Our analyses are presented in Table T1 for Site 1218 and Table T2 for Site 1219. We find a factor of 2–4 variability in opal and Corg at each site but order of magnitude changes in CaCO3. We observe significant production changes at each site, particularly in the time interval between 38 and 42 Ma, with large additional changes in CaCO3 superimposed. Figure F2 shows the 33- to 55-Ma time series of CaCO3, Corg (expressed on a carbonate-free basis), and biogenic SiO2 (carbonate-free) at Sites 1218 and 1219. Corg and biogenic SiO2 are shown on a carbonate-free basis to remove the dilution effects of carbonate on the other biogenic time series. A better way to remove these dilution problems is to convert percentage data to mass accumulation rates (Fig. F3). Deposition of particulate material at the seafloor to generate the sediment column is unlike adding solutes to a dilute solution; on the contrary, addition or removal of any one sedimentary component during deposition changes the percentage of all others present. In contrast, dilution effects among solutes in solution are usually negligible because the concentration of solutes is extremely small relative to the solvent. MAR time series better show the real variation in sediment deposition than percentage data, provided that the age model is reasonably accurate.

The age models we use are the best presently available for the Eocene. Age models for both Sites 1218 and 1219 have been orbitally tuned and intercorrelated to the base of Site 1218 (42 Ma) (H. Pälike, unpubl. data). Because we were able to APC core the Eocene section at Sites 1219 and 1220 and reliably measure paleomagnetic reversals, we were able to base the middle Eocene (42–50 Ma) age model on paleomagnetic chronostratigraphy (Cande and Kent, 1992, 1995). Age models are less reliable for the lower Eocene section below the ubiquitous chert layers near the lower/middle Eocene boundary; for this section there are no improvements on age control beyond the shipboard work reported in the Leg 199 Initial Reports volume (Lyle, Wilson, Janecek, et al., 2002).


CaCO3 time profiles (both weight percent and MAR) illuminate a series of events in the middle and late Eocene. The ages, magnetochrons, and revised meters composite depths (rmcd) of these events are displayed in Table T3 using the chronostratigraphy by H. Pälike (unpubl. data). Carbonate events occur about every 2 m.y. in the middle–late Eocene up to the Eocene–Oligocene transition. We refer to these as carbonate accumulation events (Table T3). The largest of these events (CAE-3) is between 40.5 and 41.5 Ma, mostly in Magnetochron C18r. We have evidence that these carbonate events are associated with polar cool conditions (see below) and are weaker analogs to the Oi-1 event of the Eocene–Oligocene transition (Coxall et al., 2005). The largest of these CAE events (CAE-3) reaches its maximum about half a million years after the warm Middle Eocene Climate Optimum (MECO) event reported by Bohaty and Zachos (2003), which ended at ~41.5 Ma. The end of CAE-3 was caused by a relatively slow decrease in productivity upon which a rapid shallowing of the CCD is superimposed, indicative of a large Corg transfer to ocean waters from an external reservoir (see below).

Site 1219 displays a basal carbonate section typical of pelagic sediments on cooling and deepening ocean crust. Basal carbonates disappear by 53 Ma (Figs. F2, F3). The section between 53 and 50 Ma was not recovered, but a continuously cored, though not complete, section of the younger sediments was recovered. We analyzed the sediment section between 50 and 38 Ma in detail but did not analyze the upper Eocene section. Based on shipboard surveys for nannofossils, the two upper Eocene cores (199-1219A-18H and 19H) were barren of nannofossils except at the very base of Core 19H (Lyle, Wilson, Janecek, et al., 2002).

The events older than 42 Ma are best delineated in the CaCO3 weight percent record (Fig. F2) rather than MAR (Fig. F3) because nonlinearities in percentage expand the low part of the percentage scale relative to MAR. The events generally have multiple peaks of CaCO3 deposition, and we can demonstrate orbital cyclicity of CaCO3 MAR in CAE-3, the best resolved event.

Site 1218, because of its younger basal age, contains only the events from CAE-3 to CAE-7. The magnitude of the CAEs thus gradually increases from 50 Ma up to CAE-3 (41.5–40.5 Ma) and decreases afterward.

Biogenic SiO2 and Corg

Corg and biogenic SiO2 vary significantly over the middle and late Eocene, but the magnitude of this variation is significantly smaller than that of the CaCO3 variation. It appears that production may be an important factor to cause a carbonate accumulation event but that changes in carbonate dissolution magnify the CaCO3 signal relative to the other biogenic components. Because high bulk MAR generally coincides with high carbonate intervals at Sites 1218 and 1219, there is a general similarity between the carbonate-free and MAR time series (Figs. F2, F3). Nevertheless, it is clear that there were distinctive styles of deposition in the middle and late Eocene (42–38 Ma) relative to modern depositional processes. Olivarez Lyle and Lyle (this volume) explore the productivity issues in detail. They note that much more of the Corg rain must have been recycled in the water column or surface sediments in the Eocene relative to the Holocene. In this paper we will briefly describe these time series and their relationship to the CaCO3 record.

