We have constructed mass accumulation rates (MARs) for the cores for the 0-140 ka period for which we have confidence in our age model. In each case the MAR is calculated from the product of the sedimentation rate, the percentage, and the dry bulk density (grams of sediment per cubic centimeter of original wet sample). For all sites except Cores W8709-13 and Y74-2-22 we estimated the dry bulk density from GRAPE wet bulk density because we had measurements at essentially all of our sample depths from MST analyses. The estimate is based on a linear correlation between wet and dry bulk density:

dry = 1.563 wet - 1.560. (1)

For both Cores W8709-13 and Y74-2-22, we estimated a change of dry bulk density with depth by comparison to nearby cores or drill sites where physical properties data were available.

We constructed MARs because MAR time series are independent of the behavior of other sedimentary components, provided that the age model is accurate. CaCO3 and Corg weight percent time series may be driven by changes in dilution by other components, especially in slope sediments dominated by terrigenous clays and other aluminosilicates. Figure 9 shows the record at Site 1019, which was most strongly affected by changes in terrigenous sedimentation. The CaCO3 MAR peak near the last glacial maximum all but disappears in the weight percent time series because of the huge increase in terrigenous sediment burial during this interval.

Fortunately, most of the cores had less extreme changes in sedimentation rates than Site 1019, and most records resembled those of Site 1020 (Fig. 10). The conversion to MAR makes little difference in the variance in the CaCO3 time series, because the CaCO3 weight percent peaks are a factor of 2-5 higher than the surrounding intervals, whereas the sedimentation rate varies by only about a factor of 2. The sedimentation rate model also has its dominant variance at lower frequencies than the carbonate events. The MAR conversion is more important for Corg, because the change in Corg weight percent is of the same magnitude as the change in sedimentation rates, and the dominant frequencies for variation are more similar in both time series.

The net effect of conversion to MAR is that all the carbon time series develop a stronger coherence to each other. Part of the increased coherence comes from use of a common age model, which, if wrong, will contaminate all the records with spurious peaks. However, more of the improvement comes from the removal of dilution artifacts in the weight percent record, as Figure 9 demonstrates.