By means of the multivariate statistical methods factor and cluster analysis, it was possible to reduce the dimensionality of LWD data from Leg 171A without loss of important information. The resulting set of factor and cluster logs make subsequent evaluations and interpretations much easier. At all sites, the analyses resulted in two or three factor logs that are consistently loaded by comparable factor loadings. We conclude that factor 1 and factor 2 are good proxies for lithology and porosity respectively. Factor 3 possibly contains additional information regarding changes in the electrical properties of the sediments.
No cores were recovered during Leg 171A. However, cores were recovered during Leg 156 and, by design, several of these sites are near Leg 171A sites. The physical properties core measurements from Leg 156 will be used in a future study to calibrate the results from the statistical evaluations of LWD logs from Leg 171A. Obviously, this will enhance the interpretation of the meaning of the factor logs and clusters. Of course, the exact number and choice of input logs for the factor analysis is varying according to the experience of the user and the geology of the logged sequence as well as to the target of the results wanted. But this preconditioning is helpful in improving the factors and making their geological meaning more explicit.
However, the preliminary results from multivariate statistics can be used for first-order interpretation. In Figure F6, all cluster logs are shown together with plots of cross correlations of ATR with ROMT and of GR with PEF. The upper row of crossplots (ATR-ROMT) dramatically shows the change of physical properties behavior above and below the décollement zone. The ATR-ROMT correlation is strong above the décollement zone because changes in porosity greatly exceed changes in lithology. The lithology is essentially homogenous and consists of calcareous clay with minor variations due to clay type, ash content, and structure (fractures, dipping beds, etc.). These small variations in lithology can also be followed up by the lower row of crossplots with the PEF-GR correlations. Following the west-east transect, no big change in the scatter plots can be seen. Therefore, above the décollement zone, the variance in both logs is mainly caused by porosity. Below the décollement zone, there is much greater variability in lithology, which contributes variance to the ATR and ROMT logs in differing manners. For instance, the ATR log is more sensitive to the grain size than is ROMT. Also, there are still effects caused by localized structure. Thus, there is poor correlation between ATR and ROMT below the décollement zone. A similar effect can be seen in the PEF log at Site 1045, for example. The PEF responds primarily to lithology and only secondarily to porosity (Rider, 1996). However, above the décollement zone, PEF shows a strong correlation with ATR. Below the décollement zone, the correlation is very weak. Thus, according to the general compaction trend above the décollement zone, both logs are responding primarily to changes in porosity. Below the décollement zone, PEF responds mainly to lithology variations, whereas ATR still responds primarily to porosity. This is verified by the strong correlation between PEF and ROMT both above and below the décollement zone. In other words, there is clear visual evidence that PEF and ROMT (and GR, Th, and K) are sensitive to lithology, whereas ATR is more sensitive to porosity.
The main objective of using the multivariate methods factor and cluster analysis in this study was twofold: