Deep-sea hiatuses track the vigor of Cenozoic ocean bottom currents

The deep-sea stratigraphic record is pervaded by hiatuses (Keller and Barron, 1983; Keller et al., 1987). These discontinuities are chiefly caused by mechanical erosion of the seafloor by currents, dissolution of biogenic carbonate associated with fluctuations of the carbonate compensation depth (CCD), or periods of nondeposition (Keller and Barron, 1983; Moore et al., 1978). Based on the stratigraphy of Deep-Sea Drilling Project (DSDP) sites, a number of pioneering investigations in the 1980s suggested that deep-sea hiatuses are largely the result of major changes in ocean circulation and the flow of bottom currents and may be linked to climatic perturbations (Keller and Barron, 1983; Keller et al., 1987). But these ground-breaking interpretations were made using relatively sparse data without consideration of the fate of the missing material and without a global tectonic model, leading to misidentification of key events. We assess the distribution of deep-sea hiatuses and their paleo–water depths based on nearly 300 deep-sea drill holes using a plate-tectonic model that includes stretching along rifted margins. We couple this analysis with a global data set of contourite drifts (Thran et al., 2018) with maximum age constraints to connect hiatus formation to potential regions of excess deposition. We use a set of regional CCD reconstructions to consider the role of carbonate dissolution in forming hiatuses. Our analysis gives new insights into the link between seafloor sediment redistribution, gateway evolution, and the vigor of ocean-bottom currents.

We have selected age-depth models for 293 deep-sea drill holes (Fig. 1; Table S1 in the Supplemental Material¹) from the Neptune Sandbox Berlin (NSB) database (http://www.nsb-mfn-berlin.de/), which is based on harmonized, updated, and accurate lists of microfossils (Renaudie et al., 2020). The age models have been calibrated to the Gradstein et al. (2012) time scale. We follow Spencer-Cervato (1998) in defining hiatuses as gaps in the stratigraphic record longer than 0.18 m.y. and in assuming that they are accurately represented in the biostratigraphic NSB data. The quality of the age models used is discussed in Renaudie et al. (2020). We disregard sites where the age-depth relationship is very poorly constrained. The number of sites increases steadily from a low of 48 at 66 Ma to a high of 207 at 2.5 Ma (Fig. 2A), yielding a total of 409 hiatuses over the Cenozoic. Our global data set across all ocean basins, including on submerged continental crust, and in variable water depths minimizes bias introduced by drilling objectives and drilling methods. This is partially expressed in the lack of bias in hiatus duration for our entire time series (Fig. S1). Additionally, Spencer-Cervato (1998) concluded that incomplete core recovery, typical at DSDP sites, did not result in false hiatuses. DSDP and Ocean Drilling Program (ODP) holes from comparable water depths, e.g., on the Walvis Ridge, have similar resolution and show good agreement in age-depth models (e.g., DSDP Site 524 versus ODP Site 1267, and DSDP Site 525 versus Site ODP 1264) (see Table S1, and data files in the Supplemental Material¹), although direct comparison is difficult because location and water depth of the holes is not the same. The time of hiatus onset is imprecise because the amount of sediment removed is unknown (Moore et al., 1978), also limiting mass-balance considerations between areas of erosion and deposition. Thus, the age of hiatus initiation is based on the age of underlying sediment and represents a maximum, while the age of hiatus cessation is determined by the age of overlying sediment. We construct paleobathymetry grids for oceanic and stretched continental crust including a correction for dynamic topography model “M7” (Müller et al., 2018a) using pyBacktrack version 1.4 (Müller et al., 2018b) and extract paleo–water depths for each drill site.

The frequency of deep-sea hiatus occurrence has fluctuated throughout the Cenozoic (Fig. 2A; Video S1), driven by regional as well as larger-scale changes in ocean circulation and sediment redistribution. The vast majority of hiatuses (72%) are <5 m.y. in duration, 23% of hiatuses are between 5 and 20 m.y. long, and 5% of hiatuses are longer than 20 m.y. (Fig. 2B). The paleo–water depth at which the hiatuses formed show a trimodal distribution—shallow (<2000 m), intermediate (2000–3000 m), and deep (>3500 m) (Fig. 2C). While shallow and intermediate hiatuses fluctuate in frequency throughout the Cenozoic, deep hiatuses show a marked increase only during the mid- to late Miocene (Fig. 2A). Only a small number of deep hiatuses occur below the CCD during the Eocene and the mid- to late Miocene, when the CCD was at its shallowest at ~3000–3500 m (Fig. 2D). This suggests that most hiatuses longer than ~0.2 m.y. are the result of mechanical erosion and transfer of sediments by currents.

