18O
values, and exhibit features similar to the oxygen isotope record.Glacial-interglacial
changes in sedimentology dominate the MST and the isotope records at Site 981.
This interpretation of the long MST records from this site (Fig.
2A-C) is supported by visual comparison with the benthic
foraminiferal oxygen and carbon isotope data over the interval from 59 to 89 mcd
(~0.9-1.4 Ma; Fig. 4). The NG and MS
values increase during glacial periods, as indicated by more positive 18O
values, and exhibit features similar to the oxygen isotope record.
To quantify the
relationships between these parameters over this interval, we resampled the MST
data at the exact depths for which we have isotope data. Because the MS data
appear to have a log-normal distribution at Site 981 (unlike all the other data;
see Fig. 5), the log of the magnetic
susceptibility data (LMS) rather than the raw data are used for the following
discussions. Correlations were calculated among the three MST data sets (LMS,
NG, and GRAPE) and the two isotope records (18O
and 13C;
given in Table 3). F-tests were
performed, and only correlations significant at the 95% confidence level are
discussed. The correlations given among the MST data are all positive and
significant at the 99% level. The LMS and the NG records correlate reasonably
well in the interval of 59-89 mcd (R = 0.82). Overall, it appears that the
components that give the sediment high MS covary with, or are the same as, the
components that contribute high NG emission. The GRAPE wet bulk density also
tends to increase in concert with MS and NG, which implies that increased
abundance of these components reduces the sediment porosity slightly.
There is moderately good
correspondence between the benthic 18O
and the NG and LMS records; the correlation is lower but still significant
between 18O
and GRAPE. The shapes of the MST and 18O
records are distinctly similar throughout the upper part of this interval
(~58-77 mcd): LMS and NG, and to some extent GRAPE, tend to increase within
glacial intervals. The similarity between the oxygen isotope record and the MST
data appears to deteriorate below ~77 mcd.
Although benthic 13C
and 18O
show only a weak anticorrelation (R = -0.18) in the 59-89 mcd interval at Site
981, the 13C
has a stronger anticorrelation with both NG and LMS (-0.55 and -0.52,
respectively), showing that the deep water tends to be lighter in 13C
when the delivery or preservation of material with high MS and high NG emission
is enhanced. Also, as shown by the positive correlation discussed previously
between the benthic oxygen isotope record and these MST parameters, LMS and NG
vary with the glacial-interglacial cycles. Therefore, we suggest that some of
the variability in the NG emission and LMS data results from
glacial-interglacial changes in terrigenous delivery (as shown by the
correlation between these data sets and the 18O),
and some of the variability is associated with changes in deep-water circulation
(as shown by the negative correlation between these two MST parameters and the 13C).
At the most basic level, the correlation with 18O
supports the hypothesis that the long-period variability visible in the MST data
is caused by Milan-kovitch glacial cycles.
To determine which
sedimentological components cause the increase in LMS and NG emission during
glacial periods, we compared the MST data to the results of discrete sample
analyses (Table 4 [see Table of
Contents, ASCII Files, this volume]). These
include clay mineral-ogy, grain size, and percent calcium carbonate, as well as
lithic and foraminiferal counts and isotope data for the subset of eight
sediment samples. Statistically significant correlations between MST and
discrete data are given in Table 5.
Percent carbonate shows a particularly strong negative correlation with both 18O
and NG.
Correlations between these
variables are further summarized by R-mode factor analysis of the correlation
matrix. Parameters that increase during glacial periods have positive loadings
on the first factor (Table 6); strong
negative loadings characterize variables that increase during interglacial
periods. The glacial-interglacial character of this factor is best demonstrated
by its positive association with 18O.
One of the parameters with high negative loading for factor 1 is percent
carbonate. This may result from some combination of increased productivity of
carbonate-secreting organisms during interglacial periods when iceberg presence
is at a minimum and dilution of carbonate material by the increased flux of IRD
during glacial episodes.
Many of the sedimentological and mineralogical changes described by the first factor are probably controlled by the delivery of IRD to the sediments. The mean and standard deviation of the Elzone grain-size distribution are both positively associated with this factor. Because the grain size is measured on the carbonate-free portion of the <63-µm size fraction of the sediment, the variability in the size distribution is caused only by shifts in the source area and transport mechanism of terrigenous sediments and is not affected by changes in carbonate productivity. An increase in the delivery of ice-rafted material to the sediments might be expected to decrease the sorting of the carbonate-free portion and increase the standard deviation of the grain-size distribution. The number of >150-µm lithic particles per gram in these samples, however, has only a weak loading on the first factor, and lithic counts show no statistically significant correlations with any other individual parameter. Therefore, although the supply or absence of IRD may influence the grain size of the sediments deposited at this site, such effects are subtle.
