The differences observed in the amplitude of the planktonic and benthic 18O and 13C variations indicate that temperature, salinity, and productivity changes occurred in the surface water and deep water masses in relation to past climatic changes. The salinity variations that are related to the global ice effect are similarly recorded by the 18O values of planktonic and benthic foraminifers. The 18O gradients between foraminiferal species are thus affected only by the vertical thermal gradients and by local salinity variations at the ocean surface due to differences in the precipitation-evaporation balance.

The export of organic carbon from the photic zone extracts preferentially 12C from the dissolved inorganic carbon reservoir, which becomes richer in 13C; there is thus a direct relationship between high 13C values and high rates of biological pumping in surface waters. However, inputs of deep nutrients by upwelling introduce 13C-depleted CO2 in the surface waters, which is characteristic of eutrophic conditions and causes the decrease of the surface-subsurface 13C gradient. Intermediate between oligotrophic and eutrophic conditions, mesotrophic situations exhibit high surface-subsurface 13C gradients due to the combination of lateral advection of deep nutrients and high surface productivity levels (Pierre et al., 1994). At depth, the contribution of CO2 derived from organic matter oxidation causes a decrease of the 13C values of deep waters. During glacial times, high 13C gradients between planktonic and benthic foraminifers indicative of high productivity were generally coeval with low atmospheric CO2 concentrations in the air trapped in antarctic ice, which was interpreted by Berger et al. (1989) as an argument on the link between the biological pump and the global concentration of atmospheric CO2. Recently, Broecker and Henderson (1998) have proposed that the ocean productivity changes are global and related to the oceanic nitrogen cycle, which is controlled by dust flux. In this scenario, increasing nitrogen fixation during glacial times would cause the increase of CO2 uptake in surface waters and the drawdown of atmospheric CO2 concentrations.

The estimation of the oxygen and carbon isotopic gradients between foraminiferal species living at different depths or at different seasons may help to reconstruct changes in surface water hydrography. Giraudeau (1993) and Giraudeau and Rogers (1994) provided information on the spatial distribution of planktonic foraminiferal species in the Benguela upwelling system. Their studies showed that planktonic assemblages are distributed according to the various surface water masses and associated hydrological fronts induced by the upwelling process. Among the three planktonic species that were used in the present stable isotope study, G. ruber is considered to be characteristic of warm surface waters; the subpolar species G. bulloides dominates the planktonic assemblages of cool, nutrient-rich, newly upwelled waters, whereas the deep-dwelling species G. inflata is associated with offshore oligotrophic waters.

In the following discussion, the surface-to-deep isotopic gradients 18O and 13C are given by the difference between the 18O and 13C values of G. inflata and Cibicides. The isotopic gradients between G. inflata and G. ruber are used as a tracer of the seasonal or vertical (depth of the thermocline) temperature gradient in the surface mixed layer, whereas the isotopic gradients between G. inflata and G. bulloides are used as an index of the variable contributions of oligotrophic offshore waters and upwelled eutrophic waters at the site location.

Oxygen Isotopic Gradient

Surface-Deep Water Gradient

The 18O G. inflata-Cibicides record displays large oscillations that roughly follow the G-IG cyclicity (Fig. F5A). Second-order, rapid, high-amplitude fluctuations of 18O around the present-day gradient (18O values enriched by up to 1.0 or depleted by up to 0.8) are indicative of the surface-to-bottom temperature and salinity gradients. High 18O values are interpreted as either decreasing surface-to-bottom temperature gradients or increasing surface-to-bottom salinity gradients.

The largest oscillations of 18O with amplitudes up to 1.8 occur between 80 and 280 ka, a period that comprises MISs 5 to 8. Such high amplitudes cannot be attributed to changes of up to 8C in the surface-to-bottom temperature gradient because such temperature changes would be extreme. We therefore consider that surface waters were also affected by important salinity variations. A comparison of the isotopic data with the planktonic foraminiferal distribution at Hole 1087A (Giraudeau et al., Chap. 7, this volume) shows that some peaks of high 18O values (terminations I and II, MISs 11 and 12) are in phase with the peaks of abundance of G. menardii, both proxies indicating maximum penetration of salty 18O-rich Indian Ocean thermocline waters. Other peaks of high 18O values (MISs 4, 5, 6, 7, 8, and 12) coincide with the maximum abundances of G. bulloides, which suggests the presence of cold, upwelled surface water. Low 18O values generally occur during cold events of glacial and interglacial stages; if these isotopic gradients are interpreted in terms of the surface-to-bottom temperature gradient, the cooling of surface waters had to be less than that of deep waters.

Surface Water Gradients

The amplitude of 18O variations through time is greatest for G. bulloides and least for G. ruber, except for MIS 12. This is reflected by the planktonic 18O variations; for example, the amplitude for G. inflata-G. bulloides is twice that of G. inflata-G. ruber (Fig. F5B, F5C). MIS 11 is poorly constrained due to the scarcity of G. bulloides and G. ruber during this time interval.

