Understanding the mechanisms and patterns of millennial-scale climate changes has until now lacked high-resolution and precisely dated marine records from the Southern Hemisphere, particularly from the Pacific sector. Therefore, one of the major goals of Leg 202 was to improve the knowledge of pattern and causes of millennial-scale climate variability in the southeast Pacific, with a particular focus on the last glacial–interglacial cycle.

Modeling studies have shown that the southeast Pacific may play a crucial role for the transmission of temperature anomalies from high southern latitudes to the tropical Pacific both through intermediate water masses (e.g., Gu and Philander, 1997; Schmittner et al., 2003) and perhaps also via the surface currents along western South America. The latter linkage between the eastern boundary current system and the equatorial Pacific seems to be important on longer glacial to interglacial scales (Feldberg and Mix, 2003; Pisias and Mix, 1997) and appears to be at least partly valid likewise on millennial scales (Lea et al., 2006).

Traditionally, it has been thought that millennial-scale changes during the last glacial interval are driven either by instabilities in the Northern Hemisphere ice sheets (MacAyeal, 1993) or feedback related to Atlantic thermohaline circulation (e.g., Rahmstorf, 2002), driven by anomalous input of meltwater into the North Atlantic (e.g., Clark et al., 2001, 2004). In spite of extensive efforts, however, attempts to track all the sources of meltwater and mechanisms for thermohaline oscillations remain controversial (Broecker, 2006). Other ideas for the origin of millennial-scale climate changes may involve rhythmic solar forcing (Bond et al., 2001, 1997) or internal resonant oscillations of the coupled ocean-atmosphere system (e.g., Alley et al., 2001; Ganopolski and Rahmstorf, 2002; Schulz et al., 2002).

Millennial-scale climatic oscillations, in particular during MIS 3, have been detected in many places around the globe (e.g., recent compilations by Clark et al., 2001, and Voelker, 2002) and have been primarily explained by oceanic and atmospheric responses to North Atlantic climate change (Schmittner et al., 2003). However, other studies suggest an important role of both the tropics (including long-term properties of ENSO) and the Southern Ocean (Clement et al., 1999; Koutavas et al., 2002; Stott et al., 2002). For example, changes in SSTs in the tropical Pacific may have large impacts on the global hydrological cycle (Clement et al., 1999) and perhaps on greenhouse gases (Palmer and Pearson, 2003; Visser et al., 2003) and are thus of global significance. A model study of the last glacial climate, which includes calibration with SST estimates from Leg 202 and other studies (Hostetler et al., 2006), suggests that relatively small sea-surface warming in tropical and subtropical oceans can shift Northern Hemisphere ice sheets from positive to negative, meaning that the low-latitude oceans could trigger early phases of deglaciation or interstadial climate oscillations within the ice age. In addition, the high latitudes of the Southern Hemisphere (including sea ice around Antarctica) (Tziperman and Gildor, 2003) may provide controls of greenhouse gases and for resumption of Atlantic thermohaline circulation after previous reductions (Knorr and Lohmann, 2003; Weaver et al., 2003).

Several Leg 202 sites reveal sedimentation rates high enough to resolve millennial-scale or even centennial- to decadal-scale variations (Figs. F11, F12). These include the Chilean margin Sites 1233–1235 that contain very high sedimentation rate sequences (often >100 cm/k.y.) (Fig. F12) as well as two sites from the eastern tropical Pacific and the Panama Basin (Sites 1240 and 1242) (Fig. F11), where sedimentation rates are on the order of 10–20 cm/k.y. Taken together, these sites provide a unique paleoceanographic north-south transect to track millennial-scale variability from the northern margin of the Antarctic Circumpolar Current (ACC) (Site 1233) along the Humboldt Current northward (Sites 1234 and 1235) toward the tropics (Sites 1240 and 1242).

