We have constructed a composite depth scale using the program SPLICER, developed by Peter deMenocal and Ann Esmay (SPLICER version 2.2 is available on the World Wide Web from the Borehole Research Group at Lamont Doherty Earth Observatory, Columbia University, at http://www.ldeo.columbia.edu/BRG/ODP). This program allows data from several holes at a site to be viewed and manipulated simultaneously. Correlations are made within the program by selecting data from one core from one of the holes and then overlying it with data from another core from a different hole. The data being compared can be shifted vertically relative to each other until a preferred correlation is made. These depth offsets are recorded in a SPLICER "affine" table. After correlating all cores between holes, the end result is an affine table that gives the depth shift that needs to be applied to the mbsf depth scale for each core to place the cores into the composite depth scale. The depth of each core can then be computed in meters composite depth, as presented in Table T2 for Site 1098.
Correlation typically begins with cores from the top of the sedimentary section and proceeds downhole. For sites with multiple holes, each cored from the seafloor to some depth, the initial core selected for the top of the composite section (0 mcd) would ideally be one in which the mudline and several meters of core were recovered with minimal coring deformation. Features from cores from the other holes would then be aligned with features from the initial core by applying depth shifts, and so on downhole. Because core expansion is common, the correlation process results in mcd scales that are commonly expanded by about 2%-20% from the true depth (e.g., Farrell and Janecek, 1991; Hagelberg et al., 1995; MacKillop et al., 1995). Hence, the composite depth scale should not be confused with a true depth scale or even a best estimate of a true depth scale (see discussion in Harris et al., 1995). Instead, it is a common depth scale that provides a stratigraphic framework for other studies.
To determine the quality of the correlation, we used visual comparison as well as the calculated correlation coefficient shown graphically within SPLICER. The correlation coefficient is determined by cross-correlating the data from two cores. Its value ranges from +1 to -1, and the window over which it is calculated is adjustable. In nearly every case, we used the default window length of ±2 m above and below a selected tie point. Occasionally, we reduced this to ±1 m to focus on specific features or to avoid features related to coring disturbance.
No expansion or compression of one core or core interval relative to others is permitted within SPLICER. Because of this restriction, it is not always possible to exactly align every anomaly with similar anomalies in other cores. Second-order correlation of that nature can be best accomplished by using programs like ANALYSERIES (Paillard et al., 1996) or by using methods like inverse correlation (Martinson et al., 1982) to adjust the first-order correlation accomplished with SPLICER.
A splice is constructed by combining discrete intervals from the cores from different holes to give one continuous sedimentary section for the site. The splice uses portions of data from different holes after the data have been converted to the mcd scale. Typically, a cored interval selected to be part of the splice will have the least drilling disturbance and the fewest gas voids, IW samples, and irregularities when compared with the same depth interval from other holes. We have produced a "splicertable," which gives the intervals that are preferred for the spliced sedimentary section (Table T3).