Clear as mud: Clinoform progradation and expanded records of the Paleocene-Eocene Thermal Maximum

The Paleocene-Eocene Thermal Maximum (PETM; 56 Ma) and attendant carbon isotope excursion (CIE) represent the largest warming event and carbon cycle perturbation of the Cenozoic. Global temperatures rose by 4–8 °C (e.g., Kennett and Stott, 1991; Zachos et al., 2003) while δ¹³C values decreased by 2‰–4‰ in marine sections (foraminiferal and organic records) and by as much as 7.6‰ in the terrestrial realm (plant lipids) (see McInerney and Wing [2011] for a full review of CIE proxies). Understanding the trigger, rate, and timing of this light-carbon injection is hindered, in part, because most marine CIE-PETM sections are thin (<1 m in 200 k.y.) and thus cannot resolve geologically rapid events (<10 k.y.).

The general structure of the CIE is similar globally, beginning with a sharp drop in δ¹³C values that is used to correlate the base of the Eocene (Paleocene-Eocene boundary) from its stratotype in Dababiya, Egypt (Aubry et al., 2007). In deep-sea sections, the δ¹³C decrease is sharp (<10 cm), while on continental margin sections, it can be >2 m thick (Figs. 1 –3). A sustained interval of low δ¹³C values (the body of the CIE) followed the δ¹³C decrease and was succeeded by a logarithmic return to near pre-CIE δ¹³C values (the recovery; e.g., Dickens et al., 1997; Röhl et al., 2007). The PETM-CIE lasted ∼200 k.y. from onset to recovery (Dickens et al., 1997; 170 ± 30 k.y. astronomical estimate of Zeebe and Lourens [2019]).

The warming associated with the PETM coincided with rapid input of fine-grained sediments to the continental shelf along the mid-Atlantic U.S. continental margin. Thick PETM deposits have been reported on the Maryland (e.g., Self-Trail et al., 2017) and New Jersey coastal plains (NJCP) (e.g., Cramer et al., 1999), which provide greater temporal resolution of the onset and body of the CIE compared to deep-marine settings. Rapid mud deposition on the mid-Atlantic paleoshelf during the PETM has been dubbed the “Appalachian Amazon” (Kopp et al., 2009), characterized by energetic, mud-laden riverine transport and subaqueous dysoxic deposition (Stassen et al., 2012). In the NJCP, uppermost Paleocene Vincentown Formation sands and silts are conformably overlain by the kaolinite-rich clays of the lowermost Eocene Marlboro Formation (Gibson et al., 2000; Cramer et al., 1999) that preserve the CIE.

We cored these sediments adjacent to NJCP–PETM outcrops (Medford Auger Project [MAP]; 39.86°N, 74.82°W; Fig. 1; described in the Supplemental Material¹) and correlate with PETM sections across a transect recovered in International Ocean Discovery Program (IODP) Leg 174AX sites at Wilson Lake (WL hole B; WL hole A was drilled by the U.S. Geological Survey [USGS]), Millville (MV), and Bass River (BR) (Figs. 2 and 3).

These sites each record different parts of the CIE (onset, decrease, body, and recovery) identified by extensive bio- and chemostratigraphic studies (Cramer et al., 1999; Gibbs et al., 2006; Harris et al., 2010; Stassen et al., 2012, 2015; Wright and Schaller, 2013; Makarova et al., 2017). Proximal sites on our transect preserve a notable “transitional unit” that is completely absent in deep-sea sections. This transitional package of sediments preserves the marked shift in grain size, carbonate content, and carbon-isotope geochemistry that signals the rapid change in facies associated with this hyperthermal. These PETM records are greatly expanded (e.g., the CIE body is >10 m thick at WL and MV; Fig. 3), requiring average sedimentation rates exceeding 12.5 cm/k.y. (with estimates as high as 2 cm/yr; Wright and Schaller, 2013), versus ∼4 cm/k.y. background sedimentation rates during Vincentown Formation deposition (Miller et al., 1997).

