Marine carbonate burial represents the largest long-term carbon sink at Earth’s surface, occurring in both deep-sea (pelagic) environments and shallower waters along continental margins. The distribution of carbonate accumulation has varied over geological history and impacts the carbon cycle and ocean chemistry, but it remains difficult to quantitatively constrain. Here, we reconstruct Cenozoic carbonate burial along continental margins using a mass balance for global carbonate alkalinity, which integrates independent estimates for continental weathering and pelagic carbonate burial. Our results indicate that major changes in marginal carbonate burial were associated with important climate and sea-level change events, including the Eocene-Oligocene transition (ca. 34 Ma), the Oligocene-Miocene boundary Mi-1 glaciation (ca. 23 Ma), and the middle Miocene climate transition (ca. 14 Ma). In addition, we find that a major increase in continental weathering from ca. 10 Ma to the present may have driven a concomitant increase in pelagic carbonate burial. Together, our results show that changes in global climate, sea level, and continental weathering have all impacted carbonate burial over the Cenozoic, but the relative importance of these processes may have varied through time.

Marine calcium carbonate (CaCO3) deposition occurs on continental shelves and slopes, and in the open ocean above the carbonate compensation depth (CCD; Milliman, 1993). At present, deep-sea (hereafter termed pelagic) carbonate burial represents ∼55%–65% of global carbonate burial, while carbonate accumulation along continental margins (including both shelves and slopes; hereafter termed marginal environments, or margins) accounts for the rest (Mackenzie and Morse, 1992). Carbonate alkalinity is delivered to the oceans mainly by silicate and carbonate rock weathering, and it is removed by carbonate burial, either in marginal or pelagic environments. These alkalinity inputs and outputs should be in balance over geological time scales (Berner, 2004; Ridgwell and Zeebe, 2005). This does not require true steady-state conditions, but it implies that input and output processes adjust sufficiently fast to changes to maintain balance at any given time on secular time scales (Boudreau et al., 2019). However, the nature and dominant location of carbonate accumulation have shifted over geological history, in association with changes in plate tectonics, paleogeography, global sea level, climate, ocean chemistry and the evolution of calcifying organisms (Mackenzie and Morse, 1992; Stanley and Hardie, 1998; Ridgwell and Zeebe, 2005). High sea levels resulted in extensive shallow-water carbonate accumulation in epicontinental seas during parts of the Phanerozoic, notably during the Carboniferous and Cambrian (Hay, 1985; Opdyke and Wilkinson, 1988; Walker et al., 2002), while pelagic carbonate burial was insignificant until the evolutionary success of coccolithophores and planktonic foraminifera in the Mesozoic (Martin, 1995; Ridgwell, 2005). An overall, long-term decline in marginal deposition since the Late Cretaceous in response to falling sea levels and a reduction in epicontinental shelf areas may have resulted in a further increase in pelagic deposition toward the present (Opdyke and Wilkinson, 1988; Walker et al., 2002).

For carbon cycle reconstructions, it is important to resolve temporal changes in marginal and pelagic carbonate burial and their respective contributions to global carbonate accumulation. Because the continental margins represent the transition zone between the terrestrial and marine realms, they may either act as a carbon sink through immediate carbonate burial, or as a carbon source to the pelagic oceans through carbonate export. Additionally, marginal carbonates are deposited as aragonite, calcite, and magnesium-rich calcite, whereas pelagic carbonates are deposited primarily as calcite (Milliman, 1993). Because of these differences in carbonate mineralogy, shifts in marginal and pelagic carbonate burial may also drive changes in ocean chemistry (Berner, 2004). Progress has been made to estimate pelagic carbonate burial rates through time (Opdyke and Wilkinson, 1988; Boudreau and Luo, 2017), but relatively little is still known about secular changes in marginal carbonate burial. Here, we employ a new approach to estimate marginal carbonate burial rates for the Cenozoic, based on a mass balance for global carbonate alkalinity that incorporates independent continental weathering and pelagic carbonate burial histories.

