Organic carbon content (org-C) of the studied samples varies between 0.30% and 3.00% with an average of 1.60 wt% (Appendix A). Org-C is relatively high and decreases downward from 3.00 wt% to 1.65 wt% with decimeter-scale low-amplitude (<0.5%) fluctuations in the Holocene (Fig. 5). The decreasing trend is interrupted by a high org-C peak at 2.2 mbsf. Org-C is low between 1.0 wt% and 1.6 wt% within the Younger Dryas and high between 1.6 wt% and 2.3 wt% within the Bølling/Allerød. Org-C is low and shows centimeter- to decimeter-scale high-amplitude fluctuations between 0.30 wt% and 1.67 wt% in MIS 2 where low org-C peaks coincide with sand layers. In MIS 3 to 5a (5.2-15.1 mbsf), org-C shows decimeter-scale fluctuations with org-C maxima (1.4 wt%-2.0 wt%) generally occurring within silty clay with opaque filaments and/or particles and org-C minima (0.7 wt%-1.4 wt%) generally occurring within clayey silt with occasional faint laminations. Within the lower part of this interval, org-C minima coincide with sand layers at 14.3, 14.45, and 14.75 mbsf.
Carbonate carbon content (carb-C) of the studied samples varies between 0.2 wt% and 2.0 wt% with an average of 0.81 wt%. Carb-C is relatively high (0.7 wt%-2.0 wt% with an average of 1.1 wt%) (Appendix A) in the Holocene, where it gradually increases downward from 0.8 wt% at core top to 2.0 wt% at 2.1 mbsf (Fig. 5). Carb-C suddenly drops to 0.6 wt% within the Younger Dryas and goes up again to 1.3 wt% within the Bølling/Allerød. Within MIS 2, carb-C varies in centimeter- to decimeter-scale between 1.2 wt% and 0.2 wt% with carb-C minima coinciding with sand layers. In MIS 3 to 5a, carb-C shows decimeter-scale fluctuations between 1.3 wt% and 0.3 wt%. Within this interval, carb-C minima below 0.4 wt% occur at 5.5, 8.2, 8.9, 11.5, 13.1, and 13.7 mbsf. No distinct sand layers are present within these intervals, but indistinct sandy zones and sharp contact surfaces are present. The intervals with carb-C maxima tend to coincide with silty clay and with opaque filaments and/or particles.
Quartz, feldspar, smectite, illite, chlorite + kaolinite, calcite, pyrite, and amorphous material are identified in all the samples (Appendix A), and we categorized these minerals into five groups based on the positive correlations of their peak heights and similarities in their vertical variation profiles (Fig. 6).
The first group is quartz and feldspar, which show a strong positive correlation (r2 = 0.64). Vertical profiles of quartz and feldspar are generally similar (Fig. 6A). They are relatively low in the Holocene, significantly high in the Younger Dryas, and the lowest within the Bølling/Allerød. They are intermediate to high, with high-amplitude variations in MIS 2 with their maxima coinciding with sand layers. Within MIS 3, they are intermediate with decimeter-scale fluctuations with their maxima tending to coincide with faintly laminated and/or sandy intervals. Within MIS 4 to 5a, they are intermediate to high with decimeter-scale fluctuations. It is interesting to note that their minima tend to coincide with sand layers at 13.8, 14.3, 14.5, and 14.75 mbsf, which is opposite to what is observed in MIS 2.
The second group is illite, chlorite + kaolinite, and smectite. Among these clay minerals, illite has a strong positive correlation with chlorite + kaolinite (r2 = 0.64), whereas smectite has weak positive correlations with illite and chlorite + kaolinite (r2 = 0.29 and 0.33, respectively). Illite and chlorite + kaolinite have almost no correlation with quartz and feldspar, whereas smectite has a weak negative correlation with quartz (r2 = 0.28). Within the Holocene, the three minerals gradually increase downward with small-amplitude, decimeter-scale fluctuations (Fig. 6B). Smectite decreases significantly within the Younger Dryas and increases to its maximum within the Bølling/Allerød, whereas illite and chlorite + kaolinite are not significantly changed within these intervals. Within MIS 2, the three minerals reach their maxima at 3.4 mbsf, then gradually decrease downward with large-amplitude fluctuations to 5.0 mbsf, with their minima coinciding with sand layers. Between 5.0 and 6.0 mbsf in the uppermost part of MIS 3, abundance of the three minerals are very low, then rapidly increases downward. Within MIS 3, they are intermediate to high with distinct maxima at 7.7, 8.3, 10.3, and 11.5 mbsf. Within MIS 4, they are relatively low, but gradually increase downward in MIS 5a. Within MIS 3 to MIS 5a, decimeter-scale, moderate- to large-amplitude fluctuations are present throughout with their minima, tending to coincide with the intervals of faint laminations and sandy zones.
