On the duration of the Paleocene-Eocene thermal maximum (PETM)



[1] The Paleocene-Eocene thermal maximum (PETM) is one of the best known examples of a transient climate perturbation, associated with a brief, but intense, interval of global warming and a massive perturbation of the global carbon cycle from injection of isotopically light carbon into the ocean-atmosphere system. One key to quantifying the mass of carbon released, identifying the source(s), and understanding the ultimate fate of this carbon is to develop high-resolution age models. Two independent strategies have been employed, cycle stratigraphy and analysis of extraterrestrial helium (HeET), both of which were first tested on Ocean Drilling Program (ODP) Site 690. These two methods are in agreement for the onset of the PETM and initial recovery, or the clay layer (“main body”), but seem to differ in the final recovery phase of the event above the clay layer, where the carbonate contents rise and carbon isotope values return toward background values. Here we present a state-of-the-art age model for the PETM derived from a new orbital chronology developed with cycle stratigraphic records from sites drilled during ODP Leg 208 (Walvis Ridge, Southeastern Atlantic) integrated with published records from Site 690 (Weddell Sea, Southern Ocean, ODP Leg 113). During Leg 208, five Paleocene-Eocene (P-E) boundary sections (Sites 1262 to 1267) were recovered in multiple holes over a depth transect of more than 2200 m at the Walvis Ridge, yielding the first stratigraphically complete P-E deep-sea sequence with moderate to relatively high sedimentation rates (1 to 3 cm/ka, where “a” is years). A detailed chronology was developed with nondestructive X-ray fluorescence (XRF) core scanning records on the scale of precession cycles, with a total duration of the PETM now estimated to be ∼170 ka. The revised cycle stratigraphic record confirms original estimates for the duration of the onset and initial recovery but suggests a new duration for the final recovery that is intermediate to the previous estimates by cycle stratigraphy and HeET.

1. Introduction

[2] The Paleocene Eocene thermal maximum (PETM) is one of the most abrupt and transient climatic events documented in the geologic record [e.g., Zachos et al., 2001, 2005]. This event was associated with pronounced warming of the oceans and atmosphere, changes in ocean chemistry, and reorganization of the global carbon cycle [Kennett and Stott, 1991; Koch et al., 1992; Thomas et al., 2002; Zachos et al., 2003, 2005; Tripati and Elderfield, 2005; Sluijs et al., 2006]. Warming of deep waters and subsequent oxygen deficiency may have been responsible for extinction of 30–50% of deep-sea benthic foraminiferal species [Thomas and Shackleton, 1996] and planktonic biota were affected by changes in surface water habitats [e.g., Kelly et al., 1996; Bralower et al., 2002; Kelly, 2002; Raffi et al., 2005; Gibbs et al., 2006a, 2006b]; global warming also may have led to a pulse of speciation or migration among mammalian groups [e.g., Koch et al., 1992, Bowen et al., 2001; Gingerich, 2003]. The PETM corresponds to a significant (∼3.5–4.5‰) negative carbon isotope excursion (CIE) recorded in marine and terrestrial sections [e.g., Kennett and Stott, 1991; Koch et al., 1992; Bralower et al., 1997; Zachos et al., 2004, 2005; Schouten et al., 2007]. The source and triggering mechanism of this event are still the focus of much debate [e.g., Lourens et al., 2005; Sluijs et al., 2007; Storey et al., 2007]. An orbital trigger for the PETM and similar (but less severe) events has been suggested [Lourens et al., 2005], but the specific orbital parameter association is still not completely resolved [Westerhold et al., 2007]. Other mechanisms that might explain the abruptness of the CIE include the input of methane into the ocean and atmosphere from the dissociation of methane hydrates in continental margin sediments or from the cracking of coal during rifting of the northern North Atlantic Ocean [Dickens et al., 1995, 1997; Svensen et al., 2004].