The initial published studies on millennial-scale climate variability based on Leg 202 focus on the southernmost Site 1233 located on the Chilean continental margin at ~41°S (Heusser et al., 2006b; Kaiser et al., in press, 2005; Lamy et al., 2004; Pisias et al., 2006) because the ~70-k.y.-old sequence recovered continuously with the APC extends over ~135 mcd, resulting in unprecedented high sedimentation rates, at least for the South Pacific (Fig. F5). In addition, the site is well located to compare both surface and deep ocean millennial-scale patterns to high southern latitudes. Today, SST gradients within the northernmost ACC are very large and intimately linked to the northern margin of the westerly wind belt, making this region very sensitive to latitudinal shifts of atmospheric and oceanographic circulation associated with the southern westerly winds. Site 1233 is located close to the southern Chilean coast (~40 km) and close to the northwestern margin of the PIS, which occupied a large area of southernmost South America during the last glacial. Thus, this unique location allows detailed comparison of various continental climate and paleoceanographic proxy records within the same archive and therefore avoids problems linked to age model uncertainties.

An important result from Site 1233 is the Antarctic timing and pattern shown in millennial-scale oscillations of alkenone-based SST (Lamy et al., 2004), as well as in radiolarian faunas and pollen assemblages (Pisias et al., 2006; Heusser et al., 2006b) that strongly support the concept of hemispheric asynchrony as seen in the ice records from Greenland and the Antarctic (Blunier and Brook, 2001; Lynch-Stieglitz, 2004) (Fig. F11). Other land-based records of glacier and vegetation changes from the Southern Hemisphere have not presented a clear picture, and some have been previously interpreted to support hemispheric synchrony with climate oscillations of Greenland, such as the Younger-Dryas climate interval (Denton et al., 1999; Lowell et al., 1995; Moreno et al., 2001). Many of these land-based studies have been performed in southern Chile directly onshore of Site 1233. Based on the high-resolution Fe content record from Site 1233 taken as a proxy record for glacial activity on land (Fig. F12), Lamy et al. (2004) suggested that part of these discrepancies might be due to a lagged response of PIS extent changes to SST changes offshore. Modeling studies show that southeast Pacific SSTs exert the dominant control on ice sheet extents in this region (Hulton et al., 2002). The Fe content record indeed shows a pattern that is very similar to that of the SST record, however with a lag of several hundred years.

Kaiser et al. (2005) extended the alkenone-derived SST record at Site 1233 back to 70 ka. They compared the Site 1233 record to the available SST records from the Southern Hemisphere mid-latitudes and conclude that the Antarctic millennial SST pattern was probably a hemisphere-wide phenomenon. Furthermore, they performed SST gradient reconstructions over the complete latitudinal range of the Pacific eastern boundary current system. This reconstruction suggests strengthened and equatorward displaced subtropical gyre circulation during cold stages (MIS 2 and 4). Conversely, oceanic circulation was probably weakened, and the ACC, and associated westerly wind belt, moved southward during relatively warm periods (early MIS 3 and Holocene Climate Optimum).

For Termination 1, Lamy et al. (2004) further presented a paleosalinity record that supports these findings, as it shows a major meltwater input from the PIS ~1000 yr after initial deglacial warming. Kaiser et al. (in press) show that the close relationship of the Fe record and the SST pattern extends into the older part of the records (70–50 ka). During MIS 4, a delay of ~500 yr of PIS retreats relative to SST increases has been found, similar to that described for the coldest part of MIS 3 and 2 (Lamy et al., 2004). During early MIS 3 (~60–56 ka), synchronous variability in both records resembles the deglacial–Holocene time interval, reflecting either a meltwater pulse or a dominant control of Fe input by rainfall changes related to a rather small PIS.

At Site 1234 (1015 m water depth), the benthic foraminiferal stable isotope record clearly records changes in subsurface ocean temperature (and/or salinity) that are in phase with temperature changes in Antarctica (Mix et al., unpubl. data; Heusser et al., 2006a; Robinson et al., 2007; Blunier and Brook, 2001). Site 1234 is also particularly well suited for study of vegetation changes, as it is offshore the modern transition zone between the rainforest-dominated systems to the south and the Mediterranean (semiarid, summer dry) climate systems to the north. Pollen assemblages at Site 1234 reveal high-amplitude oscillations of rainforest vegetation, although lagging the benthic 18O and Byrd ice core record by a few hundred years, suggesting either a ~1000-yr exponential response time that may reflect the time constant of vegetation response to regional climate change (Heusser et al., 2006a) or perhaps the influence of lagging ice sheets as noted by Lamy et al. (2004). However, some pollen assemblages from Site 1233 (Heusser et al., 2006b), interpreted as indicative of latitudinal shifts of the westerly wind belt, in general follow SST changes as reconstructed by radiolarian census data (Fig. F12). Pisias et al. (2006) performed statistical analyses suggesting that both the radiolarian and pollen data co-vary with Antarctic temperature, at least on timescales >3000 yr. They also confirm the systematic lag of changes in ice sheet extent based on the Fe content data (Kaiser et al., in press; Lamy et al., 2004).

