Introduction

Pelagic calcifiers, such as foraminifera, pteropods, and coccolithophores, compose up to 70% of the modern global marine carbonate export to the seafloor1,2, and their calcification rates fundamentally link to the inorganic carbonate chemistry of seawater3,4. Large and rapid increases in atmospheric CO2 concentrations cause ocean acidification (OA), which decreases seawater pH, carbonate ion concentrations, and carbonate mineral saturation states while increasing the concentration of total dissolved inorganic carbon5,6. Studies of modern planktic foraminifera impacted by OA reveal decreased biocalcification rates and other signs of stress, such as shell thinning7,8. Reduced biocalcification buffers OA by increasing surface water alkalinity3, similar to other negative feedback mechanisms, such as chemical carbonate compensation and silicate weathering5,6, although biological carbonate compensation is geologically instantaneous by comparison. Increased surface ocean alkalinity due to biological carbonate compensation enhances atmospheric CO2 drawdown3, which can mitigate climate warming but further compound biocalcification stress6. Understanding the dynamic relationship between pelagic calcification and atmospheric CO2 concentrations thus has key implications for quantifying the marine inorganic carbon pump and other controls on the global carbon cycle, including anthropogenic forcing9. It also directly underlies efforts to advance ocean alkalinity enhancement, a promising strategy for mitigating fossil fuel CO2 emissions10,11 and stabilizing ocean water pH12,13.

Geologic OA events provide a means to examine the biocalcification response to carbonate chemistry feedbacks during times of high background CO2. The Cretaceous (145–66 Ma) was a hothouse period when atmospheric CO2 may have averaged over 500 ppm (some 230 ppm higher than pre-industrial levels)14, with some intervals of significantly higher concentrations related to massive volcanic eruptions5. Cretaceous black shale deposits reflect short-term global carbon cycle perturbations and widespread, albeit not global, oceanic anoxia within this high-CO2 greenhouse climate. The geologic record preserves eight oceanic anoxic events (OAEs), which are marked by organic-rich deposits, in several cases bearing large carbon isotope (δ13C) excursions (CIEs), as well as geochemical evidence for large igneous province (LIP) emplacement15,16. For example, recent geochemical studies of the Cenomanian-Turonian OAE2 (~94 Ma), which is defined by a 3–4‰ increase in organic carbon isotope values (δ13Corg)17,18, indicate that eruption of the Caribbean LIP and/or High Arctic LIP drove transient increases in atmospheric CO2 levels19.

Several studies have implicated LIP induced-OA as the cause for faunal turnover events that occur in association with many OAEs, as well as additional morphometric changes in foraminifera and coccolithophores6,16,20. Interestingly, plankton turnover events slightly precede OAEs21, along with size reductions of heavily calcified nannofossil taxa, such as nannoconids and schizosphaerellids22. Such is the case for OAE2, where nannofossil extinctions22 and placolith size reductions20 occur just prior to the event, which is now referenced in Tethyan sections as the lower critical interval (LCI)23,24. Within the LCI, foraminifera from multiple localities show evidence for reduced test size18, as well as other morphometric indicators of stress, such as reversals in coiling direction25 and changes in chamber number and shape18. The detailed records of morphometric changes across OAE2 offer an excellent opportunity to evaluate the hypothesis of biocalcification stress. Test density measured by area-normalized shell weights7 and bulk density measured by micro-CT scanning8 have emerged as optimal morphometric methods for examining OA-driven biocalcification stress in modern foraminifera, but the signals are highly sensitive to dissolution and thus cannot be applied in the geologic record where foraminifera are less pristinely preserved. Thus, to study OA-biocalcification stress in deep time, we pair existing morphometric observations with high-precision calcium isotope(δ44/40Ca) measurements, which offer a promising technique for constraining paleo-biocalcification.

Experiments focused on inorganic calcite precipitation show that slower precipitation rates yield smaller isotopic offsets between calcite and fluids \(({\Delta}{{{{\rm\scriptsize{CaCO}}}}_{3}-{{{\rm\scriptsize{H}}}_{2}\scriptsize{{O}}}}=\delta^{44/40}{{{\rm\scriptsize{{Ca}}}\scriptsize{CO_3}}-\delta^{44/40}{{{{{\rm{Ca}}}}}}_{{{{{{\rm{H}}}}}}_{2}{{{{{\rm{O}}}}}}}})\) and thus higher δ44/40Ca values compared to faster precipitation rates26. Modeling studies further illustrate that rate-dependent shifts in \(\Delta_{{{{{{\rm{CaCO}}}}}}_{3}-{{{{{\rm{H}}}}}}_{2}{{{{{\mathrm{O}}}}}}}\) correlate with carbonate system parameters, such as carbonate ion concentration and carbonate mineral saturation states27,28. Carbonate chemistry control of δ44/40Ca values for some foraminifera (Globigerinoides ruber) has been partially demonstrated in culture (Supplementary Fig. 1), consistent with assertions that other foraminifera (O. universa) fractionate Ca isotopes similarly to inorganic calcite29. Moreover, four geologic δ44/40Ca records representing multiple candidate OA events and different marine carbonate archives, including foraminifera, mollusks, and bulk sediments, all show positive excursions consistent with reduced precipitation rates30,31,32,33.

