First record of stable isotopes (δ13C, δ18O) and element ratios (Mg/Ca, Sr/Ca) of Middle to Late Jurassic belemnites from the Indian Himalayas and their potential for palaeoenvironmental reconstructions

Middle to Late Jurassic belemnites from the Spiti and Zanskar valleys in the Indian Himalayas were used for stable isotope (δ13C, δ18O) and element (Mg/Ca, Sr/Ca) analyses. Although the Himalayan orogeny deformed and altered a large portion of the collected fossils, cathodoluminescence and scanning electron microscopy in combination with analyses of iron and manganese contents allowed the identification of belemnites believed to still retain their original chemical composition. Results indicate a long-term temperature decrease from the Middle Callovian–Oxfordian to the Tithonian, which is proposed to have been caused by a concomitant drift of eastern Gondwana into higher palaeolatitudes. Reconstructed absolute temperatures depend on the used equation and assumed δ18O value of seawater, but most likely varied between 17.6 °C to 27.6 °C in the Kimmeridgian and Tithonian with average values between 22 °C to 24 °C. This way, temperatures were similar to slightly warmer than today at comparable latitudes. The reconstruction of absolute temperatures for the Middle Callovian–Oxfordian was hindered by a larger number of poorly preserved belemnites representing this time interval.


Introduction
Since many decades, stable oxygen isotope (δ 18 O) analyses of fossil shells have been used to reconstruct the temperature development throughout Earth's history, including the Jurassic period (e.g. Epstein et al. 1951;Urey et al. 1951;Bowen 1963;Stevens and Clayton 1971). The temporal resolution of these palaeoenvironmental reconstructions has increased considerably, but the available data are still dominated by results from European localities (e.g. Korte et al. 2015;Martinez and Dera 2015). Originally, these European δ 18 O records seemed to reflect a long-term warming throughout the Late Jurassic (compare Dera et al. 2011), but it was not clear whether this represented global or local changes due to a scarcity of data from other regions. In the meantime, clumped isotope analyses suggested that Late Jurassic temperature fluctuations reconstructed for European material might be overestimated (e.g. Wierzbowski et al. 2018). Consequently , it seems worthwhile to focus more energy on yet understudied regions outside Europe, such as parts of Gondwana, to identify truly global climate patterns of the past. Although data are still sparse, recent studies suggest that low-latitudinal regions of western Gondwana (e.g. Egypt, Algeria) showed comparatively stable temperatures throughout the Middle and Late Jurassic, while eastern Gondwana experienced a concurrent temperature decrease (Fürsich et al. 2005;Alberti et al. 2017Alberti et al. , 2019Alberti et al. , 2020aSadji et al. 2021). The latter has been found in fossil records from Madagascar and western India (Kachchh), but data are still sparse particularly in the Late Jurassic. The present study therefore uses fossil material from the Indian Himalayas with Callovian to Tithonian ages to examine its potential for palaeoenvironmental reconstructions.

Geological overview
In the Himalayas, marine sedimentary rocks with a Jurassic age (Tethyan strata ;Pathak 2007;Pandey et al. 2013) form a belt stretching from northern Pakistan through India and Nepal until southwestern China. Three regions within the Indian part of this belt are described in greater detail in the literature: Zanskar, Spiti, and Kumaon (e.g. Blanford 1864;Stoliczka 1865;Uhlig 1903;Holdhaus 1912;Brookfield and Westermann 1982;Krishna et al. 1982;Baud et al. 1984;Jain et al. 1984;Jadoul et al. 1985Jadoul et al. , 1990Gaetani et al. 1986;Oloriz and Tintori 1991;Pathak 1993Pathak , 2007Garzanti et al. 1995;Tiwari et al. 1996;Bhargava and Bassi 1998;Vijaya 2002;Suttner 2005, 2021;Pandey et al. 2013). While the Kumaon outcrops (also referred to in the literature as Malla Johar, Niti, and Laptal) are situated in a restricted border-region and were not accessible during the present project, sections within the Spiti and Zanskar valleys were measured and sampled (Fig. 1).
In general, the examined successions of the Spiti and Zanskar valleys start with more than 600 m of wellcemented, thick-bedded limestones which constitute the Para and Tagling formations of the Kioto Group with a Triassic to Middle Jurassic age (Figs. 2, 3a, b;e.g. Jadoul et al. 1985;Garzanti et al. 1995;Bhargava 2008). These rocks formed in a comparatively shallow setting exemplified by common cross-bedding. Many horizons contain shell fragments, but due to their strong cementation, it is often impossible to collect complete, identifiable fossils. Consequently, dating by index fossils is complicated and the exact position of the Triassic−Jurassic boundary is unknown. The strata show an increase in siliciclastic content towards the top, the latter being  Oloriz and Tintori 1991). Near its top there are a number of hardgrounds characterized by encrusting oysters and bivalve borings, which indicate phases of sediment starvation and sedimentary gaps (Fig. 3c).