[3] Identifying potential triggering mechanisms for the PETM, as well as understanding the relationship between forcing and consequences requires a very precise and high-resolution chronology. For example, quantifying the climate sensitivity requires robust estimates of the mass of carbon released, and hence the rate of the CIE. Until recently, however, estimates of the absolute age of the onset and the duration of the event were poorly constrained, varying between 54.88 and 55.50 Ma, and 100 and 250 ka, respectively [e.g., Kennett and Stott, 1991; Koch et al., 1992; Aubry et al., 1996; Röhl and Abrams, 2000; Röhl et al., 2000; Farley and Eltgroth, 2003; Giusberti et al., 2007]. By using an astronomically calibrated but floating timescale, the age of the onset (54.93 to 54.98 Ma) and the duration (150 to 220 ka) of the CIE were initially determined at Ocean Drilling Program (ODP) Site 1051 [Norris and Röhl, 1999] then refined using combined records from Sites 690 and 1051 [Röhl et al., 2000]. However, because the onset of the PETM in pelagic sequences is marked by a pronounced dissolution layer or condensed interval and the recovery by a lithologically uniform carbonate-rich interval, an alternative constant flux age model was developed [Farley and Eltgroth, 2003]. This model is based on the concentrations of extraterrestrial He (3HeET) and the assumption that the flux of this isotope to the Earth remained constant during the PETM. Both age models are in agreement for the duration of the main body of the PETM (70–80 ka for the “core”, the onset, peak, and initial recovery phase (rapid rise in δ13C, but low carbonate; here termed phase 1)), but diverge for the final recovery phase of the CIE (slow rise in δ13C, high carbonate; here termed phase II), with orbital age models producing 140 ka for this interval and He age models 30 ka. Identification of cycles in the Ca (or Fe) records in the recovery interval of the Site 690 section is complicated due to the high and uniform carbonate content of the sediments.

[4] A new era in Cenozoic paleoceanography was launched with the recovery of Paleogene sediments in multisite depth transects during Ocean Drilling Program Legs 198 (Shatsky Rise, Pacific Ocean [Bralower et al., 2002; Westerhold and Röhl, 2006]) and 208 (Walvis Ridge, Southeast Atlantic Ocean [Zachos et al., 2004]). These expeditions yielded the first high-quality, stratigraphically complete sedimentary sequences of the early Paleogene, recovered in offset, multiple-hole sites. The lithologic and geochemical records generated from these cores exhibit the highly cyclic nature of early Paleogene climate, while also demonstrating that the early Eocene Greenhouse World was punctuated by multiple transient global warming events, or hyperthermals [Thomas et al., 2000; Zachos et al., 2004]. The occurrence of multiple hyperthermals within the late Paleocene–early Eocene suggests a repeated trigger as their cause. Recently, X-ray fluorescence (XRF) core scanning records from ODP Leg 208 sites and from ODP Site 1051 spanning a ∼4.3 million year interval of the late Paleocene to early Eocene were used to establish a longer time series and to develop a robust and improved chronology of magnetochrons [Westerhold et al., 2007] which is consistent with records from the Bighorn Basin [Wing et al., 2000; Clyde et al., 2007].

[5] One of the obstacles to developing age models for PETM sections is providing a exact definition of the termination of the CIE on a global scale, e.g., at Site 690, the location of the termination is somewhat subjective because of the asymptotic shape of the CIE. In addition, the low signal-to-noise ratio of the XRF Ca concentrations in this high-carbonate interval has made cycle extraction difficult and somewhat subjective.

[6] Here we develop a revised chronology for the PETM using high-resolution geochemical data from the ODP Leg 208 depth transect in combination with new Barium (Ba) XRF intensity data of the expanded section at ODP Site 690 from the Weddell Sea, Southern Ocean (Figure 1). The Barium (Ba) records, in combination with Fe, Ca, and carbon isotope data from the Leg 208 sites and Site 690, show similar patterns that allow for refinement of correlation and age calibrations. These new data provide much better constraints on the durations of each phase of the CIE, particularly the recovery phases (I and II). These records will also allow for a more accurate recalibration of the He isotope chronology from Site 690 [Farley and Eltgroth, 2003]. Moreover, we propose that the definition of the termination of the CIE be based on a combination of cyclostratigraphic proxies derived from XRF scanner and other methods rather than carbon isotopes which gradually become uniform, thus making it difficult to define a globally recognizable termination point for the recovery.

Figure 1.

