Rainfall seasonality on the Indian subcontinent during the Cretaceous greenhouse

The Cretaceous greenhouse climate was accompanied by major changes in Earth’s hydrological cycle, but seasonally resolved hydroclimatic reconstructions for this anomalously warm period are rare. We measured the δ18O and CO2 clumped isotope Δ47 of the seasonal growth bands in carbonate shells of the mollusc Villorita cyprinoides (Black Clam) growing in the Cochin estuary, in southern India. These tandem records accurately reconstruct seasonal changes in sea surface temperature (SST) and seawater δ18O, allowing us to document freshwater discharge into the estuary, and make inferences about rainfall amount. The same analytical approach was applied to well-preserved fossil remains of the Cretaceous (Early Maastrichtian) mollusc Phygraea (Phygraea) vesicularis from the nearby Kallankuruchchi Formation in the Cauvery Basin of southern India. The palaeoenvironmental record shows that, unlike present-day India, where summer rainfall predominates, most rainfall in Cretaceous India occurred in winter. During the Early Maastrichtian, the Indian plate was positioned at ~30°S latitude, where present-day rainfall and storm activity is also concentrated in winter. The good match of the Cretaceous climate and present-day climate at ~30°S suggests that the large-scale atmospheric circulation and seasonal hydroclimate patterns were similar to, although probably more intense than, those at present.

October to December. The δ 18 O of the water in the Cochin estuary close to the shell collection site ranged between −3.5‰ VSMOW during the entire period encompassing both the monsoons (June-December) and −0.5‰ VSMOW during the dry season (January-May) 19 . The corresponding carbonate growth bands had an average δ 18 O of −4.5‰ VPDB for the period of rain and −2.3‰ VPDB for the drier months. The record of daily rainfall at Cochin was obtained from the 'World Weather' TuTiempo Network (http://en.tutiempo.net) (Suppl. Information 2). It shows that rainfall, and hence freshwater input into the estuary, peaked in the period May through September, which  45 (grey shading denotes amplitude of signal). δ 18 O values for V. cyprinoides measured by continuous flow IRMS at 1 mm resolution (grey squares) are also plotted. Slower shell growth during cooler periods reduced the temporal resolution for the temperature estimates. (B) Observed daily rainfall (grey bars, http://www.tutiempo.net/) compared with the reconstructed percentage contribution of freshwater to the Cochin estuary (blue squares) and cumulative river discharge (purple diamonds, discussed in detail in Section S4 of the Suppl. Information 1). The freshwater contribution was calculated by subtracting the influence of temperature from the shell δ 18 O record, and applying a mixing model with end-member δ 18  corresponds to the period of thunderstorm activity during the pre-monsoon months (March-May) and increased rainfall during the monsoon (May-September) months. The analysis of CO 2 clumped isotopes (CO 2 isotopologues) in aragonite growth bands of modern V. cyprinoides and bio-calcite of Cretaceous Phygraea (Phygrea) vesicularis makes it possible to reconstruct seasonal ambient water temperatures with an uncertainty of ±2 °C 20,21 . Seasonal temperatures calculated from clumped isotope ratios (Δ 47 ) measured on the aragonite growth increments in the V. cyprinoides shell ranged between 20.4 and 32.5 °C (Fig. 2), with a mean value of 28 °C. These temperatures are close to the directly-measured temperatures of the estuarine water (Suppl. Fig. S5). As seen in Suppl. Fig. S5, a poor correlation (R 2 of 0.36, P = 0.04) is seen with the measured temperature. We suspect the factors like uncertainty in the determination of water temperature and the difference in the surface and bottom water temperature could have influenced the correlation. Monthly mean temperatures for the year 2009-2010 approached a minimum of 26 ± 1 °C during the winter months (December-February), whereas the regional maximum of 31 ± 1 °C was recorded during the summer months of March and April 22 (Fig. 2, Table 2).