The third group is calcite, which shows a weak positive correlation with amorphous material (r2 = 0.25) and weak negative correlations with quartz and feldspar (r2 = 0.33 and 0.32, respectively). Calcite has a strong positive correlation with carb-C (r2 = 0.75), suggesting that calcite is the dominant phase of carbonate minerals. Consequently, the vertical profile of calcite is basically similar to that of carb-C (Fig. 6C).
The fourth group is amorphous material. The area of amorphous hump has a weak positive correlation with calcite as described above and weak negative correlations with quartz and feldspar (r2 = 0.39 and 0.36, respectively). The area has a strong positive correlation with org-C (r2 = 0.69). Consequently, its vertical profile is basically similar to that of org-C (Fig. 6C).
The fifth group is pyrite. Pyrite shows no correlation with other mineral phases. Pyrite is almost absent in surface sediments, but increases rapidly between 5 and 10 cmbsf (Fig. 6D). Within the Holocene, pyrite is intermediate to high with indistinct maxima at the middle of the interval. Pyrite is slightly low in the Younger Dryas and increases abruptly to form a distinct maximum in the Bølling/Allerød. It decreases abruptly at the top of MIS 2, gradually increases downward toward 4.2 mbsf, and gradually decreases again toward 5.5 mbsf at the top of MIS 3. Within MIS 2, moderate-amplitude centimeter- to decimeter-scale fluctuations are present with its minima occasionally coinciding with sand layers. Within MIS 3 to 5a, pyrite shows large-amplitude decimeter-scale fluctuations with its maxima coinciding with homogeneous silty clay with opaque filaments or disseminated particles. Distinct maxima of pyrite abundance are present at 7.7, 8.8, 9.5, and 10.6 mbsf in the middle of MIS 3, at 13.5 and 13.8 mbsf in the lower part of MIS 4, and at 14.8 mbsf in MIS 5a.
The ten major elements analyzed by XRF (including LOI; Appendix B) are categorized into five groups based on their interrelationships and similarities in their vertical variation profiles (Fig. 7). The first group (group 1) is SiO2, Na2O, and K2O, which show strong positive correlations with each other (r2 = 0.62-0.82). These elements are generally low and gradually decrease downward to a minimum at 2.2 mbsf within the Holocene (Fig. 7A). An exception is Na2O, which shows a small maximum at 1.8 mbsf. Group 1 elements increase drastically in the Younger Dryas, which then decrease drastically within the Bølling/Allerød. Within MIS 2, group 1 elements are intermediate to high and show high-amplitude fluctuations with their maxima coinciding with sand layers. Group 1 elements are high in the uppermost part of MIS 3, gradually decrease toward the middle part, and remain about the same in the lower part. They are relatively high within MIS 4 to 5a. Superimposed on this trend within MIS 3 to 5a is moderate-amplitude, decimeter-scale fluctuations with higher values tending to coincide with intervals with faint lamination and sandy zones. The maxima in MIS 5a correspond to sand layers.
The second group (group 2) is Fe2O3 and MgO, which have a strong positive correlation (r2 = 0.73). Group 2 elements show moderate to strong negative correlations with SiO2 (r2 = 0.52 and 0.60, respectively) and moderate to weak negative correlations with Na2O and K2O (r2 = 0.44 to 0.24, respectively). Vertical profiles of these elements are basically mirror images of those of SiO2 (Fig. 7B). Exceptions are at 10.6, 11.4, and 14.65 mbsf, where either Fe2O3 or MgO varies in phase with SiO2.