Location of ODP Site 690 in the Weddell Sea, Southern Ocean, and ODP Leg 208 Sites 1262 to 1267 on the Walvis Ridge, South Atlantic. For comparison, ODP Site 1051 located on the Blake Nose in the Western North Atlantic, the Bighorn Basin section in Wyoming, USA, and the Forada section in northern Italy are also shown.

2. Material and Methods

[7] ODP Leg 208 recovered stratigraphically complete and undisturbed PETM sections at four sites between 2.7 and 4.8 km water depth (Sites 1262, 1263, 1266, 1267 [Zachos et al., 2005]). At each site, the PETM sequence is characterized by an abrupt transition from underlying carbonate rich ooze to a dark red “clay layer” which then grades upward into ooze. The most expanded PETM succession was recovered at the shallowest site, Site 1263. All Leg 208 sites are condensed during the PETM because of severe carbonate dissolution [Zachos et al., 2005] with CaCO3 content <1 wt% in the clay layers, compared to >80 and 90 wt% in the underlying and overlying oozes. The thickness of the clay layers increases with water depth, from 5 cm at the shallowest Site 1263 (2717 m; paleodepth ∼1500 m) to 35 cm in the deepest Site 1262 (4755 m; paleodepth ∼3600 m) [Zachos et al., 2005]. We scanned longer core sections covering the late Paleocene-early Eocene interval from ODP Sites 1262 (27°11.150S, 1°34.620E) and 1263 (28°31.980S, 2°46.770E), as well as short (1 to 2 m) intervals spanning the PETM at Sites 1265 (28°50.10′S, 2°38.35′E, 3083 m), 1266 (28°32.55′S, 2°20.61′E, 3798 m), and 1267 (28°5.88′S, 1°42.66′E, 4355 m) and acquired XRF Fe, Ca, and Ba data. In addition, we collected new XRF Ba data from Core ODP 690B-19H, but also XRF Fe, Ba and Ca data for longer core sections (160 to 176 meters below seafloor, mbsf) covering the late Paleocene–early Eocene interval (Leg 113, Maud Rise, 65°9.629′S; 1°12.296′E, 2914 m; Figure 1) [Barker et al., 1988]. Upper Paleocene sediments at ODP Site 690 also show oscillations of CaCO3 and clay content as expressed as white to pale brown banding, similar to the cyclicity in the Leg 208 sections.

[8] Relatively fast nondestructive core logging methods enable continuous measurements at much finer scales (down to millimeter scale) than are practical for discrete sampling methods [Röhl and Abrams, 2000]. The X-ray fluorescence (XRF) core scanner acquires bulk-sediment chemical data from split core surfaces [Richter et al., 2006; Tjallingii et al., 2007]. Although elemental intensities are dominantly proportional to concentration, they are also influenced by the energy level of the X-ray source, the count time, and the physical properties of the sediment [Röhl and Abrams, 2000]. Ba data were collected every 1 cm down-core over a 1 cm2 area with a generator setting of 50 kV and a sampling time of 30 seconds directly at the split core surface of the archive half with XRF Core Scanner II at the University of Bremen. The split core surface was covered with a 4 μm thin SPEXCerti Prep Ultralene® foil to avoid contamination of the XRF measurement unit and desiccation of the sediment.

3. Results and Discussion

[9] All records (Fe, Ca, Ba) show distinctive and correlative elemental patterns in the clay layer of the basal CIE, despite the condensed nature of this interval. The Fe cyclicity at Site 690 disappears in the upper half of the event where sediments are characterized by uniformly high (>85%) CaCO3 contents [Röhl et al., 2000], whereas at Site 1263 the Fe records still show distinct variations (Figure 2). The Ba intensity data at Site 690 almost perfectly correlate to the Ba concentration and accumulation record of Bains et al. [2000], who interpret their data as biogenic barium indicative of export paleoproductivity (Figure 2). Dickens et al. [2003], on the other hand, noted that changes in the Ba and barite fluxes, if global, must be largely a consequence of changes in the supply of Ba to the ocean, either from rivers or methane hydrates. Moreover, some of the total Ba in sediments is likely recycled during early diagenesis: adsorbed on mineral surfaces, co-precipitated with Fe-Mn oxyhydroxides or precipitated as barite [Paytan and Kastner, 1996].

Figure 2.