The δ 18 O of precipitated carbonate (mollusc shell) depends on both the isotopic composition and temperature of the water. Once the temperature has been determined independently by CO 2 clumped isotopes (Suppl. Cretaceous climate reconstruction using Phygraea (Phygraea) vesicularis. The concentrations of atmospheric greenhouse gases were significantly higher in the Cretaceous than they are today, therefore the global temperature was higher than it is now. The average equatorial sea surface temperature is estimated to have been ~31 °C 26 , compared to ~27 °C at present. This higher temperature is likely to have induced changes in the hydrological cycle. During the Early Maastrichtian, the Indian plate was positioned further south at mid-latitudes (~30°S) relative to its present position in the northern hemisphere. The tropical to sub-tropical climate favoured the formation of coal and limestone, and the higher concentrations of greenhouse gases led to enhanced terrestrial productivity 27 .
A fossilized shell of the mollusc Phygraea (Phygraea) vesicularis, with well-preserved distinguishable carbonate growth bands (Suppl. Fig. S6), was recovered from Late Cretaceous (Early Maastrichtian) shell-bearing strata near the village of Ottakoil, Tamil Nadu, India. Siliciclastic sediments (Ottakoil Formation) immediately above the fossiliferous carbonate layer and a conformable off-lap of much younger fluvial sand deposits (Kallamedu Formation) represent a regressive marine sequence. An overlying conglomerate bed has structures indicative of deposition in a shallow marine environment.
Oxygen isotope analysis of the carbonate in the growth bands of P. vesicularis revealed a record of seasonal changes in seawater temperature and salinity. The δ 18 O and Δ 47 of the shell carbonate, and a Ce anomaly in the adjacent Kallankurichchi limestone are indicators of evaporative, anoxic environmental conditions at the time of deposition 28 .
The studied P. vesicularis shell had distinct prismatic and nacreous layers, no gross differences in Fe and Mn concentrations as measured by electron microprobe (Suppl. Information 3), and a lack of cathodoluminescence, indicating good preservation of its original composition 29,30 . Electron Backscattered Diffraction (EBSD) images of its growth structures, texture, crystal size and crystal orientation showed features similar to those characteristics of seasonal growth in modern day estuarine oysters (Suppl. information 1 Section S7, Figs S7, S8), consistent with there having been little or no diagenetic alteration 28 . The shell also had a large and systematic range of δ 18 O (>3‰, Fig. 3) as measured by both gas-source isotope ratio mass spectrometry (GIRMS) and Sensitive High Resolution Ion Micro Probe (SHRIMP). The analysed section of shell recorded two cycles of δ 18 O, reflecting repetition of the seasonal change in temperature and salinity (Fig. 3). The temporal variation in δ 18 O followed a sinusoidal pattern, with higher δ 18 O in summer and lower δ 18 O in winter. The GIRMS measurements of aliquots of calcite micro-drilled from the shell showed a seasonal change in δ 18 O of ~3‰ (−1.5 to −4.6‰) with a mean value of −2.7‰ (n = 21). The individual SHRIMP microanalyses (n = 125) had a larger range (~6‰). The differences in δ 18 O between adjacent 25 µm spots were particularly large for the periods of summer growth (up to 3‰), much larger than can be explained by analytical uncertainty (±~0.3‰) or temperature fluctuations (Table 3). This variability coincided with changes in the crystal structure of the shell (Suppl. Fig. S6).The coarsely crystalline growth bands, mostly denoting summer growth, were more heterogeneous than the finely crystalline bands formed during winter. The Δ 47 determined from the clumped isotope analyses ranged from 0.66 to 0.73‰, corresponding to a temperature range of 21 to 37 °C, with a mean value of 30 °C. This temperature range accounts for about 3.5‰ of the variation in the measured δ 18 O. All remaining variation (~2.5‰) is due to changes in the δ 18 O of the water in which the mollusc grew.

