RESULTS

The absolute timing and assigned ages of biostratigraphic datums often change with the refinement of successive chronostratigraphic studies. In the past decade or so, biostratigraphers have assigned significantly different ages with differences of as much as 2 m.y. to the same biostratigraphic datum (Fig. 7). It is, therefore, imperative to either identify the biostratigraphic time scale used when referring to ages of different age models or simply refer to the biostratigraphic zone within which an event lies. Fortuitously, the biostratigraphic datums of Raffi and Flores (1995) were also used in the age models for ODP Legs 138 and 154, placing their ages in a common biostratigraphic time frame with that of Leg 165. After establishing that the coccolith datums were reliable and corroborated by the magnetostratigraphy and foraminiferal datums, age models were generated for each site (Sites 998, 999, and 1000) (Fig. 8). The depth and calculated age are displayed in Figure 8B for every sample analyzed in this study.

Despite numerous turbidites observed in Hole 998A on the Cayman Rise (Yucatan Basin), the sedimentation rate for the middle to upper Miocene segment in Hole 998A is very linear with essentially a constant sedimentation rate of 0.94 cm/k.y. (Fig. 8B). Based upon this sedimentation rate, the lowest among the three studied Caribbean sites, the 50-cm-spaced samples correspond to a temporal resolution of ~53,000 yr.

The average sedimentation rate in the section of Hole 999A on Kogi Rise (Colombian Basin) is not as linear as Site 998 and shows the most variation of the three sites. Although the average rate is 2.09 cm/k.y., the sedimentation rates range from a low of 0.9 cm/k.y. (10.39-9.36 Ma) to 4.5 cm/k.y. (13.57-13.19 Ma) (Fig. 8B). The observed variability can be attributed to variable siliciclastic input at Site 999 because of the relative proximity of the Magdalena River mouth on the northern coast of Colombia. Because of a higher sedimentation rate in Hole 999A, the 50-cm-spaced samples in this study yield on average a temporal resolution of ~24,000 yr.

Among the three locations, the highest average sedimentation rate (4.35 cm/k.y.) is observed from ~13.3 to 10.7 Ma in Hole 1000A (Fig. 8B). This high sedimentation rate is likely related to the bank-derived neritic component added to the pelagic carbonates. This average sedimentation rate yields an average temporal resolution of 12,000 yr for the 50-cm-spaced samples of this study. As in Hole 998A, the sedimentation rate remains relatively constant. However, during Zone CN6 (10.39-10.71 Ma), the sedimentation rate was only 2.4 cm/k.y., about half the average rate observed for the middle to upper Miocene in Hole 1000A. Though Hole 998A displays the lowest resolution of the three sites, the samples available in that hole extend to 16 Ma as opposed to 14 and 13.8 Ma for Holes 999A and 1000A, respectively.

Extensive induration of the lower reaches of Hole 999A prevented complete disaggregation and determination of the coarse fraction, and for similar reasons, no coarse-fraction data exists for Hole 1000A. In samples that could be disaggregated and sieved, comparing the mass ratio of the sand-sized fraction to the bulk sample is used as a proxy for reduced carbonate preservation (Berger, 1970b; Bassinot et al., 1994, and references therein). Because foraminiferal calcium carbonate tests become fragmented with dissolution, the coarse-fraction weight percent of a given sample is expected to give some indication of its degree of preservation (i.e., decreasing sand content with increasing dissolution).

Most of the coarse-fraction data from Hole 998A varies within a narrow range (0.2%-3.5%); occasionally it exceeds 10% at the beginning and end of the record (Fig. 10A). Times of reduced coarse sediment occur as both short intervals and distinct low points. The coarse-fraction percent remains generally low from 13.6 to 9.5 Ma. A relative preservation index, assigned to Hole 998A planktonic foraminifers during the cruise, reveals a significant decrease in preservation between 150 and 160 mbsf (~10.5-12.0 Ma), bracketing the interval of the carbonate crash. At 9.5 Ma, the coarse-fraction percent quickly increases to a level similar to that prior to 14 Ma. Observations for the interval between 16.4 and 13.8 Ma are not as well supported by the data because of the low sample resolution (200-400 k.y.).

