Little can be said about the relative reliability of the two chemical methods used because of the absence of absolute sediment standards for determining the accuracy of any biogenic opal technique. In any case, most of the differences could be due to inhomogeneities among samples rather than reflecting the accuracy of the techniques. Paradoxically, OLin values were consistently higher, even though a weaker digestion solution (0.5-N NaOH vs. 1-N NaOH in OPer method) was used.

One source of the discrepancy between OLin and OPer results may be the extraction of silica from coexisting marine clays and authigenic silicates in the OLin technique. Experimental and field studies (Berner, 1981; DeMaster et al., 1983) have reported the preferential dissolution of ultrafine particles, such as clays, during the initial extraction period. Therefore, excessive grinding of the samples might have erroneously increased the opal content in OLin samples, especially those with low opal contents. In addition, OLin samples were extracted for 5 hr, whereas in OPer samples, the extraction time was dependent on the sample type (Fig. F3). Opal-poor samples were dissolved within 1 hr, whereas 2 hr were required for opal-rich samples. The 5-hr dissolution in OLin samples could have resulted in silica being leached from clay minerals. However, we cannot rule out the possibility of incomplete digestion of diatoms in OPer data. Extraction curves show an early spectral absorbance increase due to the preferential dissolution of opal followed by a slower, linear increase representing the dissolution rates of the silicate minerals present. Although dissolution curves appeared linear after 1 and 2 hr, longer periods might have been required to extract 100% of the biogenic opal in the samples.

Despite all difficulties, the overall patterns between OLin and OPer data are strikingly similar. The results from the two techniques are comparable and provide the opportunity to combine the two records by applying a linear transform (Fig. F4A, F4B). Thus, longer and more detailed opal records for subsequent time series analysis could be obtained by splitting samples among laboratories and merging the results after a proper calibration and statistical treatment.

Interestingly, the substantial increase in the concentration of opal at 2.6 Ma (Fig. F5A) is similar to that found at the northern and southern boundaries of the BC system. In the equatorial Atlantic (ODP Sites 662, 663 and 664), Ruddiman and Janecek (1989) reported abrupt upper Pliocene increases in opal and terrigenous dust fluxes near 2.5 Ma. Also, in the subantarctic region (ODP Site 704), Froelich et al. (1991) showed a marked increase in the accumulation rates of biogenic silica at ~2.5 Ma. Cieselski and Grinstead (1986) identified the latest Gauss/early Matuyama (2.67-2.47 Ma) as the time of greatest change in Neogene climate in the northern antarctic and subantarctic regions.

The comparison of opal content with DAI at both sites (Fig. F5B, F5C) indicates that the DAI is a good proxy for opal concentration in sediments, at least in coastal upwelling areas. Studies on sediment traps at Walvis Ridge also showed that diatoms closely parallel variations in opal fluxes (Treppke et al., 1996). On the other hand, it is noteworthy that DAI has a logarithmic nature, which derives from the convention used in the smear-slide analysis (see "Methodology and Stratigraphy").

The evidence of high opal and diatom deposition in the upper Pliocene and the specific composition of the diatom assemblage lead us to suggest that for the area of the Benguela Current system, a combination of increased upwelling and intensified advection of silica-rich subsurface waters must have existed at MDM time. The merging of different data sets describing the variations of opal deposition in and around the MDM will make more detailed studies of this feature possible.