DISCUSSION

The ¹⁸O record of low-latitude Atlantic Ocean Site 1006 appears to be much more variable than the low-latitude Pacific Ocean Site 806 record from the Ontong Java Plateau throughout its entire 1.4 m.y. (Fig. 6). The small interglacial/glacial amplitude changes in the Site 806 record may not be representative, however, of the western Pacific. Sedimentation rates are relatively low at Site 806, implying that bioturbation may have smoothed the isotope record. Other isotope records from the western Pacific show large amplitude changes as summarized in Martinez et al. (1997). Some western Atlantic sites exhibit the large amplitude changes such as shown at Site 1006. For instance, Curry and Oppo (1997) record a 2.1 change across the Holocene/glacial transition at Ceara Rise, very similar to the amplitude change recorded at Site 1006. High-resolution cores taken from drift sediments are most likely less affected by bioturbation, revealing the maximum Holocene/glacial amplitude.

Broecker (1989) noticed that the amplitudinal change is much larger for the last glacial maximum/Holocene transition in the Atlantic, and he concluded that a larger salinity contrast existed between the Atlantic and Pacific during the glacial periods than today, assuming that temperature change was on the same order of magnitude in both tropical oceans. The Atlantic-Pacific contrast appears to have been significantly larger during several glacial periods such as glacial Stages 6, 8, 14, 20, 22, and also during several glacial periods before 900 ka (Fig. 6). The glacial ¹⁸O difference between the Pacific and Atlantic is on average ~1.5. The salinity contrast (3.3 psu) during the cold periods must have been larger than the salinity contrast today, taking the average 1.5 difference in ¹⁸O and assuming a similar temperature regime in both oceans. This would imply increased water vapor loss from the Atlantic, making the waters around the Great Bahama Bank extremely salty. The salinity is difficult to establish because of potential differential temperature changes in the Atlantic and Pacific, and because Site 806 probably does not record the maximum interglacial/glacial amplitude in ¹⁸O in the western Pacific.

Differential temperature changes between the Pacific and Atlantic during glacial periods are not unlikely when taking into account potential past changes in global oceanic circulation patterns. Today, warm Pacific surface waters return to the southeastern Atlantic via the low-latitude Indian Ocean and the Agulhas Current. The Benguela system transports the warm waters farther north via the South Equatorial and the North Brazil Currents through the Caribbean into the Gulf Stream. One of the bottlenecks in this general oceanic circulation system is the transport of warm waters around the Cape of South Africa before the warmer waters can be transported northward. Past northward movements of the subtropical convergence may have prevented or reduced the influence of Agulhas waters coming around, thus reducing heat delivery to the north. Reduction of transport of warm water around the Cape would have happened most certainly during glacial periods. A reduction in warm-water planktonic foraminiferal abundances during the last four major glacial periods indicates cooler temperatures around the Cape (Imbrie et al., 1973; Berger et al., 1985; Little, 1997) during maximum northerly expansion of the Antarctic Polar Front. Another cooling mechanism of the surface waters in the southern Atlantic relates to the Benguelan upwelling system. The southeastern trade wind system was stronger during the cooler periods forcing increased upwelling along the southwestern coast of Africa which made the Benguelan surface waters cooler (Little et al., 1997). The glacial surface waters in the Benguelan system would have moved northward, eventually affecting the warm pool waters of the Caribbean through cross-equatorial flow.

A consequence of cross-equatorial flow of glacial cool water is that glacial northward heat transport was reduced when a mixture of colder waters of subantarctic origin flowed from the west through Drake Passage, replacing the warmer waters from the east. Cross-equatorial flow of this relatively cool water would lead to heavier ¹⁸O values in planktonic foraminiferal shells in the western equatorial Atlantic Ocean. However, the relatively low salinity values of southern origin would make the ¹⁸O values lighter. Assuming that generally cooler waters crossed the equatorial Atlantic, the temperature difference between the Atlantic and Pacific Oceans would increase and the salinity difference would decrease because the present low-latitude Atlantic is saltier than the Pacific; therefore, at least a large part of the glacial ¹⁸O difference between the Atlantic and Pacific could have been controlled by temperature. Temperature is the more likely factor in the glacial ¹⁸O difference between the Atlantic and Pacific because sufficiently increased salinity in the Gulf Stream area would lead to a strong Atlantic conveyor circulation. In contrast, glacial deep-sea ventilation in the Atlantic was reduced (Boyle and Keigwin, 1982; Curry and Lohmann, 1983; Curry et al. 1988; Sarnthein et al., 1994).