The third group (group 3) is CaO and LOI, which show a strong positive correlation (r2 = 0.68). CaO does not show positive correlations with any other elements, whereas LOI shows moderate positive correlations with Fe2O3 and MgO (r2 = 0.41 and 0.50, respectively). CaO and LOI show strong negative correlations with group 1 elements (r2 = 0.66-0.77 and 0.65-0.92, respectively). CaO also shows a strong correlation with carb-C (r2 = 0.82), suggesting that this element is contained primarily in carbonate minerals, whereas LOI shows a strong correlation with org-C (r2 = 0.78) and a moderate correlation with carb-C (r2 = 0.59), suggesting that it dominantly represents organic matter and carbonate minerals. Calculation of weight loss due to ignition of carbonate mineral and organic matter by assuming CaCO3 and CH2O stoichiometry shows that ~80% of LOI can be explained by these two phases. Vertical profile of CaO and LOI is basically similar to that of carb-C and org-C, respectively (Fig. 7C).
The fourth group (group 4) is TiO2, MnO, and P2O5. TiO2 and MnO show a strong positive correlation (r2 = 0.73), whereas P2O5 shows only weak positive correlations with MnO and TiO2 (r2 = 0.36 and 0.21, respectively). Group 4 elements do not show any correlation with other elements. Vertical profiles of TiO2 and MnO are very similar, and P2O5 is also similar in many respects (Fig. 7D). Namely, these elements vary with only small amplitude with depth throughout the sequence except for several samples. Those samples showing the lower concentrations correspond to sand layers and sandy intervals, whereas those samples with the higher concentrations occur at the base of some sand layers and a sandy interval. Although vertical variations are small, there are slight variations with depth in the case of MnO and P2O5. MnO is slightly low within the Holocene to Younger Dryas and MIS 4 and slightly high within the Bølling/Allerød to MIS 3 and MIS 5a. In the case of P2O5, it is intermediate and slightly decreases downward in the Holocene to Younger Dryas, slightly high in the Bølling/Allerød to MIS 2, slightly low and gradually decreases downward in MIS 3 to MIS 4, and slightly increases in MIS 5a.
The fifth group (group 5) is Al2O3, which does not show correlations with any of other elements (r2 < 0.20). Al2O3 does not vary significantly, but it is slightly low within the upper Holocene to MIS 2 (Fig. 7E). It increases abruptly at 5.6 mbsf in the uppermost part of MIS 3 and stays at that level downward throughout the rest of the sequence. Superimposed on this general trend are decimeter-scale fluctuations. Within the Holocene and MIS 3, most of Al2O3 minima correspond to carb-C maxima, suggesting that they reflect dilution by carbonate; the minima within MIS 2 and MIS 4 to 5a, however, correspond to sand layers, suggesting dilution by sand. In general, decimeter-scale fluctuations of Al2O3 are similar in phase with group 2 elements.
Results of mineral and major element analyses suggest that variations in sediment composition can be explained by mixing of 4 or 5 end-member components. To make this point more clear, we applied Q-mode factor analysis to the data set of major element composition to identify and characterize common factors that control compositional variations. This method allows us to partition the sediment composition into the end-member components in the sediments.
Factor analysis provides a description of a multivariate data set in terms of fewer common factors that account for variance within the data set. Factor score and factor loading represent importance of each variable within each end-member and of each end-member within each sample, respectively. Factor scores and loadings generated by simple factor analysis are not scaled as (percent) fractions of variables within end-members and of end-members within samples, respectively. This fact prevents us from direct comparison of factor loadings with abundance of minerals, org-C, and carb-C. However, if the data are closed by constant total sum of variables in each sample, the factor scores and loadings can be converted into values expressed as (percent) fractions (Miesch, 1976). For this reason, we conducted Q-mode factor analysis for all the samples using concentrations of all the major elements including LOI (11 variables), of which total sum is constant at 100%. The largest number of variables and smallest analytical error for the major element composition data set allow us to examine the factors that explain only small fractions of variances in the data set.
Varimax rotation of factors was conducted to obtain geologically meaningful (preferentially positive) scores and loadings of components (Imbrie and van Andel, 1964). Analytical procedure is similar to Miesch (1976) and Leinen and Pisias (1984) except that we used all major elements and did not apply oblique rotation of the axes. Extracted factors are considered as representing end-member components (e.g., detrital, biogenic, and/or authigenic). Relative enrichment or depletion of elements between factors are expressed by factor scores, whereas relative contribution of factors between samples is expressed by factor loadings. Because we did not apply oblique rotation of axes (Miesch, 1976; Leinen and Pisias, 1984), scores and loadings are only semiquantitative expressions of composition and content of factors, respectively.