In this study, we utilize the Ca isotope proxy to comprehensively test for biocalcification stress preceding OAE2. We generated paired high-resolution and high-precision δ44/40Ca, δ13C, and δ18O records for bulk carbonate and planktic foraminifera (Rotalipora cushmani) from the Bottaccione section in Gubbio, Italy, along with δ44/40Ca and δ13C records for bulk carbonate from the Aristocrat Angus (AA) core, Denver Basin, Colorado, USA (Supplementary Fig. 4). All records reveal that δ44/40Ca values increased prior to the onset of the positive CIE that marks OAE2. Moreover, the pre-CIE positive δ44/40Ca shifts coincide with decreased test diameters of R. cushmani and correlate with decreases in the osmium isotope composition (Osi) of seawater widely attributed to massive LIP inputs34,35,36. Considered together, these observations reveal OAE2 as a geologic OA event. A return to lower δ44/40Ca values during the main phase of OAE2 provides evidence for surface ocean alkalinization in the aftermath of a biocalcification crisis, which in turn suggests that the Plenus Cold Event37 could reflect a resulting period of transient CO2 drawdown. Results from this study highlight the dynamic interplay between atmospheric CO2 perturbations and biocalcification feedbacks, and further have implications for implementing ocean alkalinity enhancement as an anthropogenic climate change intervention strategy.

Results and discussion

Within the LCI of the Bottaccione section at Gubbio, bulk carbonate (n = 31) δ44/40Ca values range from −1.56‰ to −1.45‰ and generally increase up-section through the LCI to the Last Occurrence (LO) of R. cushmani (Fig. 1); directly above this level, a sharp decrease in carbonate content prohibits further δ44/40Ca analysis. The δ44/40Ca values of coeval R. cushmani specimens (n = 18) start around −1.60‰, concomitantly increase through the LCI to −1.41‰, and then decrease to −1.48‰ just below the LO of the taxon at 110.85 msl (Fig. 1). The maximum R. cushmani δ44/40Ca value of −1.41‰ occurs at 110.72 msl. This pattern of change in δ44/40Ca relative to the LO of R. cushmani is identical to that observed in the Portland core, Western Interior Basin (WIB)30 (Supplementary Fig. 5).

Fig. 1: Stratigraphic calcium and carbon isotope data from the Bottaccione section.
figure 1

Relative to stratigraphy of the Bottaccione section, a calcium isotope values (δ44/40Ca) of planktic foraminifera Rotalipora cushmani (red triangles), Whiteinella archaeocretacea (pink cross), and bulk carbonate (gray circles) where the error bars correspond to the analytical external reproducibility of ±0.05 (2σSD), and b carbon isotope values of carbonate (δ13Ccarb), including planktic foraminifera Rotalipora cushmani (blue triangles), Whiteinella archaeocretacea (blue cross), and bulk carbonate (gray circles). Oceanic Anoxic Event 2 is denoted by the positive δ13Corg excursion73 just below the black shale interval, here expressed at the Bonarelli Level. The last occurrence of R. cushmani is denoted by the downward pointing arrow. The open circles next to the stratigraphic column represent sample locations. LCI lower critical interval, ASW Atlantic Seawater; VPDB Vienna Peedee Belemnite.

Full size image

Observed changes in δ44/40Ca values for R. cushmani significantly correlate with the specimens’ Foraminiferal Abnormality Index (FAI; p = 0.004, R2 = 0.50), which is defined as the percentage of malformed and/or deformed specimens18, as well as their Feret diameters (p = 0.006, R2 = 0.45), but do not correlate with changes in coiling direction (p = 0.14, R2 = 0.17; Fig. 2). At one stratigraphic interval (110.43 msl), measurements were made for R. cushmani and another planktic foraminifer, Whiteinella archaeocretacea. The δ44/40Ca values are identical (Fig. 1). Bulk carbonate (n = 32) δ44/40Ca values from the AA core, WIB, which has an expanded succession comprising the onset of OAE2 relative to the Portland core and the Bottaccione section, range from −1.58‰ to −1.40‰, with the maximum value well preceding the level of bentonite A, corresponding to the LO of R. cushmani (Supplementary Fig. 4). The positive δ44/40Ca excursion spans both calcareous mudstone and limestone beds. Near the end of OAE2 in the Western Interior, stable and relatively negative δ44/40Ca values similarly persist across multiple lithologies. All results presented here closely mirror a previously generated δ44/40Ca record from the WIB30.