One of the implications of changes in global circulation at the glacial/interglacial time scale is that the interglacial/glacial temperature range in the low-latitude Atlantic was larger than in the Pacific. The coupled oxygen isotope records of Sites 1006 and 806 would suggest a larger temperature range: Site 1006 shows a much larger isotope variability than does Site 806. Martinez et al. (1997), however, documented that the fluctuations in ¹⁸O on the interglacial/glacial time scale are much larger at other sites in the western Pacific, which would indicate that the oxygen isotope records are much more dependent on regional evaporation-precipitation cycles and temperature changes. It is, therefore, impossible to know how valid an Atlantic-Pacific ¹⁸O comparison based on Sites 1006 and 806 is. Another problem is that bioturbation may have affected the records depending on sedimentation rates. We conclude that if global circulation changes have modified the oxygen isotope record of Site 1006, temperature changes largely explain the isotopic variability.

The interglacial ¹⁸O values of Site 1006 are mainly more positive than at Site 806. The interglacial ¹⁸O values should be heavier in the low-latitude Atlantic than in the Pacific, assuming that a salinity contrast similar to that of the present day existed during all interglacial periods. There is no temperature difference between the two tropical oceans during the summer, although a temperature contrast may exist during winter as the Bahamian area cools ~5°C and the Ontong Java Plateau area does not. G. ruber mainly calcifies its shell during the summer (Cifelli and Smith, 1974), although Deuser (1986) shows that G. ruber is present year round in subtropical areas. A combination of slightly cooler temperatures and higher salinity explains why the Site 1006 oxygen isotope curve is heavier in general than the Site 806 profile. In addition, the ¹⁸O record of Site 1006 was generated using G. ruber, while the ¹⁸O profile of Site 806 used G. sacculifer. G. ruber tends to generate slightly lighter ¹⁸O values than G. sacculifer, which may further explain why the Site 806 curve shows lighter ¹⁸O values than the Site 1006 curve. This general feature can be seen in all interglacial periods except for the Holocene and Stage 9. We don't know why the ¹⁸O difference between G. ruber (Site 1006) and G. sacculifer (Site 806) disappears during the Holocene and Stage 9 interglacial periods.

The ¹⁸O difference prior to Stage 11 was in general larger (on average 0.7) during the sea-level highstands. A larger salinity contrast by increased vapor export from the Atlantic to the Pacific may possibly explain this, or an increased temperature difference. A temperature difference seems to be the more likely scenario in light of the above discussion. Unfortunately, there are no paleotemperature data off South Africa prior to Stage 11 available to determine whether somewhat cooler waters came around the Cape during the sea-level highstands before Stage 11. The oxygen isotope record of Site 1082A, drilled in the Benguelan system, recently produced by Jahn et al. (1999) shows trends very similar to the Site 1006 isotope record with lighter ¹⁸O values during the interglacial periods since isotopic Stage 11. The similarity between the two records shows that global circulation patterns are involved in modifying the temperature and salinity regimes that influence the oxygen isotope records. The ¹⁸O record of Site 806 also shows a slight trend toward lower values in isotopic Stage 11 (Fig. 6). Another western Pacific example comes from the Great Barrier Reef area where Sites 820 and 823 were drilled. Both records show a ¹⁸O decrease during isotopic Stage 11 (Peerdeman et al., 1993; Alexander, 1996). The similarity between the Pacific and Atlantic stable isotope records of a ¹⁸O decrease since isotopic Stage 11 suggests that temperature increased in both areas, possibly as a result of strengthening of the warm pool of surface ocean water in both areas.

