The 18O record reproduces well the contrast between Stages 2 and 1, and between Stages 6 and 5 (Fig. F2A). Stage 4 is not well represented in our data. Not enough isotope determinations are available for Stage 7 to define this period. Comparisons of the Globobulimina spp. 18O record with the BFAR and PFAR indices reveal higher accumulation rates of foraminifers during glacial periods. Transitions of BFAR and PFAR between glacial and interglacial regimes are rapid. Furthermore, foraminiferal abundance maxima tend to occur during the transitions. The original abundance index (BF/g and PF/g) shows very similar patterns (Fig. F2B, F2C). The close correspondence of these parameters indicates that BFAR and PFAR records are not an artifact of sedimentation rate estimates.

The benthic foraminiferal assemblages represent an integration of foraminiferal shell production and preservation in the seafloor, and a portion of the variability in benthic foraminifers abundance may well have to be ascribed to fluctuations in preservation. Several methods can be used to assess the state of preservation, such as (1) percentage of CaCO3 (e.g., Luz and Shackleton, 1975), (2) percentage of sand fraction (>63 µm) (e.g., Johnson et al., 1977), (3) percentage of fragmented planktonic foraminifer tests (fragmentation) (e.g., Thunell, 1976), and (4) ratio of benthic to planktonic foraminifers (B/P) (e.g., Parker and Berger, 1971). Because each of these approaches may partly be controlled by other factors (e.g., productivity, dilution, and ecology), we look at them jointly (Fig. F3). The carbonate and sand content do not correlate well with BF/g (r = 0.45 and r = 0.31, respectively), as might be expected if preservation were the controlling factor for all three series. On the other hand, the overall trends in B/P and fragmentation point to increased dissolution during interglacials, with better preservation during glacial periods. Gaps in the fragmentation record (mostly during isotopic Stages 5 and 7) represent intervals with very few planktonic foraminifers and planktonic foraminifer fragments. We attribute the lack of planktonic foraminifers and fragments not to dilution but dissolution, because the sedimentation rate was not particularly high during those intervals. Nonetheless, benthic foraminiferal tests are generally in good condition and do not show a high degree of fragmentation. Consequently, although a preservational effect is likely, the evidence suggests that productivity is the dominant factor in producing the BF/g record.

To further test that the benthic foraminifers at Site 1079 are responding to productivity fluctuations, we compare the BFAR record with independent evidence of variations in productivity. The planktonic foraminifer G. bulloides is known to favor nutrient-rich upwelling areas in the tropical ocean (Prell and Curry, 1981). It also occurs in the coastal upwelling area off Namibia (Giraudeau, 1993) and in the eastern equatorial upwelling region (Kemle-von Mücke and Oberhänsli, 1999). Although only a few samples were analyzed for their abundance in planktonic foraminifers, it seems that increased abundances of G. bulloides are consistent with the idea that productivity was high during glacial intervals and low during interglacials (Fig. F4A). Moreover, its relative abundance shows a pattern similar to that of sea-surface temperature (SST) (Fig. F4B), based on alkenone analyses in core GeoB1016 (Schneider et al., 1995). Of course, G. bulloides is not a particularly dissolution-resistant species, and it could be argued that reduced abundances during interglacials result from dissolution.

We also compare the BFAR record with the Corg (in weight percent) record in core GeoB1016 (Fig. F4A). Given the strong variations and relatively high pollen content (Ning and Dupont, 1997), we cannot exclude varying terrigenous portions of Corg in core GeoB1016. However, we believe Corg (weight percent) in core GeoB1016 is a better indicator for marine productivity than that of Site 1079, more strongly influenced by the presence of continental-derived material (Shipboard Scientific Party, 1998b). We use the percentage of Corg, instead of the accumulation rate (Corg AR), because it has been shown that in core GeoB1016 the record of Corg concentration is more consistent with records of nutrient proxies than the Corg AR (Schneider et al., 1994), probably biased by problems with the oxygen isotope stratigraphy (Bickert, 1992). Despite a few mismatches, the records of Corg (weight percent) and BFAR show a similar temporal variability. It is been suggested that BFAR may not generally be used as a productivity tracer. Naidu and Malmgren (1995) showed that BFAR did not record the productivity signal in the upwelling and oxygen minimum zone along the Oman Margin. They argued that the low oxygen concentrations might instead be controlling the benthic foraminifer abundance. Our data suggest that, similar to the results of Guichard et al. (1997) off northwest Africa, at Site 1079 oxygen was not a limiting factor, and the benthic foraminifers responded primarily to changes in the flux of organic matter.

There is also a good correlation between the BFAR record and the influx of pollen and spores measured in core GeoB1016-3 (Ning and Dupont, 1997) (Fig. F4C). If we interpret the pollen flux as a measure of offshore winds (Ning and Dupont, 1997), the relationship suggests that during glacial periods enhanced winds increased the upwelling of cold and nutrient-rich waters, supporting high productivity levels.