Compilation of Fe (red, counts per second, cps; total counts, area), Ca (blue, counts per second, cps; total counts, area), Ba XRF data (green, total counts), Ba accumulation rates [from Bains et al., 2000], and bulk carbon data for Sites 690 [from Bains et al., 1999] and 1263 [from Zachos et al., 2005] across the PETM interval. Numbers indicate precession cycle assignments and letters indicate horizons as identified by Zachos et al. [2005]. Please note that precession cycles 6 to 8 can be identified in the Ba curves, whereas the Fe shows less clear and the Ca records exhibits no clear cyclicity at all for this interval (recovery phase of the PETM).

3.1. Interpretation of New High-Resolution XRF Scanning Ba Records From the South Atlantic and Southern Oceans

[10] The Leg 208 data along with the new Site 690 data both complement and improve upon the original Site 690 records of Röhl et al. [2000] which was compromised by several factors. First, the records were generated with a first generation XRF Core Scanner [Röhl and Abrams, 2000], which had a restricted elemental range between K and Sr. As a consequence, clear identification of cycles in the Ca and Fe records in the phase II of the recovery interval was complicated due to the very high carbonate content of the sediments and the low signal-to-noise ratios [Röhl et al., 2000; Kelly et al., 2005]. The addition of Ba yields a record with a much higher signal-to-noise ratio than Ca and Fe. Because the total Ba analyzed, potentially consists of both marine sourced Ba in barite (productivity), as well as Ba derived from terrigenous sources, it is most useful in intervals in which the carbonate content is either exceptionally high (very low Fe intensities) or low (extremely low Ca intensities). Second, only a single core was originally analyzed at Site 690 and thus it was not possible to undertake any further detailed time series analysis for deciphering longer-term trends like modulation of precession cycles by orbital eccentricity. The long continuous cores from Leg 208, though of lower resolution, are continuous over millions of years and thus provide the necessary framework for more precise identification of cycle periodicity.

[11] In general, it appears that sedimentary Ba content is not significantly affected by changes in redox conditions as minor changes in sulfate concentration are not an issue [Torres et al., 1996; Dickens et al., 2003]. Thus the XRF Ba data provide a means of fine-tuning the high-resolution correlation between individual holes of each site (auxiliary material Figure S2 and Tables S1 and S2) which were originally defined using mainly magnetic susceptibility data [Zachos et al., 2004]. The new site-to-site correlation is straightforward as the Ba intensity curves show a high signal-to-noise ratio. It is particularly effective in the high carbonate recovery interval where the Ca and Fe cycles are somewhat less prominent. The high-resolution Ba data provide a more refined inter-hole correlation than can be achieved with bulk carbon isotope data alone, especially in the asymptotic part of the recovery interval (auxiliary material Figure S2 and Table S2).

[12] Though Ba concentrations should primarily reflect the flux of Ba to the seafloor which tends to be dominated by barite, extreme changes in redox conditions may also affect concentrations. The spike in Ba at the base of the CIE, for example, might reflect a geochemical front that lead to diagenetic barite precipitation [e.g., Dickens et al., 2003] rather than original biogenic Ba concentration as reducing conditions set in during the PETM. An investigation of both excess and biogenic barium data obtained from discrete samples from these sites supports this interpretation [Chun et al., 2006]. Regardless, the pattern can be used to assess relative completeness of individual sections. For example, the lack of this spike at the base of the clay layer at Site 1265 (auxiliary material Figure S1), confirms the presence of an unconformity that truncates the CIE at this site [Zachos et al., 2005].

3.2. On the Duration of the PETM

[13] We used the high-resolution Ba based site to site correlations to assess the tempo of sedimentary cycles through the PETM interval (Figure 2). This included an analysis of longer-term trends of cyclic patterns, e.g., modulation of the precession cycles by eccentricity within Chron C24r (see Westerhold et al. [2007] for more details) (Figure 3). Sites 690 and 1051 were previously correlated using precession cycles in Fe- and Ca-intensity data along with high-resolution stable isotope records [Röhl et al., 2000]. For Site 690 ∼ 11 precessional cycles were identified between the base of the CIE and the point where δ13C reaches post excursion values; 4 cycles lie within the “core” of the δ13C excursion from the initial decrease to the level where values begin to rebound, and another 7 cycles lie within the recovery phase. The duration of the entire CIE was thus estimated at ∼220 ka [Röhl et al., 2000].

