15N values indicative of lesser nitrate limitation within the system at these times (Fig. F3C). This suggests that surface productivity is primarily controlled by the nutrient budget of the upwelled waters, rather than the rate of overturning.
Because oceanic water is undersaturated in silica, diatom frustules will start to dissolve following bacterial degradation of their organic coating (e.g., Barron, 1993). Consequently, the diatom fossil record is not likely to represent the totality of the surface-dwelling living assemblage (e.g, Barron and Baldauf, 1989; Barron, 1993), and therefore, their relationship is not direct (Schrader and Schuette, 1981). However, the abundance of planktonic marine diatoms in surface and core sediments is directly related to surface water productivity (e.g., Barron and Baldauf, 1989), where highly productive surface waters are able to sustain more diatom growth than poorly productive waters. For example, during bloom events, the abundance of diatoms in a given volume of seawater may increase one hundredfold (e.g., Barron, 1993). In these situations, opal export production rates may exceed the rates of dissolution within the marine environment, thus permitting better representation of surface water productivity within the sediment (Schneider et al., 1997, and references therein). Indeed, previous studies of the upwelling areas (e.g., Bárcena and Abrantes, 1998; Treppke et al., 1996) have shown a close correspondence between the species found at the sediment/water interface and those found within the photic waters above.
The bulk of diatom dissolution takes place within the upper 1000 m of the water column ("silica corrosion zone"), and within the upper 1-2 cm of the sediment (e.g., Schrader and Schuette, 1981). Therefore for good preservation within the record (little breakage and abundant whole valves [e.g., Burckle and Cirilli, 1987]), both descent through the water column and the SR must be rapid. A high SR is recorded at Site 1083 (average 9.45 cm/k.y.). The descent through the water column for individual diatoms may be of the order of 100 yr (e.g., Burckle and Cirilli, 1987). Sinking is greatly facilitated through clumping via incorporation into zooplankton fecal pellets and/or marine snow, or through flocculation with other diatoms and clay particles. Descent within zooplankton fecal pellets has been measured at up to 30-400 m/day in the Southern Ocean (Burckle and Cirilli, 1987), which in addition may serve to limit degradation by enclosing diatoms within protective organic "capsules" (Lalli and Parsons, 1997).
Descent through flocculation can occur in Chaetoceros resting spores where an adhesive marine gel is secreted (Grimm et al., 1996) to stimulate clumping and rapid "self"-sedimentation. By possessing both a highly silicified (and therefore relatively dissolution-resistant) frustule-adopting and this self-sedimentation strategy, Chaetoceros resting spores have the potential to be well preserved within marine sediments.
From the data presented in this study, it is proposed that intense upwelling-driven productivity was not a significant feature of the time period prior to the INHG at ~2.54 Ma, with comparatively little fluctuation between interglacial and glacial periods. HIRM data from the same sediments, which can be used to infer wind intensity, also reveal little change in the atmospheric systems operative within the two climatic states strength (shown in Fig. F2 for comparison). This is consistent with the concept that the ITCZ was in a more northerly position at this time, and a strong trade wind system had not yet developed (Hay, 1993; Hay and Brock, 1992). Similar results were found from diatom studies at Site 1084, where the presence of diatom mats (not recorded at this site) were used to infer strong frontal systems in operation at this time (Berger et al., 1998). Although it would still have occurred, upwelling would have been weaker, a suggestion that is supported by Uk´37 sea-surface temperature data (A. Rosell-Melé and M.A. Maslin, unpubl. data), which show sustained warmer temperatures, with little fluctuation. This record is considered to be reliable, as the Uk´37 index is unaffected by passage through the food chain and diagenesis (Treppke et al., 1996). Consequently, it has been used very successfully in paleoclimatic studies of modern glacial/interglacial sediments (e.g., Rosell-Melé et al., 1997; Schneider et al., 1995; Kennedy and Brassell, 1992).
The data presented here would suggest that following the INHG a productivity regime with strong interglacial-glacial contrast developed. Previous studies have proposed the development and intensification of the trade winds about this time (e.g., Hay and Brock, 1992; Hay, 1993), where, as a result of ice accumulation up in the Northern Hemisphere, the ITCZ shifted southward, prompting alteration of the thermal gradient. This would have caused the longshore winds (brought about by the low-pressure cell over the Namib Desert) to intensify through the constraints created by the escarpment of the Kalahari Plateau, (Dowsett and Willard, 1996), thereby initiating development of a strong upwelling regime. Upwelling is therefore inferred to have intensified during the glacial periods (Stages 100, 98, 96, 94). This is supported by the Uk´37 sea-surface temperature record (Rosell-Melé and Maslin, unpubl. data), whereby glacials correspond with marked cold periods that are attributed to the intensified upwelling of colder waters. In addition, the HIRM data suggest that in comparison to the interglacial periods, the wind intensity increased and remained pronounced throughout the glacials following the INHG. Such a model of the wind regime is illustrated in Figure F3A.
