Ocean Drilling Program (ODP) Site 1090, on the Agulhas Ridge in the South Atlantic sector of the Southern Ocean, is ideally located to capture changes in Southern Ocean circulation patterns. Using samples taken from cored sediments, we construct multiproxy records of productivity (biogenic barium (Baex), opal, and CaCO3 mass accumulation rates (MARs)), nutrient and organic carbon burial (reactive phosphorus (Pr) MARs), and redox conditions (U and Mn enrichments) to investigate hydrographic conditions associated with climatic shifts from the Oligocene through the early Miocene. Orbitally induced cyclicity in U and Mn enrichments (100 kyr) suggests shifts in deepwater characteristics. However, CaCO3 dissolution coincident with low U and Mn enrichments does not indicate low-oxygen, corrosive waters similar to modern conditions. These observations indicate that a well-developed “modern-type” Antarctic Circumpolar Current (ACC) did not yet exist over the period from 30 to 20 Ma, with two potential consequences: The Southern Ocean was not functioning as a silica trap, permitting a broader distribution of silica that may have facilitated organic carbon burial in the ocean in general, and the lack of a deeply mixing ACC may have facilitated organic carbon burial in the Southern Ocean. Both the relative (high opal MARs coincident with low CaCO3 MARs) and absolute (high Pr MARs) burial of organic carbon suggest a powerful mechanism for pCO2 drawdown.
 Organic carbon burial is commonly called upon to explain climatic shifts through the drawdown of atmospheric pCO2. High-resolution benthic foraminiferal carbon isotope records from the equatorial Atlantic (Ceara Rise, Ocean Drilling Program (ODP) Site 926) suggest an intriguing link between climate change and carbon cycling in late Oligocene to early Miocene [Zachos et al., 2001b]. The relationship between δ13C and δ18O (Figures 1a and 1b) during the first Miocene glacial event (Mi-1) suggests that carbon cycling played a role in climate change. (The age for the Oligocene-Miocene boundary is contentious. This paper uses the Shackleton et al.  age modified to the new Laskar orbital parameters (23 Ma, J. E. T. Palike, personal communication, 2003) for the C6Cn.2n. This places the Mi-1 event within the Oligocene.) Zachos et al. [2001b] put forward hypotheses for two, probably interrelated mechanisms linking climate and carbon cycling: reorganization of ocean circulation patterns independent of, or as a consequence of, tectonic events (presumably resulting in increased burial of organic carbon, in the absolute sense or in relation to inorganic carbon); and increased sensitivity of the system resulting from decreased atmospheric CO2 concentrations.
 An open Drake Passage has significant physical and chemical consequences for modern Southern Ocean waters. Southern Ocean winds, unfettered by landmasses, may provide up to 70% of the work energy to the global ocean [Wunsch, 1998], driving meridional overturning. Furthermore, strong winds combined with topography within the Scotia Sea have been described as an oceanic “blender” aiding surface-to-deep mixing [Heywood et al., 2002]. Finally, waters within the ACC circle Antarctica four to six times [Döös, 1995] and are a combination of waters from the Pacific, Indian (with a major contribution from the Pacific via the Indonesian Throughput [Sloyan and Rintoul, 2001]), and Atlantic oceans [Döös, 1995; Sloyan and Rintoul, 2001]. As a result, modern Southern Ocean deep waters appear “old” (low oxygen, nutrient-enriched, alkaline, and light δ13C) and as a result behave corrosively (dissolve carbonates). With a Drake Passage closed to deep flow, different chemical characteristics were likely. Even with an open Drake Passage, during the Oligocene and early Miocene, Northern Component Water (NCW) could have flowed directly into the Pacific Ocean via the Central American Seaway [Nisancioglu et al., 2003], and the flow through the Indonesian Seaway into the Indian Ocean could have had different chemical and physical characteristics [Tsuchi, 1997; Cane and Molnar, 2001; van de Flierdt et al., 2004].
 The timing of the opening is quite controversial and interpretations from Site 1090 alone vary from 32.8 Ma [Latimer and Filippelli, 2002] to 27.5–23 Ma [Diekmann et al., 2004] (modified to our age model). A detailed study of thermal structure of waters to the east of the Drake Passage (ODP Site 516 [Pagani et al., 2000]) corroborates earlier conclusions that the opening to deepwater passage did not occur until the middle Miocene [Barker and Burell, 1982; Woodruff and Savin, 1989; Pagani et al., 2000]. However, considering the very complex deepwater flow in the Drake Passage, Scotia Sea, and Argentine Basin in the modern ocean [Arhan et al., 1999], determining the relationship between hydrographic dynamics and the depth of the Drake Passage will always be speculative. When determining the Southern Ocean's role in climate change, a more important question may be how particular hydrographic characteristics facilitate increases in the burial of organic carbon relative to inorganic carbon.
 Modern Southern Ocean circulation with a fully developed Weddell gyre and Antarctic frontal system has significant impacts on the global hydrography and climate, and thus initiation of the present circulation pattern is of significant interest. The Antarctic Circumpolar Current (ACC) blocks surface water from entering the southern ocean, redirecting heat and saline waters back into the Atlantic as far north as Iceland [Toggweiler and Bjornsson, 2000]. The salt imbalance is further enhanced by the diversion of warm saline waters within the Agulhas Current into the Atlantic and the ACC shunting of fresh waters from the Pacific to the Indian Ocean, bypassing the Atlantic [Gordon, 1996; Toggweiler and Bjornsson, 2000]. Thus the ACC has far-flung consequences including the characteristics of Northern Component Water [Gordon, 1996], nutrient enrichment, and the density structure of upwelling systems such as the Benguela and Peru-Chile Current systems (ventilated thermocline [e.g., Boccaletti et al., 2004]).
 Modern circulation limits the ability of the Southern Ocean to bury organic carbon, despite high seasonal productivity, because there is no permanent thermocline [Antia et al., 2001]. However, intensification of the ACC probably did not occur until the collision of Australia-New Guinea with Asia, which blocked transport of water between the Indian and Pacific Oceans [Lawver and Gahagan, 2003], and a modern-type frontal system may not have been evident until the Pliocene [Abelmann et al., 1990]. It is, however, conceivable that prior to the full development of the ACC, the vertical thermal structure in the Southern Ocean could have been significantly different than the present. This difference may have allowed the increased burial of organic carbon in this region, drawing down atmospheric pCO2, and sensitizing the environment to facilitate the growth of ice sheets [DeConto and Pollard, 2003].
