclerochronology for palaeoceanographic studies.

Geological settings, material and methods
In the present study we sclerochronologically analyzed the shells of two Inoceramus sp. (ID: SNSB-BSPG 1973 VII ?60 A, B) and one Hippurites resectus (ID: SNSBBSPG1991X) specimens (Fig 2) stored at the Bavarian State Collection for Paleontology and Geology of Munich. Permission for all analysis performed in this study was granted by the curator of the collection and specimens were permanently stored (under the above mentioned ID-numbers) and made publicly accessible at the Bavarian State Collection for Paleontology and Geology of Munich (Richard-Wagner-Str. 10, 80333 Munich, Germany). The rudist shell was 15 cm long and exhibited distinct growth lines in its outer shell layer (OSL) (Fig 2A). The shell originates from the mixed siliciclastic-carbonate deposits of the 'Atzl Reef' (Gosau Group) exposed near the town of Brandenberg (North Calcareous Alps, Austria). According to paleontological and petrological studies, the limestones of the 'Atzl Reef' were deposited in a shallow subtidal environment [63,64], and previous δ 18 O-based temperature reconstructions suggest very warm conditions compatible with a Late Cretaceous shallow water habitat [22]. To estimate the full seasonal temperature amplitude that prevailed in this environment, the stable isotope data of our specimen were combined with those of previously published Vaccinites inaequicostatus shells [22,65] from the 'Atzl reef' and the strata exposed near the town of Theresienstein (Salzburg, Austria).  [62]) showing the patterns of the main surface oceanic currents (black arrows). The location of the deep water formation site in the North Atlantic (white circle) [14] and inferred ingression paths of high latitude cold waters onto the Central European shelves and the northwestern Tethys (white arrows) are shown. (B) Paleogeography of the northwestern Tethys The inoceramid specimens consisted of two internal molds partially covered by fragments of the original OSL ( Fig 2E). The specimens were collected from the deposits of the Scaglia Rossa Formation near the town of Sant´Anna d´Alfaedo (Lessini Mountains, northern Italy). The pinkish nodular limestone surrounding the inoceramid allows to constrain the stratigraphic position to the 'lastame' lithozone [66][67][68][69][70]. The studied succession was deposited from the late Turonian to the Coniacian on the Trento Plateau, a hemipelagic rise located at the northern margin of the Tethys few hundreds of km south of the 'Atzl Reef' [66]. Existing estimates of the paleobathymetric range have yielded contrasting results. The occurrence of globotruncanids and the absence of benthic foraminifera in the sediment have been interpreted as suggesting bathyal depths well below the neritic zone [71]. Some authors postulated that the Trento Plateau was situated above the local oxygen minimum zone and suggested water depths between 400 and 700 m [72]. In a more recent study, the presence of radiolitid rudists in the 'lastame' deposits has been interpreted as indicative for water depths near the lower photic zone (ca. 100-150 m) [67].
The studied Gosau deposits belong to the late Turonian ammonite Subprionocyclus neptuni and Prionocyclus germani zones [73] and strontium isotope values of well-preserved rudist shells from the two localities revealed an upper Turonian-lowermost Coniacian age for the 'Atzl Reef' and an uppermost Turonian-lower Coniacian age for the Theresienstein material [74]. The 'lastame' deposits span from the upper Marginotrucana scneegansi to the lower Dicarinella concavata planktonic foraminifera zones and are comprised in the nannofossil zone CC12 or UC8 [69], which are temporally equivalent to the S. neptuni zone [75]. Accordingly, the studied time interval can be constrained to about 2 Ma, between 89 and 91 Ma [75].

Shell preparation, sampling and stable isotope analysis
Fragments were first coated with WIKO metal epoxy resin to avoid breakage during cross-sectioning. Then, a 1 cm-thick slab was cut from the specimen along the direction of growth (Fig  2A, 2B, 2E and 2F). The slabs were finally ground with SiC powders (F300, F800, F1200 grit) on glass plates and polished with a 1 μm Al 2 O 3 powder on a Buehler G-cloth.
The state of preservation of the material was evaluated visually, by means of reflected light microscopy as well as using cathodoluminescence (CL) and scanning electron microscopy (SEM). CL microscopy is a non-invasive technique, which uses a high-energy electron beam to identify the distribution of trace elements in minerals [76,77]. For example, low-Mg calcite (LMC) shells with high concentrations of Mn emit a bright orange to reddish CL and are typically regarded as diagenetically altered [53]. SEM analysis was performed on hand-broken fragments of the OSLs to evaluate the preservation of the shell ultrastructures and possible signs μm-scale recrystallization and precipitation of neomorphic material [36,78]. Finally, the stable isotope values of different biotic and abiotic carbonate phases were used to evaluate the effect of diagenesis on the original δ 13 C and δ 18 O record of the studied bivalve shells.
Carbonate powders (50-120 μm) for δ 13 C and δ 18 O measurements were collected from the biogenic LMC, from the micrite, and from sparite encrusting the shells and infilling voids (S1 Fig) using a Rexim Minimo dental drill equipped with a 300 μm-diameter conical SiC bit (Gebr. Brasseler GmbH & Co. KG, model number H52 104 003) mounted on a stereomicroscope. To ensure an adequate temporal resolution for seasonal environmental reconstructions, showing marked annual growth lines in the outer shell layer (OSL). (B) At higher magnification, microgrowth increments were easily recognizable. Their number within one annual growth increment (ca. 49) suggests weekly shell growth cycles for the studied H. resectus individual. (C) SEM imaging revealed the preservation of the original fibrous prismatic ultrastructure of the OSL and the presence of intra-prismatic voids inside the biomineral units. (E) One of the studied Inoceramus sp. fossils from the 'lastame' deposits of Sant´Anna d´Alfaedo (Italy). The inoceramid steinkern is partially covered by original OSL. The dotted line indicates the location of the shell cross-cut depicted in (F). In cross-cut very faint annual growth lines were recognized (F), however, these became more visible after increasing the contrast of the images (G). Under the SEM, the low-Mg calcite prisms of the studied inoceramids displayed smooth surfaces and sharp edges (H) and the absence of neomorphic precipitates between adjacent prisms. (J) The inoceramid prisms also displayed intra-prismatic voids, which were originally filled with organic matrix. ISL = inner shell layer. https://doi.org/10.1371/journal.pone.0238040.g002

PLOS ONE
Late Cretaceous NW Tethy Ocean: Upwelling or steady thermocline? the inoceramid LMC was sampled from the outer to inner surface of the fragment [54], whereas the rudist OSL was sampled following the direction of growth.
Stable carbon and oxygen isotope measurements were performed at the Institute of Geoscience of the University of Mainz. Carbonate powder were digested with a water-free phosphoric acid in He-flushed exetainers at 72˚C. The liberated CO 2 gas was then measured in continuous flow mode with a Thermo Finnegan MAT 253 gas source isotope ratio mass spectrometer coupled to a Gas Bench II. Data calibration was performed against a NBS-19 calibrated Carrara Marble distributed by IVA Analysensysteme GmbH (δ 13 C = +2.01 ‰; δ 18 O = -1.91 ‰). Stable isotope values are given in the δ-notation in parts per mil (‰). The 1σ accuracy (based on long-term analysis) and average 1σ internal precision (based on 8 injections per analyzed sample) for δ 13 C values were better than 0.04 and 0.02 ‰, respectively, and better than 0.05 ‰ and 0.03%, respectively, for δ 18 O values.

Estimation of water temperatures and δ 18 O signal
Since bivalve shells are formed near isotopic equilibrium with the ambient water, their δ 18 O values can be used to reconstruct paleowater temperatures [79]. The OSL of rudists and inoceramids was composed of fibrous and prismatic LMC, respectively. Both taxa also produced an aragonitic inner shell layer, which was not preserved in studied specimens. Therefore, in the present study, we employed a paleothermometry equation [80], which is typically used for rudist and inoceramid LMC [22,45]:  [81], whereas, climate model simulations suggest an average value of +0.5 ‰ for the Cretaceous Tethyan surface waters [82]. Furthermore, we constructed a model to assess the possible effects of the δ 18 O w changes in the water column on the reconstruction of the vertical temperature gradient. We simulated four scenarios by fitting polynomial functions to observational δ 18 O w data from the first 900 m water depths of tropical Indian, Pacific and Atlantic Ocean as well as the Gulf of Mexico [83]. Additionally, we employed the Late Cretaceous vertical δ 18 O w gradient simulated using the GENESIS-MOM numerical climate model [82]. The modeled gradients were used to calculate the vertical temperature amplitude (ΔT) required to justify the shell δ 18 O shell difference between the surface (10 m) and bottom environments (900 m water depth).

Rudist material
The OSL of the Hippurites resectus specimen exhibited marked dark brown major growth lines, which divided the shell growth pattern into 4-5 mm broad growth increments (Fig 2A).
In turn, major growth increments were subdivided by thin microgrowth lines into 40-50 microgrowth increments of 20-100 μm width ( Fig 2B). Under the CL, the OSL was non-luminescent, and SEM imaging revealed the preservation of the original fibrous calcitic ultrastructure of the shell. Occasionally, μm-scale neomorphic precipitates were observed on individual calcitic fibers. The presence of intra-crystalline voids, however, suggest that recrystallization remained only very superficial (Fig 2C and 2D). Orange to bright red luminescence was emitted by the sparitic cements, which replaced the original aragonitic shell layer of the rudist and exhibited the most negative δ 18 O values (-9.38 ‰) (  3A) and exhibited five consecutive oscillations, with the major growth lines coinciding with the δ 18 O and the δ 13 C minima ( Fig 4A).
Stable isotope values are given for the sediment, the sparitic calcite and the shells of the Inoceramus sp. and Hippurites resectus specimens.

PLOS ONE
Late Cretaceous NW Tethy Ocean: Upwelling or steady thermocline?

Inoceramid material
The inoceramid shells exhibited sharply defined polyhedral prisms, and faint major growth lines were identified when viewed under a reflected-light microscope (Fig 2F and 2G). Under CL, the studied inoceramid shell fragment was non-luminescent to dark blue luminescent, whereas, orange luminescence was emitted by the micrite and the sparite infilling the  Table 1). Under the SEM, LMC prisms were tightly packed, and no neomorphic precipitates were observed within the interprismatic space ( Fig 2I). The individual biomineral units displayed smooth surfaces and sharp edges, and preserved intra-crystalline voids, visible in cross-sections ( Fig 2H and 2J). The inoceramid LMC yielded the most positive δ 13 C and δ 18 O values, which were, on average, +2.83 ± 0.13 ‰ and -0.61 ± 0.13 ‰ (± 1σ), respectively (Fig 3B; Table 1). Shell δ 13 C values fluctuated between +2.04 and +3.38 ‰, whereas the δ 18 O data oscillated within a much narrower range, i.e., between -0.99 and -0.42 ‰ (Fig 4B). The δ 13 C and δ 18 O data were positively correlated (R = +0.83; R 2 = 0.68; p < 0.01) (Fig 3B), and the faint growth lines were located shortly after the local maxima ( Fig 4B).  Water temperatures were compiled using novel and published skeletal δ 18 O data. Temperatures were estimated assuming an average water δ 18 O signature of -1.00 ‰ [81]. Brachiopodderived temperature should be considered with caution, because these organisms may have not fractionated oxygen isotopes in equilibrium with the ambient water.

State of preservation
The reconstruction of paleowater temperatures from δ 18 O values requires a detailed diagenetic screening of the biogenic carbonates. This is because accidental sampling of diagenetically altered shell portions can lead to an overestimation of the actual paleowater temperature. The OSL of both analyzed bivalve species consisted of LMC, which is a carbonate polymorph considered to be fairly resistant against diagenetic alteration. As suggested by the presence of original macrogrowth and microgrowth lines, the studied shells did not undergo major recrystallization during burial. Likewise, at a microscopic scale we did not observe mineral modification. The SEM analysis of the rudist and the inoceramid OSL revealed well-defined fibrous prismatic (Fig 2C and 2D) and prismatic (Fig 2H-2J) aggregates, respectively. Particularly, the presence of intra-prismatic voids (Fig 2D and 2J), which are supposed to have been originally filled with intra-crystalline organic matter [86,87], suggest that alteration of the biomineral units (if present) remained only very superficial. In previous sclerochronological studies, this feature was considered to be indicative for very good preservation and a discriminant factor for δ 18 O-based temperature reconstructions [31,32, 86,87].
The well-developed δ 13 C and δ 18 O cycles measured in the Hippurites resectus specimen ( Fig  4A) allowed us to confidently exclude a resetting of the original geochemical signal during burial [88]. Hence, the stable isotope values were considered to reflect environmental seasonality. The narrow δ 13 C (1.0 ‰) and δ 18 O (0.40 ‰) fluctuations measured in the inoceramid shells required a more detailed evaluation. In fact, such reduced stable isotope amplitudes can also result from geochemical resetting. The sparite, which exhibited the lowest δ 13 C and δ 18 O values, is typically formed during burial and directly precipitates from pore fluids [89]. Therefore, its stable isotope values can be considered as the isotopic endmember of the diagenetic alteration. Accordingly, increasingly altered carbonate material should show a progressive trend toward the δ 13 C and δ 18 O values of the sparite [90]. Such trend was visible in the micrite (Fig 3B), which exhibited δ 13 C values (+2.60 ± 0.11 ‰) coherent with the bulk stable carbon isotope data of other late Turonian-Coniacian Scaglia Rossa successions from Central Italy [91]. The relative uniformity between the isotopic signal of the micrite and the sparite suggests a closed diagenetic system, in which the stable isotope composition of the pore fluids was dominated by the dissolution of the host rock [92]. Such a condition greatly increases the preservation potential of the original stable isotope signature of biogenic calcite [92]. This is also suggested by the fact that inoceramid δ 18 O and (partially) δ 13 C values were more positive than those of the micrite. Accordingly, isotopic exchange between LMC and pore fluid during burial was likely subordinated, and possible minor changes in the original stable isotope signature remained within the analytical uncertainty. Furthermore, the measured stable isotope oscillations were similar to those observed in other Late Cretaceous bivalves (including inoceramids) [43,53,93] living in deeper settings. In the light of the above, the inoceramid δ 13 C and δ 18 O values were considered to reflect the near-original signature and, therefore, suitable for paleoenvironmental reconstructions.

Stable isotope seasonality in the studied environments
The shallow water environment. The striking feature of the stable isotope record of the studied Hippurites resectus shell is the robust correlation between the δ 13 C and δ 18 O values (Fig 3A). It is highly unlikely that this stable isotope pattern resulted from kinetic fractionation effects, because rudists precipitated their shells in isotopic equilibrium with the ambient water [32]. Correlation between δ 13 C and δ 18 O has been observed in many rudist species [33,65] and is a typical feature observed in modern photosymbiont-bearing bivalves, e.g., tridacnids [45,46]. In photosymbiotic bivalves, such a trend results from the enhanced uptake of 12 C from the ambient water by the symbiotic algae during periods of high insolation and warm water conditions (i.e., more negative shell δ 18 O values) [45]. Although controversially debated in the past [28], recent studies provided morphological [44] and sclerochronological [32] evidence supporting a photosymbiotic lifestyle for at least some species of elevator rudists. A further hint for a photosymbiotic lifestyle for H. rectus is given by the growth increment counting. The number of microgrowth increments between the annual growth lines (~48) is almost two-fold the number of fortnight cycles (~25) expected to have occurred during a Late Cretaceous year [94]. This suggest that the studied H. resectus bivalve grew with a weekly periodicity and formed its shell almost year-round. These two features are also usually observed in modern algae-bearing tridacnids [45] and in other supposedly photoautotrophic rudists [33].
The cuspate shape of the δ 18 O chronology and the occurrence of the growth lines at the seasonal δ 18 O minima indicates that most of the shell production occurred during the cold part of the year [95]. Shell accretion rates strongly decreased with the onset of the seasonal warming and possibly shortly ceased during the summer extremes. Thus, the measured summer δ 18 O values are strongly affected by time-averaging and do not allow to properly resolve the information on the seasonal temperature maxima that prevailed in the studied environment. Late Turonian-early Coniacian summer temperature information of the Gosau environment are provided by published δ 18 O data of sequentially sampled Vaccinites inaequicostatus shells [22,65]. In the frame of the present study, it not possible to precisely determine the timing of shell formation of these rudists, because the position of the growth lines was not reported by the authors, and congeneric bivalves show a large inter-specific variability in the timing of annual growth line formation [30,32,95,96]. However, as shown in Fig 4A, the δ 18 O values of V. inaequicostatus follow an opposite trend as observed in the H. resectus shell indicating a predominant summer growth season. The reduced overlapping of the δ 18 O values with those of H. resectus can be interpreted as a prolonged shell growth stop during the colder part of the year. However, because of the lower sampling resolution as employed here, the reduced overlapping of the two datasets can also be explained by the larger time averaging of the δ 18 O values near the annual growth lines. Nevertheless, the data clearly indicate the existence of contrasting shell growth strategies between the two rudists, a phenomenon also known from modern bivalve cohorts [97]. Thus, by combining the δ 18 O data from the two taxa it is possible to estimate a seasonal isotopic amplitude of 2.25 ‰ and seasonal extremes of -5.10 ‰ and -2.85 ‰, for the studied shallow water setting.
The hemipelagic environment. The validity of inoceramid stable isotope data for paleoclimate reconstructions has been questioned in the past. This is because some taxa likely harbored symbionts, which affected their stable isotope data [43,52,54]. Because of the low or even absent light intensity excepted at the inferred water depth of the Trento Plateau, a photosynthetic lifestyle for studied Inoceramus sp. is unlikely. However, even if present, a photoautotrophic metabolism does not represent an obstacle to the reconstruction of paleowater temperatures, because algal activity does not affect the δ 18 O values of the bivalve hosts [47,48]. The pervasive presence of authigenic hematite [15][16][17] in the studied deposits clearly indicate the existence highly oxygenated waters [18][19][20]. Therefore, the presence of high levels of H 2 S and dysoxic conditions near the sea floor can be confidently excluded. Such conditions, however, were indispensable to sustain a thioautotrophic metabolism. Thus, studied Inoceramus sp. shells were likely not subject to chemosymbiotic disequilibrium fractionation effects.
Furthermore, isotopic equilibrium fractionation is also suggested by the good agreement of δ 18 O values with those reported in the literature for other carbonate secreting organisms. The δ 18 O data of the studied inoceramid shells (-1.44 --0.42 ‰) fall well into the δ 18 O range of late Turonian/early Coniacian North Atlantic benthic foraminifera (-1.5 --0.6 ‰), for which equilibrium fractionation can confidently be assumed [4]. At the same time, our δ 18 O values data were slightly more positive than those measured in coeval mollusk LMC from the shelf deposits of Central Europe. Late Turonian inoceramid and belemnite skeletons from the deposits of the Hitch Wood Event (i.e., the coldest interval of the late Turonian) in the Bohemian Basin (western Czech Republic) exhibit δ 18 O values between -2.46 and -1.06 ‰ [55] as well as -0.81 and -0.62 ‰ [49], respectively (Fig 5). The belemnite data were derived from the rostrum of a Praeactinocamax bohemicus specimen, a belemnite species that exclusively lived within inner shelf waters of the Bohemian Basin [49]. The stable isotope signature of the Inoceramus sp. also compares well with the average δ 18 O values of brachiopods (-1.33 ‰) from the Hitch Wood Event deposits of the Anglo-Paris Basin and the Lower Saxony Basin [12]. A similarity between the δ 18 O values of these two organisms, however, should be taken with caution, because brachiopod δ 18 O values can be affected by non-equilibrium isotopic fractionation [12,98,99]. Our Tethyan inoceramids also display more positive δ 18 O values than those of subtropical oysters from younger deposits of the Western Atlantic for which isotopic equilibrium with the ambient water are very likely [100].
The measured seasonal δ 18 O fluctuations (< 0.5 ‰) are also compatible with the reduced seasonality expected for marine waters at the postulated depths (lower photic zone or deeper) of the studied environment [67,71,] and compare very well with the stable isotope fluctuations recorded by mollusks from other Late Cretaceous hemipelagic benthic environments. Seasonal δ 18 O oscillations of 0.5 ‰ were recovered from Turonian-Coniacian inoceramids from the southern North Atlantic, for which equilibrium isotopic fractionation was inferred [53]. Our δ 18 O range also compares very well with the seasonal δ 18 O fluctuations reconstructed from deep water arcoid bivalves from the Coniacian [101] and Campanian [43,93]. Proxy-development studies on modern arcoids clearly indicate equilibrium isotopic fractionation for this taxon [97,102]. Hence, in the light of the above, it can be assumed that the studied Inoceramus sp. bivalves formed their shells near equilibrium with the oxygen isotope signature of the ambient water, as postulated for other inoceramid species [52][53][54][55].
The analysis of the inoceramid growth patterns revealed that faint microgrowth lines were not uniformly distributed across the shell transect, but rather occurred as sporadic bundles. This suggests the absence of major seasonal variations in shell growth rate. Although not common in modern bivalves [103], year-round shell growth has also been suggested for other inoceramid species from Central Europe [55], the Western Interior Seaway of the USA [54] and the North Atlantic Ocean [53]. Furthermore, the observed growth lines were located shortly after the seasonal δ 18 O minima indicating that shell accretion rates of the studied Inoceramus sp. slowed down during decreasing water temperatures. Hence, the individuals captured in their shells information on both seasonal environmental extremes [95]. In other words, unlike H. resectus, the time averaging did not affect the record of the seasonal δ 18 O minima and maxima, which allows the estimation of the full annual temperature amplitude for the benthic environment of the Trento Plateau.
The fluctuation of the δ 13 C values (1.34 ‰) indicates the presence of sub-annual variations in the incorporation of carbon isotopes. Symbiotic activity is known to have a major impact on the δ 13 C of bivalve shells [45]. However, as discussed above, a symbiotic lifestyle for the studied inoceramids can be excluded. In asymbiotic bivalves, changes in δ 13 C values of bivalve shells are regarded as reflecting variations in trophic conditions and have been employed to reconstruct past primary production [104][105][106][107][108]. Particularly for inoceramids, δ 13 C variations were suggested to indicate changes in food availability in Cretaceous benthic environments [53,54]. This is because the particulate organic matter in seawater has a very negative δ 13 C value (< -20‰) [109,110] and contributes up to 10% of the carbon incorporated into the bivalve shells [45]. Additionally, due to organic matter remineralization, increasing concentrations of organic matter in the water column can increase 12 C in the dissolved organic carbon (DIC) pool. The DIC, in turn, is the major source (up to 90%) of carbon for bivalve shell production [45]. Therefore, the observed δ 13 C fluctuations of the studied inoceramid shell most likely indicate seasonal fluctuations in the availability of food at the seafloor [43,45,54]. For the Temperature estimates for the shallow and upper ocean waters derived from stable oxygen isotope data of (a-b) rudist shells [22,65,84] and (c) shark teeth [21]. Temperature information for deeper environments were derived from (d) inoceramid [55] and (e) brachiopod shells [12] as well as from a (f) belemnite rostrum [85]. The Hitch Wood Event represents the coldest interval of the late Turonian. Temperatures were computed assuming an average ocean δ 18 O water value of -1.00 ‰ [81]. https://doi.org/10.1371/journal.pone.0238040.g005

PLOS ONE
benthic environment of the Demerara Rise it was postulated that food supply was provided by primary producers snowing from surface waters [53].

Water temperature estimates
Estimates of water temperature from biogenic δ 18 O data strongly depends on the assumption of a plausible δ 18 O w value for the studied environment [79]. For Late Cretaceous marine waters, this assumption can be particularly challenging. Most of the existing δ 18 O-based paleothermometry studies employed an average δ 18 O w value of -1.0 ‰ [22,[30][31][32][33]43,53,93] assuming a Late Cretaceous ice-free world [81]. Accordingly, our shallow water δ 18 O data suggest a seasonal temperature amplitude of 11.0˚C and seasonal extremes of 24 and 35˚C ( Fig  5). Such reconstructions, however, might underestimate the actual temperatures that prevailed in the studied surface environment. In fact, it should be kept in mind that the δ 18 O w value of -1.0 ‰ is inferred on a global average and does not account for possible spatial heterogeneity in water isotope signatures of the Cretaceous oceans. Late Cretaceous climate simulations suggest an increase of δ 18 O w values towards the equator and a δ 18 O w signature ranging between +0.3 and +0.5 ‰ for the tropical Tethyan surface waters [82].  (1), the estimated paleotemperatures increase by up to 6˚C. The resulting annual mean SST of 37˚C agrees with existing SST estimates from δ 18 O w -independent proxies. TEX 86 data from tropical-equatorial paleolatitudes suggest annual average SSTs of ca. 35˚C during the late Turonian-Early Coniacian interval [10,111].
Seasonal SST oscillations of 11˚C (assuming an invariant δ 18 O w signature) have been considered to be coherent with the mid-Cretaceous greenhouse interval [22], but are larger than the maximum annual SST range observed at modern tropical latitudes [112] (Fig 6A). However, it can be presumed that late Cretaceous surface waters experienced seasonal changes in the δ 18 O w signature, which should be considered in δ 18 O-based temperature estimates. Assuming a tropical monsoon climate similar as today, the cumulative effect of enhanced precipitation (lowering of the δ 18 O w signal) and increasing SSTs, and vice versa enhanced evaporation (rising δ 18 O w signal) and decreasing SSTs, would result in a dampening of the estimated annual temperature ranges (Fig 6A). For example, assuming a seasonal δ 18 O w fluctuation of 1.5 ‰, as observed in some modern tropical seas [83], our carbonate δ 18 O data would yield into an annual SST amplitude of 3.5˚C (Fig 6A).
Assuming a δ 18 O w signature of -1.0 ‰, the inoceramid δ 18 O data translate into temperatures ranging between 13-14 and 16˚C for benthic waters on the Trento Plateau, which fall in the lower range of temperatures estimated for the Central European shelves during the Hitch Wood Event (14-21˚C) [12,55,85]. Climate simulations, however, indicate that δ 18 O w could have increased by up to 0.4 ‰ in the first hundreds of meters below sea surface [82]. Assuming a similar trend for the Tethys ocean, δ 18 O w values in the upper water body of the study region might have been as high as 0.9 ‰, shifting our δ 18 O-based estimates to 22-26˚C. Bivalve Δ 47 data from younger Late Cretaceous deposits of North America suggest similar temperatures (22-23˚C) for open marine waters of the Gulf of Mexico. Therefore, it is possible that Tethyan deep waters δ 18 O w were more positive than previously assumed. On a seasonal scale, temperature oscillations of < 2˚C are in line with data from other Late Cretaceous hemipelagic environments [43,53,93,101]. Furthermore, our ΔT estimates fall within the annual temperature amplitude observed in modern tropical oceans between 150 and 400 m water depth (Fig 6B), where sub-annual variations in δ 18 O w are very constrained [83]. Therefore, it can be assumed that short-term variations of the water oxygen isotope signature had only a minor effect on δ 18 O-based temperature estimates.
In the context of the present study, we focused on the assessment of vertical temperature changes in the water column, and less on absolute reconstructions. However, the uncertainties in the temperature reconstructions highlighted above outline the importance to implement climate research with other proxy data.

Vertical temperature gradient and paleoclimate implications
The reconstructed seasonal δ 18 O amplitudes suggest that rudists and inoceramids were exposed to waters with very different physical properties, e.g., different δ 18 O w signatures, temperatures, or a combination of both factors. As mentioned above, vertical changes in the δ 18 O signal of the water could have played a role during the incorporation of oxygen isotopes. However, this effect was likely not dominant. In fact, to fully justify the 2.00 ‰ difference between the two shell δ 18 O datasets (Fig 5), it is necessary to assume implausible δ 18 O w water signals for the two environments. According to Eq 1, the most positive and the most negative δ 18 O values of the inoceramids and the rudists, respectively, would yield the same paleowater temperature if the surface δ 18 O water equaled -3.00 ‰ (or, otherwise, a benthic δ 18 O w = 1.00 ‰).

Fig 6. Comparison between inferred late Turonian/Early Coniacian water temperatures and observational data from modern tropical oceans.
(A) Annual water temperature amplitudes (ΔT) of the studied shallow water environment of the Tethys (dashed lines) reconstructed from the sclerochronologically sampled δ 18 O values of the rudist shells (Fig 4A) are compared to the latitudinal changes of the seasonal near-surface (2.5-7.0 m) temperatures observed in modern oceans (gray shadow) [112]. When seasonal changes in δ 18 O w (Δ 18 O w ) are taken into account, the reconstructed ΔT ranges become narrower (see text for detailed explanation). (B) Similar to (A), the estimated seasonal ΔT amplitude for the hemipelagic environment was compared to observed ΔT ranges in modern oceans at 150 m (light gray) and 400 m (dark gray) water depth [112]. (C) Results of our model on the effect of bathymetrical δ 18 O w changes on the reconstruction of the vertical temperature gradient (see method for detailed explanation). Vertical ΔT gradients were calculated from the rudist and inoceramid data assuming the vertical δ 18 O w changes modeled for the Late Cretaceous oceans [82] as well as observed in the modern Indian (yellow), Pacific (pink), and Atlantic Oceans (blue) and the Gulf of Mexico (green) [83]. The gray dotted line depicts zonally averaged vertical ΔT gradients. Bathymetries (black bars) of the Trento Plateau inferred in the present and previous studies [67,71,72] are depicted to the right. https://doi.org/10.1371/journal.pone.0238040.g006 According to the Late Cretaceous freshwater mixing line developed the western Interior Seaway waters [100], such a scenario implies an extraordinary input of freshwater/precipitation to the shallow environment. Accordingly, surface salinity should have been as low as 26 PSU, whereas, normal marine salinity conditions (34 PSU) would have existed at the bottom of the sea. Surface hypohyaline conditions, however, contrast with the highly diverse fully marine Gosau fauna at the studied locality [63,64]. Furthermore, such strong vertical gradient in salinity may have resulted in a density-driven stratification of the water column and reduced levels of oxygen in the benthic environment, as in the case of Western Interior Seaway in North America [100]. Stagnation and benthic hypoxia, however, are not consistent with the diverse macrofossil community of the 'lastame' lithofacies [66][67][68][69][70][71][72] and the presence of authigenic hematite, which indicate well-oxygenated conditions at the seafloor [18]. Therefore, it is reasonable to assume that the observed δ 18 O difference between the studied rudists and inoceramids resulted from a combined effect of changing water temperature and δ 18 O w signature.
Assuming modern-like vertical δ 18 O w changes for the upper 900 m of the studied region, the proposed model yielded temperature gradients ranging between ca. 14 and 19˚C. Among the modeled scenarios, the most conservative ΔT estimates (13-14˚C; Late Cretaceous and Indian Ocean settings) were very close to the ΔT value (14-15˚C) computed with a homogeneous δ 18 O w value. Therefore, we interpret the vertical effect of changing δ 18 O w on the measured δ 18 O shell amplitude as minor, and ascribe the isotopic differences between the rudist and inoceramid data to a heterogeneous thermal structure of the water body ( Fig 5). The precise estimation of the vertical ΔT gradient, however, also depends on the assumed water depth for the Trento Plateau. The oxygen-rich water conditions postulated for the 'lastame' lithofacies [67] suggest that the studied environment was located above the oxygen minimum zone [72], which in tropical oceans can be generally positioned between 150 and 500 m water depth [113]. This range fits well with the shallower paleodepths assumed for the Trento Plateau in previous studies [67,72] and make a bathyal depth postulated by others unrealistic [71]. The presence of a diverse ratiolitid rudist fauna in the Scaglia Rossa deposits of Sant´Anna d Alfaedo, however, provide a strong line of evidence that the studied nodular 'lastame' lithofacies was placed in the upper part (~150 m or slightly below) of this bathymetric range [72]. At a water depth between 150 and 200 m, the model suggested ΔT gradients between 13-14 (Late Cretaceous and Indian Ocean), 14-15˚C (Atlantic Ocean and Gulf of Mexico) and 16-17˚C (Pacific Ocean), which compare well to the vertical ΔT values observed in the modern tropical latitudes 10˚N (Fig 6C). Previous studies have postulated a poleward shift of ca. 5˚-10˚for the tropic/sub-tropic belts during the Late Cretaceous [114]. The annual ΔT amplitudes inferred for the shallow water setting (ca. 30˚N) compare well with those observed in modern tropical waters (20˚-25˚N) and seem to fit in this frame. However, even in the context of the Late Cretaceous greenhouse world, it seems very unlikely that the near-equatorial scenario could have extended up to studied latitudes. Therefore, to explain the δ 18 O difference between the rudists and inoceramid shells, we hypothesize the existence a heterogeneous structure of the water column at the northwestern Tethyan margin with a bottom source of cold hydrological masses. To enable an easy comparison with the literature data, in the following, temperature estimates will be reported assuming a seasonally constant δ 18 O w value of -1.0 ‰.
The temperature estimates obtained from rudist shells (24-35˚C) are in the range of those reconstructed from fish tooth apatite of the region [21] (Fig 5). This suggests that surface conditions and atmospheric variability affected both the shallow water biotope and the upper layer of the Tethys Ocean (Fig 5). Such warm surface Tethyan-like conditions also likely pervaded the shallow water and shelf environments at higher latitudes at least during part of the Turonian, as suggested by the δ 18 O data of brachiopods [12], rudist [84] and inoceramid shells [55] from Central European Basins. Short intervals of warm water conditions in the European mid-latitudes are also supported by the northward extension of the biogeographic province of Tethyan ammonites [13]. On the contrary, the benthic water temperatures of the Trento Plateau are close to the conditions reconstructed for the shelf environments of Central Europe during the Hitch Wood Event (Fig 5). The Hitch Wood Event repents a short time interval of the Late Turonian during which boreal water masses spread into Central European basins and caused a continental-wide cooling of shelf sea waters [13,115]. Furthermore, an influence of water masses with boreal affinity on the Trento Plateau during the late Turonian-early Coniacian is compatible with the more or less coeval begin of deep water production in the North Atlantic [9,10], the inflow of oxygen-rich waters in the western sub-basins of the Tethys [115,116] and the increased uptake of unradiogenic neodymium isotopes in apatite of fish teeth of the Tethys [14].
