Biosedimentological data from the studied platform succession were examined from the cores of both Holes 1196A and 1199A, using the onboard visual core descriptions (Shipboard Scientific Party, 2002b). Microfacies analysis with additional multivariate analysis of bioclasts was conducted on 84 thin sections originating from Holes 1196A, 1196B, and 1199A. A total of 24 categories of bioclasts were examined, including invertebrates, coralline algae, and foraminifers. Conventions from Coleman (1963), Adams (1968), Loeblich and Tappan (1988), Betzler and Chapronière (1993), Chapronière and Betzler (1993), and Boudagher-Fadel et al. (2000a, 2000b) were used for large benthic foraminifer determination, generally at a generic scale. Coral taxa and their ecological significance were determined following conventions from Beauvais et al., (1993), Boichart et al., (1985), Cahuzac and Chaix (1996), Chaix et al., (1986), Ditlev (1980), Vaughan and Wells (1943), Veron (2000), and Wells (1956). The coral assemblages are shown in Table T1. ADE-4 statistical programs from the University of Lyon I were used to analyze the data. A total of 59 thin sections with sufficient bioclasts were point-counted with at least 300 counts using a conventional grain solid technique. The quantitative values were entered into a contingency data matrix with thin sections in rows and categories of bioclasts in columns. These values were transformed to percentages and then ranked in value ranging from 0 to 9, using the logarithmic formula
where X = percentage. The transformation reduces large discrepancies between quantitative values. It allows multivariate analysis to take variables into account that have low values but possible high environmental significance (see Sokal and Rohlf, 2003, for more details). The obtained encoded data matrix was treated by ascendant cluster analysis (ACA) and by correspondence factor analysis (CFA).
The ACA permits ordering thin sections within groups of both increasing hierarchy and decreasing correlation, according to biological variables. Relationships between thin sections and groups are evaluated on a dendrogram. These groups are believed to represent microfacies. This microfacies matrix is obtained from the percentage matrix by rearrangement of rows, according to ACA ordination. The microfacies and the percentage of the biological variables are shown with lithostratigraphic units in Figure F4.
The CFA simultaneously computes correlations within and between thin sections and variables. It represents, within a multidimensional space of N factorial axis, the structure of data as two clouds of N + 1 variable points and M thin section points, respectively (M > N). The CFA runs an ordination of axes from 1 to N, according to the decrease of variance or inertia of both clouds. Complex relationships between thin sections and/or variables and their distribution were examined on factorial planes defined by the first axes, which collect most of the variance. Each variable and each thin section provide an absolute contribution (AC, part of variance) and a relative contribution (RC = 10,000 x r2, where r = coefficient of correlation) to each axis. Only discriminative variables (AC > 10,000/N+1 and RC > 2500, i.e., |r| > 0.5) were examined (see Benzécri et al., 1980, and Etter, 1999, for more details). The variables are given in Table T2.
The application of both ACA and CFA on the same encoded matrix allows the definition of microfacies that possibly contain specific environmental significance. Moreover, the application of both ACA and CFA allows us to define and explain the interrelationships between the microfacies regarding environmental trends (i.e., Cugny and Rey, 1981; Hennebert and Lees, 1991; Nebelsick, 1992). The statistical results are then used to develop three-dimensional depositional reconstructions that take into account information about seismic profiles, lithostratigraphy, biosedimentology, and microfacies analysis.