Results of the analysis show that variance explained by each of the first five factors is significant (exceeding the range of analytical error) and they together explain 99.6% of the total variance. The first two factors alone explain nearly 95% of the variance, whereas factors 3 through 5 explain only 5% of the variance. However, concentration of certain elements in some of the samples can be explained only after inclusion of these factors in factor model. In other words, element concentrations not explained by factors 1 and 2 exceed analytical error of the elements. Therefore, we adopted a five-factor model to explain our data set. Extracted five factors are named as factor 1 through factor 5 in descending order of variance explained by each varimax factor. We also examined the meaning of extracted factors through correlation analysis between the factor loading and abundance of minerals, org-C, and carb-C of the samples. Composition scores of these elements for each factor are shown in Table 1, and correlation coefficients between org-C, carb-C, mineral composition, major element composition, and factor loadings are shown in Table 2.
Factor 1 explains 49.8% of variance and has larger (than average) scores for group 2 and 3 elements (Fe2O3, MgO, CaO, and LOI), and smaller (than average) scores for group 1 elements (SiO2, Na2O, K2O). This factor shows strong to moderate positive correlations with org-C, carb-C, calcite, and amorphous material (r2 = 0.64 - 0.52), a weak positive correlation with smectite (r2 = 0.24), and strong negative correlations with quartz and feldspar (r2 = 0.81 and 0.68, respectively). Thus, this factor is characterized by higher contents of organic matter, carbonate, smectite, and amorphous material. Because the samples with larger loadings of this factor generally contain abundant nannofossils according to smear-slide observation, carbonate marking this factor is most likely of biogenic origin.
Factor 2 explains 45.1% of variance and has larger scores for group 1 elements and smaller scores for group 2 and 3 elements. This factor shows moderate positive correlations with quartz and feldspar (r2 = 0.53 and 0.54, respectively) and strong to moderate negative correlations with carb-C, org-C, calcite, and amorphous material (r2 = 0.82-0.43). Thus, this factor represents a detrital component marked by quartz and feldspar.
Factor 3 explains 4.0% of variance and has larger scores for SiO2, CaO, and LOI, and smaller scores for Al2O3, Fe2O3, MnO, and MgO. This factor shows faint positive correlations with quartz and calcite (r2 = 0.18 and 0.12, respectively) and weak to faint negative correlations with illite, chlorite + kaolinite, and smectite (r2 = 0.21 - 0.13). Detailed examination of loading of this factor revealed that factor 3 tends to covary with factor 1 except for the intervals between 2.3 and 7.6 mbsf and 10.75 and 14.8 mbsf where frequent intercalations of sand layers and sandy zones occur (sandy intervals). This point is more clear on a scatter plot of factor 2 vs. factor 3 loadings in Figure 8A. In this figure, the two factors show a positive correlation for samples from the sandy intervals whereas the two factors show a negative correlation for samples from the intervals with hemipelagic silty clay to clayey silt without intercalations of sand layers (hemipelagic intervals). Within hemipelagic intervals, samples with higher factor 3 loadings tend to contain abundant planktonic foraminiferal skeletons (and nannofossils) according to smear-slide observation. On the other hand, within the sandy intervals, the samples with higher factor 3 loadings coincide with sand layers. Such samples contain abundant coarse-silt-size bioclasts that are composed of planktonic and benthic foraminiferal skeletons occasionally filled with pyrite and shell fragments. Thus, factor 3 seems to represent coarse-silt-size calcareous bioclasts of probable allochthonous origin within the sandy intervals and planktonic foraminifers of probable autochthonous origin within the hemipelagic intervals. The relation between factor 3 loading and quartz peak intensity in Figure 8B shows that samples from the sandy intervals show a positive correlation, whereas the samples from the hemipelagic intervals show a negative correlation, which further supports this explanation.