The ¹⁸O records of benthic foraminifers and also the SPECMAP curve show slightly heavier values for the pre-Stage 11 highstands, implying slightly lower sea levels and perhaps a slightly larger Antarctic ice sheet during these periods than today (Kuijpers, 1989). Although it is difficult to imagine why the Antarctic ice sheet may have been slightly larger, there is ample evidence that subantarctic waters were cooler before Stage 11 (Keany and Kennett, 1972). Cold subantarctic waters extended farther northward and displaced the Antarctic Polar Front to the north, possibly reducing return flow of warmer waters from the Indian Ocean. Jansen et al. (1986) postulated that the climatic belts were displaced northward by ~5° latitude, which would have hindered the intrusion of warm waters into the Atlantic. Therefore, in this context, we interpret the Stage 11 ¹⁸O decrease as a warming and not as a result of reduced salinity. The pre-Stage 11 interglacial ¹⁸O values in Site 1006 are on average ~0.6 heavier than post-Stage 11, which would imply ~3°C warming.

Late Pleistocene warming during the mid-Brunhes Chron seems to coincide with a period when the earth's climatic system was mainly dominated by 100-k.y. cycles, although warming of surface waters as recorded by a decrease in stable oxygen isotope values in sediments from isotopic Stage 11 occurred much later than the onset of 100-k.y. cycles as defined by the mid-Pleistocene revolution at ~1 Ma (Berger et al., 1993). It is not known which phenomenon triggered the increase in amplitudinal change in climate around isotopic Stage 11 characterized by very cold glacial and very warm interglacial stages. Direct solar insolation was minimal at this time because of weakened precessional cycles when eccentricity was at a low (Imbrie and Imbrie, 1980). Jansen et al. (1986) speculated that the mid-Brunhes changes in temperature were triggered by the 400-k.y. cycle, but other possibilities such as ocean circulation changes cannot be excluded. It is simply not known which feedback mechanisms controlled the increase in global climate change since isotopic Stage 11. We speculate that increased return flow of surface waters from the Pacific to the Atlantic may have been involved, or perhaps partly involved, as one of the mechanisms because the Sites 1006 and 806 oxygen isotope curves converge during the sea-level highstands, pointing to a more homogeneous world ocean with similar surface conditions in the low latitudes.

This trend seen in the ¹⁸O record of amplification of climate change during the mid-Brunhes (younger than isotopic Stage 12) is also developed in the carbonate mineralogy and grain-size distribution profiles at various other Leg 166 sites. There is a marked increase in the average concentrations of aragonite (from 66 to 71 wt%) and in HMC (from 2 to 4 wt%), while LMC decreases at Site 1003 (from 29 to 15 wt%; Rendle et al., Chap. 6, this volume). These values are calculated from the extreme values within each stage for both sedimentary units. Quartz shows continuous representation until the Stage 11 transition; above that its occurrence is sporadic. This change is probably related to a change in the ocean circulation pattern, with a reduced input of southern material (Lidz and McNeill, 1998). This change accompanies a general coarsening of the sediments at Site 1003 (Rendle et al., Chap. 6, this volume). Regional changes in circulation patterns may be responsible for this, and thus may have contributed to the mid-Brunhes oxygen isotope changes.

We conclude that the Site 1006 isotopic variability can at least in part be explained by global circulation changes affecting the temperature regime in the western equatorial Atlantic, although regional changes in circulation may have affected the area as well. The glacial/interglacial amplitude is large throughout the record and is proposed to be due to temperature variability. The interglacial periods since isotopic Stage 11 show the lightest ¹⁸O values, which we interpret as mid-Brunhes surface-water warming. This trend is contemporaneous with changes in the mineralogy and grain-size distribution as described by Rendle et al (Chap. 6, this volume), suggesting that regional climate changes and ocean circulation changes could have altered the stable isotope record of Site 1006 as well.