Calcium carbonate and organic carbon were determined by coulometry at Boise State University (Boise, Idaho, USA) using a UIC, Inc., model CM-5012 CO2 coulometer attached to our modified version of a CM-5120 combustion furnace. Approximately 30–70 mg of dried, homogenized sediment sample was combusted in an ultra-high-purity oxygen (>99.994% O2) atmosphere at 1000°C. The reported CaCO3 data were calculated as the difference between two independent analyses of carbon from each sampled interval: total carbon and organic carbon. Inorganic carbon is calculated as the difference between the total carbon and the organic carbon fractions, and weight percent CaCO3 is the product of inorganic carbon (wt%) and 8.33.
Organic carbon was measured after the sample was pretreated with acid to remove carbonates. Approximately 70 mg of dried sample was placed inside a fused quartz combustion boat and wetted with water and 10 drops of 10% (v/v) concentrated HCl. The slurry was stirred and heated at ~110°C until sufficient solution evaporated to accommodate a second treatment with the acid. After ~1 hr, samples were oven-dried and allowed to cool before analysis with the coulometer. We emphasize the importance of using fused quartz combustion boats. Previously, we had used a glazed porcelain boat (Coors #60032) but discovered that the glaze deteriorated after about a dozen sample runs. Absorption of the sample slurry into the porcelain produced erratic results and unacceptably high background readings.
Accuracy was estimated by including two independent standards in each sample run, and precision was estimated by repeating the analyses of a subset of the unknown samples. Standards included reagent-grade calcium carbonate, reagent-grade sucrose, and an in-house standard "Midway" marine sediment from the Pacific Ocean. Standards were run at the beginning, middle, and end of each sample set. Summary statistics for the Midway standard are given in Table T1 (average total carbon = 2.64 ± 0.02 wt%, n = 523; average Corg = 0.85 ± 0.01 wt%, n = 570). Precision of the unknowns was estimated by repeating the analysis of every fourth sample in the sample run. The average difference between the repeated unknown samples is <0.01 wt% carbon. In general, samples were reanalyzed if the difference between the repeated analyses exceeded 0.02 wt% carbon. This level of precision was necessary because the average concentration of Corg at both Sites 1218 and 1219 was 0.03 wt%.
Here we argue the importance of adopting our protocol for measuring organic carbon in pelagic sediments by elucidating the inadequacies of the JOIDES Resolution standard technique. Specifically, JOIDES Resolution Corg measurements are inadequate for studies of carbonate sediments in pelagic environments when Corg falls below ~0.3 wt%. Corg values are calculated indirectly as the difference between total carbon and inorganic carbon. Note that we measure organic carbon directly, not as the difference between two parameters. Because both shipboard measurements are obtained by different analytical methods, negative Corg values often result when carbonate content is high and Corg is low, a situation that typifies marine sediments. The analytical precision (±0.2 wt% Corg) is greater than the concentration in many samples, as demonstrated for duplicate analyses of samples from ODP Hole 847B (Shipboard Scientific Party, 1992).
To test the "general adequacy" of shipboard organic carbon measurements, Meyers and Silliman (1996) conducted a shore-based analysis of 95 samples for direct comparison with JOIDES Resolution shipboard values for identical sample intervals. They concluded "little error is introduced" using the shipboard procedure for measuring organic carbon "over a wide range of concentrations." For the range that Meyers and Silliman are interested in (>1 wt%), their conclusion is correct. Unfortunately, this result has been generally assumed (e.g., Shipboard Scientific Party, 1998) without closer examination of the data, especially for Corg values in the range typical of pelagic sediments (0.2 wt% Corg, based on the compilation of Deep Sea Drilling Project [DSDP] Legs 1–33 by McIver [1975]). Whereas direct Corg measurements of Meyers and Silliman (1996) are "positively correlated" to shipboard values, our examination of the two published data sets reveals a high degree of variability and unexplained deviations directly relevant to paleoceanographic studies. For example, we note that the sample-to-sample differences between the onshore direct Corg measurements and the shipboard indirect Corg measurements range between –100% and 2,000% for all 95 samples reported. (Corg content in the samples range between 0.06 and 3 wt%, using the Meyers and Silliman results). The median difference between the two analyses, or median deviation, is high: 34% for all 95 samples, which increases as the concentration range of Corg decreases. Specifically, the median deviation increases from 40% to 60% to 75% for samples containing 0.5 wt% Corg, 0.4 wt% Corg, and 0.3 wt% Corg, respectively. In contrast, our analytical protocol routinely produces direct Corg measurements with high precision (±0.01% Corg) at very low concentrations: <0.1 wt% Corg.