We evaluate CIE heterogeneity on the mid-Atlantic paleoshelf transect and correlate using δ¹³C_bulk, percent CaCO₃, percent coarse fraction (%CF; >63 μm), and nannofossil and benthic foraminiferal biostratigraphy across a paleoslope dip profile spanning 45 km (30–50 m of deepening; Makarova et al., 2017). We do not address the trigger of the PETM, focusing instead on the relationship between rapid warming and sediment input. The differential expression of the CIE across the shelf (shape, thickness, and preserved section at each location) provides a framework of relative time that can later be used to assess proposed mechanisms.

In 2016, the MAP continuously cored 10 holes at six closely spaced (<1 km) sites in Medford, New Jersey, targeting updip PETM sections (Fig. 2). This study focuses on two adjacent holes, MAP 3A and MAP 3B (<5 m apart), drilled ∼700 m downdip of the Marlboro Formation outcrop and the IODP Leg 174AX site at Medford (Sugarman et al., 2010), to construct a complete record for site MAP 3 (Fig. 1). A modified depth scale (i.e. meters composite depth; mcd) is applied to adjust for core expansion and is used for correlation between sites MAP 3A and MAP 3B.

We measured δ¹³C_bulk and percent CaCO₃ on a Fisons Optima mass spectrometer with an attached Multiprep device in the Rutgers University (New Jersey, USA) stable isotope lab. Sediments taken from MAP cores were washed through a 63 μm sieve to determine %CF. Grain-size analysis was initially conducted via laser diffraction on a Malvern Mastersizer 3000, and later via traditional pipette methods (see the Supplemental Material).

The cores from site MAP 3 recovered uppermost Paleocene to lowermost Eocene sands and muds from 14.5 to 19.8 m mcd (Fig. 1). The Vincentown Formation (17.4–19.8 m mcd) is a silty sand interpreted to reflect a lower shoreface facies deposited below fair-weather wave base (see the Supplemental Material). The Marlboro Formation (14.5–16.5 m mcd) is a kaolinitic silty clay (mean grain size <2 μm; Fig. 1; Fig. S2 in the Supplemental Material) deposited in a prodelta setting in middle neritic paleodepths (∼30–50 m; see the Supplemental Material). The transitional unit (16.5–17.4 m mcd) between these two formations is defined by rapidly fining-upward sediments (%CF decreases 72% to 2%). This transitional unit also preserves the rapid change in percent CaCO₃ (Fig. 1). At site MAP 3, the transitional unit captures the CIE onset and part of the subsequent CIE decrease. This CIE decrease is extremely sharp in open ocean sites, which, in conjunction with the biostratigraphic correlations, suggests far higher rates of sedimentation on the NJCP: 2.7 m of sediment preserves the CIE onset and initial CIE decrease at hole MAP 3B, versus <10 cm in open-ocean sites.

We hang our cross-shelf correlation on the CIE onset, which is coincident with the initial decrease in percent CaCO₃ (Fig. 2). Our transect shows clear cross-shelf patterns (Fig. 2). The updip MAP and WL sites preserve distinct transitional units and an expanded view of the onset and start of the CIE decrease (Fig. 3). We cannot make quantitative inferences on sedimentation rates, though δ¹³C correlations suggest that the onset and decrease sediments recovered at MAP 3 are expanded compared to those at WL (Fig. S1). However, the data sets we have available—comparison of δ¹³C_bulk records, lithology (uniform silty clay), and benthic assemblages (see below)—argue that site MAP 3 preserves the early part of the CIE decrease and body while the second, more gradual step of the δ¹³C decrease (seen elsewhere on the NJCP; Fig. 3; Fig. S1) is absent.