The early Eocene at Site 1219 (54.5–53 Ma) has relatively high Corg MAR but low biogenic SiO2 MAR (Fig. F3). Because of the lack of good age control in this interval, we note only that the low opal content in the early Eocene is distinctive of the interval, consistent with similar work on the early Eocene section from Site 1221 (Murphy et al., this volume), and also consistent with the difficulty in obtaining radiolarians for biostratigraphic studies from this interval. (T.C. Moore Jr., pers. comm., 2002).

For whatever reason, the early Eocene did not support conditions for radiolarians to grow and prosper. Except for certain distinctive intervals, essentially all the biogenic SiO2 found in Leg 199 samples is derived from radiolarians. The most prominent of these diatom intervals occurs at the top of CAE-3, where the biogenic silica fraction is >50% diatoms. Diatoms never compose >5% of the total biogenic SiO2 fraction in the rest of the Eocene based on smear slides (Lyle, Wilson, Janecek, et al., 2002).

The Eocene interval (50–34 Ma) at Sites 1218 and 1219 is marked by a long-term background trend in which biogenic SiO2 MAR decreased by about a factor of 2, from ~0.14 to 0.07 g/cm2/k.y. (Fig. F3). We interpret this to indicate a trend toward lower productivity from the beginning of the middle Eocene to the end of the Eocene. There is no long-term trend in the Corg MAR profile. A middle Eocene interval (42–38 Ma) of variable but much higher deposition is the most prominent feature in the Eocene records, coinciding with CAE-3 and CAE-4. We observe little response in Corg or biogenic SiO2 MAR profiles across CAE-1 and CAE-2, possibly because there was not a strong change in productivity associated with these events.

The magnitude of biogenic SiO2 MAR is comparable at both drill sites, within events as well as in the background intervals between events (Fig. F3). However, Corg MAR tends to be higher in the more carbonate rich sediments of Site 1218 relative to Site 1219 (Fig. F3). Site 1218 was significantly closer to the equator at ~40 Ma (Site 1218 = 0° paleolatitude vs. Site 1219 = 2°S), so higher MAR at Site 1218 might be expected at this time from differences in productivity. However, preservation of Corg is linked to oxygen exposure time (Hartnett and Devol, 2003), and the higher sedimentation rates at Site 1218 relative to Site 1219 may be the most important factor in the difference between the two sites. In the late Eocene, the two sites were roughly equal distances north and south of the equatorial region and the offset between sites should have disappeared. Unfortunately, we did not analyze Corg beyond 38 Ma at Site 1219 because of the lack of CaCO3.

The middle Eocene interval of high but variable biogenic SiO2 and Corg MARs overlaps the CAE-3 and CAE-4 carbonate events, but the interval of low CaCO3 between CAE-3 and CAE-4 is relatively high in Corg MAR and biogenic SiO2 MAR compared to the interval before CAE-3 or after CAE-5 (Fig. F4). The loss of CaCO3 was disproportionate to the drop in MAR of other biogenic sedimentary components between CAE-3 and CAE-4. In other words, the Corg and biogenic SiO2 MAR time series indicate that the sudden drop in CaCO3 cannot be ascribed to a drop in production alone but must represent a major increase in CaCO3 dissolution as well.

Estimating the Carbonate Compensation Depth

The CCD is defined as the shallowest depth at which the particulate rain of CaCO3 is entirely dissolved before burial or, in other words, the shallowest depth at which CaCO3 disappears from sediments. The location of the CCD can strongly constrain ocean carbon cycles because it depends on balancing carbonate production in the surface ocean and dissolution caused by ocean carbon chemistry in the abyss (Berger et al., 1976). Typically, CaCO3 weight percent is used to estimate the CCD. However, the strong nonlinearities of percentage with respect to flux (Fig. F5) make CCD estimates based on percentage poor. Instead, CaCO3 MAR should be used because the effects of dilution are factored out by MAR.