Figure 3.

Compilation of Fe (red, total counts), Ba intensity (green, total counts), and bulk carbon data (Vienna Peedee belemnite, vPDB) for the late Paleocene–early Eocene interval of Sites 690 [from Cramer et al., 2003] and 1262 and 1263 [from Zachos et al., 2005]. 1, top Fasciculithus spp. [Westerhold et al., 2007]; 2, decrease Fasciculithus spp. [Aubry et al., 1996].

[14] The large number of precession cycles in the upper recovery interval at Site 690 (and implied low sedimentation rate), however, seemed inconsistent with the shift to high carbonate content which would indicate an acceleration of accumulation rates. This suspicion was reinforced by the He isotope chronology developed for Site 690 which suggested that the upper carbonate rich recovery interval was relatively short compared to the underlying dissolution and lower recovery intervals yielding a total duration for the CIE of about 120 ka [Farley and Eltgroth, 2003]. The overestimation of cycles was clearly a consequence of the low signal-to-noise ratios in the Ca records. However, the He record also has a high degree of uncertainty over this interval, in part, because of the relatively low number of samples [Farley and Eltgroth, 2003], and also because of possible errors in sedimentation rates based on magnetochron durations [Westerhold et al., 2007]. As a consequence, the He model likely underestimates the duration of the carbonate rich layer.

[15] The PETM section from Site 690 remains one of the most expanded and intensively studied reference sections for the CIE. Yet, location of the termination of the recovery phase of the PETM is somewhat subjective because of the asymptotic shape of the CIE. A conservative placement of the top of the recovery at 167.10 mbsf [Röhl et al., 2000] (Figure 2) lies at the top of ODP-113-690B-19H [Kelly, 2002], which is overlain by a coring gap (Figures 2 and 3) rendering this site less than ideal for defining the termination of the event. Fortunately, a continuous section is available at Site 1263 where multiple holes were drilled across the boundary. Thus, using new Ba based cycle constraints on Sites 690 and 1263, we reassessed the duration of the entire PETM. The termination of the CIE at Site 1263 and the “reference section” at Site 690 were defined (Tables 1 and 2), by identifying an inflection point in the bulk δ13C curve. Although the inflection is minor, it appears to lie at a similar level from a cyclostratigraphic point of view [Zachos et al., 2005], in the middle of a precession cycle. The Ba record from Site 1263 and for Site 690 show 5 cycles within the clay layer of the CIE and 3.5 cycles in the recovery interval phase II (Figure 2), which suggests that the 7 cycles in the Ca record from the recovery interval phase II at Site 690 as described by Röhl et al. [2000] are probably half precession cycles. We assigned the precession cycle to the modern mean precession period of 21 ka (average of the 19 and 23 ka precessional bands [Herbert et al., 1995; Westerhold et al., 2007]), resulting in a new astronomically calibrated estimate of ∼170 ka for the duration of the PETM.

Table 1. Carbon Isotope and Ba Data Tie Points From ODP Site 690 Used for Correlation to the Leg 208 P-E Boundary Sections
Tie Points690, mbsf1263, rmcd
Fe intensity tie point cycle 8167.20332.79
Onset CIE170.64335.27
Fe intensity tie point cycle -1170.99335.54
Fe intensity tie point cycle -2171.40336.04
Table 2. Carbon Isotope Tie Points From ODP Site 690 and Assigned Ages Used for Correlation and Dating the ODP Leg 208 PETM Sections
Tie Pointsa690, mbsf690 Age, ±ka (Röhl00)b690 Age, ±ka (F&E03)b690 Age, ±ka (This Study)
B-172.81−125.00 −102.53

