Summary statistics of the planktonic foraminiferal census counts are given in Table T4. Throughout the studied part of Hole 1087A, species assemblages bear a clear transitional to subpolar character. A total of 17 taxa (excluding the G. menardii complex) were identified in the >125- µm fraction. Of these, eight occur with a maximum frequency >5% and an average frequency >1%. They are, in order of decreasing average abundance, N. pachyderma (s) (d), G. inflata, G. bulloides, G. quinqueloba, Globigerinita glutinata, G. truncatulinoides, Globigerinoides ruber, and N. pachyderma (s).
This ranking, in terms of species dominance, is consistent with the distribution of planktonic foraminiferal species in surface sediments in the vicinity of the studied site. Site 1087 is presently located below the main Benguela drift, seaward of the boundary of active coastal upwelling, within a mixing domain of old upwelled and oligotrophic waters (Fig. F1). This ill-defined frontal region is characterized by a typical modern assemblage of G. inflata and G. bulloides, with common N. pachyderma (d), which extends equatorward up to Walvis Ridge, with a tendency for seaward deflection off Namibia related to a seaward extension of the filamentous regime north of 27°S (Giraudeau, 1993). N. pachyderma (d) and G. bulloides presently define this filamentous regime of mesotrophic, old-upwelled waters, whereas G. inflata is the indigenous species of the transitional faunal zone, which in the southeast Atlantic, extends in an equatorward direction along the Benguela drift (Niebler and Gersonde, 1998).
Within the interval representative of the last ~250 k.y., both dominant and subordinate taxa display a general glacial-interglacial trend in abundance changes (Fig. F3). Peak interglacials are associated with maximum abundances of G. inflata, close to Holocene values, with significant contribution of G. bulloides, whereas glacial intervals and cold substages within interglacial periods are characterized by an association dominated by the cold end-members of the planktonic foraminiferal assemblage, namely N. pachyderma (d) and, to a lesser extent, G. quinqueloba. This last association contains a significant contribution of G. glutinata. This general trend down to MIS 8 suggests a glacial-interglacial alternation of either upwelling strength and associated seaward extension of the belt of mature upwelled waters, advection of subantarctic water along the path of the Benguela drift, or a combination of both. Considering that N. pachyderma (s) is presently abundant in continental shelf sediments below the main upwelling cells (Giraudeau, 1993), the near absence of this species throughout this 250-k.y. interval suggests that any glacial pulses in upwelling strength were not sufficient to induce the presence of newly upwelled waters at the location of Site 1087.
The interval from MIS 8 to MIS 12 shows a series of complex faunal distributional trends as well as short-term events that are clearly unrelated to the general pattern observed during the past 250 k.y. (Fig. F3). Abundance changes of the dominant three taxa, N. pachyderma (d), G. inflata, and G. bulloides, do not follow the glacial-interglacial pattern described earlier. G. inflata displays a single abundance peak at the MIS 10/11 boundary, as well as a general tendency for highest concentrations in glacial intervals (with the exception of MIS 12). Similarly, peak abundances of N. pachyderma (d) seem to be restricted to isotope stage boundaries (i.e., MIS 9/8 boundary), as well as to the early phase of interglacial stages (i.e., early MIS 11). Finally, G. bulloides abundance fluctuates only slightly around Holocene values throughout MIS 8 to 11 but shows an abrupt increase during glacial MIS 12. Among the subordinate species, the subpolar and coastal upwelling species G. quinqueloba, which is a minor component of the faunal assemblages representative of the last 200 k.y. at Site 1087, reaches high concentrations well above the Holocene level throughout MIS 7 to 12, with significant peaks irrespective of the glacial-interglacial succession. Perhaps the most dramatic short-term event recorded within the lower half of the studied section is the single abundance peak of N. pachyderma (s) during early MIS 9. This maximum concentration of the cold end-member of the planktonic foraminiferal assemblage at Site 1087 is associated with high abundance of G. quinqueloba, which in the southern Benguela region forms with N. pachyderma (s) the pair of dominant taxa in surface sediments of the inner and middle shelf below coastal upwelling cells (Giraudeau and Rogers, 1994).
Contrary to the major species discussed above, G. truncatulinoides and G. ruber display a rather straightforward distributional trend with regard to the glacial-interglacial cyclicity throughout the last 460 k.y. Both species show highest concentrations at or above Holocene values during warm climatic stages. G. truncatulinoides is a deep-dwelling species that, in the Benguela region, is usually associated with G. inflata as the pair of taxa representative of oligotrophic offshore water (Giraudeau and Rogers, 1994), both species defining the transitional faunal zone in the deep South Atlantic (Bé and Tolderlund, 1971). G. ruber is the warm end-member of the Benguela faunal community and rarely exceeds 2% of the total foraminiferal assemblages in sediments of the southern Benguela region (Giraudeau, 1993). The common maximal occurrence of G. ruber and G. truncatulinoides during warm climatic stages should point to the presence of stratified, warmer surface waters over Site 1087. Although this interpretation may hold true for the last 250 k.y., an interval when other major species suggest a relaxation of the upwelling process during interglacials, this interpretation is too simplistic for the time period prior to MIS 7 when the overall planktonic foraminiferal assemblage shows contradictory patterns.