The following unit is the Ferruginous Oolite Formation, which shows a variable lithology with several crossbedded sandstones, siltstones, and limestones with ferruginous ooids. Its rocks are softer and previous authors have described Bathonian and Callovian ammonites from this unit (e.g. Krishna et al. 1982;Jadoul et al. 1985;Cariou and Enay 1999). The Ferruginous Oolite Formation changes in thickness and lithology laterally. While it is relatively well exposed in the Zanskar region with a thickness of more than 20 m (also compare Jadoul et al. 1985), it is less conspicuous in the Spiti Valley (see Fig. 3b, h, i;compare Gupta and Kumar 1975).
The Ferruginous Oolite Formation is overlain by the Spiti Shale Formation of the Lagudarsi Group (Fig. 2), which is dominated by blackish-grey silty clay and ferruginous concretions and represents a deeper-water environment, generally below wave base. The soft rocks locally contain abundant fossils, including ammonites and belemnites (predominantly Belemnopseidae, such as the abundant Belemnopsis grantianus = B. gerardi; Cariou et al. 1996). Benthic groups are commonly restricted to individual horizons and include bivalves, few brachiopods, and rare gastropods. Changes in the benthic communities indicate occasional anoxic to dysoxic conditions at the sea floor (e.g. Baud et al. 1984). The Spiti Shale Formation is generally subdivided into three members, which can be separated by comparatively minor differences such as the abundance of concretions and changes in fossil content (Figs. 2,3d,f). Due to the formation's soft nature, good outcrops are rare and tectonic faulting and folding are prominent. This complicates the measurement of detailed continuous sections and different thicknesses have been reported for the unit in the Spiti Valley by previous authors, e.g., 300 m by Pathak (2007) and 92 m by Pandey et al. (2013). Towards its top, more and more sandstone beds produce a gradual lithological change to the overlying Giumal Formation (Figs. 2,3g). In contrast, outcrops in the Zanskar range exhibit only 20-60 m of Spiti Shale Formation with an apparent increase in thickness from west to east (Jadoul et al. 1985;Gaetani et al. 1986;Oloriz and Tintori 1991). Previous authors suggested a fault contact with the overlying Cretaceous sandstones of the Giumal Formation in Zanskar (compare Fig. 3b and Fig. 3h;Jadoul et al. 1985;Oloriz and Tintori 1991).
Geographically, the Spiti Valley with its administrative capital Kaza is situated in northeastern Himachal Pradesh (Fig. 1a). The eastern/northern side of the valley is formed by a steep cliff consisting of the massive limestones of the Para and Tagling formations (Fig. 3a). The lithological change to the much softer Spiti Shale Formation leads to a dramatic change in landscape, with the latter forming plateaus and gentle slopes and constituting seasonal pastures for livestock (Fig. 3b). In the present study, fossils for geochemical analyses were collected at fresh outcrops near the villages of Demul, Hikkim, Langza, Gete, Kibber, and Chichim (Fig. 1a). All of these sections are located at altitudes above 4000 m and can be reached easily via car from Kaza. Geographic coordinates for the individual sample locations are given together with analytical results (in Table 2 below).
The Zanskar Valley with its administrative capital Padum is situated in northwestern Ladakh (Fig. 1b). The  Krishna et al. 1982;Cariou et al. 1996;Pathak 1997Pathak , 2007Suttner 2005, 2021;Bhargava 2008;Lukeneder et al. 2013;Pandey et al. 2013 Jurassic strata are exposed along a belt north of the actual valley, forming part of the Zanskar range. In this region, the Ferruginous Oolite Formation is much more conspicuous compared to the Spiti Valley (Fig. 3h, i). Unfortunately, the Jurassic strata in the Zanskar Valley contain only few fossils. Sections were measured and belemnites for geochemical analyses were collected at the hill slope of a river valley 5.5 km northeast of Rangdum Monastery, above a cliff reachable through a narrow gorge 3.5 km north of Zangla, and at the hill side directly north of Zangla Village (Fig. 1b). Further Jurassic outcrops in the region were described in literature near Lingshed Monastery (Oloriz and Tintori 1991), but these sections are only reachable on foot via several days of trekking and could not be accessed during the present study. Geographic coordinates of the individual sample locations are given together with analytical results below (also see Table 2).

Biostratigraphy
Dating of the Spiti Shale Formation is possible with the help of ammonites as guide fossils (e.g. Krishna 1983;Enay 2009;Pandey et al. 2013;compare Fig. 2). Cariou et al. (1996) recorded species of Idiocycloceras, Kinkeliniceras, Hubertoceras, Obtusicostites, Grossouvria, and Macrocephalites of an early Middle Callovian age from the base of the Lower member in the Spiti Valley. The authors collected most of these specimens from a single oolitic limestone bed and suggested the repeated occurrence of condensed beds or sedimentary gaps in the succession, which are, however, not visible in the monotonous sediments. Interestingly, most other authors have restricted the age of the basal Spiti Shale Formation to the Oxfordian (e.g. Krishna et al. 1982;Pathak 1997Pathak , 2007, based on records of mayaitids as well as Prograyiceras and perisphinctids (compare also Cariou et al. 1996;Enay 2009;Pandey et al. 2013). The occurrence of Callovian ammonites in the basal Spiti Shale Formation could not be confirmed during the present field surveys. While measuring the studied sections three specimens of Epimayaites were collected in the middle part of the Lower member, which would indicate an Oxfordian age (compare Alberti et al. 2015). Consequently, samples from the Lower member are considered here to be Middle Callovian−Oxfordian in age.
The lower part of the Middle member of the Spiti Shale Formation is characterized by the presence of Pachysphinctes, Glochiceras, and Torquatisphinctes and has been assigned to the Kimmeridgian by Pandey et al. (2013;also see Pathak 1997). The upper part of the Middle member is already Early Tithonian in age based on the presence of ammonites such as Aulacosphinctoides, Virgatosphinctes, and Hybonoticeras (Pathak 1997;Pandey et al. 2013). These earlier age assignments were supported in the present field surveys based on additional ammonite records. Consequently, fossils for geochemical analyses collected from the lower part of the Middle member were assigned to the Kimmeridgian.
The Upper member of the Spiti Shale Formation has been assigned to the late Early−Late Tithonian and partly Early Cretaceous (compare Krishna et al. 1982;Oloriz and Tintori 1991;Pathak 1997Pathak , 2007Pandey et al. 2013). Ammonites are relatively common in these strata including Tithonian forms such as Himalayites, Blanfordiceras, Uhligites, and Corongoceras, and Early Cretaceous genera such as Neocosmoceras and Spiticeras (compare Krishna et al. 1982;Pathak 2007;Enay 2009;Pandey et al. 2013). Results of the present field surveys largely agree with these earlier assignments, which also allowed the dating of the collected belemnites to the Early and Late Tithonian.

Material and methods
The fossil material used in the present study was collected during two field surveys to the Spiti and Zanskar valleys in September 2016 and September 2018. In addition to the collection of samples for geochemical analyses, a considerable number of ammonites was retrieved and used for age assignments in combination with literature data (see above). The majority of the collected material constitutes Middle Callovian-Oxfordian to Tithonian belemnites (Belemnopseidae) from the Spiti Shale Formation, while oysters and fossils from the Bathonian to Callovian Tagling and Ferruginous Oolite formations were rare.
The collected fossils were prepared following standard methodology (compare Wierzbowski 2002Wierzbowski , 2004Alberti et al. 2012a;Danise et al. 2020). Sections of 5-10 mm thickness were cut in the laboratory and their surface was ground. Subsequently, all selected fossils were examined for signs of diagenetic alteration with a cold cathodoluminescence microscope at the GeoZentrum Nordbayern of the Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany. Predominantly non-luminescent shell areas were selected for sampling, while luminescent (and therefore potentially altered) specimens were excluded from analysis or used only for comparison. Selected specimens were additionally examined using a scanning electron microscope at the Institut für Geowissenschaften of the Christian-Albrechts-Universität zu Kiel, Germany, after etching the surface with 5% HCl for ca. 40 s. Carbonate samples were then extracted with a hand-held dental drill in the case of thick shells or with a computer-controlled micromill at the GeoZentrum Nordbayern. In total, 83 belemnite rostra, seven oyster shells, and three sediment samples were analysed for their stable isotope (δ 13 C, δ 18 O) and element (Mg/Ca, Sr/Ca) composition. In addition, iron and manganese contents were measured to check for signs of alteration of the chemical composition.
Subsequently, samples were dissolved in diluted nitric acid and analysed for their elemental composition by inductively coupled plasma optical emission spectrometry using a Spectro Ciros SOP instrument at the Institut für Geowissenschaften, Christian-Albrechts-Universität zu Kiel, Germany. Average measurement uncertainty for Mg/Ca was around 1.3‰ and for Sr/Ca around 1.2‰ (1SD). Reference materials Coral JCp-1, Tridacna JCt-1 and limestones ECRM-752, Cal-S were used as secondary standards, and measurement results and reference values are compiled in Table 1.
Statistical correlations between datasets were examined using the Spearman correlation coefficient (rs) and linear trend lines were plotted based on reduced major axis (RMA) regression. Potential temporal trends were illustrated using LOESS smoothing with a smoothing factor of 0.5.

Results
The analytical results of all samples are listed in Table 2.

Fossil preservation
Attempts to reconstruct original environmental conditions by using geochemical analyses of fossils require well-preserved material. This can pose problems in many geological settings, and can be expected to be particularly difficult in the Himalayas, where the fossils experienced an intense orogeny with tectonic uplifts of more than 4000 m as well as strong folding and faulting at potentially high temperatures and pressures. Indeed, a large number of fossils in the present collection show macroscopic signs of fracturing, deformation, and recrystallization ( Fig. 4). Especially, the material of the Zanskar region was poorly preserved. This apparently poor preservation was also visible in cathodoluminescence microscopy and scanning electron microscopy ( Fig. 5). Specimens from the outcrops in the Zanskar Valley were therefore excluded from temperature reconstructions and only eight belemnite rostra were analysed for comparison. In contrast, the state of preservation was more variable in the Spiti Valley and in some cases improved considerably when moving a few meters away from fault zones. Nevertheless, luminescent belemnite rostra were encountered regularly and had to be eliminated from the study. Unfortunately, all oyster shells collected during the field surveys were found to be strongly recrystallized and only seven shells were analysed for comparison.
Since well-preserved belemnites generally show very low iron and manganese contents (e.g. Veizer 1980, 1981;Marshall 1992;Price and Sellwood 1997;Wierzbowski and Joachimski 2007), cut-off grades can be determined to exclude potentially altered specimens from further interpretations. Consequently, all belemnites with an iron content above 250 μg/g and a manganese content above 50 μg/g were excluded from further interpretation (for similar cut-off grades see Wierzbowski et al. 2009;Nunn Alberti et al. 2012bAlberti et al. , 2020bArabas et al. 2017). Analyses of sediment samples showed high iron and manganese contents pointing to the presence of both elements in the system. However, in the analysed belemnites, manganese contents were mostly negligible and below the detection limit (even for the poorly preserved samples from Zanskar), but elevated iron contents led to the exclusion of a number of samples from interpretation. Of the 83 belemnite rostra analysed, 61 specimens showed acceptable iron and manganese contents. Only data of these specimens were used for interpretations in the following sections (compare Table 2).
Another potential indicator for diagenetic alteration is a positive correlation between δ 13 C and δ 18 O values in the dataset (Fig. 6a) since burial generally leads to a decrease in both values (Hodgson 1966;Hudson 1977;Nelson and Smith 1996;Ullmann and Korte 2015). This can be seen particularly well in the extremely negative δ 18 O values of all bivalve shells and some of the discarded belemnite rostra. At the same time, δ 13 C values of the potentially altered specimens are not lower than those of the more well-preserved specimens.
Furthermore, some of the belemnite rostra with low iron and manganese contents still show surprisingly negative δ 18 O values, but not necessarily very low δ 13 C values. It seems therefore likely that even some of the accepted specimens that passed all tests for diagenetic alteration exhibit some (limited?) change in their chemical composition. Such a comparatively light influence of diagenetic alteration on some of the accepted samples might also explain the weak positive correlation between δ 13 C and δ 18 O values in the dataset (rs = 0.43; p < 0.05; Fig.  6a). In this regard, the presented data have to be considered with some caution.

Results of the geochemical analyses
While Table 2 lists all analytical results, Table 3 lists average stable isotope values and element ratios for the 61 accepted belemnite rostra representing Middle Callovian−Oxfordian, Kimmeridgian, and Tithonian ages. The δ 13 C values of the accepted samples fluctuate between − 1.43‰ and + 1.82‰ (Table 2; Fig. 7a). While the measured δ 13 C values vary around an average of + 0.05‰ in the Middle Callovian−Oxfordian, they  (Table 3).
The measured δ 18 O values vary between − 4.74‰ and + 0.65‰ (Table 2; Fig. 7b) and show a general increase from the Middle Callovian−Oxfordian through the Kimmeridgian to the Tithonian (Table 3). This longterm trend is overlain by small-scale fluctuations, which seem to indicate a rather weak increase in δ 18 O values from the Middle Callovian−Oxfordian until the mid-Kimmeridgian, followed by a stronger increase towards the Kimmeridgian-Tithonian boundary, and a slight decrease in values from the Early to the Late Tithonian. However, these short-term fluctuations might be exaggerated by the limited number of samples for some of the stratigraphic levels.
The Mg/Ca ratios fluctuate between 3.42 and 16.39 mmol/mol (Table 2; Fig. 7c). In general, the values are relatively high throughout the Middle Callovian −Oxfordian, drop significantly within the Kimmeridgian, and increase only slightly again in the Tithonian (Table  3). This way, the Mg/Ca ratios show an inverse trend compared to the δ 18 O values and both values are indeed negatively correlated (rs = − 0.73; p < 0.05; Fig. 6b).
The Sr/Ca ratios vary between 1.08 and 1.76 mmol/ mol (Table 2; Fig. 7d) and fluctuate only very slightly through time. While average values indicate a weak, but steady increase from the Middle Callovian−Oxfordian to the Tithonian, the magnitude of this change is almost negligible ( Table 3). The Sr/Ca ratios show no significant correlation with δ 18 O values (rs = 0.21; p = 0.11; Fig. 6c).

The δ 18 O value of seawater
The ratio between the stable oxygen isotopes 18 O and 16 O within a shell (represented by the δ 18 O shell value) is mainly a function of the original ratio within the ocean   Zachos et al. 1994;Roche et al. 2006;Alberti et al. 2020a). It is therefore necessary to consider the palaeogeographical position of study areas for temperature reconstructions. Since the Himalayan strata were strongly compressed laterally during orogeny, it is difficult to reconstruct the exact palaeolatitude for the sample localities, but it seems most likely that the study area was situated somewhere between 20°S to 30°S during the Late Jurassic ( Fig. 8; compare data from the so-called Paleolatitude Calculator of van Hinsbergen et al. 2015). These intermediate latitudes show a steep incline in the δ 18 O sea gradient with potential absolute values between 0.0 and − 0.4‰ following approximations of Zachos et al. (1994) and Alberti et al. (2020a). Because the Late Jurassic was characterized by the drift of eastern Gondwana into higher latitudes, it additionally seems likely that the palaeolatitude changed by up to 10°during the analysed time interval (Fig. 8).
Consequently, the δ 18 O sea value for the study area might have been higher in the Middle Callovian−Oxfordian compared to the Tithonian. Recently, clumped isotope analyses have been used in attempts to independently reconstruct water temperatures and δ 18 O sea values for selected regions and time intervals in the Mesozoic (e.g. Wierzbowski et al. 2018;Vickers et al. 2020;Fernandez et al. 2021). However, such data are not yet available for Indian study areas and the available approximations for other regions often show very wide ranges from below − 2‰ to above + 2‰ depending on the exact locality, palaeolatitude, time interval, and methodology (compare Wierzbowski et al. 2018;Vickers et al. 2020;Fernandez et al. 2021). While some of these values are comparable to previously proposed δ 18 O sea values, others would lead to the calculation of unrealistically warm or cold water temperatures. While some studies were successful to capture large-scale trends (e.g. Wierzbowski et al. 2018;, it seems that more research has to be done until clumped isotope analyses can be used systematically to improve temperature reconstructions at high temporal and regional resolutions (compare also Bajnai et al. 2018;Davies et al. 2021). Due to these reasons, temperatures in the present study were reconstructed for a traditionally and commonly used δ 18 O sea value of − 1‰ and additionally for a δ 18 O sea value of − 0.2‰ representing a subtropical latitude with comparatively high evaporation (following the method of Alberti et al. 2020a).

Different temperature equations
Since the first researchers started to use stable oxygen isotope analyses for palaeoenvironmental reconstructions, several equations were proposed to translate Fig. 8 a, b Palaeogeographic maps for parts of Gondwana for the Middle and Late Jurassic (modified after Enay and Cariou 1997;Alberti et al. 2015). Note that the position of western Gondwana did not change considerably, while eastern Gondwana drifted into higher latitudes; c Changes in palaeolatitude through time for selected study areas in western and eastern Gondwana from the Jurassic to the Early Cretaceous (data from the Paleolatitude Calculator; Besse and Courtillot 2002;Kent and Irving 2010;Torsvik et al. 2012;van Hinsbergen et al. 2015). The position of the Spiti Valley is approximated since the extent of lateral compression during the Himalayan orogeny is unknown δ 18 O shell values of calcitic fossils into water temperatures (e.g. Epstein et al. 1951Epstein et al. , 1953Craig 1965;O'Neil et al. 1969). Research on the exact relationship of δ 18 O shell values and temperatures is still on-going and more equations are used and developed until today (e.g. Anderson and Arthur 1983;Kim and O'Neil 1997;Leng and Marshall 2004;Takayanagi et al. 2013). Considering their application in temperature reconstructions in deep time (which by nature has to deal with a number of uncertainties), most of these equations result in comparable temperatures with only minor differences. Hence, the majority of studies on Jurassic temperature reconstructions uses the equation of Anderson and Arthur (1983), thereby also enabling an easier comparison of datasets: However, all of the aforementioned equations cannot explain the striking phenomenon that belemnite rostra generally show higher δ 18 O shell values than, for example, co-occurring calcitic bivalve and brachiopod shells (e.g. Prokoph et al. 2008;Mutterlose et al. 2010;Dera et al. 2011;Alberti et al. 2012aAlberti et al. , 2019. This almost systematic difference translates into reconstructed temperatures for belemnites, which are commonly around 5°C lower than those of other taxa. A number of processes have been suggested to explain this discrepancy including differences in life habits (e.g. a migratory behaviour of belemnites) or shell precipitation (e.g. Mutterlose et al. 2010;Hoffmann and Stevens 2019). Most recently, comparisons between the chemical composition of exceptionally preserved belemnite rostra and phragmocones as well as new clumped isotope data support the theory that previously suggested temperature equations are not applicable for δ 18 O shell values of belemnites and in fact lead to a consistent underestimation of temperatures (Price et al. 2015;Vickers et al. 2020). In contrast, these authors suggest that δ 18 O shell values of belemnites should be translated into water temperatures by using equations of low-precipitation rate experiments (e.g. Coplen 2007;Kele et al. 2015;Daëron et al. 2019). Since research in this area is still debated (compare Bajnai et al. 2018;Davies et al. 2021) and in order to show the magnitude of differences between both methods, the current study presents results based on Anderson and Arthur (1983) In addition to stable oxygen isotopes, a number of element ratios were proposed as temperature proxies. In particular, Mg/Ca ratios are considered to allow temperature reconstructions in a number of taxonomic groups (e.g. Eggins et al. 2003;Cléroux et al. 2008). Based on earlier research on benthic foraminifera by Lear et al. (2002), Nunn and Price (2010) proposed the following temperature equation for Mg/Ca ratios of belemnites: This equation has been used in the meantime by a number of researchers studying Mg/Ca ratios in belemnite rostra (e.g. Price et al. 2018;Alberti et al. 2020b), but the validity of this proxy is still debated. While some authors have found a negative correlation between δ 18 O shell values and Mg/Ca ratios in belemnites and proposed temperature as the common cause (e.g. Bailey et al. 2003;McArthur et al. 2007), others have failed to document a correlation and disputed the use of Mg/Ca ratios as a temperature proxy (Li et al. 2013). In the current dataset, there is a significant negative correlation between δ 18 O shell values and Mg/Ca ratios for the 61 accepted belemnite rostra (rs = − 0.73; p < 0.05; Fig. 6b), which seems to support a temperature influence. However, the calculation of absolute temperatures based on Mg/Ca ratios faces another problem. Experimental research on a number of molluscan taxa showed that water temperatures indeed determine the Mg/Ca ratios in their shell, but the exact relationship differed drastically between species (e.g. Surge and Lohmann 2008;Mouchi et al. 2013;Bougeois et al. 2016;Tynan et al. 2017). This implies that temperature equations developed for a certain species cannot be easily used for another. The applicability of the equation of Nunn and Price (2010) is therefore uncertain.
Since there is no significant correlation between δ 18 O shell values and Sr/Ca ratios (Fig. 6c), it is believed that this element ratio does not indicate temperature conditions in the analysed belemnites, but is instead influenced by the general ocean water chemistry (compare Steuber and Veizer 2002;Ullmann et al. 2013). Sr/Ca ratios of Jurassic seawater were previously reconstructed using a Sr distribution coefficient of 0.32 for belemnites, but this value is difficult to test for these extinct organisms (Ullmann et al. 2013(Ullmann et al. , 2016. In any case, it is noteworthy that the measured absolute values of the Himalayan belemnites largely agree with those of Ullmann et al. (2016) and also show a similar long-term trend with a slight increase from the Middle Callovian −Oxfordian towards the Tithonian in average values and then a slight decrease again in the Late Tithonian. Similarly, there is only a weak correlation between δ 13 C and δ 18 O shell values in the dataset believed to be caused by few specimens exhibiting a limited degree of alteration (Fig. 6a). In contrast to the largely temperature-dependent δ 18 O shell values, δ 13 C values can reflect a number of environmental factors and processes from regional changes in primary productivity or upwelling to global events such as the release of isotopically light methane into the atmosphere (e.g. Wierzbowski 2015;Ait-Itto et al. 2017). In the current study, the low number of specimens and low biostratigraphic resolution prevent further interpretations of the δ 13 C data or the identification of carbon isotope excursions, but it can be noted that absolute values are comparable to results from adjacent western India and Madagascar (Fürsich et al. 2005;Alberti et al. 2012aAlberti et al. , 2012bAlberti et al. , 2019.

Reconstructing regional water temperatures
The discussions in sections 6.1 and 6.2 show that the choice of the adequate δ 18 O sea value and temperature equation is crucial for calculating realistic watertemperature estimates. Figure 9 summarizes the relationship between temperature, equation, and choice of δ 18 O sea value, thereby illustrating that absolute temperatures can differ by more than 10°C between the different methods. For example, a δ 18 O shell value of − 1‰ corresponds to a water temperature of 16.0°C when using the equation of Anderson and Arthur (1983)  Similar differences in absolute temperature estimates can be found for the Jurassic dataset from the Indian Himalayas (Fig. 10a−d). Table 3  Temperatures then drop more prominently towards the Kimmeridgian-Tithonian boundary and remain relatively low in the Early Tithonian, followed by a slight temperature increase into the Late Tithonian. However, the reconstruction of temperature fluctuations at higher temporal resolution might be unreliable due to the scarcity of samples for some horizons and potential diagenetic alteration influencing some specimens.
By following recent research and considering the palaeolatitude of the study area, it seems advisable to use the equation of Daëron et al. (2019) and a δ 18 O sea value of − 0.2‰ for the reconstruction of absolute water temperatures (Fig. 10d). While this leads to largely realistic temperature estimates for the Kimmeridgian (average: 23.2°C) and the Tithonian (average: 22.8°C), the values for the Middle Callovian−Oxfordian are surprisingly high with an average of 33.0°C. Several Middle Callovian−Oxfordian belemnites lead to temperature estimates above 35°C with a maximum of 46.1°C (SP18-012). While this is a particular problem for this time interval, there is also a prominent outlier in the Late Tithonian with a rather unlikely temperature of 40.5°C (SP16-064). Such high values suggest diagenetic alteration of some of the specimens and obviously influence the calculation of average temperatures (compare Fig.  6a). Another approach could be to concentrate on the lowest reconstructed temperatures for each time interval: 23.7°C in the Middle Callovian−Oxfordian (SP16-006), 17.7°C in the Kimmeridgian (SP16-056), and 17.6°C in the Tithonian (SP16-069). Even though these minimum temperatures cannot be considered to represent average temperatures of the three time intervals, it is striking that they also reflect a general temperature decrease through time.
Finally, it is likely that the drift of eastern Gondwana into higher latitudes had an effect on the δ 18 O sea values. A change in the palaeolatitude from 20°S to 30°S throughout the studied time interval would result in a decrease of the ambient δ 18 O sea value by around 0.2‰  (Zachos et al. 1994;Alberti et al. 2020a). Such a change would influence temperature reconstructions and increase the magnitude of the reconstructed cooling by around 1°C.
Apart from temperature reconstructions based on δ 18 O shell values, the current dataset allows the application of the equation proposed by Nunn and Price (2010) on the measured Mg/Ca ratios (Fig. 10e). Since δ 18 O shell values and Mg/Ca ratios were found to show a strong negative correlation (Fig. 6b), it is not surprising that both proxies show similar trends. The temperature curve based on Mg/Ca ratios shows a very slight decrease in temperatures throughout the Middle Callovian−Oxfordian, which turns into a stronger cooling in the Kimmeridgian before again showing a slight temperature increase in the Tithonian. This development is also reflected in the average values (Table 3). Nevertheless, the reconstructed absolute temperature values are relatively low and comparable only to results from the equation of Anderson and Arthur (1983) at a δ 18 O sea value of −1‰. Consequently, this suggests that the equation of Nunn and Price (2010), which was developed based on data of benthic foraminifera (Lear et al. 2002), is not applicable for belemnite rostra. This potential problem was already noted by Nunn and Price (2010) and Price et al. (2018), but these authors suggested that at least the magnitude of temperature changes through time might be captured correctly. Following this reasoning, the present data suggest a temperature decrease from the Middle Callovian−Oxfordian to the Kimmeridgian/ Tithonian by around 6°C, which is almost identical to the temperature differences based on the maximum δ 18 O shell values mentioned above. The supposed drift of eastern Gondwana from 20°S to 30°S in the studied time interval would account for a cooling of around 5°C or more assuming latitudinal temperature gradients similar to today (compare Alberti et al. 2019Alberti et al. , 2020a. In summary, it seems reliable to interpret the Middle to Late Jurassic δ 18 O shell and Mg/Ca record as reflecting a significant long-term cooling in the Himalayan study area by several degrees Celsius. While this temperature decrease seems to be strongest in the Kimmeridgian, the identification of further short-term fluctuations might be less reliable due to the scarcity of samples for some stratigraphic intervals and diagenetic alteration of some specimens in the tectonically highly active region. While the latter process complicates the reconstruction of absolute temperatures for the Middle Callovian−Oxfordian, it seems likely that the Kimmeridgian and Tithonian experienced water temperatures mostly between 17.6°C and 27.6°C (= 87% of all samples) with averages between 22°C to 24°C (compare Table 3). This way, average temperatures were largely comparable to or slightly higher than today's temperatures for the same latitudes (Shea et al. 1992), which fits in the general understanding of the Jurassic world (compare Alberti et al. 2020a). The scatter of around 10°C in results from individual samples from the same stratigraphic intervals might either be caused by species-specific differences among the analysed belemnite taxa (e.g. different ecology) or shortterm temperature fluctuations. Alternatively, it might represent a minimum estimate for seasonal temperature changes with some samples recording summer and some winter conditions.

Supra-regional comparisons
Reconstructions of absolute water temperatures for northeastern Gondwana in the Late Jurassic are still scarce, but the available data are comparable to the results from the Himalayas. Fürsich et al. (2005) proposed a climate change in the Kachchh Basin in western India from the Bathonian to the Oxfordian connected with a slight temperature decrease, more humid conditions, the disappearance of tropical taxa, and a change in sedimentology. This temperature decrease was later confirmed by Alberti et al. (2012aAlberti et al. ( , 2012b to continue until at least the Late Oxfordian. More recently, Alberti et al. (2019) reconstructed a long-term cooling of about 5°C from the Callovian to the Oxfordian and Kimmeridgian for southern Madagascar based on stable oxygen isotope analyses of oyster shells. These authors suggested the concomitant rifting of Gondwana as the reason for the regional cooling, a view which is followed here. While eastern Gondwana drifted into higher latitudes, western Gondwana did not change its palaeogeographic position significantly (Fig. 8). Consequently, fossil material from northwestern Gondwana points to comparatively stable temperature conditions throughout the Middle and Late Jurassic as shown by Alberti et al. (2017Alberti et al. ( , 2020a. Sadji et al. (2021) confirmed comparatively stable temperature conditions based on stable oxygen isotope analyses of belemnite rostra from Algeria with Oxfordian, Kimmeridgian, and Tithonian ages. These major long-term trends of northwestern and northeastern Gondwana might be modulated by short-term fluctuations caused by global climate changes, but these could not be identified for the mentioned study areas due to a scarcity of samples for some stratigraphic intervals and a commonly low biostratigraphic resolution.
In contrast to the Gondwanan data, a considerable number of stable oxygen isotope analyses from European localities compiled by Dera et al. (2011) and Martinez and Dera (2015) point to significant warming from the Late Callovian almost throughout the entire Late Jurassic. However, this European temperature increase might be overestimated as suggested by clumped isotope data, which point to more stable Late Jurassic temperatures with changing δ 18 O sea values instead (e.g. Wierzbowski et al. 2018;Vickers et al. 2020).
Considering absolute values, it might be worthwhile to note that the Kimmeridgian δ 18 O shell data of the present study from the southern hemisphere are very similar to results from the northern hemisphere at a comparable palaeolatitude (Wierzbowski 2015;Arabas 2016;Sadji et al. 2021).

Conclusions
Fossils collected from the Middle to Upper Jurassic succession of the Spiti and Zanskar valleys in the Indian Himalayas were analysed for their stable isotope (δ 13 C, δ 18 O) and element (Mg/Ca, Sr/Ca) composition. Cathodoluminescence and scanning electron microscopy in combination with the determination of iron and manganese contents were used to differentiate seemingly well from poorly preserved fossils. Consequently, all collected oyster shells as well as fossils from the Zanskar Valley had to be excluded from the interpretation as they showed strong signs of chemical alteration. Finally, 61 belemnite rostra from the Middle Callovian−Oxfordian, Kimmeridgian, and Tithonian of the Spiti Valley were used for temperature reconstructions. The results indicate a long-term temperature decrease of several degrees Celsius (around 5°C or more), which can be connected to a synchronous drift of eastern Gondwana into higher latitudes. Absolute temperature estimates depend on the methodology, but based on the most recent research (Vickers et al. 2020), temperatures for belemnite δ 18 O values were considered most realistic when calculated with the equation of Daëron et al. (2019) and a seawater δ 18 O value of − 0.2‰. Following this approach, temperatures are proposed to vary mostly between 17.6°C and 27.6°C for the Kimmeridgian and Tithonian with average values being between 22°C to 24°C. Thereby, conditions would have been similar to slightly warmer than at comparable latitudes today. A relatively high number of poorly preserved specimens hinders the calculation of reliable absolute temperatures for the Middle Callovian −Oxfordian.