Factor 4 explains 0.5% of variance and has larger scores for SiO2, TiO2, MnO, and P2O5 and smaller scores for Al2O3, Fe2O3, MgO, CaO, Na2O, and LOI. This factor does not show correlation with any mineral phases and org-C. As is the case for group 4 elements, samples with extremely high factor 4 loadings occur at the base of sand layers and a sandy zone. However, it should be noted that not all sand layers have the basal part with high factor 4 loading—only the sand layers with very low org-C. It is possible that these high loadings reflect early diagenetic precipitation of MnO and P2O5 under early diagenetic oxic conditions that were temporarily caused by deposition of turbidite sand deficient in labile organic matter. But early diagenetic precipitation under temporal oxic environment cannot explain high concentration of TiO2. On the other hand, TiO2, MnO, and P2O5 in turbidite sand samples from MIS 2 show strong positive correlations each other (r2 > 0.82) and a moderate to weak positive correlation with Fe2O3 (r2 > 0.37), and stoichiometry estimated from these correlations is consistent with chemical composition of titanomagnetite. In addition, smear-slide observation suggests that the sample with highest factor 4 loading contains abundant heavy minerals including zircon, spinel, and magnetite. For these reasons, we prefer the interpretation that this factor represents condensation of heavy minerals probably including titanomagnetite that are condensed in the basal part of some of turbidite sand layers. The fact that factor 4 loadings are lower in other turbidite sands implies either different source(s) or sorting effect during transportation. Even after excluding these sand samples with exceptionally high or low factor 4 loadings, there still remains a variation in factor 4 loadings that is larger than analytical error. Namely, the loading is higher within MIS 2 and lower within MIS 5a. We suspect that this variation may reflect changes in composition of the detrital material. To test this possibility, we examined the relationship between factor 4 loading and Na2O/(SiO2 - 14.15), a measure of deviation from the Na2O vs. SiO2 regression line, for the samples excluding turbidite sands (Fig. 9). The figure shows a clear negative correlation (r2 = 0.61). Because both SiO2 and Na2O are considered as dominantly held in detrital material, this correlation strongly suggests that factor 4 represents a third detrital component.
Factor 5 explains 0.3% of variance and has larger scores for Fe2O3, MgO, and K2O and smaller scores for TiO2, MnO, and P2O5. This factor shows a faint positive correlation with pyrite (r2 = 0.12). Detailed examination of factor 5 loading revealed that samples with high factor 5 loading are restricted to three samples from a thick turbidite sand layer at 14.7 mbsf. This sand layer is characterized with abundant glauconite grains. Because glauconite is characterized with higher contents of Fe2O3, MgO, and K2O, it is likely that this factor represents glauconite. Smear-slide observation suggests that those samples with slightly high loadings of factor 5 also contain small amounts of glauconite grains.
Because grain-size distribution may reflect relative contribution of different transportation mechanisms and/or different detrital sources, we analyzed grain-size distribution of lithogenic fraction for selected samples, quantified relative amount of different size classes, and determined position of the principal mode. Seventy nine selected samples consist of 37 (every 6-cm interval) from the Holocene, two from the Bølling/Allerød, two from MIS 2, 37 (including 33 samples from every 6-cm interval) from MIS 3, and one from MIS 4. High-resolution sample sets were analyzed in two intervals to characterize decimeter-scale variations in sediment composition during the Holocene and MIS 3.
Grain-size distribution of these samples shows either unimodal or bimodal distribution with a shoulder on the finer side of the principal mode, and the distribution is resolved into two or three log normal grain-size distributions. The principal mode is located between 8 and 15 µm with average of 12 µm (termed "fine silt") and the subordinate modes are located between 39 and 73 µm with an average of 57 µm (termed "coarse silt") and between 2 and 4 µm with an average of 3 µm (termed "clay;" Fig. 10). The relative amount of each mode is different among samples and one of the subordinate modes is absent in several cases. Examples of the result of curve fitting are shown in Figure 10, and the relative amounts of each mode for analyzed samples are listed in Appendix C.
Fine silt is always a major fraction of the lithogenic component and its volume ranges from 59 to 86 vol%, whereas the volumes of coarse-silt and clay fractions range from 0 to 28 vol% and 5 to 27 vol%, respectively. The volumes of coarse- and fine-silt fractions have strong negative correlations (r2 = 0.74), whereas volumes of clay, coarse-silt, and fine-silt fractions do not show a clear correlation between one another. This is partly due to the larger error associated with volume estimation of the fine-silt fraction.