However, in the case for productivity, the three independent sets of data all suggest that there is an initial surge during the glaciation periods, after which levels fall to values similar to those of the previous interglacial stage. This is shown in Figure F3B, which charts the productivity signal chiefly derived from the record of Chaetoceros resting spores. Such a divorce between inferred wind strength (and therefore upwelling) and productivity would suggest that glacial productivity within the waters above Site 1083 was, in fact, not predominantly controlled by wind-driven upwelling but rather by some factor influencing productivity over Site 1083.
During upwelling, nutrient-rich waters are vertically advected from depths of 200-330 m, from within the poleward-flowing undercurrent (Dowsett and Willard, 1996). The nutrient pool of this water body is derived from the ventilation by the underlying AAIW (see "Modern Hydrography"). Downhole variations in abundance of the AA diatom group reflect changes in the nature of the influence of AAIW (and thus its corresponding contribution to the nutrient pool) on upwelled waters. The diatom data show that during highly productive glaciation periods following the INHG there are often concomitant increases in the abundance of Southern Ocean species. This suggests changes in the nutrient regime of the system at these times.
Experimentation has shown that nitrogen deficiency is the single factor that consistently induces spore formation (Leventer, 1991, and references therein; Schuette and Schrader, 1979). Therefore, resting spore formation should take place as soon as NO3- becomes a limiting factor within the environment. Providing that there is representational preservation, the abundance of spores could potentially reflect the initial level of NO3- available within the system, with high abundances of Chaetoceros resting spores indicating periods when nutrients were in plentiful supply. Such enhanced supply could explain the periods of high productivity (reflected in all three productivity indicators) during the initial glaciations, and the corresponding lighter 15N values indicative of lesser nitrate limitation within the system at these times (Fig. F3C). This suggests that surface productivity is primarily controlled by the nutrient budget of the upwelled waters, rather than the rate of overturning.
Figure F3 is a simplified summary interpretation of the wind strength-productivity-nutrient availability relationship, as revealed by data in this study, for the time periods both before and after the INHG. Prior to ~2.54 Ma, a transition from interglacial to glacial conditions initiates a response within all three proxies in the same direction. However, following the INHG, the system response differs in that where the productivity and nutrient-content signals remain coupled, they are not necessarily linked with wind intensity (and therefore upwelling intensity).
The question arises, therefore, as to what is exerting a control over the concentration of the nutrient pool over the site. Upwelling of a more intense nature or the upwelling of more fertile waters, would both act to increase the supply of nutrients to the photic zone. In the first scenario, an increase in upwelling intensity would have the effect of increasing the rate at which nutrients were delivered to the surface waters, therefore increasing the concentration of the nutrient pool per unit time. This could be reflected by simultaneously elevated abundances in both the AA group (via the enhanced upwelling of AAIW-derived waters) and Chaetoceros resting spores (due to elevated productivity). In the second scenario, factors affecting the initial nutrient content of the upwelled waters per unit time (i.e., the fertility of the ventilating AAIW) must be addressed: AAIW forms to the south of the APFZ, where upwelled NADW mixes with Antarctic surface water. Because its density is greater than that of the water to the north, this water sinks to a depth of between 100 and 800 m (Dowsett and Willard, 1996). The extent of the Antarctic sea ice affects the fertility of the AAIW that is released to the wider oceans. For example, during the early Pliocene, the extent of sea ice was at a minimum, and the APFZ was displaced toward the Antarctic continent (Dowsett and Willard, 1996). This created a much larger surface water area within which planktonic oceanic Antarctic diatoms proliferated and depleted the water of nutrients. By the time this water was upwelled in the Benguela system, it would have been already depleted of nutrients (and relatively richer in AA diatoms), thus limiting surface water productivity at Site 1083. When the ice volume of the Antarctic continent expanded again, it created a narrowing of the diatom belt, leading to underutilization of the nutrients in the surface waters of the Southern Ocean. In this case, the AAIW reaching the BC would have been comparatively enriched in nutrients (and relatively depleted in AA diatoms) stimulating productivity at Site 1083. Therefore, the concentration of the nutrient pool per unit volume of water affected productivity in this case.
The fertility-related AA component of the diatom assemblage may therefore reflect either a change in upwelling intensity or a change of the concentration of AA diatoms within the same unit volume of upwelled water. Changes in productivity over Site 1083 can therefore reflect three scenarios:
Although proxies indicative of wind intensity, productivity, AAIW influence, and mode of nutrient utilization are presented in this paper, it is unlikely that such data can suggest to what extent the productivity is responding to each scenario. However the "combination" scenario seems most probable, because oceanographic and atmospheric systems are both known to alter in response to climate change.
Increases in primary production within the early glacials could have important implications for global climate, since they represent periods of enhanced CO2 drawdown—a well-known feedback mechanism in the reduction of global temperatures. Because increased production is coeval with more rapid positive excursions in the 18O signal (especially during Stage 100), an important link exists between oceanic productivity and the decline to more pronounced glacial conditions following the INHG.