An example of the strong nonlinearity in weight percent data is shown in Figure F5. A hypothetical sediment composed of 90 wt% CaCO3 and 10% clay (which is inert) is subjected to a linearly increasing dissolution gradient. About 80% of the CaCO3 particulate rain must be dissolved to change carbonate from 90 to 65 wt%. The problem is even more acute if the secondary component in the sediment is also subject to dissolution, as is the case in equatorial Pacific sediments. Biogenic silica is typically the second most abundant sedimentary component in the equatorial Pacific, and it dissolves readily (Archer et al., 1993). If both the carbonate and residual fractions of the sediment dissolve at similar rates, huge amounts of CaCO3 dissolution may occur with little or no change in CaCO3 weight percent. The use of CaCO3 MAR eliminates the nonlinearity problem. Both the change in sediment composition and the change in mass are monitored with MAR. Provided that a common age model is used for sites from different water depths, a MAR-estimated CCD will be accurate even if the age model used to estimate MAR is wrong. Although the estimated absolute flux would be more inaccurate with a poor age model, the change in MAR with depth remains accurate because the mass change over the same time interval is being compared.

The MAR estimate of CCD can have errors because the CCD formulation assumes that the mass of CaCO3 particulate rain is the same at all sites. One must therefore evaluate whether the vertical rain and any horizontal sediment focusing appear to be the same at different sites in order to test the validity of the CCD estimate.

CCD Changes in the Eocene Tropical Pacific

We used Sites 1218, 1219, and 1220 to estimate the CCD in the tropical Pacific. The time series for biogenic sediments are shown in Figures F2, F3, and F4 for Sites 1218 and 1219 and provide the basis for detailed studies. We did not run new analyses on Site 1220 sediments but instead used shipboard data. The entire middle to upper Eocene section at Site 1220 is barren of carbonate, as evidenced by extensive unsuccessful searches for nannofossils for biostratigraphy. No new stratigraphic data for Site 1220 are available to us to refine the age model. Nevertheless, the data available suggest that Site 1220 biogenic SiO2 MAR was similar to that at Sites 1218 and 1219, though perhaps lower. To the first order, the data at Site 1220 show the same trends in SiO2 MAR with time. Thus, Site 1220 can help define the CCD as well, since it is 80–100 m deeper than Site 1219 throughout the Eocene (Table T4).

The CaCO3 MAR data reported in Table T4 were interpolated from our original data set (Tables T1, T2) to 0.02-m.y. spacing and then averaged over 0.1-m.y. increments. The CCDs shown in bold on Table T4 are linear extrapolations from combined Site 1218 and 1219 data and are more robust than the other data. The data shown in italics represent extrapolations of the CCD assuming a decrease of roughly 0.001 g CaCO3/cm2/k.y. per additional meter of water depth near the CCD. This estimate, although not very accurate, allows us to better identify very large shifts in CaCO3 MAR that occurred as the CCD passes one of the sites. In addition, this estimation allowed us to make some estimate, however poor, of the intervals covered by only one of the prime sites (Sites 1218 and 1219). Ideally, additional sites should be drilled to constrain CCD changes more fully.

For CAE-4 (39.3–38.6 Ma) (Table T3), estimates of the CCD are consistent among Sites 1218, 1219, and 1220, where we have more than one site with carbonate. Extrapolations from Sites 1218 and 1219 estimate that the CCD should be, within error, shallower than the depth of Site 1220 for certain of the CAE intervals. No CaCO3 was preserved at Site 1220 throughout the middle and upper Eocene, however. During CAE-3 (42.2–40.3 Ma), the CCD estimated by the 1218–1219 pair is significantly deeper than the paleodepth of Site 1220, which has no CaCO3. The overshoots deeper than Site 1220 average 100 m but reach 400 m at the peak of CAE-3 between 41 and 40.9 Ma. They indicate that at least part of the CCD estimate is caused by differential CaCO3 rain between sites or that Site 1220 preferentially lost sediment flux relative to the other two sites.

Figure F6 illustrates the general situation for CAE-3. We expect, assuming the present provides a key to the past, that the two sites on the equator (1218 and 1220) should have similar CaCO3 production and should best define the CCD. If Sites 1218 and 1220 are used to define the CCD, ~60% of the MAR at Site 1219 has to be excess (i.e., caused by sediment focusing at Site 1219 or by higher off-equatorial CaCO3 production and burial at Site 1219 relative to Site 1220). We suggest that the excess may be due to higher CaCO3 production at Site 1219 relative to Site 1220. Because Sites 1219 and 1218 have similar biogenic SiO2 MAR, a more conservative measure of productivity, they are the best pair to use to estimate CCD. Even with the uncertainties, the CCD during CAE-3 gradually increased by ~600 m between 43.5 and 41 Ma and abruptly shallowed by at least 800 m between 40.8 and 40.7 Ma.