Coarse-fraction percent for Hole 999A is limited to an interval between 11.55 and 9.0 Ma (Fig. 10B). In this time span, a series of seven intervals characterized by a low coarse-fraction percent (<2%) was observed (Fig. 10B). The timing of these intervals generally corresponds to times of low coarse-fraction values in Hole 998A (Fig. 10A).

Reduction in the calcium carbonate weight percent is seen in the three sites (Fig. 9, Fig. 11). The carbonate content reflects proportional changes in the amount of carbonate and noncarbonate sediments. The greatest carbonate reductions are observed in Holes 998 and 999 during an interval between ~12.1 and 9.8 Ma, and referred to as the Caribbean carbonate crash. Hole 998A, drilled at 3101 m water depth and, therefore, the deepest site in this study, displays the largest amplitudes (0-80 wt%) in the carbonate-content variation. Prior to the Caribbean carbonate crash, carbonate-content values averaged ~80 wt% in Hole 998A (Fig. 9A). During the carbonate crash, carbonate-content values display several high-amplitude fluctuations. Eight samples within the carbonate-crash interval contain <5 wt% carbonate. Carbonate content returns to pre-carbonate-crash levels by 10 Ma.

Hole 999A, drilled at the slightly shallower depth of 2839 meters below sea level (mbsl), has carbonate-content reductions of a comparable scale to those in Hole 998A. Carbonate of the samples averages ~75 wt% before the carbonate crash, decreases to <5 wt% during the crash interval, and increases to 62 wt% by 9 Ma (Fig. 9B).

The carbonate content of Hole 1000A displays smaller variations than Sites 998 and 999 through the middle to upper Miocene transition. The carbonate is >80 wt% before the crash, decreases below 70 wt% during the crash, and recovers above 80 wt% after the crash (Fig. 9C). Although the magnitude of the change is not on the same scale as that of Sites 998 and 999, the 30 wt% decrease is significant because the water depth of Site 1000 is shallower than the modern calcite lysocline.

Figure 11 shows carbonate contents scaled to the range exhibited in each site. It is clear that the timing of intervals characterized by dramatic or significant carbonate reduction is remarkably similar at all three sites (Fig. 11). Carbonate weight percent in Hole 998A shows a very gradual 10% decrease from 16.4 to 12.4 Ma prior to the carbonate-crash interval. Then high-amplitude swings of carbonate values are encountered during a period initiated at ~12.1 Ma. During an interval lasting 1.9 Ma, carbonate values switch five times between 5 and 65 wt%. The incidents of minimum carbonate weight percent occur at 12.0-11.8, 11.6-11.4, 11.0-10.8, 10.6-10.5, and 10.3-10.1 Ma. These five episodes, characterized by carbonate minimum values within the Caribbean carbonate crash, occurred at a periodicity of 400-500 k.y. and are emphasized in Figure 11 by the shaded bars. Carbonate content recovers to pre-crash levels by 10 Ma.

The overall pattern of the carbonate change observed in Hole 999A is surprisingly similar to the one described earlier in Hole 998A. However, there are some subtle differences between the two holes. As in Hole 998A, a gradual but more rapid decrease in carbonate content is observed before the onset of the carbonate crash. The slow decline of carbonate content ranges from 75 wt% at ~13.9 Ma to 55 wt% at the onset of the crash at ~12.1-12.0 Ma (Fig. 11B). However, this precursor to the crash is more conspicuous in Hole 999A because the carbonate decrease prior to 12 Ma is steeper and punctuated by a series of carbonate values lower than 40 wt%. This series of significant carbonate reductions are approximately spaced at a periodicity of 100-200 k.y., particularly in the interval from 13.6 to 12.5 Ma. Within the carbonate-crash interval, high-amplitude carbonate fluctuations are observed in Hole 999A (Fig. 11B). The ages at which the carbonate contents dip below 10 wt% within the highly variable carbonate-crash interval are 12.0, 11.6-11.4, 11.2-11.1, 11.0-10.8, and 10.1 Ma, similar to the five episodes observed in Hole 998A. Carbonate content recovers from the crash beginning at 10.0 Ma with the exception of one more interval of reduced carbonate weight percent at 9.4-9.2 Ma.