Thus, the land-based studies that previously suggested linkage to Northern Hemisphere climate oscillations may be explained as Southern Hemisphere climate responses; however, some areas experienced a small local lag due to the response times of forest and ice systems. These studies highlight the importance of studying terrestrial proxies in the marine cores, thereby avoiding potential offsets induced by age model uncertainties.

These surface ocean data suggest a close coupling of the high southern latitudes and the tropical Pacific, which is consistent with the previous conclusions of Feldberg and Mix (2003) and Pisias and Mix (1997). Furthermore, Lea et al. (2006), based on a high-resolution Mg/Ca SST record from the Galapagos region, show that parts of the millennial-scale SST pattern of the eastern equatorial Pacific appear to be similar to the southern record from Site 1233. SST records currently being constructed at Site 1240 in the eastern tropical Pacific seem to confirm this finding (Cacho et al., 2006). The "Antarctic pattern" appears to be clearer in subsurface ocean records. Cacho et al. (2006) present preliminary data that suggest a clear Antarctic signal in an oxygen isotope record of a thermocline foraminifer species that primarily reflects variations within the EUC fed by Antarctic Intermediate Water (AAIW) (Fig. F11). These data support the concept of an "oceanic tunnel" mechanism for transmission of high-latitude climate changes to the tropical oceans (Liu and Yang, 2003), rather than solely by atmospheric forcing associated with greenhouse gases (Lea et al., 2006), at least on millennial scales.

Similarly, 15N from Site 1234 suggests that denitrification in the southeast Pacific oxygen minimum zone follows an Antarctic climate pattern (Robinson et al., 2007; De Pol-Holz et al., 2006). Episodes of reduced denitrification recorded at Site 1234 represent times when more oxygen was present in the subsurface ocean. An increase in oxygen can be achieved through (1) lower temperatures, (2) higher ventilation rates, and/or (3) reduced oxygen demand in the low-latitude subsurface. Mechanisms that may account for the observations include changes in the chemical composition of SAMW, formed in the Subantarctic Zone of the Southern Ocean and ventilating the low-latitude thermocline (Robinson et al., 2007), or by local salinity control of Eastern South Pacific Intermediate Water (De Pol-Holz et al., 2006).

These preliminary data show that variability of Southern Ocean properties may play an important role in the transfer of climate signals in the Pacific realm during the last glacial. Sites 1233–1235 are particularly well located for assessing variations in the strength of AAIW through time (Fig. F3). Site 1235 monitors the boundary between the Gunther Undercurrent (GUC) (the poleward undercurrent, a relatively low oxygen water mass) and AAIW (a relatively high oxygen water mass). Site 1233 is located roughly in the core of AAIW and thus should provide the best available record of AAIW properties relatively close to its source in the Antarctic Subpolar Front. Finally, Site 1234 monitors the deeper boundary of AAIW in its zone of mixing with Pacific Central Water (PCW) (a relatively low oxygen water mass). Preliminary benthic and planktonic oxygen isotope records from Site 1233 show that changes in intermediate water properties closely track those of the surface water masses even on centennial to millennial timescales, albeit with a reduced amplitude (Ninnemann et al., 2006). These changes generally follow Antarctic timing as shown by alkenone SST records (Kaiser et al., 2005; Lamy et al., 2004). However, some of these changes, in particular during MIS 3, appear to be more abrupt in the ultra-high-resolution isotope records.