[16] Our estimated 170-ka duration for the PETM can be further evaluated by considering it within the context of longer-term cyclostratigraphic trends of the upper Paleocene–lower Eocene sediments at Sites 690, 1262 and 1263. Detailed time series analysis of the records spanning magnetochrons C25n and C24r were undertaken by Westerhold et al. [2007] indicating that Chrons C24r and C25n contain 148 and 23 precession cycles, respectively. The eccentricity cycles (405, 100 ka) were also extracted by filtering. This new Leg 208 cyclostratigraphic framework provided the context for assessing our proposed cyclostratigraphy for the PETM clay layer. Starting at the PETM (Figure 3) we identify the same number of precession cycles in Site 690 and Leg 208 sites with respect to the eccentricity related modulation pattern in the Fe and Ba precession cycles. For example, precession cycles 11 to 15 in both Site 1262 and 690 can be correlated to a short eccentricity cycle (“3” [see Westerhold et al., 2007, Figure 4]) above the PETM. The congruent modulation pattern of Leg 208 and Site 690 precession related sedimentary cycles lead to consistent stratigraphic features which are global in nature. In particular precession cycles 16 and 17 (Figure 3) located in a pronounced eccentricity cycle minimum are prominent features which should also be present in other sections around the globe. The consistency of our cyclostratigraphy is corroborated by biostratigraphic features. For example, the position of the abrupt decrease in abundance of Fasciculithus spp. at Site 690 [Aubry et al., 1996] (cycle 17 in Figure 3) and at Site 1262 [Westerhold et al., 2007] (top of cycle 19 in Figure 3), respectively, is roughly correlative between the sites.

3.3. Sediment Accumulation Rates

[17] The new estimate for the total duration of the PETM (170 ka) is shorter than that derived from the original orbital model and longer than that derived from the He age model (about 90–140 ka) [Farley and Eltgroth, 2003]. How does this affect accumulation rates (auxiliary material Figure S3), particularly that of carbonate through the recovery interval when rates should have accelerated [Dickens et al., 1997; Farley and Eltgroth, 2003; Kelly et al., 2005; Zachos et al., 2005]? To estimate accumulation rates, we used the orbitally tuned, but floating age model from ODP Leg 208 [Westerhold et al., 2007], which provides tuned ages from the termination of Chron 25n to the onset of the CIE (1.113 Ma), and for the total duration of Chron C24r (3.118 Ma). For the duration of the PETM/CIE, we simply adopt our 170 ka estimate. Sediment accumulation rates are estimated for the uppermost Paleocene and lower Eocene at Site 1263 and Site 690 cores using the depths, age model, and durations as described above (Table 3 and Figure 4). For both sites, the sedimentation rates are close to 2 cm/ka before the onset of the CIE, drop dramatically (≤1.0 cm/ka) during the body (clay layer) of the CIE/PETM, and then increase in the recovery interval (3 cm/ka). Although these estimates possess minor uncertainties it is clear that sedimentation rates did increase during phase II of the recovery, but not tenfold as derived from He isotopes.

Figure 4.

Sedimentation rates (top) for Site 1263 and (bottom) for Site 690 plotted versus Fe (red, total counts), Ba (green, total counts) and δ13C (black, Vienna Peedee belemnite, vPDB) data. For Site 690 the sedimentation rates from Röhl et al. [2000] and Farley and Eltgroth [2003] are also shown for comparison.

Table 3. PETM Age Model for Sites 690 and 1262–1267a
CycleDepth mbsf Site 690Depth mcd Site 1262Depth rmcd Site 1263Depth mcd Site 1265Depth mcd Site 1266Depth rmcd Site 1267Relative Age (pre = 21.0 ka) to Cycle 1Relative Age (pre = 21.0 ka) to ELMO (= 1000)Relative Age (pre = 21.0 ka) to Onset CIE
  • a

    Notes: mcd, meters composite depth rmcd, revised meters composite depth. Relative ages are in ka.

  • b

    Correlation of Site 1263 to Site 690.

  • c

    Correlation to Site 1263 by Ba data.