 Our goal in this study is to use a suite of proxies to characterize hydrographic factors that facilitated increases in organic carbon burial in Southern Ocean sediments (in the absolute sense or in relation to inorganic carbon) from the early Oligocene through the early Miocene. Toward this end, we will use reactive phosphorus (Pr) mass accumulation rates (MARs) to assess organic carbon burial; biogenic barium (Baex) MARs to evaluate export productivity; and their ratio to explore the relative efficiency of organic carbon burial to the export productivity. We will use biogenic components (CaCO3 and opal MARs) to surmise upwelled water characteristics, and the redox chemistry of sediments (U and Mn enrichment) to evaluate deepwater characteristics (details given by Faul et al. , Nilsen et al. , and Anderson and Delaney ). The Agulhas Ridge (ODP Site 1090), situated just off the tip of South Africa (Figure 1), lies below the Agulhas Current, which transfers heat and salt in the modern ocean from the equatorial Indian Ocean to the Benguela Current and eventually to the North Atlantic Ocean [Berger and Wefer, 1996; Lutjeharms, 1996]. The modern site depth is at the junction of North Atlantic Deep Water (NADW) and Circumpolar Deep Water (CDW). The site is ideally located to monitor the impact of hydrographic shifts in the Southern Ocean on the evolution of sub-Antarctic circulation.
2. Site Description
 Site 1090 (Figure 1, ODP Leg 177), located on the southern flank of the Agulhas Ridge in the South Atlantic sector of the Southern Ocean, was chosen for its relatively high sedimentation rates (6–30 m composite depth m.y.−1) at deepwater depths (current depth = 3702 m), with good core recovery for the early Oligocene through the early Miocene. In the modern ocean, the Site is located within the northern sub-Antarctic zone at a water depth close to the boundary between the NADW and CDW and above the modern calcite compensation depth (CCD) [Gersonde et al., 1999]. Assuming that the ridge is not a result of a hot spot (see below), the Site was likely to have been above the CCD throughout the period of interest [e.g., Van Andel, 1975].
 Several formation hypotheses exist for the Agulhas Ridge with the paleodepth correction differing depending on whether it is related to a hot spot or to thrusting of the South American plate over the African plate prior to the Cretaceous [Gersonde et al., 1999]. The former requires a depth correction of the site. The latter does not, and was assumed to be the case for our interpretations, as there is insufficient tectonic data to argue to the contrary [Gersonde et al., 1999] and the trend in CaCO3 MARs during the early Oligocene is opposite of what would be expected if the site was deepening through time [Anderson and Delaney, 2005]. Paleomagnetic reconstructions suggest that the site has moved slightly to the east but has had only minimal latitudinal movement and has maintained its relationship to Africa over this time period (Ocean Drilling Stratigraphic Network plate reconstruction service, http://www.odsn.de/odsn/services/paleomap/paleomap.html, magnetic reference frame, orthogonal polar projection).
 We collected geochemical data from depths of 110 to 210 m composite depth (mcd), representing ages from ∼20 to ∼30 Ma. A significant hiatus occurs from ∼33 Ma to ∼30 Ma. The minimum sampling resolution was 100 kyr with increased resolution of 20 kyr at the interval spanning the Oligocene-Miocene (O/M) boundary transitions (24.5–22 Ma).
3.1. Age Model
 The age model for Site 1090 (ODP 177) is based on excellent shipboard magnetostratigraphy, and shore-based “U channel” and individual sample magnetostratigraphy tied to the geomagnetic polarity timescale (GPTS) [Channell et al., 2003]. Chron unit ages have been cross-checked using nannofossils and strontium isotopes [Billups et al., 2002; Marino and Flores, 2002; Channell et al., 2003]. For the section from ∼25 to ∼20 Ma, we used the ages from the benthic foraminiferal δ18O record tuned to obliquity [Billups et al., 2004]. The resolution of the benthic foraminifera record is 5 kyr [Billups et al., 2004], but there is the potential for missing orbital cycles within any record. In the worst case, a missed obliquity peak could shift the timescale by 40 kyr. The integration by Channell et al.  of a suite of age indicators with the orbital tuning of the section younger than 25 Ma makes this age model particularly robust.
3.2. Calculations of Mass Accumulation Rates
 Mass accumulation rates (MARs) are critical to interpreting geological shifts in biogenic components. This particular site has three major sedimentary components: detrital, biogenic opal, and biogenic calcium carbonate. Percent opal has a negative correlation with percent CaCO3 (r2 = 0.33), suggesting that, at least in part, the weight percent opal is controlled by the generally more abundant CaCO3. There is also a significant detrital component in some of these samples (up to 95 wt %; see auxiliary material); therefore the percentages of the biogenic components are not a good indication of changes in productivity, because the percentages will be controlled in large part by the dilutant. (Auxiliary material, including specific figures of analytical merit, is available from the World Data Center for Paleoclimatology.)
 To minimize dilution effects, mass accumulation rates (MARs, in mol cm−2 kyr−1) were determined for all of the biogenic components using the following formula: [component concentration (mol g−1 sediment)] [density (g cm−3)] [sedimentation rate (cm kyr−1)]. Density values were based on gas pycnometry which measures the mass and volume of only the solids and liquid, thus giving an in situ determination [Blum, 1997]. Because meters composite depth (mcd) is constructed from expanded cores, the in situ density measurement will differ from an expanded core density. The total range in dry density measurements is 0.36 to 0.87 g cm−3, so even an overestimate of 20% due to core expansion will introduce only a minor portion of the total error in the MAR calculation. Errors associated with MARs come primarily from errors in sedimentation rate determinations resulting from poorly constrained age models. However, as noted above, this age model combines a variety of age indicators to cross-check ages derived from GPTS.
