DISCUSSION

Several different types of data have been used as indicators of foraminiferal preservation, including CaCO₃ content, B/P ratio, coarse fraction, and fragmentation (e.g., Arrhenius, 1952; Corliss and Honjo, 1981; Thierstein and Roth, 1991; Haug and Tiedemann, 1998). Close examination of these data in the PETM interval at Shatsky Rise and comparison with visual preservation will help establish the most reliable proxy, or combination of proxies, for preservation. Lithology and the CIE allow precise correlation of data between sites. The onset of the PETM is marked by the abrupt onset of the bulk carbonate ¹³C excursion which lies at or just below a sharp lithologic contact between more carbonate rich ooze overlying clay-rich ooze (Figs. F5, F6, F7, F8). In the upper part of the PETM interval, the clay-rich ooze gradually becomes more carbonate rich and ¹³C values of bulk carbonate gradually increase. Whereas no pristine foraminiferal preservation is observed in the PETM interval on Shatsky Rise, there is an abrupt deterioration in preservation at the basal lithologic contact and the base of the CIE at each site. This change includes a substantial increase in pore infilling and sparry calcite in foraminiferal chambers and on muricae (Pls. P1, P2, P3, P4, P5, P6, P7). The quality of preservation recovers gradually above the PETM with a steady decrease in secondary calcite and pore infilling. Preservation generally deteriorates with water depth, with the deepest location (Site 1211) showing evidence for stronger dissolution, indicating more corrosive waters than in the shallower sites. In the interval of poorest preservation at all sites, euhedral calcite blades ~10 Mm in length and blocky rhombs of calcite indicate that initial dissolution led to at least temporary supersaturation of CaCO₃ in pore waters.

Dissolution results in enlarged pores and breakage of foraminiferal tests (e.g., Broecker and Peng, 1982). Changes in fragmentation mirror the poorest preservation in samples near the basal lithologic contact at each site and the overlying gradual recovery of preservation. Although maximum fragmentation occurs at or just above the onset of the CIE and lithologic contact, fragmentation begins to increase slightly below this level at Sites 1209, 1210, and 1212. The sharp lithologic contact indicates little mixing and thus shows that dissolution of foraminifers occurred during early burial.

Fragmentation is typically high at each site for several centimeters above the lithologic contact, corresponding to the interval of poor preservation and greatest dissolution. The thickness of the fragmentation interval decreases with depth. At Site 1209 (2387 meters below sea level [mbsl]), fragmentation is high over an 8-cm interval. Fragmentation is elevated over 7 cm at Site 1210 (2573 mbsl) and 5 cm at Sites 1212 (2681 mbsl) and 1211 (2907 mbsl) (Figs. F5, F6, F7, F8). Background fragmentation levels also increase with depth, ranging from ~20% at Sites 1209 and 1210 to ~30% at Sites 1211 and 1212 (Figs. F5, F6, F7, F8). This also indicates increased dissolution with depth.

Calcium carbonate content has often been used as a proxy for changes in preservation during burial diagenesis (e.g., Arrhenius, 1952). Previous work (Thierstein and Roth, 1991) showed that nannofossils and foraminifers in samples with CaCO₃ contents between 60 and 80 wt% are characterized by little diagenetic overgrowth or dissolution. Samples containing <60 wt% CaCO₃ generally show evidence for dissolution and those >80 wt% often show evidence for secondary calcite overgrowth. Shatsky Rise PETM CaCO₃ contents range from 70 to >95 wt%, which under normal circumstances would cause minor overgrowth during burial. However, predicted PETM CO₂ input should result in a rapid increase in carbonate solubility at the seafloor with corresponding changes in CaCO₃. At all four sites, significant decreases in CaCO₃ content mirror intervals of poor preservation, suggesting dissolution as the main control on CaCO₃ values (Figs. F5, F6, F7, F8). At Sites 1209, 1210, 1212, and 1211, CaCO₃ content decreases by 10% over 4 cm, 7% over 3 cm, 15% over 3 cm, and ~25% over 4 cm, respectively. Although the PETM was a transient event, carbonate production in surface waters likely responded to warming and ocean circulation changes (Thomas et al., 1999; Bralower, 2002; Stoll and Bains, 2003). Consequently, decreased CaCO₃ content may also be a partial result of decreased surface water productivity.