Fig. 2: Regression analyses of Bottaccione section δ44/40Ca values and morphometric data.
figure 2

Planktic foraminifera Rotalipora cushmani δ44/40Ca values across the lower critical interval (LCI) regressed against R. cushmani morphometric data published in Coccioni et al.18 shown in black triangles, including a Foraminiferal Abnormality Index or FAI, b Feret Diameter, and c Right coiling direction. Statistically significant relationships shown by black lines are observed between δ44/40Ca and FAI (p = 0.004, R2 = 0.50), as well as δ44/40Ca and Feret Diameter (p = 0.006, R2 = 0.45); however, the weak negative relationship between δ44/40Ca and right coiled specimens is not significant (p = 0.14, R2 = 0.11). ASW Atlantic Seawater.

Full size image

Bulk δ13Ccarb measurements agree with previously generated records for the Bottaccione section (see Supplementary Fig. 5) and the AA core. In addition, comparisons between bulk and foraminiferal δ44/40Ca, δ13Ccarb, and δ18O values at Gubbio show that bulk values are on average heavier for all systems (Fig. 3). These differences are most visible within the lower section of the LCI, where bulk carbonate measurements are −0.18‰ heavier than R. cushmani for δ44/40Ca, at most 0.64‰ heavier for δ13Ccarb, and ~1‰ heavier for δ18O. All data can be found in Supplementary Data 1.

Fig. 3: Comparison of bulk carbonate and foraminiferal δ44/40Ca values.
figure 3

Bottaccione section a foraminiferal δ44/40Ca (Rotalipora cushmani in red triangles and W. archaeocretacea in pink crosses) against bulk carbonate δ44/40Ca values, where the black line represents a 1:1 relationship between the different archives, b foraminiferal δ13C (Rotalipora cushmani in blue triangles and W. archaeocretacea in a blue cross) against bulk carbonate δ13C values, where the black line represents a 1:1 relationship between the different archives, and c foraminiferal δ18O (Rotalipora cushmani in green triangles and W. archaeocretacea in a green cross) against bulk carbonate δ18O values, where the black line represents a 1:1 relationship between the different archives. ASW Atlantic Seawater; VPDB Vienna Peedee Belemnite.

Full size image

Controls on the δ44/40Ca system over short timescales

This study is primarily focused on trends in δ44/40Ca values within the LCI preceding OAE2 (Fig. 1). Whereas a previous study focused on the long-term decrease in δ44/40Ca values after OAE2 began38, the LCI spans a timescale of only ~60 kyr preceding the positive shift in δ13C. Given the large inventory of Ca in seawater (~22 mmol/kg for Cretaceous seawater39), combined with the element’s long residence time (τ ≈ 1 Myr), input–output flux imbalances cannot substantially shift the Ca isotope composition of seawater over such short timescales31,40. Rather, only two mechanisms can produce the observed trends, either addition of secondary Ca derived from seawater or changes in \(\Delta_{{{{{{\rm{CaCO}}}}}_3}-{{{{{\rm{H}}}}}_{2}{O}}}\) during primary carbonate formation. We consider these hypotheses below.

Addition of secondary Ca

Due to slow reaction rates, little to no fractionation occurs when primary calcite recrystallizes and/or authigenic carbonate precipitates41,42. If these diagenetic reactions are pervasive and occur under open system (fluid-buffered) conditions, then secondary incorporation of seawater Ca could elevate δ44/40Ca values, thereby simulating effects from reduced precipitation rates43. Conversely, if diagenesis is minimal or occurs under closed system conditions (sediment-buffered), then primary signals can be preserved43.

Multiple observations show that the sediment piles at both sites experienced minimal fluid interactions. Modern diagenetic studies indicate low sediment-water interactions in depositional environments typifying the Bottaccione section (pelagic44) and the AA core (hemipelagic45)41,43. Accordingly, bulk carbonates from both sites have low δ44/40Ca values compared to carbonates altered under fluid-buffered conditions46. The positive δ44/40Ca excursions at the Bottaccione section underlay the organic-rich Bonarelli Level. Organic matter (OM) respiration could have caused carbonate dissolution, thereby increasing fluid flow, resulting in authigenic carbonate precipitation42. Such a mechanism is expected to lower bulk δ13C values, but in the LCI, bulk δ13C values negligibly vary while δ44/40Ca values clearly increase. Therefore, we find no evidence supporting a fluid-buffered diagenetic regime at our study sites.

Our study is the first to analyze foraminiferal and bulk δ44/40Ca values from one location. The foraminiferal samples from the Bottaccione section show textural evidence for recrystallization and possible calcite infilling (see Supplementary Fig. 6). However, their δ44/40Ca values are toward the lower range for calcite47, and the values are either the same or lower than those for corresponding bulk carbonate (Fig. 3). The mostly unique δ44/40Ca values of bulk carbonate and foraminifera indicate that the archives were not geochemically homogenized and are therefore high-fidelity. Furthermore, as R. cushmani composes a minor fraction (0.1–4.3%) of the bulk carbonate48, differences between the archives suggest the infilled calcite contains Ca internally derived from test chambers. Apparent convergence for a few samples could represent homogenization, but presumably under sediment-buffered conditions, as these samples display both low and high δ44/40Ca values. As discussed below, it is equally, if not more probable that the pattern reflects primary signals.