In our Eocene data set, all samples with Corg values of 0.10 wt% were repeatedly analyzed, yet the tests produced similarly high results. To further test the interpretation of the measurement results, we pursued two testable hypotheses that could explain the high values: (1) sample contamination and (2) inherent limitations of the analytical procedure.
We tested the anomalous samples for contamination by drilling muds and oils when drilling through chert horizons at both sites. Such horizons are typically difficult to drill, and contamination is possible through use of excess drilling muds and oils during subsequent retrieval and processing. Suspect samples were tested for contamination by comparing their proximity to identified chert horizons with their Corg content. Fifteen distinct chert horizons (occurring as fragments, nodules, or layers) at Site 1218 are relevant for comparison with our sampled intervals. Of the 15 horizons, nine samples coincide with the location of a chert horizon. If contamination is the cause, then these samples should exhibit high organic carbon values. In fact, these nine samples fall in the range of average values, between 0.01 and 0.06 wt% Corg. Using a second approach, we tested this theory by measuring the distance between a sample with high organic carbon and its proximity to the closest chert horizon. We expected to find increasing organic carbon with decreasing proximity to the chert horizon if drilling resulted in sample contamination. As shown in Figure F2, no such relationship was found between these two variables. Based on these two tests, we find no evidence of sample organic carbon contamination from drilling through chert horizons.
High organic carbon concentration could reflect the presence of dolomite, which is found in the basal sections of Leg 199 sites, or other dissolution-resistant phases. Because the organic carbon measurements are made after acidification to remove solid calcium carbonate, dissolution-resistant carbonates would artificially contribute to the Corg fraction when combusted at 1000°C. Our laboratory experiments have shown that dolomite-bearing samples respond well to a simple modification of the acid treatment (adding more acid and increasing the reaction time) to produce organic carbon values that are lower and consistent with the downcore trend. Because dolomite is found in Leg 199 samples, we were suspicious of all samples exhibiting spikes in the downcore trends of organic carbon; all were reanalyzed several times using this modified treatment (excepting the insufficient samples for six samples from Site 1218). All four samples from Hole 1219A responded to the modified treatment by falling to average levels for the site. These samples were taken from carbonate chalks (90 and 95 wt% carbonate) at the base of Hole 1219A. Although dolomite was not noted in smear slide studies (see "Site 1219 Smear Slides" in Lyle, Wilson, Janecek, et al., 2002), the samples response to the modified treatment suggests the presence of a dissolution-resistant phase.
For Site 1218, only 3 of the 19 reanalyzed samples produced lower values which are in line with average levels (0.05 wt% Corg). Most samples remained high after the modified treatment. Ten of these samples remained in the anomalously high range (>0.10 wt%), and six samples resulted in lower values that were still in the elevated range between 0.06 and 0.09 wt% Corg. No common factor explains the results; there is no significant correlation with carbonate content, lithology, or location downcore. Furthermore, the "null hypothesis" was not confirmed (i.e., samples from sections known to contain dolomite did not necessarily exhibit high Corg values, even when the "weaker" acidification method was used). For example, dolomite contents between 15% and 20% are reported from smear slide analysis of sections at the base of Hole 1218A (Cores 199-1218A-29X and 30X; 266–274 meters below seafloor (mbsf); see "Site 1218 Smear Slides" in Lyle, Wilson, Janecek, et al. [2002]). These sections, described as nannofossil chalks, contain high carbonate contents (64–90 wt%). We analyzed 18 samples from this interval but only 2 produced moderately elevated Corg values using the standard method (0.09 and 0.11 wt%). These values fell to 0.02 and 0.06 wt%, respectively, after the harsher acidification treatment was used. More intriguing is the fact that the other 16 samples from the dolomite zones reflect the site average of 0.03 wt% Corg. Therefore, the presence of dolomite or chalk is not a predictor of the high Corg values we measured. With regard to data presentation and interpretation of the results, we included only the Corg values (Table T2) from the repeated analyses using the stronger acid treatment.