3.3. Geochemical Extractions
 For all geochemical extractions, the samples were freeze dried, lightly ground, and, to assure a limited grain size distribution, passed through a polypropylene, 150-μm sieve. We used two methods for the total digestion of the sediments, which differed solely in the ratio of hydrofluoric to nitric acid [Murray and Leinen, 1996; Environmental Protection Agency, Method 3051: Microwave-assisted acid digestion of sediments, sludges, soils, and oils, 1996, available at http://www.epa.gov/SW-846/pdfs/3051.pdf]. We placed 50 mg of sediment in an enclosed Teflon container (7 mL) in either a four-to-one or one-to-four mixture of concentrated hydrofluoric and concentrated nitric acid (hereafter designated as the high and low HF:HNO3 techniques) in the microwave at 12% of total microwave power (this will vary for different microwaves) for 90 min. Samples were taken to dryness and then followed by consecutive drying of 1 mL of concentrated nitric, hydrochloric, and nitric acid. Samples were dissolved in 1 mL of concentrated HNO3 with 0.5 mL of hydrogen peroxide, 5 mL of glass distilled water, and 100 μL of concentrated HF. Any samples with Al/Ti ratios widely deviant from the average crustal abundance (greater than 40 or less than 15 mol mol−1) were rerun (see auxiliary material).
 Digestions of known weights of samples were analyzed by Inductively Coupled Plasma Mass Spectrometry (ICP-MS, Finnegan, Element I) for barium (Ba), manganese (Mn), and uranium (U). Relative standard deviations (RSDs) on analyses of long-term solid consistency standards were 6%. Aluminum (Al) and phosphorus (P) were run on Inductively Coupled Optical Emission Spectrometry (ICP-OES, Perkin Elmer, Optima). RSDs were 2% and 9%, respectively.
 We measured P concentrations in sediment samples (0.1 gram sediment) using a sequential extraction technique that isolates four sedimentary P components: oxide-associated (sorbed to oxides or oxide coatings), authigenic (carbonate fluorapatite), detrital (terrestrial silicates and apatite), and organic P [Anderson and Delaney, 2000]. P concentrations in known volumes of extractants (and thus in the solid samples) were determined using an automated spectrophotometric flow injection analysis system (Lachat QuickChem 8000). The relative errors (1s) on the long-term means for authigenic P, the dominant sedimentary component, were 5%, and for total P (propagation of error from all components), 4%. The efficiency of the P extraction was determined by cross comparing a subset of total sediment digestion P with the sum of the extracted components. Total digestions of P compare well with the totals from the four-step extraction (auxiliary material, m = 1.01 ± 0.035, b = −0.86 ± 0.44, r2 = 0.89, n = 110). The four-step extraction is capturing all of the P.
 Carbonates were run by coulometry (UIC, Inc., Coulometrics model 5012). Relative standard deviations on the means for multiple determinations of a pure calcium carbonate standard, samples run in duplicate within a given analytical run, and replicate analyses of the consistency standards were always less than 1%. The effective detection limit for weight percent CaCO3 depended on the sample size; for typical sample sizes of 5–10 mg, the detection limit is 0.5–1.0 wt %.
 Weight percent biogenic opal was determined from a base extraction of a known sample weight (2M Na2CO3 [Mortlock and Froelich, 1989]). Analyses were run on an ICP-OES. Relative error in long-term measurements of solid consistency standards was 8%. The detection limit for biogenic opal, defined from three times the standard deviation of replicate measurements of a blank, was equivalent to 0.9 wt % in a typical size sample (0.03–0.1 g). Incomplete digestion of biogenic opal by the methodology of Mortlock and Froelich  has been observed in sediments of this age [Lyle and Lyle, 2002]. Smear slides before and after digestion suggest that 5–15% of the diatoms were not digested, resulting in an underestimate of the biogenic opal.
 The detrital component is the total weight minus the weight percents of biogenic opal (as SiO2) and CaCO3. An underestimate of the weight percent biogenic opal would result in an overestimate of the detrital component by an equivalent amount.
3.4. Spectral Analysis
 Sampling resolution for the long-term, low-resolution record from ∼20 to ∼30 Ma is 100 kyr. We increased sampling resolution to ∼20 kyr in the interval spanning the Mi-1 glacial event (24.5 to 22 Ma). The spectral analysis was carried out using AnalySeries [Macintosh Program Performs Time-Series Analysis, 1996]. Spectral periodicities were evaluated using the multitaper method [Dettinger et al., 1995]. Time steps for the low- and high-resolution records were 100 and 20 kyr, respectively. Significance was determined relative to red noise using a standard F test [Dettinger et al., 1995]. Cross-spectral analysis on the high-resolution record was carried out using standard Blackman-Tukey methods [Macintosh Program Performs Time-Series Analysis, 1996].
 All results are archived at the World Data Center-A for Paleoclimatology. Late Eocene data are included in the figures and discussion for comparative purposes, and detailed analysis of these data is the subject of another paper [Anderson and Delaney, 2005].
4.1. Nutrient (Organic Carbon) Burial
 Pr is defined as the biologically derived P in the sediments and includes all of the measured extracted P components except the detrital component (Appendix). Concentrations of Pr are from 5 to 48 umol g−1 sed; Pr MARs range from 4 to 30 μmol P cm−2 kyr−1. Pr MARs cycle through highs and lows from the early Oligocene to the middle Miocene (30–20 Ma, Figure 2c), with a slight increasing trend toward the early Miocene.
4.2. Productivity Proxies
 Baex is calculated as (Batotal)sample − (Ba/Al)bulk continental crustAlsample. This calculation assumes that the average continental crust abundance is representative of the detrital Ba component (1.43 mmol mol−1 [Turekian and Wedepohl, 1961; Wedepohl, 1995]). Baex concentrations range from 4 to 23 μmol Ba g−1, and the MARs range from 2 to 22 μmol cm−2 kyr−1 (Figure 2d). Similar to Pr, Baex cycles through highs and lows with a slightly increasing trend toward the early Miocene. Pr and Baex MARs correlate and have a slope of 2.1 ± 0.12 and an intercept of −1.5 ± 0.05 (r2 = 0.55, n = 258).