Additional clues on the course of productivity variations can be obtained from the changes in composition of the benthic foraminifer assemblage. Loubere et al. (1993) and Loubere (1996) showed that the link of taphonomic process to organic carbon flux, combined with benthic response to organic flux, produces distinctive fossil assemblages linked to carbon flux gradients. At Site 1079, the assemblage is dominated by calcareous taxa with infaunal microhabitat preferences and ability to tolerate persistent oxygen depletion resulting from the oxidation of organic matter. In contrast, epifaunal species, which are less tolerant of oxygen depletion (Corliss and Emerson, 1990), are mainly present during warm periods, probably due to lower productivity and higher dissolved oxygen concentrations in bottom waters (Fig. F5A). Important genera represented at Site 1079 include Bolivina, Bulimina, Cassidulina, Epistominella, Globobulimina, and Uvigerina. The sum of Bolivina pseudopunctata and Bolivina dilatata (Fig. F5B, F5C) reflects BF/g (and infauna) because of their dominance within the assemblages. Abundances of these two species differ within glacial periods. B. pseudopunctata appears to be better correlated with marine Corg, represented by the Corg record of core GeoB1016, than B. dilatata, which tends to covary with the Corg record of Site 1079. This suggests that B. pseudopunctata prefers labile organic matter, whereas B. dilatata does well where influx of terrigenous organic matter dominates the environment. The fact that Corg (weight percent) at Site 1079 shows maxima during interglacials could be related to increases in terrestrial nutrient input due to stronger monsoonal flow and precipitation. Alternatively, high interglacial Corg (weight percent) values could result from decreased dilution by terrigenous input during sea level highstands. Also, the seasonal pattern of the flux of organic matter to the sea bed seems to have influenced the character of the benthic assemblage. Distributions of some recent deep-sea benthic foraminiferal species are sensitive to seasonal pulses of phytoplankton detritus (Gooday, 1993). Corliss and Silva (1993) showed a rapid seasonal growth response in benthic foraminifers along the California borderlands, probably in response to increased organic matter flux to the seafloor. As for the fossil record, Hermelin and Shimmield (1995) used benthic foraminiferal assemblages recovered from the Arabian Sea to interpret productivity events during the last 150 k.y. Likewise, Thomas et al. (1995) studied benthic foraminifers from the North Atlantic for the last 45 k.y. and interpreted changes in assemblages in terms of abundances of phytodetritus-linked species. The spikiness in the abundance of B. pseudopunctata (up to 5000 specimens/g) leads us to believe that this species may be opportunistic and may respond rapidly to threshold values in environmental conditions. Very few data are available for this species in modern environments.

Additional support for this hypothesis comes from the results of the cluster analysis. The dendrogram in Figure F6 shows two major clusters, each represented by one of the two most abundant species. B. pseudopunctata associates with Bulimina exilis and Bulimina marginata. According to previous studies (Caralp, 1984; Jannink et al., 1998), B. exilis flourishes under conditions of pulse-like accumulating organic matter of high nutritive quality. On the other hand, species that group with B. dilatata, such as Uvigerina auberiana and Fursenkoina mexicana, may require high but more sustained rather than strongly pulsed flux of organic matter. Nonionella spp., Cassidulina laevigata, Chilostomella oolina, Globobulimina spp., Epistominella smithi, and Bolivina subaenariensis are grouped in a cluster, representing the third main contributor to the abundance of benthic foraminifers. These species typically occur in areas with high organic matter input and can tolerate reduced oxygen concentrations (Sen Gupta and Machain-Castillo, 1993; Bernhard et al., 1997). It is worth noting that Uvigerina peregrina/hispida and Bulimina aculeata are present in samples where U. auberiana and B. marginata are low, respectively (primarily during interglacials). This pattern could be explained by different ecological preferences related to factors that vary with water depth (e.g., temperature, oxygen, and grain size) and/or resistance to dissolution.

Spectral analyses of the log-transformed BF/g (not presented here) and BFAR records yielded similar results. Variations in the benthic foraminifer abundance and accumulation rates at Site 1079 occur at periods of 100 k.y. (eccentricity) and 23 k.y. (precession), whereas the 41-k.y. cycle (obliquity) is poorly defined (Fig. F7). The distinct 100-k.y. cycle corresponds to the major glacial-interglacial contrast represented by the relatively short record. The predominance of the 23- and 100-k.y. periods over the 41-k.y. period is in accordance with spectral results from SST and productivity records of core GeoB1016 and other eastern South Atlantic cores (Schneider et al., 1996) as well as with previous results from equatorial Atlantic records (e.g., McIntyre et al., 1989). The record of BFAR at Site 1079 also shows strong power at the 52-k.y. period, which may correspond to the difference tone between 41 and 23 k.y., whereas the presence of modest power near 15 k.y. may be related to the sum frequency of the 23- and 41-k.y. cycles, as indicated by Jansen et al. (1996). These authors documented a strong 15-k.y. signal in the record of the Angola-Benguela Front paleopositions during the last 180 k.y. This suggests that migrations of the Angola-Benguela Front may have influenced the productivity fluctuations at Site 1079 and, ultimately, the record of benthic foraminifers. Unfortunately, effects from dissolution on the productivity indices used cannot be entirely excluded; this also may influence the resulting spectrum, making it more complex than it would be for either productivity or dissolution alone.