Environmental control of \(\Delta_{{{{{{\rm{CaCO}}}}}_3}-{{{{{\rm{H}}}}}_{2}{O}}}\)

Absence of evidence for authigenesis and fluid-buffered early-diagenetic control points to a reduction in the magnitude of \(\Delta_{{{{{{\rm{CaCO}}}}}_3}-{{{{{\rm{H}}}}}_{2}{O}}}\) as the most plausible hypothesis for identical positive δ44/40Ca excursions observed in two different carbonate archives deposited in two globally separated locations. For both inorganic and biogenic carbonates, two primary environmental parameters influence \(\Delta_{{{{{{\rm{CaCO}}}}}_3}-{{{{{\rm{H}}}}}_{2}{O}}}\): temperature and precipitation rate. In general, foraminiferal δ44/40Ca values weakly relate to temperature49,50, with the magnitude of \(\Delta_{{{{{{\rm{CaCO}}}}}_3}-{{{{{\rm{H}}}}}_{2}{O}}}\) decreasing at higher temperatures29 (Supplementary Fig. 1). While proxy temperature records with sufficient resolution do not exist from either the WIB or the Bottaccione section prior to OAE2, comparison of foraminiferal δ18O values (a canonical paleo-thermometer) and δ44/40Ca values at the Bottaccione section help constrain the influence of temperature on \(\Delta_{{{{{{\rm{CaCO}}}}}_3}-{{{{{\rm{H}}}}}_{2}{O}}}\). No statistical relationship exists between the two proxies through the LCI (p = 0.16, R2 = 0.09) (see Supplementary Fig. 7), and δ18O values through the section only vary by 0.17‰, suggesting minimal temperature change during the LCI.

The data appear consistent with kinetic control of \(\Delta_{{{{{{\rm{CaCO}}}}}_3}-{{{{{\rm{H}}}}}_{2}{O}}}\). In other words, carbonate chemistry dominates both inorganic26 and foraminiferal calcite precipitation rates7,8. Inorganic carbonate studies have demonstrated that decreased precipitation rates reduce the magnitude of \(\Delta_{{{{{{\rm{CaCO}}}}}_3}-{{{{{\rm{H}}}}}_{2}{O}}}\)26,41, and some biogenic carbonates fractionate Ca isotopes similarly to inorganic calcite29,32,51,52 (Supplementary Fig. 1). Application of the stable Ca and Sr isotope (δ44/40Ca-δ88/86Sr) multi-proxy to biogenic carbonates spanning OAE1a offers particularly strong evidence for precipitation rate control51. Across this paleoclimate event, δ44/40Ca and δ88/86Sr values correlate with a slope identical to the one predicted from kinetic theory and observed in inorganic calcite precipitation experiments53, further confirming that most carbonates form kinetically27.

Following from the above interpretation, the coincident foraminiferal and bulk records sampled in the Bottaccione section present an opportunity to examine how, if at all, different organisms responded to carbonate chemistry stress. As noted above, R. cushmani composes a very small fraction of the pelagic bulk carbonate representing the LCI. Instead, nannofossils, particularly the coccolithophores Biscutum constans and Zeugrhabdotus22, dominate. While both foraminiferal and bulk carbonate archives show positive δ44/40Ca excursions through the LCI, Fig. 3 shows that bulk carbonate has higher δ44/40Ca, δ13C, and δ18O values than foraminifera. Previously studied Cretaceous sediments show that coccolith fractions have higher δ13C and δ18O values compared to deeper dwelling planktic foraminifera54. As R. cushmani dwelled deeper in the thermocline55, it follows that the higher bulk δ13C and δ18Ο values are weighted towards surface ocean calcifiers, such as coccoliths. Thus, we interpret the bulk δ44/40Ca data as a record of surface ocean biocalcification stress, noting that the surface ocean signal may be muted, as bulk carbonate comprises both surface and thermocline dwelling species. Figure 1 may suggest that coccolithophores and other surface ocean calcifiers experienced biocalcification stress before R. cushmani, consistent with rapid surface OA and downward propagation of acidified waters. If our archives record species-specific responses, then the scatter observed in the Bottaccione bulk record may represent shifts in species abundances leading up to OAE222 because the bulk records average numerous pelagic species. Changes in species abundance may also explain scatter in the AA record, particularly considering evidence for Tethyan and Boreal water-mass mixing, and resulting ecosystem shifts, before and during OAE256.