 Biogenic opal weight percent ranges from 2 to 28% (computed as SiO2, formula weight = 70 g mol−1), and the MARs range from 0.3 to 4 mmol Si g−1. Opal MARs cycle through highs and lows with no evidence of change at the O/M boundary (Figure 2e). CaCO3 weight percent ranges from nondetectable to 87% and the MARs from nondetectable to 11 mmol cm−2 ky−1. The CaCO3 MARs are relatively low through most of the Oligocene interval but have increased amplitudes starting about 1 m.y. prior to the O/M boundary (Figure 2g).
 Pr/Baex ratios have a long-term oscillation of about 5 m.y. and except for higher values in the early Oligocene samples, average about 2 mol mol−1 (Figure 3a). Pr/CaCO3 ratios are generally lower than 5 (umol wt %−1 CaCO3), but with peaks at 29.8 Ma, 29.5 Ma, ∼28.5–27.5 Ma, ∼26–24 Ma, and ∼22–21 Ma (peak ratios up to 60, Figure 3b). Implications of these ratios are discussed in section 5.1.
 Average crustal abundances of U/Al and Mn/Al were used to calculate enrichments of U and Mn (1.29 μmol mol−1 and 8.6 mmol mol−1, respectively [Turekian and Wedepohl, 1961; Taylor and McLennan, 1995; Wedepohl, 1995]). The U enrichment values for this site fell above one with the exception of a single point (Figure 3b, Appendix). U enrichments (relative to average crustal abundance of 1.29 μmol mol−1) were generally close to one with an average of 1.6 ± 0.4 μmol mol−1 and a range from 0.7 to 4 (Figure 3c). Mn enrichments ranged from 2 to 14, with the highs generally correlating with periods of U enrichment (Figure 4 and auxiliary material).
4.4. Spectral Analysis
 For 30–20 Ma (100 kyr resolution), spectral power occurs for opal, Pr, Baex, and detrital MARs at about 1.2 and 0.4 m.y. (Table 1). U enrichment and Pr/Baex have power at 0.4 m.y. (Table 1). CaCO3 shows no significant spectral power over this time period. The 1.2 m.y. and 400 kyr periodicities are amplitude modulation of obliquity and eccentricity, respectively [Olsen and Kent, 1999].
Table 1. Spectral Analysis of Site 1090 Using the Mulitaper Method [Dettinger et al., 1995] With Macintosh Program Performs Time-Series Analysis a
Number is the significance of periodicity relative to red noise. All data were detrended and resampled evenly at the resolution of the records (100 kyr for the 30–20 Ma interval and 20 kyr for the 24.5–22 Ma interval).
 Spectral analysis of the high-resolution section spanning the O/M boundary (24.5–22.5 Ma, 20 kyr resolution) shows spectral power at 124 and 95 kyr for most of the proxies (Table 1), indicating significant eccentricity forcing [Olsen and Kent, 1999]. Cross-spectral coherence is evident in the high-resolution records for Pr, Baex, opal, and δ13C at 400, 125, and 95 kyr, although the phase relationships are variable (Table 2). Most of the phase offsets are not greater than the record resolution except for eccentricity periodicities for Pr, Baex, and opal MARs, which precede δ13C by about 80 kyr.
Table 2. Cross-Spectral Coherence for the Time Period of 24.5–22 Ma, Using Blackman-Tukey Method With AnalySeriesa
 U and Mn enrichments indicate the redox state at the sediment-water interface [Calvert and Pedersen, 1993; Morford and Emerson, 1999; Mangini et al., 2001], which is important for describing water mass characteristics, as well as assessing the stability of barite (see below). U is enriched in sediments at redox potentials just below Fe but prior to sulfate reduction [Klinkhammer and Palmer, 1991; Rosenthal et al., 1995]. Thus U enrichment provides another tool for assessing whether depositional conditions favored barite mobilization, thus compromising the Baex measurements (see below). Low U enrichments persisted throughout the Oligocene and early Miocene samples.
 Because U and Mn enrichments occurred by two different mechanisms, their combined records provide additional information. The U came from the water column, whereas Mn was mobilized within the sediments and reprecipitated as a result of a redox boundary within the sediments or the overlying water column (for additional details see [Nilsen et al., 2003; Anderson and Delaney, 2005]). Small coincident enrichments in both U and Mn for Site 1090 samples (Figure 4) are strongly suggestive of an oxidized water column above moderately reduced sediments. This result contrasts with measurements in the late Eocene sediments from Site 1090 [Anderson and Delaney, 2005], where U enrichment persisted with values of 4–6 (Figure 3c), and Mn enrichments were not coincident with U enrichments, suggesting that the overlying water column was oxygen-depleted.
 The Pr/Baex ratio can indicate two different processes: (1) mobilization of Ba under reducing conditions or (2) the fraction of organic carbon buried (as indicated by Pr) relative to the export productivity delivered to the sediments (as indicated by Baex). The likelihood of mobilization can be assessed using sulfate pore water concentrations, the sedimentary enrichment of U, and by comparing the values to ranges typical of the present-day ocean. Pore water sulfate concentrations are all close to seawater (25–27 mM [Gersonde et al., 1999]) and as noted above U enrichments are low. Expected P/Ba ratios can be calculated from ratios of organic carbon to Ba taken from the literature [Francois et al., 1995; Dymond and Collier, 1996; McManus et al., 1998], using the C:P Redfield ratios of 117:1 [Anderson and Sarmiento, 1994]. P/Ba ratios calculated in this way range from 2 to 12 mol mol−1 and are based on two assumptions: that the Redfield ratio has not changed through time and the range of burial conditions for barite in the modern ocean is similar to the past. The ratios in this study are less than 10 (Figure 3a), and this result, combined with redox indicators suggesting oxic to mildly reducing conditions, indicates there has been no Ba remobilization.