17-136.56328.08no data-227.893362491339
11-138.12no data312.31303.52229.572102617213
10166.24138.38no data313.01304.10229.901892638192
1170.60 335.24c315.88b306.75b231.76b028273
Onset CIE170.64140.12335.27315.91306.78231.79  0
−3171.88140.53-end of splice-232.27−632890−60
−4172.40-- -232.45−842911−81
−5172.80140.85336.86 308.10232.63−1052932−102
−6173.20141.04337.11 308.43232.85−1262953−123
−7173.49141.27337.42 308.80233.10−1472974−144
−8no data141.52337.82 309.13233.35−1682995−165
−9no data141.74338.20 309.55233.65−1893016−186
−10no data141.95338.47 309.87233.88−2103037−207

[18] We also compared our new orbital chronology with one of the more expanded marine sections, the Forada section in northern Italy which was deposited in a hemipelagic, near-continental setting. In this section, the Paleocene succession of limestone-marl couplets is sharply interrupted by a ∼3.30-m-thick unit of clays and marls representing the “clay layer” of the PETM [Giusberti et al., 2007]. The CIE main excursion at Forada is about 3.4 m long and therefore is almost 3 times more expanded than the corresponding interval at ODP Site 690 (Figure 5). Several parameters (hematite, carbonate, δ13C, radiolarians) oscillate in a cyclical fashion and are interpreted to represent precession cycles. The “core” of the excursion interval spans 5 complete cycles (∼105 ka). The 1.4-m thick recovery interval in the Forada section contains six distinct limestone-marl couplets, and is interpreted to represent six precessional cycles with a duration of ∼126 ka [Giusberti et al., 2007], providing a total CIE duration of ∼231 ± 22 ka, 5–10% longer than the estimate of Röhl et al. [2000].

Figure 5.

Latest high-resolution correlation of Site 690 Fe (total counts) and δ13C data (Vienna Peedee belemnite, vPDB) to the Forada section (CaCO3 and δ13C data) in northern Italy (modified from Giusberti et al. [2007]). Letters indicate horizons as identified by Zachos et al. [2005] for Site 690. PFDI, Planktonic Foraminiferal Dissolution [from Luciani et al., 2007]; BEE, Benthic Extinction Event.

[19] For the core of the PETM, only 4 cycles were identified at Site 690 and 1263, while 5 were identified at Forada over the same time interval. This difference is just an artifact of the way in which the cycles are defined, using Fe (or clay) for the ODP records (Fe data reflecting the terrigenous component) (Figures 2, 3, and 4) whereas for the Forada section the cycle count is based on the carbonate record (Figure 5). The hematite record at Forada [Giusberti et al., 2007] also exhibits 4 cycles that vary roughly inversely with carbonate and would be the equivalent (as same number) of the 4 Fe cycles identified at the ODP sites.

[20] With regard to the recovery interval, Giusberti et al. [2007] also noted that the seven precession cycles from the corresponding recovery interval (phases I and II) at Site 690 [Röhl et al., 2000] exceeds that at Forada by one cycle. They correlated cycle by cycle using the numbering scheme for Site 690. Unfortunately, the sample resolution for the Forada record is not high enough, especially in the recovery interval, to precisely correlate details in records for this part of the CIE. However, with our new data we were able to revise the correlation of Forada and the deep-sea sites. One possible additional correlation point is the LO/decrease of Fasciculithus spp. which is associated with precession cycle 16. The pronounced minimum in the eccentricity modulation of precession cycles 16 and 17 aligns closely to the thick limestone bed (marker bed, cycle 16) at Forada (Figure 5) consistent with the proposed detailed correlation within and around the CIE between the sites. If we use this event as a tie point, then the number of precessional cycles at Forada closely matches that of the deep-sea sections.

[21] Given the massive global carbonate dissolution at the onset of the PETM, could we have underestimated the number of precession cycles in the base of the clay layer in the pelagic site? This is possible, though it seems unlikely given that the extent of carbonate erosion probably differed between Maud Rise and Walvis Ridge, and that no pure clay layer was produced at Site 690 [Zeebe and Zachos, 2007]. However, Site 690 does show evidence for a significant decrease in carbonate content. As such, it might be possible that one precession cycle was removed in the middle of the onset of the PETM between cycle −1 and 1 (Figure 2) and that the equivalent of 10 ka is missing. Our study and the cycle identification at Site 690 and correlation to the Leg 208 sites result in an estimate of 5 precession cycles. The Forada section seems to exhibit 6 cycles for the same interval. However, we are confident that the new cycle-based age model is robust as the new Leg 208 records show individual precession cycles of the recovery interval that are perfectly modulated by shorter eccentricity cycles. Therefore we can exclude the possibility that the number of peaks or precession cycles has been underestimated or overestimated. At this point a few words about modulation are warranted (auxiliary material Figure S4): the long term record shows that the PETM is located between a 405-ka maximum and 405-ka minimum [Westerhold et al., 2007, Figure 6]. This constraint demonstrates that the duration of the clay layer of the PETM cannot be longer than the equivalent of 7 precession cycles, which would already be the maximum estimate. If we assume that the clay layer is longer then the equivalent of 7 precession cycles than the 405-ka filter will move out of phase. The Leg 208 cyclostratigraphy clearly demonstrates that the clay layer represents a minimum of 5 and a maximum of 7 precession cycles (auxiliary material Figure S4). In the future, we recommend that sediment models simulate the effect of cyclical sediment flux to determine if the expression of the precession cycle signal can be obscured by a combination of dissolution and bioturbation.