The disparity between CCD estimates from Sites 1220 and 1219 illustrates the sensitivity of the CCD to differences in CaCO3 production and sedimentation. We determined that the Site 1218–1219 pair gives a better CCD estimate because the biogenic SiO2 MAR at each site is nearly the same over the intervals for which we have common data (Fig. F3). Because biogenic SiO2 was not measured at Site 1220, we cannot compare Site 1220 to the other two sites in detail. If we use the magnetic anomalies at Site 1220 and assume an average of 70 wt% biogenic SiO2 in the sediments, biogenic SiO2 MAR has a pattern similar to but somewhat lower than the average from Sites 1218 and 1219. Because the original CaCO3 rain should be proportional to the rain of other biogenic components out of surface waters, these data indicate that initial CaCO3 deposition at Site 1220 was generally somewhat lower than at the other two sites and the CCD estimate using Site 1220 should be skewed shallow.

The difference in CCD estimated by the Site 1218–1219 pair vs. the Site 1218–1220 pair appears to be more than a local difference caused by sedimentation. No site on the 56-Ma transect north of Site 1219 (Sites 1220, 1221, 1217, 1216, or 1215) contained any CaCO3 in its middle and upper Eocene sections. There appears to be strong gradients in CaCO3 production in the Eocene both longitudinally and latitudinally. We also found, for example, that CaCO3 in lower Eocene sediments was better preserved away from the equator rather than near it (Lyle, Wilson, Janecek, et al., 2002).

Oxygen Isotopes

A set of samples from Site 1219 with CaCO3 contents >10 wt% was run for bulk oxygen isotope content at Stockholm University to test the hypothesis that the CAEs are indicators of Eocene polar glaciation (Fig. F7). This hypothesis derives from the rapid deepening of the CCD at the Eocene–Oligocene transition (Lyle, Wilson, Janecek, et al., 2002), now shown to be associated with the Oi-1 glaciation (Coxall et al., 2005). A second set of samples from CAE-3 at Site 1218 was analyzed for oxygen and carbon isotopes including bulk carbonate, benthic foraminifers, and planktonic foraminifers to study both surface and deepwater variations (Tripati et al., 2005). Heavier oxygen isotopes in bulk carbonate should indicate either a local cooling of surface waters (because bulk carbonate is primarily of nannofossil origin) or a whole-ocean change in oxygen isotope composition caused by the withdrawal of 16O into continental ice sheets.

Limiting bulk oxygen isotope analysis to Site 1219 samples containing >10 wt% CaCO3 results in data that represent only the CAE events (Fig. F7). Analysis of these CAE samples revealed differences between the carbonate events. The most striking of these is that CAE-4 appears to be offset toward significantly heavier oxygen isotope values for equivalent CaCO3 relative to the earlier CAE events measured, perhaps associated with the long-term isotope trend toward enrichment in 18O noted for the Eocene oceans (Zachos et al., 2001). It would be worthwhile to pursue these isotopic variations in a better preserved record elsewhere to differentiate between trends and events.

Within carbonate events (e.g., CAE-2 and CAE-3) there is a general correlation between heavy oxygen isotopes and high CaCO3. Within CAE-3 there is also a tendency toward heavier oxygen isotope values as the event progressed. The lightest isotopic values are found at the beginning of CAE-3 (~42 Ma), and the isotopic values increase as the CaCO3 content increases to 41 Ma. At the end of CAE-3, bulk oxygen isotopes at Site 1219 do not drop as fast as CaCO3 values do.

These results mirror the isotope excursions observed at Site 1218. The oxygen isotope shift to lighter values is distinctly younger than the initial loss of CaCO3, both in bulk carbonate isotopes as well as in benthic foraminiferal oxygen isotopes, as shown in Figure F8A. The first drop in CaCO3 MAR occurred at 41 Ma, whereas oxygen isotope values lighten only after 40.7 Ma. The isotopic data shown in Figure F8 are confined to one core from one hole (Core 199-1218C-21X), so we know that the rapid changes we observe are not caused by an error in the sampling splice. The benthic foraminiferal isotope results thus implicate the high latitudes as the source of the signal because both of the causal mechanisms for oxygen isotope variation (ice and temperature) are set in high latitudes.

Although the oxygen isotopes do not align with the drop in CaCO3 MAR, the first major drop in CaCO3 coincides with an apparent 0.5 carbon isotope event in the deep water (Fig. F8B). Unfortunately, the base of this event is at the base of the studied core, so is as yet poorly defined. Tripati et al. (2005) more fully discuss the implications of these stable isotope records. We observe that high 18O correlates with high CaCO3 and supports the link between increased carbonate burial at the equator and global cooling. The failure of glacial events to maintain themselves prior to the Eocene/Oligocene boundary suggests that strong negative feedbacks stabilized warm conditions in the Eocene but that this stabilization failed at the Eocene/Oligocene boundary.