In order to investigate the complex history of hiatus occurrence, we focus on hiatus frequency as a function of paleo–water depths (Fig. 2C) in five ocean basins with sufficient data coverage (Fig. 3) spanning times of major oceanographic change. Sites in the Southern Ocean have been assigned to the South Atlantic, the South Pacific, or the Indian Ocean due to paucity of data in that region. We reconstruct the location of hiatuses and non-hiatuses in a paleobathymetric context (Video S1), highlighting hiatuses that likely formed due to carbonate dissolution below the CCD (Fig. 4).

In the Paleocene, most of the hiatuses occur on topographic highs in all ocean basins (Fig. 4A) at depths above the regional CCD (Fig. 3), which is unlikely to have been shallower than 3 km (Fig. 2D). These hiatuses are significantly longer (>>1 m.y.) than the relatively brief late Paleocene carbonate dissolution event that has been linked to changes in Pacific Ocean circulation (Hancock and Dickens, 2006). The Paleocene was warm (Fig. 3A; Westerhold et al., 2020), with deep water forming in the North Atlantic and Southern Ocean (Corfield and Norris, 1996) and the South and North Pacific (Thomas et al., 2008). The thermohaline circulation driven by these sites of deep-water formation most likely caused the long-lived hiatuses we observe at this time, especially at sites located on marginal and oceanic plateaus acting as obstacles to deep-water flow and hence susceptible to sustained erosion by currents. Sparse data for this period preclude an assessment of potential sites of deposition associated with the hiatuses. However, topographic features such as seamounts and plateaus are known to impinge as well as enhance current speeds (Rebesco et al., 2014), which is supported by the regional co-occurrence of hiatuses and conformities throughout the ocean basins (Fig. 3).

A hiatus peak in the South Atlantic in the mid- to late Eocene (ca. 43–39 Ma) (Fig. 3C) is marked by the appearance of hiatuses at intermediate paleo–water depths around the Rio Grande Rise–Walvis Ridge region (Fig. 4B; Video S1). The CCD in the central South Atlantic was substantially shallower (at ~3.5 km) than in the rest of the Atlantic at 43 Ma (Fig. 2D), suggesting that the Rio Grande Rise–Walvis Ridge hiatuses were most likely caused by carbonate dissolution, possibly shifting carbonate deposition to regions of deeper CCD in the ocean basin. However, it is impossible to establish a carbonate mass balance from the marine sedimentary record and CCD fluctuations alone because carbonate abundance cannot be used to uniquely infer causal mechanisms of deep- to shallow-marine carbonate fractionation (Boss and Wilkinson, 1991).

Widespread hiatuses occur in the South Atlantic at depths of 2000–3500 m and in the South Pacific and southern Indian Ocean at depths of <2000 m (Figs. 3C, 3E, 3F, and 4C) between ca. 34 Ma and 30 Ma. This hiatus peak is accompanied by the appearance of giant contourite drifts in the South Atlantic (Fig. 4C; Video S1) such as those along the Argentine continental margin (Hernández-Molina et al., 2010). The Eocene-Oligocene transition at ca. 34 Ma marks a dramatic shift from a Warmhouse to a Coolhouse climate (Fig. 3A; Westerhold et al., 2020) and a deepening of the CCD in all ocean basins, notably in the North Pacific (Fig. 2D) where hiatuses are relatively uncommon at this time. Most of the Oligocene hiatuses are likely the result of major changes in ocean circulation triggered by the combined widening and deepening of the Drake Passage (Eagles and Jokat, 2014) and the opening of the deep Tasman Gateway connecting the South Pacific and Indian Ocean at ca. 33.5 Ma (Scher et al., 2015). In a recent ice-sheet climate simulation, the opening of these gateways together with the onset of Antarctic glaciation results in increased atmospheric pressure gradients and westerly winds ~60°S, cooling surface waters, and intensifying Antarctic deep-water formation (Kennedy-Asser et al., 2019). Together, these modeled changes predict the onset of the modern Antarctic Circumpolar Current (ACC) at ca. 30 Ma, supporting a previous inference by Scher et al. (2015) based on the Southern Ocean neodymium isotope record. The enhanced overturning circulation was most pronounced in the Southern Hemisphere, as reflected in our observed increase in the frequency of South Atlantic, Indian, and South Pacific ocean hiatuses (Figs. 3C, 3E, and 3F) and the appearance of widespread contourite drifts in the South Atlantic (Hernández-Molina et al., 2010).