The relatively low number of taxa identified at Site 1087 is in part due to the lack of key tropical species, such as Pulleniatina obliquiloculata, Sphaeroidinella dehiscens, and Globigerinoides conglobatus, in the local environment of the southern Benguela region. The absence of these species, which are otherwise part of the extant and fossil assemblages of the northern Benguela region (Ufkes et al., 1998; Giraudeau, 1993), as well as of the nearby South Atlantic and South Indian subtropical gyres (Bé and Tolderlund, 1971; Bé and Hutson, 1977), suggests that dramatic hydrological situations associated with extreme surface warming did not occur at the location of Site 1087 during the past 460 k.y. In other words, neither long-term eastward shifts of the subtropical water masses associated with a near-complete relaxation of the coastal upwelling process and Benguela Current drift nor massive inflows of Indian Ocean water around the southern tip of Africa are detectable from our record at the studied time resolution of 2 to 4 k.y.
We do know, however, that substantial leakage of Indian tropical water past the southern tip of Africa, due to large Agulhas rings and filaments, is part of the present circulation pattern of the South East Atlantic (e.g., Lutjeharms, 1996), and that this mechanism is believed to be a key modulator of the Atlantic heat conveyor (Gordon, 1985). The near absence of tropical species at Site 1087 is not so anomalous in that remote sensing studies clearly show that the path of Agulhas-shed rings past the Cape of Good Hope lies along a southeast-northwest track, well offshore of the continental margin (Lutjeharms and Gordon, 1987).
G. menardii, as other tropical species, is negligible in terms of contribution to the foraminiferal assemblages at Site 1087 (<0.5 wt% throughout the studied interval). A separate observation of the whole, unsplit >125-µm fraction, however, indicates that specimens of G. menardii, although rare in abundance, are near-continuously present at Site 1087 throughout the last 460 k.y. (Fig. F4). This presence has major implications. If, as suggested by others (Berger and Vincent, 1986; Charles and Morley, 1988), this species is a reliable tracer of interocean exchange south of Africa, it is likely that the "Cape Valve" was hardly closed throughout the last four climatic cycles, or alternatively, that regardless of the volume of Indian Ocean waters transferred to the Atlantic, leakages to the southern Benguela have not stopped since MIS 12.
The downcore pattern of accumulation rate of G. menardii at Site 1087 (Fig. F4A) is enlightening in view of the late Quaternary succession of this taxon in the tropical and subtropical Atlantic (Ericson and Wollin, 1968). Although the extinction of G. menardii in the tropical Atlantic during certain glacial intervals (i.e., during the last glacial = biozone Y, or during the upper part of MIS 6 = biozone W) is still an enigma, it has been proposed that Indian to Atlantic Ocean leakage of surface water is a plausible mechanism for reseeding this taxon during warm periods (Berger and Wefer, 1996). To our knowledge, our record is the first evidence for such a mechanism. Accumulation rates of G. menardii show short-term distinct peaks that are synchronous with the position of the U/V, W/X, and Y/Z zonal boundaries. If one assumes empirically that the accumulation rate of this species is a function of the volume of Indian Ocean water leakage in the southeast Atlantic, then we might surmise that G. menardii reseeding of the tropical Atlantic is indeed directly tied to maximum interocean exchange south of Africa.
Besides events of peak accumulation of G. menardii at the above mentioned zonal boundaries, an interval of high concentrations of this taxon is evident from MIS 10 to MIS 12 (Fig. F4A). Here, long-term intervals of maximum interocean exchange are expressed particularly during MIS 11 and 12 and are associated with distinct foraminiferal indications of extended warming periods. The ~20-k.y. interval of peak accumulation of G. menardii during MIS 12 is marked by the single occurrence throughout the last 460 k.y. of dominantly dextral G. truncatulinoides at Site 1087 (Fig. F4B). Dextral G. truncatulinoides is found preferentially in tropical water masses of the Indian and Atlantic Oceans (Bé and Tolderlund, 1971), whereas surface sediment studies indicate that the 21°C surface summer isotherm roughly defines the northernmost boundary of dominantly sinistrally coiled specimens in the South Atlantic (Niebler and Gersonde, 1998). The co-occurrence of G. menardii and dextral G. truncatulinoides during MIS 12 therefore traces the presence of tropical waters over Site 1087 throughout what is commonly considered to be one of the most extreme glacial intervals of the late Quaternary in the Southern Ocean (Hodell, 1993; Howard and Prell, 1994).
The following interval of high G. menardii accumulation at Site 1087 during MIS 11 (Fig. F4A) is associated with a high contribution to the total assemblage of the subtropical species G. ruber (Fig. F3). Contrary to the previous MIS 12 period, G. truncatulinoides coiling is predominantly sinistral (Fig. F4B). In the absence of additional information from independent proxies, we can only assume that mechanisms responsible for the MIS 12 and 11 intervals of mixed-layer warming over Site 1087 were different, these differences being in part driven by the extreme gradients in surface-water temperature and by associated latitudinal shifts of hydrological and atmospheric boundaries in the Southern Ocean across these two climatic stages (Howard and Prell, 1994; Hodell, 1993).