 If there has been minimal Ba remobilization, the ratio of Pr/Baex provides an assessment of the burial of organic carbon (Pr) relative to export productivity (Baex) [Nilsen et al., 2003; Anderson and Delaney, 2005]. This proxy is based on two other assumptions: no mobilization of Pr and no change in the burial efficiency of Baex because of shifts in deepwater concentrations. The former assumes the efficient sedimentary transfer of organic P to authigenic P, with little to no preferential regeneration within the sediments. This has generally been found to be the case even with anoxic overlying waters [e.g., McManus et al., 1997; Anderson et al., 2001; Filippelli, 2001], although there appear to be some exceptions [e.g., Ingall et al., 1993; McManus et al., 1997; Slomp et al., 2002]. Oceans are undersaturated with respect to Ba, and recent literature indicates that a number of factors control Ba burial [e.g., Schenau et al., 2001]. However, globally distributed, deep-water, sediment trap data yield a consistent relationship between export production and deep-ocean barium accumulations irrespective of seawater Ba concentrations [e.g., Francois et al., 1995], lending confidence to at least a first-order relationship between export productivity and Ba burial.
 Pr/Baex ratios in the early Oligocene samples are similar to the late Eocene samples (Figure 3a). A shift to lower ratios of Pr/Baex occurred at about 28.5 Ma (Figure 3a), suggesting that the efficiency of organic carbon burial relative to the overlying export productivity decreased. This signals a potential change in the hydrography of this region that impacted organic carbon burial.
 Increased Pr/CaCO3 ratios highlight times of carbonate dissolution or significant biological shifts to noncalcareous producers. This ratio may be used in conjunction with cross plots of Pr with CaCO3 to distinguish these factors (Figures 3d and 5). (Any minor component normalized to CaCO3 should yield similar trends. Baex gives the same result.) With the dissolution of CaCO3, Pr per gram of sediment will appear to increase while in fact the diluting component, CaCO3, is dissolving. The strong negative correlation between the weight percent CaCO3 and Pr is illustrated for the interval from ∼26–24 Ma and is best explained by dissolution (Figure 5 and Table 3). In conjunction with high Pr/CaCO3, three significant periods of dissolution are suggested by the negative correlation between Pr and CaCO3: ∼28.5–27.5 Ma, ∼26–24 Ma, and ∼22–21 Ma (Figures 3d and 5 and Table 3). There are also apparent dissolution peaks at 29.8 Ma and 29.5 Ma, but the limited data points make regressions impossible.
Table 3. Correlation of Pr to Calcium Carbonate During Periods of High Pr/CaCO3a
Number of Samples
A strong negative relationship suggests dissolution of CaCO3 during that interval. Units are mmol wt %−1 (to convert to mmol mol−1, multiply by 10).
−0.50 ± 0.07
44 ± 1.0
−0.23 ± 0.02
25 ± 0.4
−0.67 ± 0.08
49 ± 1.4
5.2. Deepwater Characteristics
 Corrosivity (regional CCD trends) and δ13C are often used to evaluate water “age” and potential paths of deep-water flow. Their use stems from observations within the modern ocean that as deep waters age, as a result of organic matter degradation and the release of isotopically light pCO2, corrosivity increases and δ13C decreases along the flow path. Imposed on these trends are global shifts due to reservoir and input changes [e.g., Broecker and Peng, 1982; Berner, 1999]. The CCD deepened in all the ocean basins throughout the Oligocene [Van Andel, 1975]. CaCO3 MARs from Site 1090 do show a substantial increase in the earliest Oligocene samples, but the MARs are low again after the hiatus (30 Ma), when the global CCD was thought to be deepest [Van Andel, 1975]. Furthermore, significant periods of dissolution from 28.5–27.5 Ma, ∼26–24 Ma, and ∼21.8–20.5 Ma, along with two short spikes at 29.8 Ma and 29.5 Ma, suggest that CaCO3 deposition was not being controlled solely by the global CCD; local CCD fluctuations and changes in calcareous organisms' productivity were also important (Figures 2g and 3d and Table 3).
 Because Site 1090 is located close to the hypothesized South Atlantic CCD during the Oligocene and early Miocene (around 3700 m water depth, [Van Andel, 1975]), dissolution events recorded in the sediments may monitor vertical shifts in the boundary between (and thus relative volume of) two different deepwater masses. This could be evidence of a water structure that was substantially different than today, with older northern component water (NCW) and younger southern component water (SCW). Note that the ages of these deepwater masses would likely have differed from modern without an ACC. The “age” of modern Southern Ocean deep waters is a consequence of the incorporation of older Pacific Ocean and Indian Ocean waters into the circumpolar current [Döös, 1995; Sloyan and Rintoul, 2001], as well as the surface-to-deep mixing [Wunsch, 1998; Arhan et al., 1999; Gille, 1999] that delivers nutrient-enriched, light δ13C waters to the shelf areas where deep water is formed.
 Benthic foraminiferal δ13C gradients between the North and South Atlantic in the early Oligocene suggest oceanic reorganization with a presumed increased influence of a northern component water (NCW) [Miller and Fairbanks, 1983, 1985]. Although sedimentological evidence suggests that NCW existed as early as the early Oligocene [Davies et al., 2001], neither its strength nor its composition (which would dictate the observed gradients) are constrained. Benthic foraminiferal δ13C data from this site (23–15 Ma, with modified age model), compared to other sites in the Atlantic, Indian, and Pacific oceans, do not indicate a gradient [Thomas and Gooday, 1996; Billups et al., 2002]. However, it is difficult to find sites with similar depths for comparison, and the source waters may have had different δ13C signatures than modern waters, further complicating the interpretations [Raymo et al., 2004]. Considering that the δ13C of the downwelled water is not at all constrained, and even small changes in the end-member δ13C would affect observed gradients [Wunsch, 2003], with δ13C alone, it would be difficult to know if NCW influenced this site. Furthermore, models indicate that NCW flowed into the Pacific via the Central American Seaway until the Miocene [Nisancioglu et al., 2003], not into the South Atlantic.
 Alternatively, SCW to the site could have shifted between a more direct route following the Agulhas Ridge (short route, Figure 1) and a more circuitous route in which the waters first traveled north, entering the South Atlantic via the Romanche Fracture Zone, and then moved south again along the African coast (long route, Figure 1). The longer SCW route occurs within the modern ocean, but only a small portion of the deep waters from the north cross the Walvis Ridge [Siedler et al., 1996]. However, there is evidence within the eastern equatorial Atlantic that the flow of SCW via the Romanche Fracture Zone was greater from the late Eocene through the late Miocene [Wagner, 2001]. The Walvis Ridge was formed by a migrating hot spot [Zachos et al., 2004] and thus could have been more of a barrier than at present. However, there are passages within the Walvis Ridge at modern depths greater than 3000 m [Zachos et al., 2004], and because the original igneous intrusions occurred in the Maastrichtian, most of the cooling and deepening of the crust would have occurred by the Oligocene.