Discussion
The well-preserved P. vesicularis shell recovered from Late Cretaceous sediments of the Cauvery Basin is an ideal specimen with which to determine the past rainfall seasonality pattern on the Indian coast when the subcontinent was positioned at latitude ~30°S. The observed variations in Cretaceous seawater δ 18 O and temperature are similar to those found in a modern coral-based proxy record from the Dampier Archipelago 27 , Western Australia (20.5°S), which shows the strong seasonal effect of evaporation on salinity. Coupled analyses of Sr/Ca and δ 18 O in the coral growth bands indicate that sea surface salinity during the period 1988-1994, as calculated from the residual δ 18 O after the temperature component was subtracted, was highest when strong evaporation in summer caused salinity (and water δ 18 O) and to increase. In contrast, reduced evaporation and increased rainfall in winter cause a decrease in seawater δ 18 O and salinity 27 . The seasonal range in Early Maastrichtian water temperature in the Cauvery Basin was 21-37 °C, so the −1.6 to −3.1‰ range in δ 18 O measured in the P. vesicularis shell growth bands translates to a range in water δ 18 O from −0.5 to 1.7‰ VSMOW ( Table 3). The seasonal change in freshwater contribution to the Cretaceous estuary can be calculated assuming the average δ 18 O of Early Maastrichtian rainfall to be about −6‰ VSMOW (inferred from the δ 18 O of contemporaneous palaeosol carbonates 7 ) and seawater in the evaporating enclosed basin to have a δ 18 O of ~2‰ (as in the modern Red Sea-global seawater δ 18 O database; Craig, 1966). Based on this two end-member mixing model, the seasonal change in freshwater contribution to the Cretaceous estuary ranged from 3 to 31% between summer and winter (Table 3).
Further, water δ 18 O and temperature are strongly correlated, with the lowest temperatures associated with the lowest water δ 18 O values, suggesting a climatological control on the amount of runoff to the estuary. Temperature and the δ 18 O of precipitation are directly correlated with latitude 7 whereby the δ 18 O of Late Cretaceous precipitation was lower while India was located in the southern hemisphere mid-latitudes. This conclusion is consistent with a Late Cretaceous palaeoclimate reconstruction from further south at Seymor Island 31 , where maximum freshwater discharge occurred during winter, suggesting warmer winters than those currently experienced in coastal regions in southern mid-latitudes.
The sedimentary record of the Kallankuruchchi Formation provides a further opportunity to probe the Late Cretaceous temperature and hydrological seasonality. Previous studies have interpreted sedimentary structures in the Kallankuruchchi Formation (cross bedding, cut and fill, hummocky cross stratification) as indicating deposition during major storm events. Evidence of storm deposition is also found in other contemporaneous strata in the Cauvery basin 32 . The weight of evidence is that, unlike the modern day, Cretaceous southern India experienced the bulk of its rainfall in winter, whereas reduced rainfall and runoff in summer increased both the salinity and temperature of the estuarine water, resulting in deposition of thick piles of evaporite and carbonate sediments in the stratigraphic succession.
The climatic conditions experienced in southern India at ~30°S latitude during the Late Cretaceous were evidently similar to those of the modern-day coastal plain of Western Australia around the same latitude, where seasonal storms and cyclones impact on the carbonate platform 33 . Our finding suggests that large-scale atmospheric circulation and seasonal hydroclimate patterns at mid-latitudes during the Cretaceous global warming interval were not substantially different from the present-day. The result is relevant for climate models designed to simulate the extent to which elevated atmospheric CO 2 levels, and the accompanying global warming, might alter the Hadley circulation and mid-latitude storms in the future.

Methods
Background. The Cochin Backwaters at Kerala, India (9°40′ N-10°08′ N, 76°11′ E-76°25′ E) is populated by several species of mollusc, the most widespread being Villorita, known for its edible value. The area occupies the northern part of the Vembanad-Kol wetlands, and covers ~255 km 2 , extending from Alleppey to Cochin before merging with the Arabian Sea via two permanent openings. The region has a modern-day tropical climate with two main rainy seasons. Most of the rainfall (~70%) occurs during the SW monsoon from June to September. Much of the remaining rainfall (~15%) occurs during the NE monsoon from October to November, while the December-to-May pre-monsoon period has sporadic rainfall accounting for the remaining 15%. The surface water temperature in the estuary reaches a maximum of 32 °C in April and drops to as low as 24 °C in August.