We derive a steady-state carbonate alkalinity mass balance for the present, where alkalinity is delivered to the oceans by continental weathering (Fweathering) and anaerobic processes in organic-rich sediments deposited along continental margins (Fanaerobic), while alkalinity is removed through marginal and pelagic carbonate burial (Fmargins and Fpelagic, respectively; Fig. DR1 in the GSA Data Repository1). We calculate changes in Fweathering for the Cenozoic based on inverse modeling of the marine strontium (Sr) and osmium (Os) isotope systems, two well-established tracers for weathering inputs to the global ocean (Ravizza and Zachos, 2003; see also the Data Repository). We use published Cenozoic seawater 87Sr/86Sr records (McArthur et al., 2012) and 187Os/188Os records (Klemm et al., 2005; Burton, 2006; Nielsen et al., 2009) in combination with seafloor spreading and/or degassing rate reconstructions (e.g., Van Der Meer et al., 2014) to estimate changes in continental silicate weathering and sediment weathering—the sum of carbonate weathering and organic-rich sediment weathering—back in time (Figs. DR2–DR6 in the Data Repository). This approach is inspired by previous modeling studies (Delaney and Boyle, 1988; Li et al., 2009; Li and Elderfield, 2013) and uses the same Sr and Os isotope data sets as Li and Elderfield (2013), but, uniquely, we do not assume a priori equilibrium between inputs and outputs in our weathering reconstructions. Rather, we consider the Sr and Os isotope systems without incorporating additional mass balance constraints (e.g., carbon isotopes), which enabled us to independently use the carbonate alkalinity mass balance to subsequently derive marginal carbonate burial. We investigate a range of scenarios for continental silicate weathering and sediment weathering based on different sets of assumptions regarding the evolution of their respective 87Sr/86Sr and 187Os/188Os compositions over the Cenozoic (either as constant values or as linear increases in one or multiple parameters; Table DR2; Fig. DR6). These continental silicate and sediment weathering estimates are proportionally added to obtain the total carbonate alkalinity delivery associated with continental weathering.

For Fanaerobic, we use the sum of alkalinity addition from net denitrification and pyrite burial, as estimated for the present-day by Hu and Cai (2011). Inverse modeling of carbon and sulfur isotope records has demonstrated that pyrite burial rates were likely higher in the early Cenozoic than at present (Rennie et al., 2018), but the resulting changes in alkalinity fluxes are rather small compared to those associated with continental weathering and carbonate burial. Therefore, we choose to maintain a constant value of Fanaerobic over the Cenozoic within the uncertainties reported by Hu and Cai (2011).

Boudreau and Luo (2017) estimated Cenozoic changes in Fpelagic independent of any continental weathering reconstructions. These estimates were derived using a set of equations that related pelagic carbonate burial rates to the positions of the calcite saturation horizon and the CCD. Based on this work, we derive two different scenarios for Fpelagic, driven by either the global Cenozoic CCD curve (Lyle et al., 2008; i.e., the Delaney and Boyle [1988] compilation with an updated age model) or the Pacific CCD curve for the past 50 m.y. (Fig. 1F; Pälike et al., 2012). The Lyle et al. (2008) CCD includes information from all major ocean basins and thus represents global trends, although coverage of Atlantic Ocean sites is relatively poor. The Pacific CCD record of Pälike et al. (2012) is only an approximation of global CCD change, but this record is still useful because the Pacific Ocean represents the largest ocean basin by volume—even more so in the early Cenozoic than at present (Lyle et al., 2008). However, the CCD may have been relatively insensitive to changes in weathering and carbonate burial during the early Eocene (Greene et al., 2019).

Finally, Cenozoic changes in Fmargins are calculated by solving our carbonate alkalinity mass balance through time, with propagation of uncertainties in Fweathering, Fanaerobic, and Fpelagic (see the Data Repository).

Our range of Fweathering scenarios displays a gradual increase from the early Cenozoic toward the present (Fig. 1A), a result consistent with previous modeling studies in terms of general shape and magnitude (Delaney and Boyle, 1988; Li and Elderfield, 2013). The large increase in Fweathering we predict in the late Neogene coincides roughly with the final stage of Himalayan uplift (Raymo et al., 1988), but the impact of Himalayan uplift on continental weathering rates remains in debate because weathering reconstructions are generally nonunique and different weathering proxies offer contrasting views (Kump, 1989; Edmond, 1992; Lear et al., 2003; Willenbring and Von Blanckenburg, 2010; Caves Rugenstein et al., 2019). By comparison, the contribution of Fanaerobic to alkalinity delivery is small for most of the Cenozoic.

The Fpelagic scenarios show an overall increase toward the present, but they diverge during the Oligocene and especially during the Miocene (Fig. 1B). This divergence is due to underlying differences between the Lyle et al. (2008) CCD and the Pälike et al. (2012) CCD. In particular, the latter may contain departures from global trends because it also responds to changes in equatorial Pacific carbonate production relative to global carbonate production, or changes in carbonate deposition fractionation between the Atlantic and Pacific basins. The general shape of the Lyle et al. (2008)–based Fpelagic scenario appears to correspond best to the reconstructions of pelagic carbonate accumulation of Opdyke and Wilkinson (1988).

If our independent Fweathering, Fanaerobic, and Fpelagic estimates are robust, secular trends in Fmargins should emerge that fit existing observations from the geological record (Fig. 1C). The Lyle et al. (2008)–based Fmargins scenario is relatively smooth for most of the Cenozoic, except for one interval of rapid change. After a gradual increase in the Eocene, Fmargins displays a major decrease between 38 and 34 Ma, followed by an interval where the rates become quite low. These low values persist during most of the Oligocene until ca. 23 Ma, after which Fmargins recovers and remains constant until the end of the Miocene. A small, final increase is observed between the Pliocene and the present. The Pälike et al. (2012)–based Fmargins scenario displays more dramatic shifts than the Lyle et al. (2008)–based scenario, but it does not extend beyond 50 Ma. Following a slight decline in the Eocene, it is characterized by an interval of very low values between 36 and 10 Ma. This interval is punctuated by distinct minima with negative burial rates at ca. 32 Ma, ca. 23 Ma, and ca. 12 Ma, which indicate net removal (weathering) of previously buried marginal carbonates and transport of alkalinity to the pelagic realm. Finally, a large increase is observed from ca. 10 Ma to the present that is not apparent in the Lyle et al. (2008)–based scenario.