The 18O G. inflata-G. bulloides is generally much higher than during the Holocene, except for a few rapid events during MIS 2, 8, and 10 that reach the Holocene level (Fig. F5B). The overall high 18O values, particularly during isotopic events 5.3 and 9.3, are caused by an increase in 18O values of G. inflata and a decrease in 18O values of G. bulloides. Conversely, the 18O decreases, such as those observed at the MIS transitions 11/10, 7/6, 5/4, and 4/3, correspond to rather constant 18O values of G. inflata and to increasing 18O values of G. bulloides. These isotopic variations integrate the variations of a multicomponent system where the seaward extension of the upwelled waters relative to the offshore oligotrophic waters is modulated by the cooling intensity in the upwelling cell and by the strength of the Benguela Current.

The 18O G. inflata-G. ruber and 18O G. inflata-G. bulloides records display very similar vertical patterns (Fig. F5C). They show no difference in their amplitude during G-IG cycles; however, the highest 18O values occur during the warm substages reflecting periods of higher temperature contrast in the surface waters. The onset of deglaciations displays a typical evolution with a sharp increase followed by a sharp decrease of 18O values; this can be interpreted either as a succession of deepening and shoaling of the thermocline or as abrupt changes from high to low seasonal contrast. Similar observations were made by Wefer et al. (1996) in the Benguela Current System, and these authors emphasized the difficulty of interpreting this proxy in terms of seasonal contrast or thermocline depth. In any case, this evolution would mean that the Benguela Current was temporarily reduced at the end of glacial stages and then reinforced at the beginning of interglacial stages.

Carbon Isotopic Gradient

Surface-Deep Water Gradient

The 13C G. inflata-Cibicides values are subjected both to sharp short-term and long-term oscillations, with a maximum amplitude of 1.7, that are not related to the G-IG cycles (Fig. F6A). The 13C variations thus are quite different from the pattern described by Berger et al. (1989). We assume that at Site 1087, latitudinal and meridional migrations of hydrological fronts dominated the surface water productivity.

The short-term periods of low 13C values, which indicate low 13C values of surface waters, suggest an influx of Indian Ocean thermocline waters. Recent 13C measurements along a transect south of Africa along 30E show that the surface waters of the Agulhas Current have a carbon isotopic signature lower by 0.5 than the subantarctic waters (Archambeau et al., 1998; Quentin, 1997). The important contribution of Indian Ocean waters can thus easily explain the decrease of the surface-to-bottom carbon isotopic gradient. The other peaks of low 13C values occur independent of glacial stages (MISs 2, 3, 4, and 6) and interglacial stages (MISs 5, 7, and 13), but they correspond to the periods of maximum abundance of G. bulloides that are also marked by 18O anomalies; in these cases, the 13C decrease at the surface is more likely caused by the seaward extension of 13C-depleted upwelled waters.

The high 13C values are mostly concentrated in late MIS 5 and during the 260- to 425-ka interval. As mentioned above, the high 13C values of planktonic foraminifers were interpreted as indicative of high primary productivity. These events, which interrupt the negative 13C excursions, indicate the installation of a different regime dominated by more oligotrophic conditions.

Surface Water Gradients

The carbon isotope gradients in the surface waters are balanced by the uptake of 13C-depleted CO2, resulting from surface productivity and by the inputs of 13C-depleted CO2 due to the vertical and lateral advection of nutrient-rich waters. The 13C variations between G. inflata and G. bulloides are twice those measured between G. inflata and G. ruber (Fig. F6B, F6C); this is explained by the largest 13C variations recorded by G. bulloides, a species that is the most sensitive to changes in upwelling strength. The overall 13C patterns show no correspondence with the G-IG cycles; however, these patterns document global trends toward increasing or decreasing values, which are inferred to represent major changes in the trophic regime and in the depth of the thermocline that are controlled by the lateral and vertical advective fluxes of nutrient-rich deep waters.

The minimum values of the carbon isotope gradient between G. inflata and G. bulloides are interpreted as representative of oligotrophic conditions, whereas the maximum values are indicative of mesotrophic conditions. Three steps may be identified following the evolution of the G. inflata-G. bulloides 13C values (Fig. F6B). Starting from 500 ka, where the 13C values are similar to the Holocene values, there is a decrease by >1 (on average) up to 400 ka; this period corresponds to the initiation of oligotrophic conditions that remain quite steady up to 350 ka. After 350 ka and up to 100 ka, this planktonic 13C increases by >2 on average, marking the progressive development of more mesotrophic conditions. After 100 ka, the 13C values decrease again by ~1 to reach the Holocene values, marking the return toward oligotrophic conditions.

The carbon isotope gradient between G. inflata and G. ruber is considered to follow the vertical movement of the thermocline: shoaling of the thermocline brings deep nutrients closer to the surface and the 13C values decrease in the surface waters, whereas deepening of the thermocline diminishes the inputs of deep nutrients to the surface and results in the increase of 13C values. The same three steps are also observed in the 13C G. inflata-G. ruber record (Fig. F6C). During the period from 500 ka to 350 ka, 13C values increase by ~0.7 on average, an indication for the progressive deepening of the thermocline. The second step from 350 ka to 100 ka corresponds to the progressive decrease of 13C by ~0.6 and is related to the progressive shoaling of the thermocline. During the last step, 13C increases again by ~0.3, suggesting that the thermocline deepened slightly to reach the present-day situation.