Sea-surface salinities in the EPWP off the west coast of Central America are closely coupled with excess precipitation relative to evaporation near the ITCZ (Benway and Mix, 2004). This region is noted for its extreme warmth (often >30°C), exceptionally low salinity (near 32), and a strong, shallow pycnocline (typically centered near 20–40 m depth). A portion of the net freshwater flux to the Panama Basin originates in the Atlantic or Caribbean (Jousaumme et al., 1986), so low salinities here partially reflect the transport of freshwater from the Atlantic to Pacific Basins via the atmosphere. The dynamics of this transport are important because this relatively small transport of freshwater helps to maintain the relatively high salinity of the Atlantic Ocean—a key parameter in maintaining the global thermohaline "conveyor belt" circulation dominated by NADW production (Zaucker et al., 1994; Rahmstorf, 1995).

Indeed, millennial variations in paleosalinity with amplitudes as high as ~4 psu are documented at Site 1242 off Costa Rica, with a dominant period of ~3–5 k.y. during the glacial–deglacial interval and ~1.0–1.5 k.y. during the Holocene (Benway et al., 2006) (Fig. F13). Paleosalinity variations recorded at Site 1242 mimic Northern Hemisphere climate variability such as that in the Greenland ice core and North Atlantic (McManus et al., 2004). Similar Northern Hemisphere climate patterns have also been documented on the Pacific margin of Mexico (Ortiz et al., 2004) and North America (e.g., Behl and Kennett, 1996; Hendy et al., 2002; Mix et al., 1999).

In the EPWP, paleosalinity changes correspond to millennial-scale climate changes in the surface and deep Atlantic and the high northern latitudes, with generally higher (lower) paleosalinity during cold (warm) events (Fig. F13). The covariance of paleosalinity between the EPWP (Benway et al., 2006) and Caribbean (Schmidt et al., 2004) from the same latitudinal band suggests that ITCZ dynamics play a role in tropical hydrologic variability. There appears to be no significant change in paleosalinity of the EPWP in glacial relative to interglacial times, suggesting that changes in the tropical system do not respond passively to events of the high northern latitudes. Following the LGM, higher than average interbasin salinity contrast occurs during warmer intervals such as the Bølling/Allerød and early Holocene climatic optimum, and lower than average interbasin contrast corresponds to colder periods such as the Younger Dryas and Heinrich event H1, although the timing of abrupt changes is not identical to the high-latitude events. Phase calculations suggest the possibility of a lead of interbasin salinity contrast over Atlantic meridional overturning circulation by as much as ~1 k.y. The EPWP paleosalinity and interbasin contrast records are also coherent with Atlantic 13C records (McManus et al., 1999, 2004; Oppo et al., 2003) at periods of 3–5 k.y. The lack of glacial–interglacial contrast in the hydrologic system and the possibility that tropical changes may lead high-latitude climate and circulation changes suggest that North Atlantic circulation is in part influenced by changes in westward vapor transport in the tropics, via control of the North Atlantic salinity budget and thermohaline overturn, as originally suggested by Weyl (1968).

Although much work on high-resolution climate changes continues, already the new materials recovered during Leg 202 have changed the view of the global distribution of millennial-scale climate change. It is now clear that millennial-scale climate and biogeochemical systems of the southeast Pacific and Chile are closely aligned with those recorded in Antarctica and the southern oceans. These climate patterns extend to the equatorial Pacific, with pathways along the eastern boundary current system and properties of SAMW being transmitted either directly through eastern boundary upwelling or indirectly via the EUC or via AAIW. Just north of the equatorial front, the climate pattern of near-surface waters shifts to that of the Northern Hemisphere. Remaining puzzles include sorting out the ultimate causes of these dominant northern and southern climate signals. One possibility is that the primary variability is initiated by North Atlantic freshwater anomalies, via thermohaline circulation influence on heat transport and a north-south "see-saw" effect, which is in turn propagated into the South Pacific via intermediate and mode waters (Schmittner et al., 2003) and biogeochemical responses of upper-ocean nutrients (Schmittner, 2005). Another allows for tropical triggering of climate change, via ENSO dynamics and influences on the hydrologic system (Clement et al., 1999). Another possibility is that the primary variability emanates from the Southern Ocean via transmission through the thermocline circulation to the tropics (Lee and Poulsen, 2006).