 In both the long and short route scenarios, the older, more corrosive deep waters might be expected to have been more reducing as well. In fact, a detailed examination of the CaCO3 MARs and the U and Mn enrichment factors (24.5 to 22 Ma, Figure 4) indicate higher, not lower, rates of deposition of CaCO3 during periods of U enrichment (more reducing conditions). This suggests that less corrosive water was more reducing. Warmer, saline deep water would have been consistent with these observations. Perhaps the predominant water downwelled within the Southern Ocean alternated between colder, well-oxygenated deep water delivered via the “long route” and warmer, more saline waters delivered via the “short route.”
 With these scenarios, fluctuations solely related to deep-water signatures are best delineated by U enrichment because CaCO3 MARs were also affected by surface water characteristics (section 5.3). Strong 400 kyr cyclicity evident in U enrichment is not apparent for CaCO3 MARs because dissolution events occurred at the beginning and end of this record (26–24 Ma, 21.8–20.5 Ma, Table 3 and Figure 3b). The strong CaCO3 MAR cyclicity evident at 100 kyr periodicities over the high-resolution section was mostly likely a result of both surface and deepwater processes (Table 1 and Figures 4 and 6).
 In the modern Southern Ocean, very cold but relatively fresh bottom waters are formed in the Weddell Sea, whereas a warmer, saltier Antarctic Bottom Water (AABW) is formed in the Ross Sea and on the Adélie coast, with the Adélie coast contributing about 25% of the total AABW [Döös, 1995; Sloyan and Rintoul, 2001]. These waters are mixed within the circumpolar current, changing their general chemical and physical character [Döös, 1995; Sloyan and Rintoul, 2001]. As mentioned in section 1, the mixing itself is a function of a lack of landmasses (allowing unrestricted winds) and bathymetric irregularities within and down flow of the Drake Passage [Wunsch, 1998; Arhan et al., 1999; Gille, 1999]. Thus the distinct character of these two water masses in the Oligocene and Miocene may have resulted from a lack of (or lesser) mixing resulting from a closed Drake Passage or a less intense ACC. The greater dominance of warmer, dense waters (more reducing) in the Site 1090 Eocene record compared to the Oligocene and early Miocene records would be consistent with a decreased influence of these waters or a greater mixing of bottom waters toward the present.
 Alternatively, it can be argued that the U enrichment associated with high CaCO3 MARs resulted from oxidation of organic matter delivered with the calcareous sediments. Two lines of evidence suggest otherwise. First, there is no evidence of CaCO3 dissolution, which would be expected with proton production during organic carbon oxidation; second, organic carbon burial (P MARs) was lower during periods of high CaCO3 MARs and U and Mn enrichment (Figure 4). Thus organic carbon delivery was not driving the redox chemistry during periods of increased CaCO3 sedimentation. Furthermore, the lower redox conditions suggested by U enrichments did not facilitate the preservation of organic carbon (Figures 2c and 3c). Higher U and Mn enrichment coincident with CaCO3 preservation and low U and Mn enrichment coincident with CaCO3 dissolution (Figure 4) suggest changes in deepwater characteristics.
5.3. Surface Water Characteristics
 Modern Southern Ocean hydrography has a profound effect on the horizontal and vertical distribution of water masses and their associated nutrients. The modern frontal system blocks the movement of nutrient-depleted surface waters into the Southern Ocean [e.g., Toggweiler, 1994; Toggweiler and Samuels, 1995]. Deep mixing of older waters brings abundant nutrients to the surface and fuels biological productivity, particularly by diatoms [e.g., Broecker and Peng, 1982]. However, because this region is Fe-limited [Archer and Johnson, 2000; Fung et al., 2000; Aumont et al., 2003; Coale et al., 2004], diatoms tend to produce more shell per unit carbon [Hutchins et al., 1998; Takeda, 1998]; thus, although siliceous sedimentation is abundant (the Southern Ocean is a sink or “trap”), less organic carbon is transported to deep waters, and as a consequence of deep mixing (i.e., no annual thermocline), even less is buried [Antia et al., 2001]. Thus the modern Southern Ocean decouples organic carbon and silica burial, and decreases the burial of organic carbon in both the absolute and relative sense. Evolution of the Southern Ocean toward a modern hydrography should be evident in siliceous (opal MARs) and organic carbon burial (P MARs), as well as their relative burial (Pr/Baex). In addition, shifts in the abundance of opaline sedimentation versus carbonate sedimentation may reveal something about the nutrient contents upwelling at Site 1090.
 The decrease in opal MARs at Site 1090 (presently located within the sub-Antarctic zone) in the Oligocene and early Miocene, as compared to the late Eocene (Figure 2e [Anderson and Delaney, 2005]), might suggest that the hydrography in this region was evolving toward a modern frontal system, with opaline sedimentation becoming concentrated south of the polar front (i.e., a silica trap). However, extensive opaline sedimentation throughout the equatorial Atlantic during both the Eocene and Oligocene argues against the Antarctic functioning as a silica trap from the early Oligocene to early Miocene [Barron and Baldauf, 1989; Wagner, 2001]. Furthermore, the latitudinal distribution of opaline sediments within the Southern Ocean does not appear similar to the modern distribution until the middle Miocene to early Pliocene [Kennett, 1977; Abelmann et al., 1990].
 Changes in the amount of organic carbon burial (P MARs, Figure 2c) and the efficiency of organic carbon burial (Pr/Baex, Figure 3a) suggest that although a modern frontal system may not have been established until after the middle Miocene, the regional hydrography was changing. The high organic carbon burial rates (P MARs) leading up to the Oi-1 glacial event [Anderson and Delaney, 2005] are not evident within the posthiatus record (Figure 2c). Furthermore, although high burial efficiencies (Pr/Baex) existed within the early Oligocene, they were never as high as the observed efficiencies at ∼37 Ma [Anderson and Delaney, 2005], and they decreased from ratios of 8 to ratios of ∼2 by 27 Ma (Figure 3a). Finally, from 30 to 20 Ma at this site, Pr and Baex are significantly correlated (r2 = 0.55, slope = 2.1 ± 0.12, intercept = −1.5 ± 0.05, n = 258); have cross-spectral power at 400, 125, and 95 kyr; and are in phase within the error of the record's resolution (Table 2), while prior to 33 Ma, they are not correlated [Anderson and Delaney, 2005].
