RESULTS

Results of the magnetostratigraphic determinations (see the "Supplementary Materials" contents list) are shown in Figures F3 and F4. The concentration of dependent magnetic parameters is given in Figure F3. Note that several hiatuses are present but that the Gauss Chron, which is the subject of much discussion, is well defined, although the top of the Gauss is missing as at many sites in the Southern Ocean. Figure F4 shows the concentration of independent magnetic parameters that can be used for mineralogical interpretations. For instance, estimates of the concentration of ferrimagnetic minerals present in our samples can be obtained from some parameters such as k (low-field magnetic susceptibility), IRM, and ARM.

Normalizing these parameters by the susceptibility compensates for variations in concentration and, hence, for example, the k_ARM (anhysteretic susceptibility)/k ratio varies inversely with magnetic particle size and is therefore a useful granulometric parameter. The interpretation of this ratio may be complicated by significant amounts of superparamagnetic or paramagnetic phases.

The S-ratio (S_-0.3) is very useful for discriminating ferrimagnetic grains (such as magnetite) from high-coercivity antiferromagnetic grains such as goethite (-FeOOH) or hematite (-Fe₂O₃). So, downcore variations of this parameter may be associated with changing mineralogy. Values close to -1 indicate lower coercivity and a ferrimagnetic mineralogy (e.g., magnetite); values closer to 0 indicate a higher coercivity, possibly an antiferromagnetic (e.g., hematite) mineralogy. Finally, with the magnetic mineralogy constrained as magnetite, the ARM/IRM ratio can be used as a magnetic grain-size indicator too, because the ARM is more effective in activating finer magnetite grains than the IRM.

Figure F5 shows the postcruise revised magnetostratigraphy, based on further laboratory work, and integration with biostratigraphy (marine diatoms and radiolarians). The essential difference from the shipboard interpretation is that the Olduvai is now the "fused" Jaramillo and Olduvai (i.e., Subchrons C1r.1n and C2n are now fused) and the reversed interval between those two subchrons is lost in a hiatus. Because the Pliocene/Pleistocene boundary occurs just above the Olduvai Subchron (Berggren et al., 1995), this boundary must lie in the hiatus (Florindo et al., in press). Additional disconformities must be present near 6.1, 14.3, 15.6, and ~16 mbsf (Florindo et al., in press). (See Fig. F3 caption for data sources).

Figure F6 shows the combined results of our granulometric analyses (see the "Supplementary Materials" contents list). Texturally, all samples fall in the mud category as follows: mud, sandy mud, slightly gravelly mud, and slightly gravelly sandy mud. One sample was gravelly mud (after Folk, 1980). The preponderance of mud is indicated by the mud percentage line (only shown for comparison purposes). Sand (area between the gravel and the gravel and sand lines is present in low abundance) (see below). It increases noticeably in the upper 15 m because of the presence of (mainly planktonic) foraminifers in the sand-sized fraction.

All samples, regardless of textural classification, have a sand-sized component (0.0625-2 mm), but many samples are without a gravel component. Below 15 mbsf, the 0.15- to 2-mm fractions of the sand-sized components consist primarily of lithogenous (terrigenous) material—biogenous components (mainly radiolarians) are rare. Lithogenous components are quartz, feldspar, mafic minerals, lithic fragments of all three rock families, coal, and so on. Quartz consists of several populations, including clear and clouded quartz, quartz with rutile and other inclusions, and quartz as component of lithic fragments.

Although some quartz grains are well rounded and a few have the characteristics of wind-blown grains, the vast majority of quartz grains exhibit the mechanical breakage features such as conchoidal fractures typical of glacial environments, steplike fractures, and so on. No, or very little, rounding is discernible on these grains. The preponderance of quartz grains with "glacial" features suggests that (intermittent) ice rafting occurred throughout the time interval studied here (see Warnke et al., 2001).

A remarkable component of the same sand-size fractions is "garnet-colored" garnet. It is optically indistinguishable from garnet recently discovered in Pleistocene samples in piston core TN0-57-6-PC4, raised in the South Atlantic (42°52.1´S, 8°57.7´E; 3751 m), suggesting that icebergs calving from marine termini in Prydz Bay may occasionally reach that far-distant site (Teitler et al., 2002). Garnet grains show mechanical breakage features indicative of glacial environments.

Sediment Brightness Determined from Spectral Data

Sediment brightness (sometimes termed lightness) is calculated as the area under a spectral curve in the visible (400-700 nm) portion of the spectrum. Brightness measured in marine sediments commonly shows a positive correlation with calcium carbonate, especially in pelagic sediments, and thus is sometimes used as a proxy for carbonate content; however, other factors can also influence brightness (see Balsam et al., 1999, for detailed discussion about interpreting brightness as a function of carbonate content). Figures F2 and F7 show the brightness for the HiRISC section (0-54 mbsf) in Hole 1165B. Figure F8 shows percent reflectance curves for six color bands. Note that in the upper 10 m of the section the brightness seems to correlate with the carbonate content (Fig. F2). However, for the remainder of the section where carbonate is essentially zero, the brightness also varies widely and thus must be related to mineralogical components other than carbonate. For example, the brightness curve seems to correlate quite well with grain-size magnetic parameters, bulk density, and natural gamma radiation curves from the HiRISC section (Figs. F3, F4, F5). In particular, a major change in sediment properties at the beginning of the late Pliocene in these parameters is clearly recorded by an overall sharp decrease in brightness just above 34 mbsf (Figs. F2, F7).