Relation to existing OAE2 records

The Ca isotope records presented in this study correlate with existing morphometric, ecological, and geochemical evidence consistent with OA and carbon cycle perturbations leading up to OAE2. For example, nannofossil abundance data from the same section show that nannofossil extinctions began and persisted through the LCI20,22. During the LCI (here defined to start ~60 kyr prior to the CIE) at the Bottaccione section, R. cushmani tests become smaller (show decreased Feret diameters), abnormal, and malformed, and their coiling directions change from right to left18. Above, we deduce that reduced precipitation rates were the principal driver of elevated δ44/40Ca values through the LCI; therefore, we interpret the significant relationships between R. cushmani δ44/40Ca values and two morphometric observations (FAI and Feret diameter) as consistent with OA-driven biocalcification stress. We note that foraminiferal morphometric proxies, particularly coiling direction18, have exhibited sensitivity to additional environmental parameters, such as salinity, temperature, oxygen content25, and toxic metal loading22. While our analysis does not indicate that these stressors impacted the δ44/40Ca records at the onset of the LCI, thallium isotope reconstructions provide evidence for global deoxygenation starting ~43 kyr prior to OAE257, which could have compounded biocalcification stress. Osmium isotope (Osi) records for the WIB show that Osi values declined prior to the onset of OAE2 and led the positive CIE by ~60 kyr36. Using an integrated timescale to compare these Osi records to our δ44/40Ca records (see Supplementary Fig. 5), we show that the positiveδ44/40Ca excursions coincide with decreases in Osi (Fig. 4). Furthermore, small decreases in the Osi records ~300 kyr prior to OAE2 align with increased δ44/40Ca values in the bulk carbonate records from the Portland core  and the Bottaccione section. The overall correlation between δ44/40Ca and Osi is statistically significant (p < 0.01, Supplementary Fig. 10) but relatively weak (R2 = 0.38), which we interpret as evidence for dynamic biocalcification feedbacks on δ44/40Ca as discussed below. The Osi decrease most likely reflects a pulse of volcanism from the CLIP and/or High Arctic LIP35,36. Previous studies of LIP eruptions across OAE2 suggest a slow rate of CO2 addition16; however, our δ44/40Ca records suggest this global CO2 pulse occurred relatively rapidly, resulting in OA and biocalcification stress. The observation that elevated δ44/40Ca values predate the CIE further indicates that calcification stress represents a principal response to increased CO2 and that oceanic anoxia followed. Our study highlights that a reconceptualization of the onset of OAE2, to include the LCI, may be appropriate. At present, the initiation of OAE2 in the Gubbio section is defined by the positive shift in δ13Corg values, whereas the Ca isotope proxy provides evidence that the volcanically-driven carbon cycle perturbation predates the CIE. If OAEs are truly global perturbations of the carbon cycle, then our results demonstrate that the perturbation began before the CIE recorded in δ13Corg. This view highlights a need to further focus on the pre-OAE2 interval, as defined by the LCI in the Bottaccione section.

Fig. 4: A compilation of geochemical data preceding and across OAE2.
figure 4

Time series of a carbon isotope (δ13C) values from two sites, the AA core in gray squares (Joo and Sageman45) and the Portland core in gray circles (Sageman et al.17), b osmium isotope (Osi) from two sites, the AA core in gray squares (this study) and the Portland core in gray circles (Du Vivier et al.35, 36), and c calcium isotope values from three sites: the Gubbio Bottaccione section with foraminiferal measurements in red triangles and bulk carbonate in black diamonds (this study), the AA core in gray squares (this study), and the Portland core in gray circles (Du Vivier et al.30). The gray shaded area marks the previously observed lag between the Osi excursion indicating enhanced volcanism and the onset of the positive δ13C excursion marking greater organic carbon burial36, which corresponds to the lower critical interval. The dashed line marks a small decrease in Osi and a small increase in δ44/40Ca at roughly 300 kyr before OAE2 began. The timing of the Plenus Cold Event, shown with blue shading, follows from O’Connor et al.37. PCE Plenus Cold Event, OAE2 Oceanic Anoxic Event 2, ASW Atlantic Seawater; VPDB Vienna Peedee Belemnite.