The beginning of the Miocene is marked by an increase of hiatuses at intermediate water depths in the South Atlantic (Fig. 3C) and at intermediate and shallow depths in the North Atlantic (Fig. 3B). These are accompanied by the initiation of contourites in the Norwegian-Greenland Sea, the equatorial Atlantic, and the Scotia Sea (Fig. 4D; Video S1), representing nearby sites of deposition of eroded material. We interpret this event as the initiation of a proto–Atlantic Meridional Overturning Circulation (AMOC) driven by the complete opening of the deep Drake Passage at ca. 23 Ma (Eagles and Jokat, 2014) and the progressive closure of the Tethys seaway at 20 Ma (Bialik et al., 2019), which has been shown to enhance the ACC and proto-AMOC (Hamon et al., 2013) based on ocean models. In addition, the early Miocene deepening of the Fram Strait and the Greenland-Scotland Ridge connected the Arctic to the northeastern Atlantic as another critical element for developing the AMOC (Straume et al., 2020). This corresponds to the appearance of new hiatuses and contourite drifts (Video S1) in the Norwegian-Greenland Sea, such as the Eirik Drift (Fig. 4D), which was initiated around this time by Northern Component Water (Müller-Michaelis et al., 2013). A reduction in the frequency of hiatuses in the North Atlantic (Fig. 3B) coincides with the plume-driven uplift of the Greenland-Scotland Ridge between 18 Ma and 15 Ma (Straume et al., 2020) causing a weakening of North Atlantic deep-water formation. The Greenland-Scotland Ridge starts subsiding again after 15 Ma while the Fram Strait becomes fully open (Straume et al., 2020), reinvigorating AMOC and increasing the frequency of hiatuses throughout the Atlantic (Figs. 3B and 3C).

Hiatus peaks since ca. 13 Ma display a distinct decreasing global trend in frequencies from 35% to ~10% in the Quaternary (Fig. 2A), tracking post–Miocene Climate Optimum global cooling (Fig. 3A). A similar decreasing trend was noted by Spencer-Cervato (1998) but only for the last 5 m.y. and which remained unexplained in terms of paleoceanographic changes. A small decrease in hiatuses may be explained by a 1 km deepening of the CCD since the late Miocene, reducing the maximum number of deep-water hiatuses such as those in the South Atlantic caused by carbonate dissolution at 13 Ma (Figs. 2D and 4E). However, the vast majority of hiatuses already occurred above the CCD before this deepening (Fig. 2D) and only increase slightly during the Miocene carbonate crash (Fig. 2D). We suggest that the decline in hiatus frequency is more likely due to a slowing of abyssal circulation since the mid-Miocene and a reduction in interoceanic deep-water flow rates as indicated by a range of proxy records including magnetic fabric analysis of contourite drifts (Hassold et al., 2009), effectively slowing bottom-current speeds and seafloor erosion. These observations are supported by ocean circulation models which suggest that deep-ocean ventilation is more vigorous in warm than in cold climates (de Boer et al., 2007). The dependence of density on temperature relative to salinity is increased at higher temperatures, reducing stratification upon warming by diminishing polar freshwater stabilization, contributing to increased convection and deep-water formation (de Boer et al., 2007). A second key process in driving the speed of the deep global ocean circulation is the intensity of surface winds, which was found to have increased with global warming over the past two decades, but this effect is difficult to validate in deep time (Hu et al., 2020). The late Miocene to recent trend of decreasing hiatus frequency we observe provides a missing observational link, given that a waning vigor of intermediate and deep circulation during global cooling would result in decreasing hiatus occurrence.

Our analysis of Cenozoic deep-sea hiatuses in a tectonic and paleobathymetric framework illustrates that the hiatus record can be used as a proxy for the vigor of deep-ocean circulation and for tracking paleoceanographic effects of the opening of key gateways. Our synthesis of deep-sea hiatuses could be used for tracking the fate of deep-sea sediments and for ground-truthing deep-ocean circulation models.

This manuscript benefitted hugely from reviews by Mitchell Lyle, Philip Sexton, Ted Moore, and Editor Gerald Dickens. We thank John Cannon and Xiaodong Qin for technical support. This research was supported by the Australian Research Council Future Fellowship grant FT190100829 to A.D. and by AuScope. All figures were made using Generic Mapping Tools (GMT) version 6.1.