 Also in contrast to prehiatus records, high P and opal MARs are associated with low CaCO3 MARs, while high CaCO3 MARs are associated with low P and opal MARs (Figures 2 and 6). This suggests that opaline sedimentation, or the hydrographic conditions that favor opal-secreting organisms, functioned more effectively in the burial of organic carbon from 24.5–22 Ma than in the modern ocean or in the late Eocene. At 100 kyr periodicities, opal and P MARs precede the δ13C fluctuations throughout the high-resolution section (24.5–22 Ma) including the Mi-1 glacial event by about 80 kyr (Table 2). Thus the associated burial of organic carbon with biogenic opal may have provided a forcing mechanism for climatic change by the drawdown of pCO2. The rapid transport and burial of diatom-associated organic carbon is commonly assumed for upwelling systems of the modern ocean, but major periods of the late Eocene did not have concurrent opal and P burial rates (Figures 2 and 3 [Anderson and Delaney, 2005]). Also, as mentioned above, organic carbon and opal burial are not correlated in the modern Southern Ocean south of the polar front because of the deep mixing [Antia et al., 2001]. Thus the correlation of export productivity and organic carbon burial during the Oligocene and early Miocene could have had significant climatic impacts that are not present in the modern ocean.
 Nutrients are supplied to the surface ocean by upwelling and riverine inputs. With the opening of passages in both the Arctic and Antarctic and closing of equatorial regions, meridional gradients would have increased, which in turn would have increased atmospheric circulation, including the monsoonal system [Kennett, 1977; Berger and Wefer, 1996; Ruddiman et al., 1997; Thiede et al., 1998; Wagner, 2001; Sluijs et al., 2003]. Detrital sedimentation (Figure 3d) fluctuates on both 400 and 100 kyr cycles, highly suggestive of precessionally modulated monsoonal inputs (Table 1). In addition, cross-spectral analysis indicates coincident fluctuations between the detrital, Pr, Baex, and opal MARs. Furthermore, highs in Site 1090 CaCO3 and opal MARs alternate, suggesting a shift between nutrient-poor and nutrient-rich surface waters. Both records have clear orbital forcing at eccentricity periodicities (124 and 95 kyr). A comparison of the records shows that peaks in carbonate are associated with lows in the opaline sedimentation that are independent of the overall sedimentation rate (Figures 2 and 6). Within the eastern tropical Atlantic, Wagner  observed alternating cyclic fluctuations in opaline and calcareous sediment deposition in the Oligocene (potentially starting in the late Eocene), which he hypothesized was related to changes in nutrient supply from tropical African runoff associated with changes in the orbitally driven intensity of the monsoon. Wagner  suggests that siliceous deposits are consistent with higher nutrient levels and greater productivity, whereas calcareous deposits are associated with lower nutrient levels. Our records would be consistent with opaline sedimentation being associated with increased weathering caused by greater monsoon intensity. However, during periods of increased meridional gradients there was likely to have been greater meridional overturn as well; thus, neither the equatorial or sub-Antarctic records can easily separate the roles of weathering (higher concentrations of nutrients within upwelled waters) and delivery (greater upwelling).
 It is the amount of organic carbon buried relative to total carbon (inorganic carbon plus organic carbon) that drives the removal of atmospheric pCO2. The observation of coincident siliceous and associated organic carbon burial, potentially associated with monsoonal strength, in two distinctly different regions (the equatorial Atlantic and sub-Antarctic zone), suggests a powerful mechanism for the drawdown of atmospheric pCO2. Others have hypothesized that eccentricity's modulation of precession could induce increased weathering and delivery of nutrients (e.g., opal and P) to the ocean, thus driving opaline sedimentation and coincident organic carbon burial [Wagner, 2001; Wang et al., 2003]. Filtered P and opal MAR records at 400 kyr periodicities (Figure 7) indicating similar increases and decreases in δ13C, P, and opal MARs variability and cross-spectral coherence of Pr, Baex, and opal MARs at 100 kyr periodicities support this hypothesis. The positive excursion of δ13C with opaline sedimentation was observed for the Pleistocene [Wang et al., 2003], and the authors suggest that because silica is primarily delivered by tropical rivers, the oceanic δ13C is related to low-latitude events such as the monsoons. However, over timescales greater than the mixing time of the oceans, the delivery of upwelled, nutrient-rich waters into the photic zone will dictate the type and amount of regional productivity. Thus hydrography is equally important. For example, hydrographic conditions that facilitate the efficient burial of opaline sediments without the burial of organic carbon, similar to the modern Southern Ocean, will have less impact on the burial of organic carbon relative to inorganic carbon. Thus, over the time period of our study, this multiproxy data set suggests that a combination of high- and low-latitude hydrographic conditions played a role in sensitizing the climate system. Even over glacial-interglacial events, the “leakiness” between the Southern Ocean and equatorial regions with respect to silica, resulting in more efficient burial of organic carbon within the equatorial regions (i.e., associated organic carbon and opal burial), has been hypothesized as a positive feedback mechanism in the drawdown of pCO2 [Sigman and Boyle, 2000; Matsumoto et al., 2002].
5.4. Hypothesized Hydrography of Site 1090
 Two hypothesized hydrographic modes, which fluctuated over 100 kyr cycles, are consistent with all the proxy data and with observed orbital forcing:
 1. Strong meridional gradients driven by precessionally modulated eccentricity orbital forcing would have created stronger monsoonal activity, which, in turn, would have been associated with higher levels of nutrients in surface waters, thus favoring opaline sedimentation. Greater north-south gradients would have been associated with vigorous cold, well-oxygenated deep waters forming in the Southern Ocean, sufficient to bathe Site 1090 with waters from the “long route.”