The described thermal decoupling between bottom and shallow water habitats contrasts with the existence of an active upwelling system at the northwestern margin of the Tethys postulated by previous authors [14]. Although the published neodymium isotope data suggest some degree of vertical exchange in water column, our seasonally resolved temperature estimates clearly indicate that bottom waters did not reach the shallow water environment of the Helvetic Shelf. Furthermore, existing studies on shallow water filter-feeding bivalves show that exposure to upwelling events usually results in a linear relationship between δ 13 C and δ 18 O values of the shells [117][118][119]. Such a relationship, however, was not detected in the stable isotope values of the V. inaequicostatus (Figs 3A and 4A), for which a heterotrophic lifestyle can be assumed. The lack of a correlation between shell δ 13 C and δ 18 O values is typical for filter-feeding bivalves from shallow water habitats, where nutrient supply, primary production and water temperatures are controlled by surface seasonality [102]. The shell growth rates of the photosymbiotic H. resectus provide a further line of evidence for the shallow water environment being mainly controlled by surface dynamics. The metabolic response to heat stress among bivalves is heterogeneous. However, taxa able to grow their shells over water temperature changes of >12˚C are not uncommon (97,102). Following tank experiments short-term temperature rises by up to 6˚C can induce death of the symbiotic zooxanthellae and bleaching [120,121], without presenting a major metabolic stress for the bivalve host [120]. Furthermore, as discussed above, the annual ΔT amplitude of the Tethyan shallow water might have been lower than 11˚C, considering the effect of the seasonally changing δ 18 O w on the δ 18 O shell records of the rudists. Hence, it is likely that fluctuations in water temperatures did not represent a major obstacle for H. resectus to biomineralize. On the contrary, nutrient availability and irradiance are the two major limiting factors for shell growth in algae-bearing bivalves [45]. Since accretion rates in H. resectus decreased during summer (i.e., maximum seasonal irradiance), it is plausible that growth was mainly controlled by nutrient limitation. In this context, increased summer insolation might have boosted environmental photosynthetic activity leading to nutrient depletion of the upper photic zone in a comparable manner as observed in modern sub-tropical zones [122].
The positive correlation between δ 13 C and δ 18 O values of the inoceramids suggests that the process governing temperature fluctuations also affected food availability and/or the stable isotopic signature of the DIC on the Trento Plateau. The concomitant increase in food availability and temperature on the seafloor resulting from snowing of organic matter and advection of water masses from the surface has been invoked to explain the observed stable isotope pattern in previous Late Cretaceous studies [43,53]. However, as observed on modern basinal highs [123], seasonal changes in horizontal transport of particulate organic matter might have also played a role in the trophic balance of the environment. Additionally, bottom currents from the North Atlantic possibly provided a steady supply of cold and nutrient-rich waters to the lower photic zone, where weakened insolation determined coincidental fluctuations in seasonal primary production and water temperatures. If this interpretation holds true, the present sclerochronogical data possibly suggest, to some extent, a decoupling also in the food webs between the shallow and bottom environments at the northwestern margin of the Tethys.

Conclusions
The present study highlights the potential of bivalve sclerochronology for Late Cretaceous paleoclimate reconstructions. Whereas most of the existing sclerochronological studies have focused on shallow water bivalves, information on the seasonality in deeper settings during this time interval is still very limited. As shown here, seasonally resolved environmental data from different taxa, which inhabited different water depths, can provide quantitative information on the internal structure of an ancient water column. Such data are often only available from numerical climate simulations and are of fundamental importance for paleoceanographic studies.
Our findings indicate the existence of a steep vertical temperature gradient of ca. 15˚C from the bottom to the surface of the northwestern Tethys Ocean during the late Turonianearly Coniacian. Such gradient is steeper than observed at modern tropical latitudes (20˚N-25˚N) under normal marine conditions and suggest the existence of a structurally heterogenous water body at the northwestern Tethyan margin. Bottom cooling was likely caused by benthic currents with old and nutrient-rich waters of North Atlantic origin. At the same time, however, strong radiative forcing at the surface resulted in the warmest conditions recorded over the entire Late Cretaceous in the shallow water habitats of the region. The substantial differences in the vertical water temperatures and their seasonal amplitudes suggest the existence of a perennial thermocline. The precise position of the latter remains unknown, however, it was likely located below the migratory depth of the ichthyofauna used for δ 18 O-based temperature reconstructions [21]. Moreover, the stable isotopic patterns of the studied bivalve shells also suggest that the supply of North Atlantic waters to the lower photic zone may have resulted in a certain degree of independence between the trophic systems of the two environments. Hence, it is unlikely that an active upwelling system persisted in the region at that time, as postulated by previous authors.