The carbonate-content values in Hole 1000A are as high as ~90 wt% at 13.8 Ma, but then drop to 70 wt% by 10.9 Ma, before increasing again to 90 wt% by 9.0 Ma (Fig. 11C). Superimposed on the general trend are highly variable carbonate contents with several minima at ~72 wt% prior to 13.8 Ma and decreasing below 65 wt% at 12.6, 12.4-12.3, 12.2, 12.0, 11.7, 11.5, 11.1, 10.9, 10.5, 10.2, and 10.0 Ma. Two additional minima occur at 9.8 and 9.4 Ma. With the exception of these latter two minimum values, the carbonate-content values following the carbonate crash do not experience the same degree of variation as seen in the carbonate content prior to the carbonate minimum.

Similarities between the pattern of carbonate-content variation among the three sites were compared statistically. Correlation coefficients and variables derived to test for the significance of the correlation coefficient are listed in Table 3. Based on the test for significance outlined by Swan and Sandilands (1995), the carbonate-content curves for Holes 998, 999, and 1000 are correlated between one another and the null hypothesis is rejected at 99% confidence.

The carbonate records in Figure 9 and Figure 11 display the variations of weight percent CaCO₃ relative to the bulk sample. In an effort to minimize the effect of carbonate dilution by noncarbonate components, CO₃ MARs were also calculated (Fig. 12). Prior to the carbonate crash, the CO₃ MARs averaged 0.75 g/cm² per k.y. and remained constant in Hole 998A (Fig. 12A). Carbonate accumulation then dropped to almost 0 g/cm² per k.y. at 12.0 Ma (Fig. 12A). Accumulation rates remained low and highly variable throughout the 2 m.y. of the carbonate crash. Within this period, the CO₃ MAR oscillated between zero and 0.7 g/cm² per k.y. Five episodes during which only trace amounts or no carbonate accumulation are observed occurred at 12.1-11.8, 11.6-11.3, 11.1-10.8, 10.6-10.5, and 10.3-10.1 Ma. By 10 Ma, the carbonate accumulation rates recovered to precrash levels.

The pattern in carbonate accumulation rates in the Colombian Basin (Site 999) is very similar to that of the Yucatan Basin (Site 998) (Fig. 12A, B). The higher sedimentation rates at Site 999 (2.09 cm/k.y.), twice as high as the rates at Site 998 (0.92 cm/k.y.), translate into overall higher CO₃ MAR at Site 999. In Hole 999A, CO₃ MARs gradually decrease prior to the carbonate crash. The decline is observed from the beginning of the data set at 14.2 Ma but especially after 13.8 Ma with a CO₃ MAR of nearly 2.0 g/cm² per k.y. to 1.0 g/cm² per k.y. prior to the onset of the carbonate-crash interval at ~12.1 Ma (Fig. 12B). The carbonate-crash interval is characterized by high-amplitude variations in the CO₃ MAR and ends at 10.0 Ma. Similar to Site 998, significantly lower accumulation rates occur at near 12.0, 11.6-11.5, 11.0-10.8, and 10.2-10.1 Ma (Fig. 12B). However, the CO₃ MAR full recovery is only reached after ~9.4 Ma, a date that also postdates the nadir of the carbonate-crash interval in the eastern equatorial Pacific. Although most of the decreased CO₃ MARs occur during the 2-m.y. period of the carbonate crash, significant drops of CO₃ MAR are already observed in the time prior to the crash at 13.55, 13.05, and 12.55 Ma. These episodes of low CO₃ MAR appear to occur at a frequency of ~500 k.y. (Fig. 12B). Miocene carbonate cyclic variations at the frequency of 400-500 k.y. have been reported in deep-sea sediments in other parts of the Atlantic Ocean (e.g., Zachos et al., 1997). These earlier precursors of CO₃ MAR reductions observed in Hole 999A do not occur in Hole 998A (Fig. 12A), possibly because of the partial isolation of the Yucatan Basin from the southern Caribbean basins.