Biogenic silica was measured using a Hach model DR/4000 spectrophotometer after sediment samples were digested in a 2-M KOH solution. The method is described in Lyle et al. (this volume) and Olivarez Lyle and Lyle (2002), where we address the problems of measuring dissolution-resistant radiolarians in Eocene and Miocene marine sediments (Moore, 1969). The problems are twofold. First, the widely used Mortlock and Froelich (1989) method, which uses a 2-M sodium carbonate bath, is ineffective at dissolving Eocene radiolarians, which they acknowledge and which we demonstrated for sediments from the Leg 199 site survey (Olivarez Lyle and Lyle, 2002). The second problem is the pervasive assumption in the scientific community that a significant fraction of clay minerals dissolve during the alkaline bath and result in serious overestimations of the reported biogenic opal content of the sediments. The Olivarez Lyle and Lyle (2002) study shows that using the harsher KOH treatment can be successfully employed to dissolve radiolarians without compromising the analysis via clay dissolution. However, either solvent (KOH or Na2CO3) overestimates biogenic silica when the volcanic glass content is high, which is not the case for the Eocene samples from Sites 1218 and 1219.
We used two standards in the opal analysis: a reagent grade dissolved silica standard (Hach 1106-49) and one of two in-house dry sediment standards—the "1218C composite standard" or the "1219A composite standard." Composite sediment standards are mixtures of sediments from each interval we analyzed and therefore reflect the full range of variation in biogenic silica, matrix materials, and dissolution-resistant opal phases. In certain cases, a diatom ooze composite sediment standard was used from ODP Leg 178 Site 1098 because good reproducibility using our laboratory method previously had been established for this standard (Olivarez Lyle and Lyle, 2002). Summary statistics for the composite sediment standards are given in Table T1. Generally, every sample tube was analyzed twice and every fourth "unknown" sample was replicated as a separate unknown in the sample run. If the difference between the replicate analyses exceeded ~7 wt% SiO2, then the sample was reanalyzed during a subsequent run. Note that our reported biogenic silica data have not been corrected for structural water content, which is variable and may be as high as 15 wt% H2O for radiolarians. Hence, the biogenic SiO2 content we report underestimates the true sediment fraction.
Table T1 also includes the results of additional tests we conducted to gauge the efficiency of our standard method for dissolving Eocene biogenic opal and to gauge the contribution of dissolved silica from clays by using reagent-grade talc as the clay test sample. The former test involved collecting the sediment residue that remained in the centrifuge tube after a standard KOH digestion. All tubes contained the same 1218C composite sediment standard, a radiolarian ooze containing 32 ± 2 wt% SiO2 (n = 87). To achieve critical mass, the sediment residue from 24 tubes was combined to make nine samples and then dried, reweighed, and prepared for a second opal analysis using our standard procedure with the exception that samples were not subjected to a second hydrochloric acid and peroxide treatment. As shown in Table T1, 1.6 wt% SiO2 was leached from the 1218C sediment residue during the second digestion. This value is ~5% absolute of the average value from the first digestion and is consistent with our smear slide analyses of the KOH-digested residues of Leg 199 Eocene sediments showing trace amounts of radiolarian fragments; it suggests that very little silica is leached from the clay mineralogical matrix. In addition, we subjected 26 samples of reagent-grade talc to our digestion procedure to show that very little silica (1.5 wt% SiO2) is leached in a 2-M KOH solution subjected to our standard protocol (Table T1). We also repeated a KOH digestion of the recovered talc residue; this second digestion produced only 0.4 wt% SiO2. Hence, the silica released from the first leach could reflect contamination from the milling process used to refine the reagent-grade clay, implying that the lower value (0.4 wt% SiO2) from the second leach accurately reflects the amount of silica released from clay-dissolution using KOH. In either case, we are confident that our biogenic silica results are not compromised by clay dissolution, that our method efficiently dissolves thick-walled radiolarians, and that our biogenic silica data are the best available for Eocene marine sediments to date. Extremely low opal environments warrant caution when the 0.5–1 wt% SiO2 that leaches from clays represents a higher proportion of the total biogenic silica present.
Here we estimate the biogenic barium to organic carbon ratio of modern surface sediments beneath upwelling regions. We used Leg 199 inductively coupled plasma–atomic emission spectrometer (ICP-AES) data for barium and other elements (Shipboard Scientific Party, 2002) to estimate the initial organic carbon content of surface sediments in the equatorial Pacific. We assumed that (1) the geologic processes and features that control biogenic barite precipitation in the water column today were also operative during the Eocene; (2) barium concentrations in the sediments are a valid proxy for barite, as demonstrated by Pfeifer et al. (2001) and specifically for Leg 199 sediments by Faul and Payton (this volume), who demonstrated a strong linear correlation (r2 = 0.9) between shipboard Ba and barite concentrations; (3) biogenic barite has been well preserved (see above); (4) our estimate of the Bio-Ba/Corg ratio of modern surface sediments for upwelling regions is valid for Eocene surface sediments beneath the equatorial upwelling region; and (5) seawater sulfate levels in the Eocene were the same. Note that recent work (Turchyn and Schrag, 2004) shows that seawater sulfate concentrations may vary by about ±10% over short timescales of 2–3 m.y.