The total XRD peak areas of both the clay and fine silt fractions have strong positive loadings on the first factor. Although peak area is not strictly equivalent to weight percent, changes in the total peak area of the X-ray patterns are representative of changes in the total crystallinity of that fraction of the sediment. Increased delivery of terrigenous material, by ice rafting or by currents, is likely to increase the proportion of well-crystallized material relative to poorly crystalline or amorphous sedimentary components (such as biogenic silica or badly degraded clays) and autochthonous materials, such as iron oxides or organic matter.
The strong associations of the LMS and the NG data with this factor confirm that they are each partially representative of glacial-interglacial changes in sedimentology at this site. The GRAPE wet bulk density also has a positive, although weak, loading on this factor. Examining the long MST records from Site 981 (Fig. 2A), we can see that the MS goes to zero deeper than 173 mcd (at ~2.7 Ma, the time of the onset of major Northern Hemisphere glaciation) except for one small pulse at ~196-198 mcd. Sulfate reduction has consumed nearly all of the pore-water sulfate by 150 mcd at Site 981 (Shipboard Scientific Party, 1996b), so it is not unlikely that reduction diagenesis is affecting the MS. Particularly in sediments older than the onset of glaciation, and thus predating any large flux of coarse magnetic material to Site 981, some combination of low supply and diagenesis has produced sediments almost completely void of magnetic material.
Of all the MST data, NG
has the clearest association with sediment mineralogy at Site 981, exhibiting a
strong negative correlation with weight percent carbonate in the sediment and
positive correlations with the total XRD peak areas of the clay and silt
fractions and fine-silt kaolinite + chlorite and quartz. The strong negative
correlation with carbonate implies that the gamma-emitting material and
carbonate are the two most important sedimentary components, and the relative
proportions of the two determine the gamma emission of the sediment. Because of
the strong correlation of the NG with the oxygen isotope data, we can deduce
that terrigenous supply in-creases, carbonate supply decreases, or both occur
during glacial periods. Neither the grain-size data nor the lithic particle
counts identify clearly whether the terrigenous material is primarily IRD or
current-deposited silt. However, a moderately strong association of the standard
deviation of the grain-size distribution with the first factor shows that
sediment sorting is decreased in glacial times. One explanation of this invokes
an increased proportion of poorly sorted IRD in the carbonate-free fine
fraction. A weak but significant positive association of the lithic counts per
gram sediment with factor 1 strengthens the argument that increased IRD delivery
is an important part of the glacial-interglacial variability at Site 981,
although no significant correlation is evident between lithic counts and 18O
independently.
The second R-mode factor (Table 6) is driven almost entirely by the strong correlation between the clay mineral abundances, particularly in the clay fraction (Table 5). The strong covariation in the clay mineralogy suggests that the source of these clay minerals is relatively constant in its composition. However, the second factor also shows weak loading for the LMS opposite to that of the clay mineral peak areas, which implies that there may be another component contributing slightly to the MS.
In summary, the sediments of Site 981 appear to be primarily a two-component system that varies on glacial-interglacial time scales. Lithic particles from a relatively homogeneous terrestrial source that is characterized by high NG and biogenic carbonates make up most of the sediment. The LMS correlates well with the NG data in the 0.9-1.4 Ma interval, and it has a high loading on the first R-mode factor in the discrete data set, suggesting that it too is controlled in large part by the same terrigenous source.
Unlike at Site 981, where nannofossils make up a large portion of the sediment (Shipboard Scientific Party, 1996b), at Site 984, the abundant sortable silt material is predominantly terrigenous (Shipboard Scientific Party, 1996c). Hence, although biogenic carbonate and IRD are present at both sites, a third independent component of current-transported fine silt with a basaltic composition is also found at Site 984. In addition, there may be a significant fraction of biogenic silica.
The MS data at Site 984
appear to be normally distributed (Fig. 6).
Visual comparison of the MST data and the stable carbon and oxygen isotopic
records of the 0.9-1.4 Ma interval at Site 984 (Fig.
7) shows that the glacial-interglacial contrast characteristic of
Site 981 is not as evident at Site 984 (Fig. 3A-C).
There is a weak positive correlation between the 18O
and MS (Table 7); NG, however, shows
no correlation with 18O.