We find that the overall patterns in Fmargins correlate with the Cenozoic evolution of global sea level and climate (Figs. 1D and 1E). For the early Cenozoic, we estimate relatively high Fmargins values corresponding to globally warm climates (Zachos et al., 2008) and high sea levels (Miller et al., 2005), while subsequent long-term cooling is generally mirrored by a decline in Fmargins values. The absolute Fmargins values we obtain are mostly lower than those reconstructed by Opdyke and Wilkinson (1988), potentially due to underestimation of Fweathering, Fanaerobic, or both in the early Cenozoic. Importantly, however, the distinct minima in Fmargins correlate with important episodes of sea-level change and loss of marginal environments. For example, the large Fmargins shift at ca. 36 Ma occurs just prior to the Eocene-Oligocene transition (ca. 34 Ma) and the onset of Antarctic glaciation (Lear et al., 2000), which was characterized by an episode of major sea-level fall and CCD deepening due to transient weathering of exposed shelf carbonates and a shift of carbonate burial to the pelagic oceans (Coxall et al., 2005; Merico et al., 2008; Armstrong McKay et al., 2016). Similarly, the minimum at ca. 23 Ma is virtually synchronous with ice-sheet expansion during the Oligocene-Miocene boundary Mi-1 glaciation (Miller et al., 1991; Zachos et al., 2001), and the minimum at ca. 12 Ma, which is only apparent in the Pälike et al. (2012)–based scenario, follows ice-sheet expansion just after the middle Miocene climate transition (ca. 14 Ma; Flower and Kennett, 1994; Shevenell et al., 2004; Armstrong McKay et al., 2014). The strong correlations between episodes of decreasing or negative Fmargins and major episodes of cooling and ice-sheet expansion suggest that these phenomena were related through changes in global sea level and carbonate compensation.

Strikingly, the Pälike et al. (2012)–based Fmargins scenario deviates from these aforementioned long-term Cenozoic trends over the past 10 m.y. The final Fmargins increase in this scenario results from a relatively invariant Fpelagic at a time of strongly increasing Fweathering. However, this Fmargins increase would be opposite to the continued decrease in platform carbonate burial that has been inferred over this period by other studies (Hay, 1985; Opdyke and Wilkinson, 1988; Walker et al., 2002) and that would be expected based on progressive Cenozoic sea-level fall (Haq et al., 1987; Miller et al., 2005; Müller et al., 2008). By comparison, the Lyle et al. (2008)–based Fmargins scenario resembles the reconstructions of Opdyke and Wilkinson (1988) better over this period because the large Fweathering increase is mostly accommodated by a simultaneous increase in Fpelagic. This reveals the potential of continental weathering as a driver of changes in pelagic carbonate burial, but this relationship may be variable through time due to changes in background climate states (e.g., Greene et al., 2019). Therefore, we suggest that future studies should focus on fully incorporating the complex interactions among climate, sea level, weathering, carbonate burial, and the CCD in models.

We present estimates of carbonate burial along continental margins for the Cenozoic, using a global alkalinity mass balance that incorporates independent estimates of continental weathering and pelagic carbonate burial. The correlation between the inferred changes in Fmargins, Fpelagic, and the interplay of global climate, sea level, and continental weathering implies that these phenomena and processes have all impacted carbonate burial over the Cenozoic, but that their relative importance may have changed over time. In summary, global cooling and sea-level fall may have driven the progressive, long-term decline in marginal carbonate burial for most of the Cenozoic, while an increase in continental weathering related to tectonic uplift in the late Neogene may have driven the final pelagic carbonate burial increase toward the present. Additionally, these long-term trends appear to be modulated by transient climate events such as the Eocene-Oligocene transition, Mi-1 glaciation, and middle Miocene climate transition. Episodes of negative marginal carbonate burial associated with such climate events may indicate large-scale weathering and erosion of previously deposited carbonate material and release of alkalinity from the continental margins to the pelagic oceans, which is consistent with the concept of shifting carbonate accumulation through time.

This work was carried out under the program of the Netherlands Earth System Science Centre (NESSC), which is financially supported by the Ministry of Education, Culture and Science (OCW) of the Netherlands. We thank Mitchell Lyle, Bradley Opdyke, Gerald Dickens, and David Armstrong McKay for constructive reviews that significantly improved the quality of this manuscript.

1GSA Data Repository item 2019362, supplementary information including detailed model descriptions and sensitivity analyses, is available online at, or on request from
Gold Open Access: This paper is published under the terms of the CC-BY license.