The CO₃ MAR in Hole 1000A (927 m of water depth) is much higher than at the two deeper Sites 998 and 999, with values ranging from ~3.0 to 5.5 g/cm² per k.y. (Fig. 12C). The lowest CO₃ MAR in Hole 1000A is, therefore, significantly higher than the highest CO₃ MAR in Holes 998A and 999A. These very high CO₃ MARs can be explained by the shallow depth of Site 1000 and by its proximity to shallow carbonate platforms that shed large volumes of neritic aragonite and perhaps magnesian calcite-rich sediment to the adjacent basins. Similar to Sites 998 and 999, the carbonate-crash interval between 12 and 10 Ma in Hole 1000A is characterized by an overall lower CO₃ MAR and a series of high amplitude (1.5 g/cm² per k.y.) variations in CO₃ MAR. As at Site 998, and in particular at Site 999, episodes characterized by significantly lower CO₃ MARs occur at 12.1-11.9, 11.6-11.5, 11.1-10.8a, and 10.2-9.9 Ma (Fig. 12C).

X-ray diffraction reveals the presence of aragonite in Hole 1000A. The aragonite component of the carbonate fraction in Hole 1000A was most likely produced on adjacent carbonate bank tops and exported offshore to the Site 1000 location. The production and, therefore, the export of bank-derived aragonite and magnesian calcite is tied to bank-top flooding and directly linked to sea-level fluctuations (e.g., Droxler et al., 1983; Schlager et al., 1994, and references therein). Because a magnesian calcite peak could not be detected from the low magnesian calcite peak, magnesian calcite is probably absent or only present in trace amounts. Because aragonite is metastable relative to low magnesian calcite, the water column becomes undersaturated with respect to aragonite at much shallower depths than does low magnesian calcite (Droxler et al., 1991). Aragonite MAR is a reliable proxy for carbonate preservation/dissolution (Schwartz, 1996). Figure 12D shows variations in aragonite MAR from 13.1 to 8.9 Ma in Hole 1000A. The points of lowest aragonite MAR occur at 13.1-12.9, 12.3-12.0, 11.8-11.6, 11.1-10.8, 10.6-10.3, and 10.0-9.3 Ma. The first four episodes of reduced aragonite accumulation during the interval of the carbonate crash correspond relatively well to the first four episodes of carbonate reduction observed in Holes 998A, 999A, and 1000A (Fig. 12). However, the youngest 0.7-k.y.-long interval (between 10 and 9.3 Ma), characterized by reduced aragonite accumulation, occurs at a time when the carbonate had already fully recovered. This interval of low-aragonite MAR could be explained by an overall sea-level lowstand at the beginning of the late Miocene (Haq et al., 1987). As shown below in the isotope result section, this interval corresponds to some relatively heavy ¹⁸O values at Sites 998 and 999 (Fig. 13), likely corresponding to a marked sea-level lowstand.

Zero to 20 tests of P. wuellerstorfi were found in samples with an average of three per sample. Some of the samples in Holes 999A and 1000A were too indurated to allow separation of coarse from bulk samples; therefore, O and C isotopes are available only for some part of the upper middle/lower upper Miocene in those holes. Diagenetic effects on measured isotopes, if present, would likely alter the oxygen isotopes. However, because the ¹⁸O values derived from these samples are quite similar to contemporaneous values of compiled ¹⁸O records of J.C. Zachos (unpubl. data), the values of both ¹⁸O and ¹³C that are included in this study appear to have been spared from significant diagenetic alteration and are thought to be valid.

¹⁸O

In waters below the thermocline, the oxygen isotope ratio incorporated into benthic foraminifer tests is influenced to a lesser extent by temperature fluctuations. Oxygen isotope values in this study, therefore, are expected to reflect changes in ice volume and serve as a good proxy for eustatic sea-level changes.