If NG is driven by the relative proportion of terrigenous material as at Site
981, then the input of that material relative to the input of low-NG material is
not being paced directly by the Milankovitch-scale glacial-interglacial cycles
that control much of the low-frequency oxygen isotope signal. In addition, NG
and MS have a weak negative correlation through this interval, which implies
that unlike at Site 981, the components controlling these two signals vary
independently. Visual inspection of the MST data shows that, although the MS and
GRAPE data vary on roughly a glacial-interglacial scale as well as at higher
frequencies, the low-frequency band is not apparent in the NG data, although
natural gamma data do have a positive correlation with GRAPE data.
A total of 16 samples from Site 984, many of which were barren of benthic foraminifers, were analyzed to help identify the sedimentological variability driving the MST data sets at the Bjorn Drift (Table 8 [see Table of Contents, ASCII Files, this volume]). Distinct differences between Sites 984 and 981 can clearly be seen in comparing the MST and discrete data from the two sites (Table 9, Table 10). Although GRAPE values are comparable between the sites, Site 981 has higher NG and Site 984 higher MS values. Compared with Site 981, MS and the number of lithic grains per gram are much higher at Site 984, whereas the NG is distinctly less (Table 10). The mean of the grain size of the carbonate-free <63-µm portion is slightly higher at Site 984, perhaps because of the greater IRD content of the sediments. The standard deviation is the same, however, which implies that the "sortedness" of the terrigenous fraction is about equal at the two sites. The Site 981 sediment is on average nearly 50% carbonate, as opposed to ~7% at Site 984; the counts of foraminifers per gram and the percent coarse fraction also are much higher at Site 981. A comparison of the means of the clay mineralogy data for the two sites (Table 11) shows that the fine, carbonate-free sediment at Site 981 is better crystallized than that at Site 984; the greater total peak area and individual normalized peak areas further suggest that there is a higher concentration of amorphous material (likely biogenic silica) at Site 984.
As at Site 981, correlation between various parameters (Table 12) is summarized by R-mode factor analysis of the data (Table 13). The first factor at Site 984 appears to be representative of IRD input to the sediment, as indicated by its strong positive association with the number of lithic grains per gram. The weight percent in both the clay fraction and the coarse fraction have positive associations with this factor, as does the standard deviation of the grain size. The fine silt fraction, however, shows a strong negative correlation with this factor. Input of poorly sorted IRD may have enhanced the abundances of the coarsest and finest size material at this site. This first factor is strongly associated with the total, illite, kaolinite + chlorite, and quartz XRD peak areas in both the clay and fine-silt size fractions. This implies that IRD input is associated with these primary and secondary minerals that are most likely derived from felsic terrigenous sources. These associations bear up well in the correlations between the individual data sets (Table 12). In addition, the NG has a positive loading on this factor and correlates well with the XRD peaks of most of the minerals associated with this factor. It appears that the more felsic portion of the IRD input at this site drives the NG signal.
By contrast, smectite and plagioclase do not have strong positive associations with the first R-mode factor. Both size fractions of smectite have only weak loadings for factor 1; the fine silt-fraction plagioclase has a weak negative association. However, the clay-fraction plagioclase has a strong negative association, and the MS also shows a negative loading on this factor. This implies that the sedimentary component associated with clay-fraction plagioclase that contributes the magnetic material is not IRD. Smectite concentration appears to be controlled independently of either the continental IRD or the high-MS source. The only significant correlation that the smectite concentration has is a negative association of the clay-fraction smectite normalized XRD peak area with the Elzone standard deviation. High smectite concentrations therefore are associated with better sorting of the sediments. This suggests that the smectite is delivered primarily by current activity.
The second R-mode factor at Site 984 emphasizes the correlation between percent carbonate and foraminifers per gram. These two data sets each have strong negative loadings on this factor. Both MS and clay-fraction smectite, however, have positive associations with the second factor. Variation in the relative proportions of biogenic carbonate and magnetic, plagioclase-, and smectite-rich silt components appear to control the second factor.
In summary, four significant sedimentary components appear to influence the MST signals at Site 984. The most distinct component appears to be well-crystallized, poorly sorted, felsic IRD, which correlates strongly with NG. Carbonate is also a distinct, independent component. Third, a plagioclase-rich component that has high MS is associated with good sorting and is probably transported from Iceland as suspended silt in a nepheloid layer; smectite is associated with this material also. As clearly shown by Blum and Richter (1997), the correlation of plagioclase peak area and MS results from the correlation of both variables with magnetite concentration. Biogenic silica probably makes up a fourth component.