Barium concentrations in surface sediments can be used to predict the amount of primary productivity in surface waters when the Dymond et al. (1992) sediment trap model is employed (Pfeifer et al., 2001). We used two data sets to determine the Bio-Ba/Corg ratio of equatorial surface sediments for this study. The first data set (Table T3) (SETPAC data; Lyle, 1992) is directly related to our study because it represents the geochemistry of modern surface sediments from the eastern tropical Pacific Ocean. The second data set (Anderson, 2003) reflects surface sediments from a high-productivity region in the Southern Ocean. Together, these data provide a first-order characterization of the Bio-Ba/Corg ratio of surface sediments from upwelling zones.
The SETPAC data set (Table T3) includes X-ray fluorescence measurements of barium, aluminum, and other elements for 153 surface sediments (upper 5 cm) and Corg measurements by acid-oxidizer methods (Weliky et al., 1983) from the eastern Pacific region. The Bio-Ba/Corg ratio varies significantly with proximity to the South American continent and ranges from 0.02 to 5.30 (average = 0.59 ± 0.66; n = 158). To make valid comparisons between the modern and paleoupwelling region, we considered only samples located within 5° of the equator and west of 100°W longitude, thus eliminating samples likely to reflect a continental or hemipelagic signature. The average Bio-Ba/Corg ratio for the restricted data set, highlighted in Table T3, decreased slightly to 0.57 (± 0.15), but the range decreased significantly to 0.35 to 1.0 (n = 13). The restricted data set is bounded by ±5° latitude and 100°–130°W longitude and spans the paleolatitude and longitude coordinates of all Eocene samples from Sites 1218 and 1219. The Bio-Ba/Corg ratio is roughly correlated with water depth: the smallest ratio represents surface sediments at 3200 m; the largest ratio is from 4500 m water depth.
The Bio-Ba/Corg ratios for the modern equatorial Pacific sediments are remarkably similar to the ratios for surface sediments beneath upwelling areas in the Southern Ocean. The Southern Ocean sediment data were obtained from Anderson (2003) and include a total of 29 samples from three cores (NBP98-02-05-2, NBP98-02-07-13, and NBP98-02-06-5). The study area is bounded by 60.24°–63.11°S latitude and 169.74°–170.19°W longitude—a high productivity region dominated by diatoms. Organic carbon contents range between 0.3 and 0.5 wt% and biogenic silica is generally high, between 30 and 80 wt%. Aluminum data, used to correct the total barium for terrigenous sources, are not reported. The amount of Bio-Ba was calculated by assuming that the nonbiogenic (terrigenous) fraction is an aluminosilicate with a SiO2:Al2O3 ratio of 3.5 and an average Ba/Al ratio of 0.0075 (Dymond et al., 1992). The terrigenous fraction was calculated as the difference between the total amount of sediment and the biogenic components (100% – [CaCO3 wt% + Biogenic SiO2% wt%]). The Si:Al ratio we chose is robust because the proportion of biogenic components is very high (up to 80%); altering the SiO2:Al2O3 ratio of the terrigenous fraction between 3 and 4 made little difference in the ultimate Bio-Ba values. Applying this model to the 29 samples, we find that the Southern Ocean Bio-Ba/Corg ratios and those of the equatorial SETPAC data set closely agree. The median Bio-Ba/Corg value is 0.66 (wt%/wt%) and ranges between 0.36 and 1.4 (n = 29) for the Southern Ocean surface sediments vs. the equatorial Pacific (median = 0.54 [wt%/wt%]; range = 0.35–1.0; n= 13). This range applies to a depth within the sediment column of 2.6 m for Southern Ocean samples (the SETPAC data set is limited to the upper 5 cm of the cores). We found no substantive difference in the Bio-Ba/Corg ratio based on sampling depth; for example, ratios for Southern Ocean samples located deeper than 0.5 m range between 0.33 and 0.87.
