METHODS

All samples were oven dried at 50°C and divided in subsamples. Bulk sediment mineralogy was determined by X-ray diffractrometry (XRD) (using a SCINTAG XRD 2000 diffractometer), based on a semiquantitative estimation and using external standards (Kübler, 1983). This method allows us to quantify a certain number of mineral species (i.e., quartz, calcite, K-feldspar, and plagioclase). Other mineral components, including amorphous minerals, Fe oxides and hydroxides, some clay minerals, and organic material cannot be quantified because of the lack of appropriate standards and technical complications (for a review, see Kübler, 1983). The relative error of the bulk rock mineralogy is ~5%.

Clay mineral analyses were based on methods by Kübler (1987). An aliquot of the sample was mixed with deionized water (pH = 7-8), and the carbonate fraction was removed by the addition of 10% HCl (1.25 N). Separation of different grain size fractions (<2 µm and 2 to 16 µm) was obtained by the timed settling method based on Stokes law. XRD analyses of oriented clay samples were made after air drying at room temperature and in ethylene glycol solvated conditions. Clay minerals for the two fractions are given in relative percent abundance. Smectite content is estimated by using the method of Moore and Reynolds (1997). The relative error does not exceed 10% of raw intensity.

Grain size analyses were performed on the insoluble and carbonate-free residues prepared for the clay mineralogy analyses using a grain size Laser Oriel CIS (Computerized Inspection System). The relative error is 10% (Jantschik et al., 1992).

The characterization of organic matter was performed on 100 mg of dried and ground sediment with a Rock-Eval 6 and using a standard whole-rock pyrolysis method (Espitalié et al., 1986; Lafargue et al., 1996). The distribution of sedimentary phosphorus phases was determined using a four-step sequential extraction technique adapted from the SEDEX method (Anderson and Delaney, 2000; Ruttenberg, 1992). For all the steps other than the iron-bound P, the ascorbic acid-molybdate blue method was employed to analyze the extracted supernatants (Greenberg, 1995). A Perkin Elmer spectrophotometer was used to determine absorbences and P concentrations. All samples were diluted 1:10 for spectrophotometric determination. Concentrations of the iron-bound P phase and ferric Fe, extracted during the first step of the procedure, were determined using an OPTIMA 3000 Perkin Elmer inductively coupled plasma-atomic emission spectrophotometer at the "Environmental Laboratory" in Neuchātel. Typical mean errors for each phase were calculated as the relative standard deviation of the consistency standards and range from 2% to 7% for authigenic P, from 3% to 5% for detrital P, from 2% to 5% for organic P, and from 3% to 6% for iron-bound P. Because of material limitation, XRF was performed only on four selected samples at the Geological Institute of Fribourg using a sequential X-Ray spectrometer PW2400. Based on the analysis of a certified reference standard, the relative average error was ±1%.

The depth of the GCLs was recorded during the shipboard description of the sediments directly on core description log sheets ("barrel sheets" produced by AppleCore software). The depth in meters below seafloor of each GCL at Sites 1143, 1145, 1146, 1147, and 1148 was extracted from the barrel sheets. Depths in meters composite depth and ages were calculated using the splices and the age models calculated on board and provided in Wang, Prell, Blum, et al. (2000) (see Table T2). Occurrences of GCLs from Site 1147 are analyzed together with the ones from Site 1148. Site 1147 was drilled to recover sediments that were missing or disturbed in the uppermost part of Site 1148 (Wang, Prell, Blum, et al., 2000). Tests for randomness, as well as autocorrelation and spectral analysis, were performed to investigate their temporal distribution over the last million years. All the records were resampled at even intervals and detrended before being analyzed. The Arand (by P.J. Howell, Brown University) and MatLab software packages were used.

NEXT