The transitional unit thins downdip at site MV (Wright and Schaller, 2013; Makarova et al., 2017) and appears to be absent at site BR (Cramer et al., 1999; Fig. 2), where the CIE onset may be diastemic (Stassen et al., 2012). The recovery is apparently completely preserved downdip at site BR, truncated at site MV (Makarova et al., 2017), thin and also incomplete at site WL, and absent updip at site MAP 3 (Fig. 3).

Our correlations are reinforced by biostratigraphy. The CIE body is associated with the Rhomboaster-Discoaster (RD) assemblage in the open ocean (Kahn and Aubry, 2004) and in New Jersey (Harris et al., 2010). At downdip sites (MV, BR), the RD assemblage appears at the base of the CIE body, increases sharply to an acme, and decreases in abundance before disappearing in the recovery (Fig. 3; Harris et al., 2010). We do not have access to quantitative data for site WL, however nannofossil biostratigraphy (compiled for hole A in Stassen et al. [2012, 2015] and hole B by Miller et al. [2017]) places the lowest occurrence of Discoaster araneus (a marker of the RD assemblage) at the base of the CIE body (Fig. 3). Benthic foraminiferal assemblages at site MAP 3 are characterized by small individual specimens (<212 μm) and dominated by Anomalinoides acuta and Ammobaculites midwayensis (Makarova, 2018), indicating equivalence to the CIE decrease or body at site WL (Stassen et al., 2015).

We explain variability of the sediments of the transitional unit and associated CIE onset and decrease across the shelf as the result of a progrational clinoform delta, with thin sigmoidal topsets, thick foresets, and thin bottomsets (Fig. 3). Though the chronology is known only within a relative time scale of several thousand years, the shift in deposition from site MAP to site WL during the CIE onset and decrease resulted in a 9 km seaward progradation (Fig. 3) in ≤ 4 k.y. (using the chronology of Zeebe and Lourens [2019]) and perhaps much faster (using the chronology of Wright and Schaller [2013]).

Evidence for rapid, mud-laden riverine transport and high sedimentation rates includes the lack of bioturbation, physical remnants (vertical sticks and leaves; e.g., Wright and Schaller, 2013), biofacies assemblages (Stassen et al., 2012, 2015), and magnetofossils at least partly indicative of dysoxic environments (Kopp et al., 2009; Wang et al., 2015).

In our model, muds originating from the Appalachians, Piedmont, and coastal plain built individual chronostratigraphic units, each preserving a different “snapshot” of the CIE. The earliest packages of fining-upward sand to mud (transitional unit) and overlying clay (Marlboro Formation) captured the CIE onset and initial part of the δ¹³C decrease (Fig. 3). As accommodation space filled, fluid mud transport drove delta progradation seaward, allowing subsequent clinoform deposition to record the CIE body and recovery (Fig. 3). Meanwhile, updip sections were bypassed and truncated as the seafloor intersected wave base. This produced a regional unconformity capping the Marlboro Formation (Fig. 3; e.g., Gibson et al., 2000). Volume scaling suggests that Amazon shelf–like conditions on the mid-Atlantic coastal plain would have required ∼25% of the modern Amazon sediment flux during the PETM (Kopp et al., 2009).

Inferring paleoenvironmental conditions via stratigraphic correlation is subject to some uncertainty due to spatiotemporal variability in sedimentation rates and autocyclical shifts in depocenters (e.g., Trampush and Hajek, 2017; Foreman and Straub, 2017). However, the biostratigraphic, chemostratigraphic, and lithologic trends demonstrated in this study (Fig. 2) are consistent with our progradational depositional model, whether that trend is predicted to disappear proximally and expand in a downdip direction (CIE body and recovery, percent RD influx [the sediments preserving the appearance of RD]) or, in contrast, disappear distally with expanded sections updip (CIE decrease, low-carbonate zone). We acknowledge that our sedimentary records may not be complete, which could have resulted in an incomplete depiction of a three-dimensional, lobe-shaped geometry on our two-dimensional cross section. However, the consistent patterns observed across the transect (Figs. 2 and 3) support this general pattern of sedimentation.