For both high-productivity regions we note the following similarities in the Bio-Ba/Corg ratio of surface sediments: the median value at both sites is between 0.6 and 0.7 and ranges between 0.35 and 1.0 and the ratio increases with water depth in the restricted SETPAC data set but does not exceed the upper limit of 1. Processes which control this ratio, both within and between regions, are not explored in detail here but may reflect (1) differences in the initial ecology of the primary producers and their effects on new production (e.g., the diatom to coccolithophorid ratio would affect new production via more or less efficient recycling of nutrients), (2) differences in water depth and the impact on the burial of organic matter, and (3) differences in the total sedimentation rate and its impact on the postdepositional exposure of organic carbon to oxygenated seawater.
We estimated the "expected" organic carbon concentration in Leg 199 sediments by utilizing the surface sediment information above along with a Bio-Ba/Corg ratio that is representative of surface sediments and barium data for Leg 199 sediments from which we estimated the biogenic barium component. Using the data from modern upwelling regions, the maximum Bio-Ba/Corg ratio of 1.02 was chosen because it is the same for both upwelling regions and would tend to minimize the predicted Corg concentrations, consistent with the average measured content (0.03 wt% Corg).
First, we corrected all shipboard ICP-AES barium data for terrigenous, authigenic, and clay sources. Dymond et al. (1992) used an average Ba/Al ratio of 0.0075 to adjust the total Ba content for terrigenous sources of barium. This correction alone left the measured barium content of Leg 199 sediments essentially unchanged, resulting in very high estimated Corg values uncharacteristic of pelagic sediments. We applied two additional corrections, for an authigenic and hydrothermal component, justified by compositional differences between sediment trap studies and Leg 199 sediments. Our estimated Bio-Ba concentration was calculated using Equation 1:
The ratio Ba/Al = 0.0075 is the correction used for the terrigenous component as explained above. Ba/Mn = 0.014 reflects the authigenic barium component (Burns and Burns, 1979), and the presence of this phase is suggested by the y = 1/x relationship between the Ba/Mn concentration ratio (y) vs. core depth (x) found for Site 1218, a profile that typifies authigenic enrichments. The Ba/Mn ratio is high at the surface in the slowly accumulating red clays and in the basal section above basement at Site 1218. The high ratio in the basal section may reflect Mn oxide precipitates from hydrothermal activity. Dymond (1981) reports that 20% of the total barium within 100 km of the East Pacific Rise (EPR) crest is of hydrothermal origin and falls to 5% of the total barium in the basins adjacent to the EPR crest. The bulk MAR of the upper sediment column of both Leg 199 sites is low, between 70 and 90 mg/(cm2 x k.y.), a condition favoring authigenic precipitation of Mn oxides. Barium has a high affinity for manganese oxides (e.g., todorokite) and forms as basal sediments near mid-ocean ridges; concentrations in todorokite and metalliferous sediments can exceed 3 wt% (Burns and Burns, 1979.) The ratio Ba/Fe = 0.02 represents the correction for barium contained in Fe-smectites and for nontronite, an Fe-enriched smectite, in particular (Corliss et al., 1978). This correction is necessary because smectites are enriched at both Leg 199 sites below 5 mbsf (Shipboard Scientific Party, 2002).
Differential compaction with depth and differences in average grain density of the cored intervals are accounted for by converting linear sedimentation rates (LSRs) to MARs. MAR is defined as the product of the LSR (length/time) and the dry bulk density (DBD; mass/volume) and has units of [mass/(unit area x unit time)]. DBD is related to wet bulk density (WBD) by the following equation:
where P = porosity and G = average grain density of the sediment interval. Porosity is the fractional volume of a sediment sample filled with interstitial water before the sample is dried for analysis. Hence, WBD is the total mass of the original wet sample divided by the total volume of the wet sample. High-resolution DBD values were calculated every 2 cm from the shipboard "low-resolution" measurements of WBD and DBD. These two parameters are highly correlated and were used to define linear regression equations at each site. The appropriate linear equation was subsequently applied to the high-resolution gamma ray attenuation WBD data set (with a resolution at 2 cm) to predict the corresponding DBD value (Pälike et al., this volume). Linear correlation coefficients were very high (r = 0.997) for the "low-resolution" regression that utilized all of the site data (i.e., were based on all lithologic units). However, better correlation coefficients were obtained for regressions based on major lithologic units identified at each site. Ultimately, we used the appropriate equation based on lithology to predict the DBD values from WBD instead of using a single equation for each site because trial and error runs produced significant differences in the predicted DBD. For example, the calculated DBD values for intervals from a radiolarian ooze section are 8%–30% greater than calculated values from a nannofossil ooze at Site 1219. Calculated DBD values also diverged with increasing porosity, and porosity is strongly coupled to major lithology; hence, we decided to use regression equations generated for the major lithologic units in a given core.