3.4. Timing of the CIE Relative to Orbital Phases

[22] The final point concerns the relationship of the PETM to orbital phases. The question of whether the CIE and other hyperthermals was triggered by the 100 or 405 ka eccentricity cycles has been explored in several papers [Cramer et al., 2003; Lourens et al., 2005] using a variety of pelagic sediment cycle records including those from Walvis Ridge [Lourens et al., 2005]. A recent reevaluation of the longer cyclic records from Walvis Ridge indicates that the CIE did not occur during a time of maximal variability in insolation (405-ka maximum) as suggested previously, but with a decreasing segment of a 405-ka eccentricity cycle [Westerhold et al., 2007, Figure 6b]. Nevertheless, there seems to be a relation to a 100-ka eccentricity cycle maximum [Westerhold et al., 2007, Figure 6c]: the amplitude modulation of the short eccentricity cycle clearly visible in Figure 3 (with maxima of the 100-ka cycles at position of precession cycles −10, −5, 2, 8, 13, 20) indicates that the onset of the CIE occurred halfway (after precession cycle −2 and −1) within the short eccentricity cycle comprising precession cycles −2 through 3, which represents a period of relatively high-amplitude climate variation and could be a mechanism for forcing climate change on 10-ka timescales.

4. Conclusions and Implications

[23] High resolution Fe and Ba intensity records were generated for the P-E boundary intervals at ODP Site 690 on Maud Rise and Sites 1262 and 1263 on Walvis Ridge. The P-E boundary section at Site 690 is the one of the most expanded in the deep sea, while the sections at Sites 1262 and 1263 lie within the longest, most stratigraphically continuous and astronomically tuned sequence of marine sediments spanning the upper Paleocene and lower Eocene. The new data highlight significant features and striking changes in response to this brief episode of extreme warmth. The cycle-to-cycle correlation between ODP Sites 1262 and 1263 in the context of the astronomically calibrated longer-term record [Westerhold et al., 2007] indicates that the cycles have durations in the precessional frequency band.

[24] The lower portion of the PETM which includes the dissolution phase and lower recovery interval contains 5 precession cycles, while the upper recovery interval contains 3.5 cycles. The total duration of the PETM is now estimated to be ∼170 ka, roughly mid-way between previous estimates based on cycle stratigraphy and He isotopes. One key implication is that the carbonate rich upper recovery layer represents a time of enhanced accumulation rates consistent with the notion of a supersaturated ocean [Farley and Eltgroth, 2003; Kelly et al., 2005].


[25] We thank two reviewers for their comments. This research used samples and/or data provided by the Ocean Drilling Program (ODP). ODP is sponsored by the U.S. National Science Foundation (NSF) and participating countries under management of Joint Oceanographic Institutions (JOI), Inc. Funding for this research was provided by the Deutsche Forschungsgemeinschaft (DFG) to U. Röhl and T. Westerhold and by NSF grants EAR-0120727 and 0628486 and ACS-PRF grant 42705-AC8 to T. J. Bralower and to J. C. Zachos. We thank Gar Esmay (ECR) for assistance with shipping Site 690 cores to Bremen, Alex Wülbers (BCR) for logistical assistance, and Vera Lukies (Bremen) for support in the XRF Core Scanners Lab. The complete data set presented in this paper is available online in the WDC-MARE PANGAEA database at http://www.pangaea.de (doi:10.1594/PANGAEA.667443).