Coarse fraction has often been interpreted as a proxy for foraminiferal preservation (e.g., Haug and Tiedemann, 1998). Increases in coarse fraction are generally interpreted as indicative of better preservation, as foraminifers are thought to be more susceptible to dissolution than nannoplankton (e.g., Schlanger and Douglas, 1974). At Sites 1209 and 1210, however, highest coarse fraction occurs in intervals of poorest visual preservation (Figs. F5, F6). At Site 1212, coarse fraction increases 2–3 cm above the basal contact, within the interval of poor preservation (Fig. F7). Only Site 1211 shows a decrease in coarse fraction in the interval of poor preservation (Fig. F8). Because coarse fraction increases with dissolution at three of the four sites, it is unreliable as a proxy for preservation. High coarse fraction at Sites 1209, 1210, and 1212 may be the result of changing environmental variables such as productivity.

Benthic foraminifers are generally more resistant to dissolution than planktonic foraminifers (e.g., Lipps, 1973; Thomas and Shackleton, 1996; Thomas 1998), so increases in B/P may also be indicative of dissolution. B/P ratios are consistent with preservational trends at Sites 1209, 1211, and 1212 (Figs. F5, F7, F8), but the B/P maximum at Site 1210 does not correspond to any visual change in preservation (Fig. F6). The abrupt B/P peaks at Sites 1209, 1211, and 1212 correspond to the most extreme dissolution intervals and fail to capture the extent of the interval of poor preservation. Benthic foraminifers were relatively rare (<6%) in all samples. Thus, counting finer-size fractions (>38 or >63 µm) may have recovered more benthic foraminifers, providing more robust and detailed B/P curves.

Percent fish debris appears to be a sensitive indicator of dissolution. At all sites, fish debris peaks at the same depth as fragmentation maxima (Figs. F5, F6, F7, F8). Assuming that fish debris is deposited at a constant rate, seafloor dissolution should result in increased concentrations. Percent fish debris also generally increases with depth: maximum fish debris is 4% at Site 1209, 17% at Site 1210, 40% at Site 1212, and 36% at Site 1211 (Figs. F5, F6, F7, F8). This trend indicates higher apparent sedimentation rates at Sites 1209 and 1210, as more carbonate is dissolved in increasingly corrosive bottom waters at the deeper sites. Abrupt short-term increases in pyrite abundance also occur near the basal lithologic contact. Higher amounts of pyrite coincide with maximum values of fragmentation (Figs. F5, F6, F7, F8).

Comparison of intervals of poor visual preservation with all of the possible preservational proxies suggests that foraminiferal fragmentation is the most reliable indicator of preservation in the PETM intervals recovered on Shatsky Rise. Visual inspection suggests that poor preservation corresponds to intervals of lower CaCO₃ content. B/P counts also faithfully record intervals of poorest preservation, but too few benthic foraminifers were recovered to identify the complete interval of increased dissolution. Coarse fraction percent responds to dissolution as expected only at the deepest location, Site 1211.

Regardless of origin, input of massive amounts of CO₂ or CH₄ into the ocean-atmosphere system would lead to a marked shoaling of the lysocline and CCD (e.g., Dickens et al., 1997). The response of the lysocline and CCD is sensitive to the location of carbon input at the onset of the PETM and its distribution through the ocean and atmosphere (e.g., Dickens, 1997; Zachos et al., 2005). In the case of a methane hydrate source, the site of oxidation of CH₄ to CO₂ is critical (Dickens, 2000). If the carbon source is oceanic (either from methane hydrate or submarine volcanism), the response of the lysocline and CCD can help determine whether, and how rapidly, the carbon entered the atmosphere as well as how it was mixed through the ocean basins (e.g., Dickens, 2000, 2001). For example, if methane hydrate is the source of carbon and oxidation took place in the Atlantic, then the lysocline would be expected to shoal more dramatically in that ocean than in the Pacific (Zachos et al., 2005). A similar result would occur if thermohaline circulation patterns during the PETM were similar to present and carbon rapidly entered the atmosphere from the ocean or originated in the atmosphere (i.e., from an extraterrestrial source or subaerial volcanism). However, if methane hydrate was input and oxidized in other locations including the Pacific, Indian, or Southern Oceans, or if these oceans were deepwater source regions (e.g., Thomas et al., 2003; Svenson et al., 2004), then they would be expected to show a more marked response. Changes in the efficiency of the biological pump during the PETM (e.g., Bains et al., 1999; Crouch et al., 2001; Bralower, 2002) also would have led to variations in the lysocline and the CCD. For example, increased supply of carbonate would tend to lead to a deepening of both levels.

A variety of evidence, including single specimen foraminiferal isotope data, indicates that a large portion of the CO₂ or CH₄ that forced the PETM was rapidly released into the atmosphere (e.g., Dickens, 2000; Thomas et al., 2002). However, the early timing of the decrease in CaCO₃ relative to biotic signals at ODP Site 690 suggests that a significant part of the CO₂, possibly derived from the oxidation of CH₄, remained in oceanic deep waters (Thomas et al., 2002). This conclusion assumes that at least part of the observed decrease in CaCO₃ was a result of dissolution and not a change in productivity.