Full size image

Our multi-proxy correlations suggest that LIP volcanism caused a rapid influx of CO2 leading up to OAE2, resulting in reduced biocalcification. Previous studies demonstrate that decreased biocalcification acts as a rapid negative feedback to OA3; however, the impact of biological compensation during continued CO2 injection into the atmosphere-ocean system has received less attention. Just as chemical compensation (burndown) increases ocean alkalinity via dissolution of carbonate sediments58, biological compensation increases alkalinity in the surface ocean, thereby buffering the effects of OA3. If atmospheric CO2 concentrations remain elevated after the initial biological compensation, then increased surface ocean alkalinity from reduced biocalcification would enhance disequilibrium between the atmosphere and the surface ocean. The disequilibrium would cause CO2 invasion to the surface ocean10, thereby creating a positive feedback loop of additional biocalcification stress and OA. The Portland core and AA δ44/40Ca records indicate that OA persisted for another ~150 kyr after the start of OAE2. Thus, while toxic metal loading from LIP volcanism offers a leading hypothesis for declining nannofossil abundances at Gubbio22, an equally plausible mechanism may be biocalcification stress from episodic and rapid OA, exacerbated by biological compensation feedbacks and continued CO2 input from LIP volcanism. Sustained suppressed biocalcification would have continued to increase alkalinity in the surface ocean causing additional atmospheric CO2 drawdown. Recovery of δ44/40Ca and Osi values in the Portland core and AA cores is consistent with a geologically rapid biocalcification response to waning CO2 inputs (Supplementary Fig. 12). Sedimentological records of carbonate weight percent provide evidence for alkalinity overshoots following other OA events, such as the Paleocene–Eocene Thermal Maximum59. Interestingly, the shift to lower δ44/40Ca values observed here coincides with an enigmatic global cooling event (Fig. 4), i.e., the Plenus Cold Event37. Transient CO2 drawdown due to biologically-mediated ocean alkalinization provides a mechanism to explain why climate cooling occurred during an otherwise hothouse period.

Evidence for biocalcification feedbacks from LIP eruptions and shifts in air-sea CO2 partitioning deduced from the Ca isotope study of OAE2 may have implications for implementing ocean alkalinity enhancement as a strategy for mitigating the anthropogenic climate crisis and OA. Our interpretations presented here support model findings that sustained and higher magnitude ocean alkalinity enhancement is needed to increase CO2 drawdown while also mitigating OA60. Observations of episodic OA during OAE2 suggest that if ocean alkalinity enhancement was lower in magnitude, and/or did not include addition of external alkalinity, then the OA mitigation co-benefits of the climate intervention strategy may be limited.

Understanding biological compensation during OAE2 further bears on the relationship between ballasting of OM and nannofossil carbonate production61,62. Future studies centered on interactions between OA, reduced biocalcification, and the biological pump leading up to OAE2 may provide more insight into the cause(s) of widespread, albeit not global, oceanic anoxia and noncalcareous stratigraphic intervals. Given that the community composition and abundance of surface ocean plankton contribute to overall OM export63, understanding how OA affects pelagic calcifiers may shed light on long-standing debates about enhanced productivity versus enhanced preservation as the cause of OAEs15. For example, if biocalcification stress on coccolithophores caused the efficiency of the biological pump to decline prior to OAE2, then thermal stratification, reduced deep sea ventilation, and enhanced OM preservation may represent the driver64. On the other hand, if coccolith biocalcification stress opened a niche for diatom proliferation, as these siliceous taxa are largely generalists65, then enhanced productivity offers a more probable explanation57. Alternatively, it is simply possible that reduced carbonate production contributes to the appearance of enhanced OM productivity, as megafauna could have continued to support the biological pump.

Conclusion

Foraminiferal and bulk δ44/40Ca records from the Bottaccione section in Gubbio Italy, as well as a bulk δ44/40Ca isotope record from AA in the WIS, all show coincident elevated values ~60 kyr prior to OAE2. A statistically significant correlation betweenδ44/40Ca and Osi, a well-established proxy for volcanism, indicates that the Ca isotope records reflect a reduction in the magnitude of \(\Delta_{{{{{{\rm{CaCO}}}}}_3}-{{{{{\rm{H}}}}}_{2}{O}}}\) resulting from slower carbonate precipitation rates due to OA. At the Bottaccione section, these results occur in tandem with morphometric indicators of biocalcification stress on foraminifera, such as test dwarfing and malformation, which provide independent evidence for OA. The relationships between δ44/40Ca, δ13C, and δ18O values of R. cushmani and bulk carbonate at the Bottaccione section further show that bulk carbonate records can capture the overall biocalcification response to OA, which could include major assemblage reorganization. While additional studies are needed to examine species-specific effects on \(\Delta_{{{{{{\rm{CaCO}}}}}_3}-{{{{{\rm{H}}}}}_{2}{O}}}\), correlating the data to existing proxies for volcanic activity suggests that elevated atmospheric CO2 levels from emplacement of the CLIP and/or High Arctic LIP decreased carbonate ion concentrations and \(\Delta_{{{{{{\rm{CaCO}}}}}_3}-{{{{{\rm{H}}}}}_{2}{O}}}\) by extension. Given that both species-specific and bulk δ44/40Ca values increase prior to OAE2, our data provide good evidence that elevated CO2 and OA preceded the positive CIE, which points to a need to redefine the timing and cause(s) of the event. Elevated surface ocean alkalinity following reduced carbonate production would have facilitated additional atmospheric CO2 drawdown, thereby sustaining biocalcification stress and creating a positive feedback loop for additional OA. The timing between elevated alkalinity and waning volcanic activity, as predicted by our records, coincides with a period of transient global cooling (i.e., the Plenus Cold Event). This study highlights OAE2 as the first geological analog for ocean alkalinity enhancement, which is a leading marine climate intervention strategy for mitigating anthropogenic climate change and ocean acidification.

