For the FA, all LWD data at each site except PSR and U were taken into account. The shallow resistivity data PSR were not used because they correlate strongly with the deep resistivity ATR and are nearly identical to ATR. This strong correlation would weigh resistivity too heavily compared to the remaining data. This identical behavior of PSR and ATR is an effect of measuring the formation only minutes after it has been drilled. There is no time that the mud can infiltrate into the formation. The deep resistivity was used rather than the shallow resistivity, because it is more likely to be representative of the undisturbed sediment away from the borehole. The U data were not used because they showed very low values over the entire borehole with little character and contained unreliable negative values (obviously due to the processing from the gamma-ray spectra). Accordingly, the computed gamma-ray CGR (GR with U portion subtracted) was also not used. Although the TNPH data showed a noisy and scattered character because of overall high porosities in the formation, a low-pass filtering of the data made them useful for the statistical analyses.
The results of the factor analysis of the LWD data with factor eigenvalues and factor loadings are given in Tables T1, T2, T3, T4, T5, T6, T7, T8, T9, and T10 are graphically presented as factor logs in Figures F2, F3, F4, F5, and F6. Factor loadings >0.4 are significant and shown in bold in Tables T2, T4, T6, T8, and T10 and also at the bottom of Figures F2, F3, F4, F5, and F6. Except for Site 1045, where two factors could be extracted, three factors were extracted from the original data sets. At Site 1045, only ~40 m was logged below the décollement (Fig. F1). This reduced data set may have caused the smaller number of factors in this case.
The results at all sites show that factor 3 is closely related to the deep resistivity ATR (factor loadings >0.75). Accordingly at Site 1045 (Fig. F3), the ATR log is shown together with the two factor logs. The gamma-ray log is shown for all sites for comparison reasons.
The factor analysis shows that the most discriminating variable at all drill sites is the GR. GR is mainly related to lithology and, in this case, in particular to the clay type and clay content. At all sites, GR has a factor loading of >0.9. Together with the Th and K content, it forms the factor 1 log at all sites. All factor loadings of factor 1 are positive and >0.8; thus the assigned physical or chemical properties show a good positive correlation. Often, the Th/K ratio is taken as an indicator for the clay type (Rider, 1996; Jurado et al., 1997). This means that the factor 1 log is mirroring the lithology with mainly varying clay type and clay content. Because illite has the highest K content among the different clay types (Rider, 1996), borehole sections with high factor 1 values may indicate higher illite concentrations, whereas sections with low factor 1 values may be characteristic of a higher smectite content (in particular within the décollement zones). However, this could also simply mean a lower clay mineral content since factor 1 has a high positive loading for Th as well as K.
At all sites (except Site 1045), the ROMT and TNPH show the highest loadings for factor 2. As expected, the signs of the factor loadings for ROMT and TNPH are opposite: high density sections have low porosity values and vice versa. Thus, factor 2 is mainly responding to the porosity of the formation. After calibration with core sample measurements (e.g., on cores from the nearby Leg 156 sites), the factor 2 log could result in a reliable and true porosity profile. In some cases (Sites 1044, 1047, and 1048), factor 2 is also loaded by PEF. The PEF is closely related to mineralogical composition and, thus, to lithology (Rider, 1996). However, because the PEF is a direct function of atomic number (to the fourth order), pore water and, thus, porosity will also have some influence on PEF in addition to changes in lithology. But PEF is influenced primarily by lithology and only secondarily by porosity. This consequence of basic physics is based on the aggregate atomic number of water, which is much lower than those of rock forming minerals and, when they are mixed together, the combined PEF is controlled by the weight concentration. This means that porosity effects on PEF are much less than seen on either the density or neutron log responses. According to Ellis (1987), kaolinite has a relative low atomic number, but other clay minerals show higher responses to PEF that reflect iron content (having a high atomic number). As ROMT is closely related to PEF, Figure F7 (upper row of crossplots) also mirrors the relation between porosity and PEF. This relation is good above the décollement, but only fair to bad below it.
The deep resistivity ATR, and to a smaller extent PEF, is the main loading for factor 3 at all sites (except Site 1045). Thus, factor 3 is likely related to changes in the electrical properties of the sediments. This might involve several influences such as grain sizes, grain orientation, the presence of conductive minerals, cementation, and varying ionic concentrations within bound water on the clay minerals.
Together with the logging units, the cluster logs are shown in color at the left side of Figures F2, F3, F4, F5, and F6. An overview of all cluster logs is given in Figure F7. The mean and standard deviation values for each cluster are given by Moore, Klaus, et al. (1998). Each cluster represents intervals where the physical and chemical rock properties are presumably similar. At all sites, four to six significant clusters could be derived by dendrogram evaluation. In this way, the clusters in the cluster logs can be seen as statistical electrofacies as defined by Serra (1986). This clustering facilitates the subdivision of the borehole into logging units, which can be compared to lithology and porosity. In all cluster logs, the décollement zone is clearly identified by cluster 1 values. As can be derived from Figure F7, cluster 1 is characterized by the lowest density values and by low gamma-ray and PEF values. Clusters 1 and 2 are all above the décollement, whereas clusters 5 and 6 are only below the décollement zone. However, because the cluster logs were calculated individually for each site (because of software constraints), a stratigraphic correlation between the wells by using the clusters has not been possible before now. In the next step, all downhole logging data of Leg 171A will be put together in one data set to perform a consistent suite of cluster logs that can be used to follow up the geological units from well to well. The results compared to the seismic profiles will be further investigated.