DISCUSSION

U Loss and Gain

The majority of samples analyzed in this study have clearly undergone some form of diagenesis which has perturbed their U-Th systematics. This is most clearly demonstrated by (²³⁰Th/²³⁸U) values in excess of secular equilibrium, and is also shown by the anomalous ²³⁴U(T) values of many samples.

The high (²³⁰Th/²³⁸U) values of Site 1005 indicate that either U loss or ²³⁰Th gain has occurred. Thorium is generally a highly immobile element, so U loss is the more likely case. Future discussion here will make the assumption that U loss has occurred, rather than ²³⁰Th gain, but if the situation were reversed much of the interpretation would stand unaltered.

Loss of U makes the apparent age too great and therefore increases the ²³⁴U(T). This loss can be investigated by constructing curves on a ²³⁴U(T)-vs.-age diagram for each sample (Henderson et al., 1993) (Fig. 3). The position of the curve is controlled by the measured ²³⁴U(0) value and samples move toward the bottom left as U is returned to the age calculation to correct for U lost from the sample. Because U loss is unlikely to cause a significant isotopic fractionation, the true age of each sample is expected to lie somewhere on a curve derived in this way.

Similar curves can be constructed for some of the samples from Sites 1003 and 1006, but here U addition seems to have occurred rather than U loss (Fig. 3). In this case, the possibility exists of adding U with an isotopic composition different from the sediment; however, the amount of U addition required for these samples is sufficiently small that, unless the isotopic ratio of the added U is extreme, it is not expected to cause samples to deviate significantly from the curves shown in Figure 3.

The percentage of U lost or gained can be calculated for each sample (although this is somewhat dependent on the diagenetic model, discussed below [see "Diagenetic Models" section]). Sample 5-2 requires a U loss of some 40%, while other Site 1005 samples require progressively less U loss downcore to a value of only 5% for Sample 5-6 (Fig. 3). Much smaller amounts of U addition are required for the Site 1003 and 1006 samples, with the only significant additions being 2% for Sample 6-3 and 3% for Sample 6-4. Note that these percentages are calculated assuming that all loss or gain occurred recently. Total U addition would need to be higher if some of it occurred in the past.

Diagenetic Models

There are two diagenetic models with which to estimate true ages for these samples. First, if straightforward gain or loss of U occurred, then the intersect between the curves and the seawater value of ²³⁴U (149) would provide the true age. This approach, however, yields ages that are unreasonable given that these high-aragonite samples are expected to come from a sea-level highstand. In particular, Sample 5-4 would yield an age of 270 ka and Samples 5-6 and 6-3 ages close to 560 ka—both times of sea-level lowstands (Fig. 3).

The second diagenetic model assumes that, in addition to the gain or loss of U, -recoil of ²³⁴U causes a continuous loss of pure ²³⁴U from these U-rich samples. If the samples are assumed to come from a highstand interval, then the intersection of the curves in Figure 3 with highstands (shown as gray bands) enables an assessment of the ²³⁴U loss and of the possible age. Such intersections for all nine samples in Figure 3 define a linear trend toward decreasing ²³⁴U(T) with age of sample (shown by the bold dashed arrow). This trend is as expected for continuous loss of ²³⁴U, and the fact that all samples fall on the same trend lends support to the idea that -recoil is systematically effecting these samples.

Such -recoil is probably an internal reorganization within the sediment. Measurements presented in this paper are on U-rich separates from the bulk sediment. This U-rich material is expected to lose more ²³⁴U by -recoil than it gains from the U-poor fraction of the sediment. The surprise is that -recoil can be so significant for 63- to 250-µm grain size separates as the recoil distance is short compared to this grain radius. A possible explanation is that U-rich organics on the grain surface play an important role in -recoil exchanges between grains.

Whatever the process, there is clear empirical evidence that -recoil is occurring for these samples and for similar samples from elsewhere in the Bahamas. For instance, deep samples from Site 1003 have ²³⁴U(0) less than zero, indicating that they have preferentially lost ²³⁴U. Measurements from the Little Bahama Bank show that aragonite in slope sediments is prone to loss of ²³⁴U (G.M. Henderson, N.C. Slowey, and M.Q. Fleisher, unpubl. data). The rate of ²³⁴U loss can be calculated from the deep Site 1003 samples and is similar to that required to explain the -recoil trend shown in Figure 3.

In general, the clear linear trend toward low ²³⁴U(T) (Fig. 3), together with independent evidence for the importance of -recoil, suggests that a diagenetic model involving both U loss/gain and continuous ²³⁴U loss by recoil is reasonable for these sediments.

Differences Between Sites

The diagenetic model favored above requires significant U loss from Site 1005 samples but slight addition of U to samples from Sites 1003 and 1006. The difference between the sites may be explained by their distance from the platform and the different fluid-flow history that they have, therefore, experienced.

Data collected during Leg 166 (Eberli, Swart, Malone, et al., 1997) suggest that fluid flow through the slopes of the Bahamas has two modes, depending on whether the banks are flooded or not (Henderson et al., 1999). This is illustrated in Figure 4 and is most clearly seen from the chlorinity data. At Site 1005, pore-water Cl concentrations show a clear inversion at depth because of lowstand circulation (Fig. 4, upper panel) and a flushed zone of seawater values above reflecting highstand circulation (Fig. 4, lower panel).

Close to the banks the direction of fluid flow, therefore, probably changes from outward flow caused by fresh-water compensation during low sea level to inward, thermally driven flow when the banks are flooded. Site 1005 sediments, therefore, experience long periods (90 k.y.) of outward flow of nonoxygenated pore water while the banks are exposed, followed by a shorter periods (10 k.y.) of downward flow of oxygenated seawater during peak interglacials (Fig. 3).

In contrast, sites situated farther from the platform have always experienced inflow of seawater. In these sites, U-rich organic material in the sediment is remineralized relatively quickly by the continuous inflow of oxygenated water. But at Site 1005, U-rich organic material may survive for considerable periods while producing radiogenic ²³⁰Th before oxidation during a highstand event. For instance, after 115 ka, sediments from Site 1005 were probably bathed in nonoxygenated pore waters while producing radiogenic ²³⁰Th. Then, at ~10 ka, the banks flooded, flow reversed, oxidation of organic material began, and U was removed while leaving behind the immobile ²³⁰Th, thus yielding apparent ages that are too old. This scenario is supported by the fact that the percentage of U loss required to explain the measured U-Th ages of the sediment decrease downcore for Site 1005. This is the result of deeper sediment packages experiencing multiple phases of U loss at previous sea-level highstands.

The small amount of U addition experienced by the samples from Sites 1003 and 1006 is probably explained by the downward advection of pore waters from oxidizing conditions to reducing conditions. As U is insoluble when reduced, the process of fluid flow moves small amounts of U downcore to precipitate in the more reducing conditions at depth.

The differences in fluid-flow regime experienced by the sediments of this study produce U diagenesis sufficiently complex that precise age information cannot be calculated for these samples. Nevertheless, there appears to be a coherent diagenetic explanation of the data such that each sample can be assigned to a single highstand event, enabling the production of age models for these cores.

Samples have been assigned to particular highstand events on the basis of their measured U-Th compositions and the consideration of diagenesis above. The sequence of highstands preserved at each site is now discussed briefly, working from the basin toward the platform edge.

Site 1006

This site has marine oxygen isotope Stages 1, 5, 9, and 11 but is missing Stage 7. This has been ascribed to a coring gap in Hole 1006A; the same coring gap does not exist in Hole 1006B (Kroon et al., Chap. 2, this volume). Sedimentation has, therefore, occurred and been preserved for each of the major highstand events back to Stage 11 for this basin site.

Site 1007

The uppermost sample in this core yields a U-Th age which is older than Holocene (24 ka). This probably reflects mixing of small amounts of older material upward in the core as this sample was taken from the transition from high- to low-aragonite-content sediment. The second sample in this core is too old to date by U/Th methods. This site has, therefore, suffered either a significant hiatus or considerable sediment erosion so that the highstands associated with Stages 5, 7, 9, 11, and possibly others are missing. This conclusion confirms the initial interpretation of the Western Seismic Line and the observations made by the Leg 166 shipboard party, both of which indicate a significant hiatus near the top of Site 1007 (Eberli, Swart, Malone, et al., 1997).

Site 1003

This site is also missing several highstand events, a fact hinted at by the abrupt change in ¹⁸O between the uppermost sediment package and deeper packages, indicating a change in sediment diagenesis (Rendle et al., Chap. 6, this volume). Stage 11 has been preserved at this site, but Stages 5, 7, and 9 are missing. The four deeper sediment packages measured in this study are older than the U/Th limit of ~450 ka.

Site 1005

This site, closest to the platform, appears to have a complete sequence of the major highstands, yielding ages from Stages 1, 5, 7, 9, and 11. The relative spacing of these ages is somewhat problematic, however. Stage 5 is very close to Stage 1, possibly suggesting sediment erosion of much of the Stage-5 sediment. Stage 7 features two distinct aragonite peaks separated by some distance. This might be explained if each package was from a different substage of Stage 7, but the U-Th dating presented here cannot distinguish these events.

Samples analyzed in this study were selected from horizons in the cores with the highest aragonite content. Such high aragonite contents are only produced when the banks are flooded and large areas of shallow water become available for aragonite production. None of these peak aragonite intervals has been dated as Stage 3, or as Substage 5a or 5c. It is possible that the absence of high-aragonite sediment of these ages reflects erosion of any such material originally deposited. Erosion of sediment is, after all, indicated by missing highstands in the sediments of three of the four sites investigated here. It is, however, more difficult to allow for removal of such sediment from Site 1006, which is situated on horizontal sea floor, some way from the platform. The lack of erosion from this site is supported by the complete sequence of major highstands recovered there.

Site 1006 features a rather monotonic decrease in aragonite concentration upward from the Substage 5e aragonite peak (Fig. 1). There is nothing approaching a return to the high values seen in Substage 5e. This pattern is similar to that reported for a core from the Tongue of the Ocean by Droxler et al. (1983). The lack of a return to high aragonite values in Substages 5c and 5a suggests that the banks were not flooded during these intervals.

The position of sea level during Substage 5c and particularly 5a has been the subject of significant controversy. Estimates based on U-Th dated corals have placed it anywhere from +2 m to -19 m compared to today's sea level (see Ludwig et al., 1996, for summary). Estimates from ¹⁸O and Mg/Ca measurements from deep-sea cores put Substage 5a sea level significantly lower (-60 m) (Dwyer et al., 1995). The banks in the vicinity of the Leg 166 Sites range from 6 to 10 m in water depth, and the Bahamas are tectonically reasonably stable. The lack of a return to high aragonite fractions in Bahamas sediments during these periods suggests that the banks were not flooded and that sea level therefore did not come within 10 m of the modern value. At this stage, such a conclusion is not robust due to the possibility of erosion of important sections of the few cores studied. However, if continuing work on Bahamas slope sediments fails to provide evidence for high-aragonite sedimentation during Substages 5a and 5c, it will increasingly suggest that sea level during these periods was lower than today.