 2. Weaker monsoonal activity would have been associated with lower levels of nutrients in surface waters, thus favoring calcareous sedimentation. Deep waters at Site 1090 would have been warmer, more saline, and less oxygenated.
 In the modern ocean, deepwater masses from the Weddell Sea are colder and fresher, whereas those formed on the Adélie coast are warmer and saltier [Sloyan and Rintoul, 2001]. The latter may contribute up to ∼25% of the modern AABW global inventory. Because of deep mixing within the ACC, waters become more homogenized. One can imagine periods of greater meridional gradients and more intense monsoonal activity causing both an increase in the winds that drive bottom water formation [Wunsch, 1998] and an increase in freshwater inputs, but without the vigorous mixing that occurs within the modern Drake Passage and Scotia Sea [Haywood et al., 2000]. Freshwater inputs to the Indian Ocean might have decreased the salinity of regional source waters to the Southern Ocean, thus decreasing the potential for deep-water formation in that area, whereas increased winds would have promoted an overall increase in the amount of bottom water, perhaps within the Weddell Sea (mode 1, Figure 1 solid line).
 In contrast, during periods of decreased meridional gradients, weaker monsoons would have created a saltier, warmer water from the Indian Ocean, which would have cooled within the Antarctic and formed a warmer saltier bottom water (mode 2, Figure 1 dashed line). As in the modern ocean [Sloyan and Rintoul, 2001], this water could have flowed northward into the Indian Ocean and then westward, south of Africa and adjacent to the Agulhas Ridge. This could have occurred whether or not a Tethyan source of saline waters existed but would have been accentuated by an open eastern Tethys.
 During mode 1, the abundant formation of deep waters would have favored the characteristics of “long route” SCW at Site 1090. The deep-water signature would have been corrosive but well oxygenated. The dominance of colder deep water throughout the Southern Ocean could have produced a shallower annual thermocline, facilitating organic carbon preservation. Furthermore, nutrient-enriched waters associated with intensified monsoonal weathering, combined with increased N-S gradients driving increased upwelling, would have promoted opaline sedimentation.
 During mode 2, deep water would have been dominated by the characteristics of “short route” SCW. Source waters for the formation of deep water would have come from more saline, warmer waters that are less corrosive (i.e., relatively young) and less oxygenated. This could have resulted in a deeper thermocline that, combined with less nutrient-rich waters, could have favored calcareous production accompanied by less efficient burial of organic carbon.
5.5. Mi-1 Glacial Event
 The Mi-1 glacial event was a convergence of minimal eccentricity and low-amplitude variability in obliquity and resulted in cool summers and ice buildup to quantities equivalent to the late Miocene [Zachos et al., 2001b]. This glaciation was unique in the early Miocene and represented the only time from 26 to 20 Ma that minimal eccentricity and low-amplitude variability coincided [Zachos et al., 2001b]. At Site 1090, the Mi-1 glacial event was a time of decreased burial of organic carbon relative to inorganic carbon, even though in the absolute sense the burial of organic carbon increased (i.e., Pr/Baex are low, while P MARs are high, Figures 2c and 3c). This is at least suggestive that deep-water conditions were not as favorable to preservation, although productivity in this region was high during the Mi-1 event. Pagani et al.  found a similar increase in productivity, albeit coccolithophorid, at the same latitude in the western Atlantic (ODP, Site 516). The coincident highs prior to and during the Mi-1 event in both the opal and Pr MARs could have driven the observed amplification in δ13C that preceded the Mi-1 event [Zachos et al., 2001b], thus facilitating the drawdown of pCO2 and sensitizing the system to orbital forcing.
 Several lines of evidence from Site 1090 suggest that the Drake Passage was probably not open to deep water until after 20 Ma, or if it was, a modern frontal system did not yet exist.
 Deepwater characteristics in the Oligocene and early Miocene alternated between oxygen-rich and oxygen-depleted (U and Mn enrichment). This potentially suggests that there was less mixing of deepwater masses because of a less intense or nonexistent ACC which affected both the depth of mixing and the residence time of deep waters within the Southern Ocean.
 Opaline sedimentation was not limited to south of the polar front and was coupled to organic carbon deposition, whereas the modern Southern Ocean functions as a silica trap but, because of surface-to-deep mixing, does not bury organic carbon.
 Fluctuation between opaline and carbonate sedimentation at Site 1090 does not appear to have been dictated by the relative position of the frontal system, as in modern sediments, but rather was driven by changes in nutrient inputs from monsoonal weathering and/or increased upwelling.
 High MARs in CaCO3 alternate with high MARs in opal and Pr and have strong eccentricity periodicities. The 400 kyr filtered records of P and opal indicate similar patterns of variability to δ13C. This is consistent with observations in the eastern equatorial Atlantic [Wagner, 2001] suggesting that intensified monsoons modulated by orbital forcing delivered both silica and other nutrients to the ocean, thus promoting opaline sedimentation at the expense of CaCO3-secreting organisms. This alone would not have resulted in the increased burial of organic carbon in either the absolute or relative sense. However, hydrographic conditions in the Southern Ocean did not function as a silica trap, as in the modern ocean. Furthermore, as mentioned above, the ACC was probably not well developed, and the Southern Ocean may therefore have been more effective in burying organic carbon. The combination of increased opaline sedimentation with organic carbon burial would have driven increased δ13C as well as draw down pCO2. These factors could have provided the hypothesized sensitizing of the climate system by the nature and rate of carbon burial [Zachos et al., 2001b].
 This paper was greatly improved by careful reading by A. Allwardt, K. Billups, K. Faul, and E. Nilsen and by discussions with C. John and K. Faul. Arthur Arcinas processed samples. M. Pagani and an anonymous reviewer provided insightful scientific and editorial comments that have significantly improved this manuscript. Final editorial polish was added by L. Miller. R. Franks provided laboratory support and analytical insights. The Institute of Marine Sciences maintains the Marine Analytical Laboratory and its Plasma Analytical facility. This research used samples and data provided by the Ocean Drilling Program (ODP). The ODP is sponsored by the U.S. National Science Foundation (NSF) and participating countries under the management of Joint Oceanographic Institutions (JOI), Inc. Funding for this work comes from NSF OCE 9819114 (to M.L.D.).