Methods

Bulk carbonate sample collection and processing

The Scaglia Bianca Formation was sampled in outcrop at the Bottaccione section in Gubbio, Italy (43°21′56.04″N, 12°34′57.56″E; 597 m above sea level) following the stratigraphic scheme outlined in Coccioni and Premoli Silva (2015)66. For each interval of interest, ~100 g of bulk rock was collected and sent to Northwestern University (NU). The Angus Aristocrat (AA) core, recovered from the north-central Colorado (40°14′32.97″N, 104°41′42.98″W; 1506 m above sea level), was sampled at NU. Prior to dissolution, bulk samples were crushed using a shatter box, or when weakly lithified, using a mortar and pestle, with a pure silica sand ‘rinse’ applied between samples to minimize potential for cross-contamination. Depending on approximate CaCO3 wt%, ~27–100 mg of powder was reacted with 10 mL of 5% HNO3, and the mixtures were placed on a rocker table for ~12 h to ensure complete dissolution. Samples were then filtered through 0.45 μm filters and stored in acid-washed LDPE bottles. Calcium concentrations were determined using a Thermo Scientific iCAP 6500 ICP-OES at NU.

Foraminiferal sample collection and processing

At the University of Urbino, ~100 g of bulk rock sampled from the Bottaccione section was treated with concentrated acetic acid following Lirer (2000)67, sieved through a 63 μm mesh, and dried at 50 °C prior to disaggregation18. After washing the sediments through a 212 μm mesh sieve, R. cushmani tests were picked from the nannofossil-rich chalk. When possible, W. archaeocretacea tests were also picked. For each sampling interval, 15–25 specimens of R. cushmani and available W. archaeocretacea were picked from this size fraction, measured for their morphometric Feret diameter, and sent to NU. In the Radiogenic Isotope Geochemistry Laboratory at NU, seven R. cushmani and/or W. archaeocretacea were subsampled, lightly crushed, and homogenized prior to cleaning with MilliQ water following the procedure outlined in Barker et al.68. While previous studies indicate that cleaning procedures produce no discernable effects for Ca isotope measurements32, the samples were nonetheless washed with MilliQ water to remove any contaminants from the disaggregation process. After cleaning, samples were dissolved in 5% HNO3 and stored in acid-cleaned LDPE bottles. Calcium concentrations were determined from diluted stock solutions using a Thermo iCapQ quadrupole ICP-MS in the Quantitative Bio-element Imaging Center at NU.

Additional foraminifera from four sampling intervals were imaged using a Hitachi S3400N-II SEM at the Northwestern University Atomic and Nanoscale Characterization Experimental Center to confirm species identification, examine the texture of specimens, and identify potential authigenic precipitates, such as diagenetic crystallites69 (Supplementary Fig. 6). Samples were mounted onto carbon-taped mounts, and multiple specimens from each sampling interval were imaged to assess inter-sample variability. The images confirm the species and also show calcite infilling of all foraminiferal shells across all sampling intervals; the implications of this infilling are addressed in the main text.

Calcium isotope ratios

Calcium isotope ratios (44Ca/40Ca) were measured at NU using a high-precision 43Ca-42Ca double-spike technique70 and a Thermo-Fisher Triton MC-TIMS equipped with 1011 Ω amplifier resistors. Data were collected periodically from June 2018 to October 2020. All procedures were conducted under clean room conditions and employed acid cleaned labware and ultrapure reagents. Blanks (n = 7) from labware, reagents, and foraminiferal cleaning procedures were quantified by isotope dilution and were found to be nearly 100 times lower than the sample size, which is negligible. Eight OSIL Atlantic Seawater (ASW) standards and two NIST SRM-915b standards were analyzed every 30 or fewer samples. Sample and standard volumes containing 50 µg of Ca for bulk carbonates and 25 µg of Ca for foraminiferal samples were spiked, and the mixtures were equilibrated on a hotplate at 60 °C for 12 h before drying down at 90 °C overnight. ASW standards and bulk carbonate samples were re-dissolved in 500 µL of 1.55 N HCl, and the solutions were passed through Teflon columns packed with Bio-Rad AG MP-50 cation exchange resin to separate Ca from K and other matrix elements. After drying down the purified fractions at 90 °C, OM was oxidized by adding two drops of 35% H2O2, and Ca was converted to nitrate form by adding two drops of concentrated 16 N HNO3. The solutions were dried and re-dissolved in 5 μL of 3 N HNO3. For 915b, aliquots containing 50 µg of Ca were processed without column chemistry, and then two drops of 35% H2O2 were added, followed by two drops of concentrated 16 N HNO3. Foraminiferal samples were not processed through column chemistry due to their lower [Ca]. For all samples and standards, ~10–16 µg of Ca was loaded onto Ta filament assemblies. The double-spike was frequently recalibrated using the ASW data for each session70. All results are reported in δ-notation relative to ASW, where δ44/40Casmp(‰) = [(44Ca/40Ca)smp/(44Ca/40Ca)ASW − 1]×1000.