Sample collection. Several specimens of Villorita cyprinoides were harvested live from the southern extremity of the Cochin Backwaters on 20 January 2010 for the present investigation. The growth rate of V. cyprinoides varies over a year and is characterised by growth increments of aragonite separated by laminar bands. The average growth rate for the species is ~8.3 mm/year 16 , determined by monitoring a population of several individuals in situ within a cage experiment. For the present study, the incremental growth bands were drilled at a spatial resolution of 1 mm for δ 18 O and clumped isotope analysis.
The Cauvery Basin, at the southern tip of peninsular India (Fig. 1), hosts a complete Cretaceous sedimentary sequence of shallow marine to estuarine deposits. The sequence consists of the Uttatur, Trichinopoly and Ariyalur rock groups, representing Early, Middle and Late Cretaceous successions, respectively. Specimens of Phygraea (Phygraea) vesicularis were collected from an exposure in the Kallankuruchchi Formation (Ariyalur group), well exposed near the PNR mine of Dalmia Cement Limited (11°7′11′′N 79°7′59′′E) 34 .

Recovery of carbonate from mollusc growth bands.
In preparation for δ 18 Oand clumped isotope analysis, internal soft body parts were discarded from V. cyprinoides and the outer carbonate shells were treated with H 2 O 2 for complete removal of organic debris, and then air dried for sectioning and drilling. Shells were dissected along the growth axis (Suppl. Figs S1a, S1b) using a section cutter and sampled along individual growth bands at 1 mm resolution using a battery operated micro mill. The recovered powder was analysed by XRD, showing aragonite to be the primary mineral phase. The shell for the Cretaceous P. vesicularis was processed in a similar manner, sectioned along the growth axis and sampled at 1 mm and 2 mm resolution for δ 18 O and clumped isotope analysis. Powder analysed by XRD suggested calcite as the primary mineral in the Cretaceous specimen. The same section was polished for in situ analysis of δ 18 O at a spatial resolution of 25 μm by Sensitive High Resolution Ion Microprobe (SHRIMP).
Conventional δ 18 O measurements. The δ 18 O (and δ 13 C) of the shell samples were measured at high-resolution using a Thermo Finnigan MAT 253 isotope ratio mass spectrometer (IRMS) coupled with a Gas bench II in continuous flow mode. About 100 μg of carbonate powder was reacted with 1 ml of H 3 PO 4 using the boat method described elsewhere 35 and the overall δ 18 O reproducibility for NBS-19 calcite was ±0.08‰. Water samples collected from the Cochin estuary were analysed for δ 18 O following the CO 2 -water equilibration method, where 100 μl of water was equilibrated with the CO 2 +He mixture for more than 18 hours (as described previously) 36 . The over-all δ 18 O reproducibility of replicate analyses of water standards was ±0.06‰.
Larger shell sample weights (~5-10 mg) required for the preparation of CO 2 gas for clumped isotope measurements were obtained by combining powders from two or more growth bands (averaging ~2-3 months of growth). δ 18 O (and δ 13 C) in the larger samples were measured on the MAT 253 IRMS in dual inlet mode along with measurements of mass-47 isotopologues of CO 2 for clumped isotope analysis. The overall δ 13 C, δ 18 O and Δ47 reproducibility for NBS-19 calcite value of ±0.04, 0.05 and 0.01‰ respectively.
Measurements of Δ 47 in shell growth bands. Δ 47 analyses were performed using the dual inlet peripherals on the Thermo MAT 253 IRMS following the preparation steps of CO 2 cleaning through use of an external GC setup 37 . The experimental procedure for sample preparation for clumped isotope analysis used the sealed vessel method of 38,39 . All carbonate samples were prepared in the experimental setup designed at the Indian Institute of Science, Bangalore. For each analysis, ~5-10 mg of carbonate powder was reacted with 1 ml of H 3 PO 4 in a sealed reaction vessel. The reaction vessel was evacuated on a gas-extraction line to a pressure of 10 −4 mbar using a combination of turbomolecular and roughing pumps. The stopcock in the evacuated vessel was then closed and the vessel kept in a water bath maintained at a constant temperature of 25 ± 0.1 °C. The reaction of carbonate with H 3 PO 4 commenced by a simple transfer of acid from the arm of the reaction vessel to the compartment with carbonate powder.