This rapid progradational mud system explains the relative distribution of biofacies, the variable expression of the CIE, and the lithology on the NJCP. The high sedimentation rates on the mid-Atlantic paleoshelf have been attributed to the warm subtropical PETM climate, analogous to that of the modern Amazon (e.g., Nittrouer et al., 1995) and other high-input rivers (e.g., Mahakam [Indonesia], Ayeyarwady [Myanmar]). Modern subaqueous clinoform deltas are associated with large mud-laden riverine systems around the globe, including the Ganges-Brahmaputra (India and Bangladesh; Kuehl et al., 2005), Fly (Papua New Guinea; Walsh et al., 2004), Yangtze (China; Chen et al., 2000), Mahakam (Storms et al., 2005), and Ayeyarwady (Liu et al., 2020) Rivers. These delta systems all share two critical conditions: high input of muddy riverine sediments, and an energetic tide- or storm-dominated environment to facilitate transport.

In particular, we identify the Holocene Amazon, Mahakam, and Ayeyarwady river systems as important case studies for drawing modern analogs for the Marlboro Formation clinoforms. Observations in these systems detail how muds coalesce and accumulate in topset depocenters under tidal and wave forces and are then episodically transported across the shelf by fluid mud processes (e.g., on the Amazon delta; Kineke et al., 1996). Using these analogs, we invoke rapid fluid mud deposition as the transport mechanism for delta mud clinoforms on the NJCP during the PETM.

Cross-shelf movement of fluid mud requires an energetic transport mechanism (i.e., tides or storms). On the Holocene storm-dominated mid-Atlantic continental shelf, clay sediments are trapped in estuaries, while mud that reaches the shelf is swept into the deep sea by energetic storms (Miller et al., 2014). Though the PETM differed climatologically (lower latitudinal gradients), the high supply of mud accumulated in the shallow embayment, where storms and tides facilitated transport to the foresets and bottomsets, as observed in modern mud-rich systems (Nittrouer et al., 1995; Storms et al., 2005; Liu et al., 2020).

Our PETM depositional model suggests progradation rates similar to those observed in Holocene high-sediment-supply river systems. For example, the Holocene Mahakam delta has prograded ∼60 km into the Makassar Strait over the past 5000 yr (∼12 km/k.y.; Storms et al., 2005), while our model suggests 45 km of progradation during the geologically brief PETM (Fig. 3). Thus, we incorporate modern studies of muddy river systems to evaluate the distribution of these PETM clays. This approach provides a mechanism that explains the variability in the preservation of the CIE in this region and a blueprint for planning future studies.

Geochemical, sedimentological, and biostratigraphic records demonstrate that deposition on the New Jersey paleoshelf during the PETM consisted of prograding depocenters. A strengthened hydrological cycle bolstered riverine transport of muds to the paleoshelf. We explain the variability of the CIE records observed on our cross-shelf transect using a progradational clinoform depositional model supported by cross-shelf correlations of multiproxy lithologic, biostratigraphic, and δ¹³C_bulk data. Proximal (updip) sites record high fluxes of sediments during the CIE onset, which rapidly filled available accommodation space, forcing clinoforms to prograde into deeper water. Our PETM progradational model illuminates the influence of transient warming events on continental shelves and presents a qualitative chronostratigraphic tool for correlating PETM sites across the mid-Atlantic paleoshelf and for selecting future drill-site locations based on CIE target intervals.

We thank the late C. Lombardi, who proposed the possibility of fluid mud deposition on the New Jersey coastal plain during the PETM; the U.S. Geological Survey drillers, particularly the late J. Grey, for their efforts; the International Ocean Discovery Program for samples from Leg 174AX sites; as well as the anonymous reviewers who, alongside Elizabeth Hajek, contributed feedback that greatly enhanced the strength of this paper. This work was supported by U.S. National Science Foundation grants OCE16-57013 (Miller) and OCE14-63759 (Miller and Browning).