The brightness for the HiRISC section derived from shipboard measurements using the Minolta CM-2002 spectrophotometer is also shown in Figure F7. During Leg 188, the Minolta instrument was mounted on the archive multisensor track (see O'Brien, Cooper, Richter, et al., 2001, for details) and all spectral measurements were automated and performed on the archive half of each core. Because of this automated process, erroneous measurements were commonly made at gaps, voids, missing intervals of core, and so on. The Minolta data shown in Figure F7 are an edited subset of the original shipboard data set from which these erroneous measurements have been deleted as far as possible (see Damuth and Balsam, this volume). Note that the shipboard brightness data fluctuations correlate quite well with the PerkinElmer-derived data, except that the brightness for the sediments measured with the Minolta is consistently lower than those values measured with the PerkinElmer. We have conducted several studies on cores from previous ODP legs comparing shipboard Minolta measurements, which by necessity must be conducted on wet cores, with PerkinElmer measurements, which must be made on dried, ground core sediments (Balsam et al., 1997, 1998, 1999; Balsam and Damuth, 2000). These studies showed that water in the sediments mutes the brightness. However, when both sets of reflectance curves are processed using a first-derivative transformation, the shipboard and shore-based analyses are quite similar and suggest that accurate, reliable spectral data can be obtained from wet cores at sea using the Minolta spectrophotometer.

Interpretation of Sediment Components and Minerals with DRS Using Factor Analysis

Interpretation of sediment components and minerals was carried out by factor analysis of the first-derivative values calculated at 10-nm intervals from the percent reflectance values. Details of the methodology, including background and advantages of VIS spectral analysis are provided in Damuth and Balsam (this volume), Balsam and Deaton (1991), Balsam et al. (1997), and Balsam and Damuth (2000). Damuth and Balsam (this volume) identified five factors from the sediments at Site 1165, and we interpret these factors as follows:

Factor 1: goethite and ripidolite, a chlorite mineral;

Factor 2: organic matter;

Factor 3: a combination of clay minerals, probably montmorillonite;

Factor 4: maghemite; and

Factor 5: hematite.

See Damuth and Balsam (this volume) for detailed explanation of these factor interpretations.

Factor interpretation can be aided by plotting the downhole distribution of each factor using factor scores, which indicate how important each factor is in each sample (e.g., Figs. F9, F10, F11, F12, F13). These downhole factor scores presumably show how the various minerals represented by each factor vary downhole. Ideally, these downhole plots are then compared to downhole distributions of various sediment components identified by other means (e.g., X-ray diffraction). There are some important similarities and differences between reflectance data and data generated by other commonly used analytical techniques such as XRD. In several ways, reflectance data derived from DRS are similar to XRD. First-derivative peak height, like the height of X-ray peaks, is a function of both the concentration of a substance (mineral for XRD) and the composition of the matrix in which it is found (Deaton and Balsam, 1991; Balsam et al., 1999). DRS data differ from XRD data in three important respects. First, DRS is not limited to crystalline material; spectra can be obtained from any substance. Second, the heights of first-derivative peaks are not only a function of mineral concentration and matrix material but also are a function of the "spectral strength" of a substance. Some substances (e.g. hematite) have a persistent spectral signal that is capable of concealing the peaks of other substances. Third, for some substances, the limits of resolution of DRS differ significantly from those of XRD. For example, depending on the matrix, hematite can be detected at a concentration of ~0.01% by weight with DRS, whereas with XRD, the limit of resolution without special sample preparation is ~1% (Deaton and Balsam, 1991). To date, limits of resolution with DRS have been determined for only a very few minerals, sediment components, and combinations of components. For XRD, a similar statement applies for many minerals and mineral combinations.