In spite of the low time resolution of the isotopic data set in Hole 998A, an overall increase of the ¹⁸O values is clearly observed from 15.5 to 9 Ma (Fig. 13A). The ¹⁸O values in Hole 998A, as light as ~0.8 at 15.5 Ma, (Fig. 13A), become progressively heavier, reaching ~2.5 at 9 Ma. In Hole 999A, the ¹⁸O record is limited to the interval between 11.6 and 9 Ma (Fig. 13B). It is reassuring that the range of ¹⁸O values (from 1.2 to 2.3) in this interval in Hole 999A is similar to the range observed in Hole 998A (Fig. 13A, B). A gradual increase in ¹⁸O values is also observed in both holes from 10 to 9.0 Ma, with some of the heaviest ¹⁸O values between 9.4 and 9.0 Ma. The heaviest ¹⁸O values in Hole 999A are observed in a short interval between 11.4 and 11.2 Ma.

As expected because of the relative shallow depth of Site 1000, the benthic foraminifer ¹⁸O values in Hole 1000A, ranging from 0.65 to 1.78, are overall lighter than those in Holes 998A and 999A (Fig. 13C). Because coccolith plates and planktonic foraminifers dominate the calcareous portion of the bulk sediment and their tests are mineralized close to the ocean surface, the bulk sample ¹⁸O values in Hole 1000A are considerably lighter than the benthic values in the same hole by ~3.0. The four ¹⁸O records in Holes 998A, 999A, and 1000A display a plateau of ¹⁸O values between 12.4 and 10 Ma, varying within a 1.0 range and systematically shifted relative to the water depth of the sites and the nature of the analyzed material.

¹³C

The ¹³C at Sites 998 and 999 ranges from 1.5 to 0.1 (Fig. 14A, B). In Hole 998A, the data set spans a 6-m.y.-long interval between ~15 and 9 Ma and illustrates an overall trend where the heaviest values are observed in the intervals older than 13 Ma and younger than 9.5 Ma, loosely bracketing the carbonate-crash interval. Most of the lighter values in Hole 998A occur during the carbonate-crash interval and reach 0.6 ± 0.1 at 13.6, 12.3, 12.05, 11.8, 10.7, and 10.5 Ma, at times when the CO₃ MARs are minimum (Fig. 14A). However, these lightest ¹³C values might not be representative of the five episodes characterized by some of the lowest CO₃ MARs because benthic foraminifers are usually absent in the samples. It may be possible that the ¹³C for these episodes decreased to values as light as the lightest two values (~0.3) in Hole 998A, surprisingly observed in a short interval between 9.8 and 9.6 Ma during which the CO₃ MARs are among the highest observed rates (~1.5 g/cm² per k.y.).

Because samples older than 11.5 Ma in Hole 999A are highly indurated, the benthic ¹³C isotopic data set for Site 999 is limited to the interval between 11.5 and 9 Ma. Within the carbonate-crash interval in Hole 999A, the ¹³C values fluctuate between 0.1 and 1.4, while the ¹³C values in Hole 998A vary only between 0.3 and 1.1. The lightest ¹³C values (~0.1-0.3) in Hole 998A are found at 11.15, 10.9, 10.55, 9.85, and 9.35 Ma (Fig. 14B). Most of these lighter values in Hole 998A occur during some episodes within the carbonate-crash interval. However, as in Hole 998A, the interval characterized by some light values between 10.05 and 9.6 Ma corresponds to a time when the CO₃ MAR had already recovered subsequently to the carbonate-crash interval.

Only a limited number of samples from Hole 1000A were analyzed for their benthic ¹³C because of their overall high degree of induration. With the exception of the lightest ¹³C value of -0.71, the range of the ¹³C values in Hole 1000A, 1.31-0.0, is about equivalent to the range observed in Hole 999A (Fig. 14C). The samples with the lightest values occur at 11.9, 11.4, and 10.8 Ma, within some of the episodes of the carbonate-crash interval (Fig. 14C). The most negative and lightest ¹³C value at 11.9 Ma can be related to the water depth (927 m) of Site 1000, probably within the oxygen minimum zone, a level usually characterized by some of the lightest ¹³C in the water column.