Correlations between modal volumes and factor loadings are examined and results are listed in Table 3. It should be noted that the correlations are calculated for selected samples that do not include turbidite sands; consequently, they are heavily biased toward the Holocene and MIS 3. Factor 1 shows a faint positive correlation with fine silt (r2 = 0.17) and a weak negative correlation with coarse-silt fraction (r2 = 0.29), respectively. Factor 2 shows a faint positive correlation with coarse-silt fraction (r2 = 0.14) and a faint negative correlation with fine-silt fraction (r2 = 0.10). Factor 3 does not show any correlation with silt and clay fractions. Factor 4 shows a faint negative correlation with fine-silt fraction (r2 = 0.10). Factor 5 shows a faint positive correlation with fine-silt fraction (r2 = 0.17). It is also noteworthy that factor 5 shows moderate negative correlation with modal size of fine-silt fraction (r2 = 0.48). The result suggests that factor 1 is preferably held in finer fractions (fine silt and clay), factors 2 and 4 in coarse-silt fraction, and factor 5 in fine-silt fraction. Factor 3 does not have special affinity to any size factions. This is partly because samples from sand layers, which commonly have higher factor 3 loadings, are not included and also because those grain-size parameters were determined for lithogenous fraction after acid treatment.
Vertical variations in volumes of each grain-size fraction and modal position of fine-silt fraction is shown in Figure 11. The volume of coarse-silt fraction is moderate and decreasing downward within the Holocene, low within the Bølling/Allerød, high during MIS 2, and moderate again in MIS 3. Variation of fine-silt fraction is basically a mirror image of the coarse-silt fraction, although a minimum corresponding to MIS 2 is not significant. The volume of clay fraction does not vary significantly except for one sample from the Bølling/Allerød where it is nearly twice as high. Superimposed on these trends are decimeter-scale variations that are obvious only within the intervals of high-resolution sample sets. Within the Holocene, the fine fraction shows three maxima at 0.95, 1.6, and 2.2 mbsf, whereas maxima are found at 7.6, 8.2, 8.4, and 8.8 mbsf within the middle of MIS 3. These maxima correlate well with the minima of factor 2 (and maxima of factor 1).
We also examined modal grain size of fine-silt fraction, which is clearly recognized on grain-size distribution diagram. Modal grain size of fine-silt fraction is relatively large (10.6-14.9 µm) in the Holocene. It abruptly decreases to 9.0 µm near the bottom of Holocene and then is small (8.0-8.7 µm) in the Bølling/Allerød, moderate around 10.5 µm in MIS 2, and moderate to large (10.0-13.5 µm) in MIS 3 (Fig. 11). Within the high resolution part of MIS 3, modal grain size shows maxima at 7.3, 7.8, 8.2, 8.5, and 8.8 mbsf that tend to coincide with the maxima of factor 2. Modal grain size of fine-silt fraction has a weak positive correlation (r2 = 0.27) with the volume of coarse-silt fraction.
The degree of pyritization (DOPT) used here is defined as the ratio of pyrite Fe within total Fe. It is commonly used to estimate relative (not absolute) levels of bottom-water oxygenation conditions with higher DOPT suggesting less oxic conditions (Berner, 1970, 1984; Calvert and Karlin, 1991; Tada et al., 1992). We estimated pyrite content from pyrite peak height on X-ray diffractograms using the regression equation derived by Tada et al. (1999) to calculate DOPT. DOPT does not show any correlation with Fe2O3 nor loadings of any factors, suggesting that it is not strongly controlled by sediment composition. On the other hand, DOPT shows a strong positive correlation with the pyrite-S/org-C ratio (r2 = 0.60), which is another indicator of bottom water oxygenation level (Berner and Raiswell, 1983), supporting the appropriateness to use DOPT as a relative measure of the bottom-water oxygenation level. An org-C vs. pyrite-S plot (not shown) indicates that most of the samples analyzed fall within the range of normal marine of Berner and Raiswell (1983) when samples from turbidite sands and the top 40 cmbsf were excluded. This implies that the bottom-water oxygenation level at the studied site did not become euxinic during the last 80 k.y. Excluding samples from the top 40 cmbsf, where active pyrite formation process seems still in progress, DOPT of the studied samples varied between 0.06 and 0.32, with an average of 0.16.
Vertical profile of DOPT is shown in Figure 12. DOPT is high within the Bølling/Allerød and middle to lower part of MIS 2, intermediate within the Holocene and Younger Dryas, intermediate to high in MIS 5a, intermediate in the upper part of MIS 2, intermediate to low in MIS 3, and low in MIS 4. Superimposed on these trends are decimeter-scale fluctuations, which are most evident within MIS 3 where distinct DOPT maxima occur approximately at 7.8, 8.8, 9.5, and 10.6 mbsf.