In the HiRISC section of Hole 1165B (0-54 mbsf), factors 1 and 2 exhibit high scores through most of the section, with the exception of a low in both factors centered at ~30 m (Figs. F9, F10). Factors 3 and 4 exhibit high values primarily below 30 m (Figs. F11, F12). Factor 5 exhibits higher-frequency variation than the other factors and, with the exception of a few points, contains primarily high values below ~10 m (Fig. F13). Despite apparent similarities in their downhole patterns, few of the factors are highly correlated to each other. The highest linear correlation is exhibited by factors 2 (organic matter) and 5 (hematite), which has r² = 0.51. This correlation is difficult to explain. The next highest linear correlation is factors 3 (montmorillonite and illite) and 4 (maghemite), which has r² = 0.36. One possible explanation for this correlation is that the maghemite is being transported with clay minerals or is produced by the erosion of clays. Factor 1 (goethite and ripidolite) and factor 5 (hematite) exhibit a linear correlation with r² = 0.23. Although hematite and goethite may form at high temperatures by thermal oxidation (Dunlop and Özdemir, 1997), they may also form at low temperatures by oxidation in soils and during chemical weathering. Under surficial conditions and especially in soils, hematite and goethite are end-members of oxidizing chemical processes, with hematite forming under dry and hot conditions and goethite under more humid conditions (Maher, 1998; Maher and Thompson, 1999). There is substantial overlap in their formation and both are commonly found together. Also formed by chemical processes taking place in soils and in rock rubble are a variety of clay minerals. The fact that this factor analysis separates hematite and goethite into two factors suggests that these minerals have different sources. Factor 3 (montmorillonite and illite) and factor 5 (hematite) exhibit a linear correlation with r² = 0.21. As above, one possible explanation of this correlation is that the hematite is being transported with clay minerals or is produced by the erosion of clays. No other factors exhibit linear correlations with r² > 0.2.

N. pachyderma (s.) ¹⁸O values range from 3.89 to 4.96 (see Fig. F14) and generally increase during the late Pliocene-Pleistocene, reflecting the transition to cooler temperatures and globally increased ice volume (see the "Supplementary Materials" contents list). There appears to be several glacial-interglacial cycles recorded in this section. Between the base and what was assumed to be the top of the Olduvai Subchron (1.95-1.77 Ma; 14.10-6.97 mbsf) during the leg, four glacial-interglacial cycles are recorded (note, however, that the top of the Olduvai is missing—see above). Based on the location of an unconformity (at ~6 mbsf) below the Brunhes/Matuyama paleomagnetic reversal (0.78 Ma; 5.37 mbsf), much of the early Pleistocene is missing. The glacial-interglacial cycle recorded above needs further investigation. An unconformity occurs at ~3 mbsf, based on the jump in values, and further age control will be necessary to identify the glacial-interglacial periods above this point.

N. pachyderma (s.) ¹³C values range from -0.344 to 0.259 and fluctuate in accordance with ¹⁸O changes. During the inferred interglacial periods ¹³C values are generally higher, indicating an increase in primary production that leaves surface waters enriched in ¹³C due to the preferential uptake of ¹²C. The small amplitude of glacial-interglacial ¹³C changes (0.4) suggests that changes in surface water primary production were relatively small in the region during the late Pliocene-Pleistocene.

Results of the clay mineral analyses (see the "Supplementary Materials" contents list) are shown in Figure F15. All results are plotted against depth. The clay mineral assemblages are dominated by smectite, illite, and kaolinite. The smectite concentration is variable, with values mainly between 0% and 30%. Illite fluctuates less; concentrations are mainly 50%-80%, and kaolinite varies mainly between 10% and 20%. Chlorite concentrations are mainly 0%-10%.

Cyclicity of Clay Mineral Assemblages during the Late Quaternary

Late Quaternary variations of clay mineral assemblages deposited on the continental rise off Prydz Bay show possible cyclicity with higher smectite concentrations during prominent interglacials. We presume that the cyclicity in clay mineral assemblages observable in Prydz Bay is the same type that was observed in the area of the Antarctic Peninsula (Hillenbrand, 2000). Smectite decreases and chlorite increases from 50 to 0 mbsf. Chlorite concentrations mirror slightly the smectite variations.

In general, during interglacial periods smectite is delivered by bottom-current transport along the continental rise. The decrease of smectite during glacial periods may be a consequence of climate-induced changes in depositional processes on the margin (see Pudsey and Camerlenghi, 1998; Rebesco et al., 1998; Hillenbrand, 2000; Pudsey, 2000; see also Grobe and Mackensen, 1992).

Pliocene to Quaternary Clay Mineral Assemblages in Sediments at Site 1165

According to the existing age model, the stratigraphic sequence of Site 1165 (uppermost 50 mbsf) extends back to ~5.0 Ma. Throughout this interval, the clay mineral content is characterized by major fluctuations of individual clay minerals, particularly smectite and chlorite. Short-term cyclic changes in clay mineral assemblages deposited at Site 1165 occur throughout the Pliocene and Quaternary independent of changes in the clay content. We note that changes in clay mineral composition are also reflected in the other proxies, described earlier. For instance, the noticeable change detected both by rock magnetic investigations and spectrographic methods at ~34 mbsf is clearly indicated in the composition of the clay mineral suite. At this level, smectite decreases whereas kaolinite increases. Illite and chlorite also show variability. In particular, there is a slight but persistent increase in chlorite.

The section studied lies in the latest Pliocene-Pleistocene, but a large section (1.70-9.25 mbsf) has not been zoned. Foraminifer faunas are very highly dominated by Neogloboquadrina pachyderma (Ehrenberg), commonly to 99.5%, suggesting that benthic productivity was generally low. This is consistent with the very low content of infaunal species. Planktonic percentages significantly less than 99.5% are due to preferential dissolution of planktonic specimens. Many residues consist very dominantly of N. pachyderma with little other content.