During the period of study, repeated measurements of ASW and 915b yielded δ44/40Ca = −0.002 ± 0.039‰ (2σSD, n = 80) and δ44/40Ca = −1.147 ± 0.055‰ (2σSD, n = 24). AA data collection began in June 2018, whereas Gubbio data collection began in May 2019; however, it is worth noting that no significant differences exist between standard values measured during the data collection periods. Since implementation of the Lehn et al.70 method, measurements of ASW and 915b have yielded δ44/40Ca = 0.000 ± 0.044‰ (2σSD, n = 706) and −1.136 ± 0.049‰ (2σSD, n = 276). Long-term data for the lab point to an external reproducibility of ±0.05‰, which is the uncertainty assigned to samples analyzed in the present study. Replicates, or reanalysis of the same dissolved sample were measured (n = 9). One duplicate sample, as well as analyses of altogether different samples taken from the same stratigraphic interval, were also analyzed. The difference between original measurements and replicates/duplicates is better than ±0.05‰.

Carbon and oxygen isotope analyses

Carbon and oxygen isotope ratios (13C/12C and 18O/16O) were measured at NU using a Thermo Scientific Delta V-Plus Isotope Ratio Mass Spectrometer (IRMS) equipped with a Gas Bench II, with a PAL autosampler. Approximately 200 μg of bulk sediment was measured into 12 mL glass Exetainers®, which were then purged with UHP He gas for 10 mins prior to acidification with 200 μL of 103% H3PO4, calibrated to a density of 1.92 g/cm3. The acidified samples were left to dissolve at 30 °C in an incubation block, for twelve hours. To optimize the isotope analyses for small samples (i.e., foraminifera), the above method was modified to maximize precision for masses ≤40 μg. Five foraminiferal samples were weighed into Sn capsules and subsequently placed into glass Exetainers®. These samples were also purged for 10 min with UHP He gas and placed in a thermostated temperature block for at least 30 min at 90 °C to ensure complete dissolution of the Sn capsule. Standards were analyzed at various size series to bracket the sample masses and to assess the linearity of corrections needed for sample δ13C and δ18O values. The resulting CO2 from each sample was measured on the IRMS, following a variable mass standard series of NU in-house Carrara Lago marble standard (CLMS) (δ13C = +2.31 ± 0.05‰, δ18O = −3.70 ± 0.10‰), one NBS18 carbonate standard (δ13C = −5.01 ± 0.05‰, δ18O = −23.01 ± 0.05‰), and a blank. Approximately 200 μg of CLMS and NBS18, along with one blank, were repeatedly analyzed every ten samples and at the end of runs to assay instrumental drift. Carbon and oxygen isotope values are reported relative to the Vienna Pee Dee Belemnite standard ascertained by repeated analyses of NBS19 (δ13C = +1.95‰, δ18O = −2.20‰). For both δ13C and δ18O values, foraminiferal and bulk samples yielded the same 2σSD of ±0.10‰.

Statistical analyses

For non-serially correlated data (Bottaccione Sectionδ44/40Ca and δ18Ο values, as well as morphometric data), regression analysis was utilized to examine covariance between variables. Perhaps due to the small sample size, a Cook’s distance plot indicated that two of the geochemical points composing the δ44/40Ca–δ18Ο correlation have high influence on the data. Thus, in order to consider the influence of each point on the regression, two alternate models were built where the data points were sequentially removed from each model. The results of each of these models suggest that no individual point influences the significance or strength of the R2 value produced by the original model with all data points included. Therefore, no justification exists for excluding the points from the linear regression. These data meet all other assumptions for application of a linear model. Morphometric data from Coccioni et al.18 were selected for regression analyses, as the high sample number (n = 4140) likely best represents the R. cushmani population at the Bottaccione Section. The morphometric correlations meet all assumptions for application of a linear model.

For the studied time-series (δ44/40Ca and WIB Osi values), data were compiled into one dataset for correlation. Given that the individual datasets represent varied sampling intervals, we used the R package Astrochron surrogateCor function71 to examine the relationship between the datasets. This tool linearly interpolates the lower density sample set and then correlates the interpolated sample set with the higher resolution record for comparison. Due to the non-normality of all datasets, we implemented the non-parametric Spearman’s Rank Correlation72. To examine whether one δ44/40Ca dataset (Bottaccione Section, AA, or the Portland core) significantly influenced the correlation, the statistical analysis was also conducted using only one of the δ44/40Ca datasets and the Osi data. No differences were observed, indicating that no one record holds leverage in the combined regression analysis.