The CO 2 generated during the reaction was cleaned using a cryogenic extraction protocol to remove contaminants responsible for isobaric interferences 37 . Purification steps involved removal of water vapour and other contaminants by a combination of liquid nitrogen trap and a dry ice and ethanol slush trap. The CO 2 , once extracted onto a cold finger, was entrained with a He stream through a capillary column (PoraPLOT Q, 25 m × 0.32 mm i.d.; Varian Inc., Palo Alto, CA, USA) and held at −10 °C for gas chromatographic separation of CO 2 from other mixtures of trace hydrocarbon and halocarbon. Eventually, the purified CO 2 sample was taken into a glass cold finger and analysed using the MAT 253 IRMS dual inlet system.
The MAT 253 IRMS was configured to analyse mass-47 isotopologues of CO 2 by simultaneously measuring mass 47, 48 and 49 (measured with 10 12 Ω resistors). Masses 48 and 49 were monitored in order to ensure that there were no isobaric interferences due to the presence of contaminants. These measurements were done in dual inlet mode with a source pressure sufficient to maintain the CO 2 mass-44 ion beam intensity at a voltage of 10-12 V. Each analysis involved 60 measurement cycles of the sample CO 2 and reference CO 2 (six acquisition lines with 10 cycles each, with a signal integration time of 8 s per measurement). The reference CO 2 gas (Linde CO 2 ) had δ 13 C and δ 18 O values of −4.41‰ (VPDB) and 24.59‰ (VSMOW), based on repeat analyses of the NBS-19 carbonate standard.
The Δ 47 analyses were standardized using in-run measurements of NBS-19, heated gas and in-house MAR J1 calcite (generated from Carrara marble) 39 . Calibration of MARJ1, which was run more frequently during the course of our measurements, was done by adopting two published values for the other two reference materials (NBS-19 and heated CO 2 at 1000 °C). The Δ 47 value of 0.392‰ for NBS-19, and 0.026‰ for heated CO 2 , were adopted to relate the heated gas scale to the CDES scale 22 . The MAR J1 Carrara marble was assigned a value of 0.395‰ for scale conversion purposes. The long-term reproducibility of MAR J1 over the period 2010-2012 (n = 59) yielded a Δ 47 value of 0.343 ± 0.01 (1σ) on the heated gas scale. All the shell carbonate samples were analysed during that period.
In order to convert the Δ 47 values to the heated gas scale, a large number of CO 2 samples were treated at 1000 °C for 2 hours upon recovery of the analysed samples. The CO 2 samples used for high temperature treatment were obtained upon transferring the sample CO 2 to an ultrapure synthetic quartz tube (6 mm o.d.), which was evacuated and sealed. The sealed tube containing the CO 2 sample was heated in a muffle furnace at 1000 °C for >2 hours and then quickly quenched to room temperature. The difference between the Δ 47 values measured in the sample CO 2 and randomized CO 2 generated upon heating at 1000 °C allowed the definition of the Δ 47-HG value in the heated gas scale 20 . All data are reported at the 25 °C reaction temperature and therefore no additional correction for the reaction temperature fractionation factor was applied. For the period of analysis, heated CO 2 yielded Δ 47 values (n = 66) of −1.47 ± 0.06‰ (1σ). The values on the heated gas scale are converted to the absolute CDES scale by using the proposed equation 22 relating the absolute value of Carrara marble and heated gas and is given here as: The analytical procedure was based on that described by Ickert et al. 43 and Long et al. 44 . In brief, a ~3 nA primary ion beam of ~15 kV Cs + was focused to a probe ~25 µm in diameter, and secondary ions of O − ( 16 O ≈ 1.9 × 10 9 c/s) were extracted at ~10 kV for isotopic analysis by dual Faraday cup multiple collection (current mode, 10 11 Ω resistors). Charge build-up on the sample surface was neutralised using a focused ~600 eV electron beam. Each analysis consisted of a 90 s pre-burn during which electrometer baselines were measured, ~2 min of ion focusing and 12 × 10 s measurements of 18