A change in the depth of the lysocline can be identified by a sharp decrease in carbonate content or a deterioration of carbonate preservation (i.e., increased fragmentation). The CCD level can be identified by the near or complete absence of carbonate in sediments. Sediments from the PETM interval at Atlantic and Caribbean Deep Sea Drilling Project and ODP sites, and in Tethyan land sections, suggest a rapid, marked shoaling of the lysocline at the onset of the PETM. In some cases, the absence of CaCO₃ suggests these sites lay below the CCD during the early part of the event (e.g., Canudo et al., 1995; Thomas and Shackleton, 1996; Bralower et al., 1997; Thomas, 1998; Thomas et al., 1999; Erbacher, Mosher, Malone, et al., 2004; Zachos et al., 2005). At Sites 690 and 738 in the Atlantic and Indian Ocean sectors of the Southern Ocean, respectively, the base of the PETM shows a less significant decrease in carbonate content (Thomas et al., 1999; C. Kelly, pers. comm., 2005). Little change in foraminiferal preservation occurs at Site 690 (Thomas et al., 2002), whereas dissolution increases at Site 738 (Lu and Keller, 1993).

Little information about the Pacific lysocline is available, largely because the PETM was either incompletely recovered (Site 865) (Bralower et al., 1995) or not recovered (i.e., during previous legs on Shatsky Rise), although Dickens (2000) interpreted the response of the lysocline in this ocean to be less than in the Atlantic. Background fragmentation levels (20%–30%) measured as a part of this investigation suggest that at least the deepest Shatsky Rise site investigated (Site 1211) was located in the upper range of the lysocline prior to the PETM and that all sites were within the lysocline during the event (fragmentation >40%).

The input of CO₂, possibly as a result of CH₄ oxidation, would result in increased dissolution of CaCO₃, especially in deeper water masses where CO₂ concentrations are elevated and CaCO₃ content is lower. This would result in a shoaling of the lysocline and CCD, as CaCO₃ would become soluble at increasingly shallower depths. As a result, foraminiferal fragmentation would increase and sedimentation rate decrease. Maximum pyrite, fish debris, and fragmentation were observed only in samples ±3 cm from the lithologic contact at all sites, consistent with a geologically instantaneous event, such as massive dissociation of methane hydrates.

If the dissolution rate of CaCO₃ exceeded the rain rate, foraminifers in the upper centimeters of the sediment column would dissolve, a concept referred to as carbonate "burn-down." Evidence for burn-down includes increased fragmentation and decreased CaCO₃ values below the basal PETM lithologic contact at Sites 1209, 1210, and 1212 (Figs. F5, F6, F7). Decreased CaCO₃ below the onset of the PETM has been noted at other PETM sites (Site 690 [Thomas and Shackleton, 1996] and the Leg 208 depth transect [Zachos et al., 2005]), but never with supporting visual observations and fragmentation data. Whereas burn-down has been predicted from models (Walker and Kasting, 1992; Dickens et al., 1997), the extent of such burn-down has not been previously constrained in a PETM section.

Increased CO₂ in deep waters in the vicinity of Shatsky Rise would cause the lysocline to shoal and increase saturation of Ca²⁺ and HCO₃^– in sediment pore waters. This would lead to secondary precipitation of CaCO₃ on foraminifers (Pls. P1, P2, P3, P4, P5, P6, P7) including overgrowth on walls and muricae and separate euhedral calcite blades in PETM sections, indicating at least temporary supersaturation of pore water CaCO₃ (aq.). There is no evidence in any of the preservational proxies that Sites 1209–1212 were situated below the CCD during the PETM. It is impossible to rule out, however, that the CCD shoaled above this depth range for a very short interval at the onset of the event (for example, ~1–2 k.y.), but leaving no substantial clay layer due to bioturbation. The magnitude of the bulk carbon isotope excursion at Shatsky Rise is smaller than that recorded at the shallowest of the Walvis Ridge sites (Site 1263), which would be consistent with a period of nondeposition of carbonate and truncation of the CIE at Shatsky Rise. Site 1208 on the Central High (3346 m) was clearly situated close to, but not below, the CCD during the PETM as there are no foraminifers in PETM sediments and only the most robust nannofossils are preserved (Bown, this volume).

We can obtain a minimum estimate for the amount of lysocline shoaling assuming that the deepest location on the Southern High (Site 1211; 2907 m) was located near the upper lysocline prior to the PETM. Our data show that even the shallowest location (Site 1209; 2387 m) lay below the lysocline at the peak of the PETM. Thus, a minimum estimate for the extent of lysocline shoaling is ~520 m. This is significantly higher than an estimate for the Pacific (~250 m) from a model run in which CH₄ is oxidized in the Atlantic (Dickens et al., 1997; Dickens 2000) but similar to a run where CH₄ is oxidized in the deep Pacific Ocean (~600 m of shoaling), which is unlikely given the CCD shoaling in the Atlantic. Leg 199 in the equatorial Pacific recovered the PETM at Sites 1220 and 1221 (3.0 and 3.5 km paleodepth), and the cores from this interval show a sharp decrease in carbonate to very low values (Lyle, Wilson, Janecek, et al., 2003). The greater depth of the Leg 199 sites compared to those studied on Shatsky Rise is consistent with the evidence for increased dissolution and suggests a rapid shoaling of the CCD in this location (Lyle, Wilson, Janecek, et al., 2003).

Although the data presented here are most consistent with results from model runs in which CH₄ (and by inference, CO₂) is input directly into the deep Pacific, we do not conclude that this was the case. Atlantic PETM sections show evidence for more significant shoaling of the lysocline, and many sites at similar depths to those studied here were situated below the CCD during the early stages of the event (Dickens, 2000; Zachos et al., 2005). In particular, a depth transect on Walvis Ridge in the South Atlantic indicates at least 2 km shoaling of the CCD over 10 k.y. at the onset of the PETM (Zachos et al., 2005). Our results suggest that the models themselves may not adequately simulate the oceans or, more likely, that the amount of CO₂ input was far greater than assumed.

The benthic foraminiferal extinction (BFE) at the PETM was the largest extinction event in this group in the last 90 m.y. (Thomas, 1998). Suggested causes of the BFE include dissolution of benthic tests or lack of food (Thomas and Shackleton, 1996; Thomas, 1998; Thomas et al., 2000). Depleted oxygen levels in deep waters resulting from increased deep-sea temperatures, oxidation of CH₄ in the water column, locally or regionally increased productivity, or a combination of these factors is a possible cause of the BFE (e.g., Boersma et al., 1998; Thomas, 1998). Deepwater dissolved oxygen levels are difficult to determine but are essential to understanding the faunal response to events during the PETM. Abundance of pyrite can suggest O₂ depletion, especially in regions of high productivity where the rapid flux of organic matter rapidly consumes deepwater O₂ (e.g., Berner, 1977; Lerman, 1979; Müller and Suess, 1979; de Baar and de Jong, 2001). Independent evidence for low O₂ includes lamination at Sites 999 (Bralower et al., 1997) and 1260 (Erbacher, Moser, Malone, et al., 2004) and the appearance of a unique low-oxygen assemblage of benthic foraminifers at all Leg 198 sites (Kahio et al., submitted [N1]). Still, all Shatsky Rise PETM sediments are homogeneous and organic matter content is minimal, indicating that bottom waters did not become anoxic.

Estimated bottom water warming of ~6°C during the PETM (e.g., Kennett and Stott, 1991; Zachos et al., 2003) would also have lowered dissolved oxygen levels. Using the equation (Broecker and Peng, 1982)

O₂ (µmol/kg) = 350 – 9T + 0.14T²,

where

T = temperature (°C),

and deepwater temperature estimates from ODP Site 865 (10° and 16°C pre-PETM and PETM, respectively) (Bralower et al., 1995), the decrease in deepwater dissolved oxygen content was from ~274 µmol/kg to ~242 µmol/kg. Dissolved O₂ in the modern western equatorial Pacific is ~260 µmol/kg (Broecker and Peng, 1982), similar to pre-PETM estimates. Unless ocean circulation was significantly different from the present causing lower Pacific O₂ levels, deepwater warming was insufficient to cause O₂ depletion. O₂ depletion resulting from CH₄ oxidation in the deep ocean may be responsible for pore water dysoxia in the absence of elevated productivity. Alternatively, increased pyrite contents may reflect highly condensed sections near the base of the PETM at the Shatsky Rise sites.

Pyrite precipitation at the PETM may alternatively be related to increased supply of Fe to the ocean. Increased chemical weathering on continents as a result of global warming would increase the Fe flux to the oceans. This is a viable explanation for the pyrite peak if Fe, not sulfate, was a limiting factor in pyrite precipitation. However, this alternative would cause increased pyrite throughout the warming interval, not just at the onset, as indicated in the PETM intervals on Shatsky Rise.