Late Quaternary climate fluctuated between extremes of glaciations, lasting ∼90 kyr on average, and interglacial episodes, lasting ∼10 kyr. Still largely unknown are the spatial and temporal evolution of these global climate states, with vigorous debate still underway on the mechanisms responsible for glacial inceptions and terminations. Though it is widely believed that the Southern Hemisphere oceans play a central role in global climate changes, few paleoclimate records exist from there. Here we present three new alkenone-derived SST records from the midlatitude Southern Hemisphere spanning the last 160 kyr, a full glacial-interglacial cycle. Our subtropical SST records from the last glacial period are characterized by (1) warming 47–23 kyr B.P., when high latitudes in both hemispheres cooled, and (2) milder temperatures during the penultimate glacial period than during the last glacial interval. These SST features are found to be of tropical- to subtropical-wide extent, implying increased thermal gradients at times of high-latitude ice sheet growth. This has implications for the vigor of atmospheric and upper ocean circulation and the transport of heat and moisture to the poles that may have been instrumental in the growth and maintenance of polar ice sheets during glacial periods.
 Polar ice core climate records, the δ18O composition of the calcareous tests of benthic and planktonic foraminifera and SST records provide the main basis for our present understanding of climatic changes of the recent geologic past. The development of two new paleotemperature proxies, the Mg/Ca ratio in planktonic foraminifera [e.g., Nürnberg, 1995; Nürnberg et al., 1996; Elderfield and Ganssen, 2000] and the unsaturation ratio of alkenones (U37K′-index) [Brassell et al., 1986; Prahl et al., 1988; Müller et al., 1998] has led to the recent increase in direct SST reconstructions. However, despite a significant number of SST reconstructions from Southern Hemisphere latitudes, the main focus has so far been on the mid- to high-latitude North Atlantic. Partly as a consequence of this, the perception of the origin of climatic changes is biased toward the North Atlantic as the main agent in triggering climatic reorganizations. Yet, the mid-southern latitudes play an important role in the communication and signal transfer between low and high latitudes and between the Southern and Northern Hemispheres. Moreover, they constitute the main moisture source for Antarctic precipitation [Delaygue et al., 2000]. SST changes in the midlatitudes therefore exert a potentially critical influence on the moisture supply to high latitudes and hence the growth and decay of polar ice sheets in the course of glacial-interglacial and shorter-term climatic changes.
 Paleo-SST records from the mid and high latitudes of the Southern Hemisphere are still sparse owing to the difficulty in obtaining continuous marine sediment profiles from this region with its carbonate-corrosive water masses and rough seas. However, available records have confirmed the global extent of glacial-interglacial climate variations and the lead of temperature over ice volume changes observed in the tropics and high southern latitudes [e.g., Howard and Prell, 1992; Schneider et al., 1995; Weaver et al., 1998; Mashiotta et al., 1999; Schneider et al., 1999; Lea et al., 2000; Calvo et al., 2001a; Sachs et al., 2001; Sikes et al., 2002]. Detailed examination of the SST records from low to mid latitudes, however, has revealed differences to the high-latitude temperature evolution in several instances. For example, Schneider et al.  presented several paleo-SST records from the tropical and subtropical South Atlantic and found, deviating from high-latitude temperatures, warmer SSTs during MIS6a than during the Last Glacial Maximum (LGM) and a temperature minimum prior to the LGM. The authors attributed these differences to a low-latitude temperature response to precession-driven insolation forcing and the transfer of this signal to midlatitudes. In contrast, a differential temperature evolution in the subtropical and high-latitude North Atlantic at the end of MIS5e has been suggested to reflect the dominant influence of mean annual insolation that is solely driven by obliquity changes and hence exhibits opposite trends north and south of ∼43°N [Cortijo et al., 1999]. Similarly, Sachs et al.  suggested a prominent obliquity signal in a sea surface temperature record from the southeast Atlantic between ∼40 and 20 kyr B.P. The authors proposed a link between the observed temperature pattern and the deuterium excess record from Antarctic ice cores that shows a dominant obliquity cycle and has previously been interpreted in terms of SST variability at the moisture source (the midsouthern latitude oceans) or shifts in the source area in response to obliquity-driven changes in the thermal gradient between mid and high latitudes [Vimeux et al., 1999, 2001, 2002]. Source temperature reconstructions for Greenland ice (GRIP ice core) and the calculated temperature gradient between Greenland and its moisture source (suggested to be the tropical to subtropical oceans) also reveal a strong parallelism with obliquity during the last glacial period [Masson-Delmotte et al., 2005]. An alkenone-derived SST record from the subtropical southeast Pacific, on the other hand, exhibits a dominant eccentricity cycle over most of the past 400 kyr, with SST changes following high-latitude obliquity variations at times of low eccentricity [Calvo et al., 2001a]. These studies demonstrate the potential of low- to mid-latitude climates to change independently from those at high latitudes. To further examine the climate response of the midlatitudes and their role in the global climate system, we have generated and extended alkenone sea surface temperature records from the midlatitudes of the southeast Atlantic and southwest Pacific that span the past 160 kyr. We find that the subtropical SST evolution is consistent with that seen in other paleo-SST records from the tropical to subtropical latitude belt of both hemispheres but distinct from that at high latitudes.
2. Materials and Methods
2.1. Core Locations and Oceanographic Settings
 IMAGES cores MD97-2121 (40°22.8′S, 177°59.4′E, 3014m water depth) and MD97-2120 (45°32.06′S, 174°55.85′E, 1210m water depth) are located east of New Zealand to the north and south of the west-east trending submarine Chatham Rise, respectively (Figure 1 and Table 1). We extended the SST record of core MD97-2120 [Sachs and Anderson, 2005] to 19.3 m (156 kyr) and measured alkenone unsaturation ratios for SST estimation along the entire 35-m-long core MD97-2121.
Table 1. Location of Sediment Cores Presented in This Study
Water Depth, m
 New Zealand's high mountain ranges and active volcanism lead to high terrigenous sediment influx into the ocean and hence sedimentation rates at the core sites of ≥30 cm/kyr (MD97-2121) [Carter et al., 2002] and 10-20 cm/kyr (MD97-2120) [Pahnke et al., 2003], with highest rates reached during past glacial periods when New Zealand's climate was dry and vegetation cover sparse [Shane and Sandiford, 2003]. Despite the high sediment load reaching the continental shelf and the high seismicity of New Zealand, neither of the core sites was affected by turbidity currents or mass failure in the past as indicated by the lack of turbidity current channels or scars [Lewis et al., 1999], the intact hemipelagic sequences along the cores and no identifiable signs of sediment disturbances [Carter et al., 2002; Pahnke et al., 2003].
 The surface oceanography east of New Zealand is controlled by the Subtropical Front (STF), a hydrographic feature that is part of the circum-Antarctic frontal system and separates subtropical from subantarctic waters (Figure 1a). The STF in the study area is deflected south and wrapped around the southern tip of New Zealand from where it continues northeast along the east coast of South Island as the Southland Current. Upon reaching Chatham Rise, the front bends to the east and follows the east-west orientation of the rise at about 44°S [Sutton, 2001]. The Southland Current splits with one branch following the STF while a small part continues its northeastward flow through Mernoo Saddle, a topographic depression at the western extent of Chatham Rise [Chiswell, 1996; Carter et al., 1998; Nelson et al., 2000].
 The surface circulation north of Chatham Rise is dominated by the East Cape Current (ECC) that flows southwest along the east coast of North Island as the extension of the East Auckland Current/Tasman Front. The ECC transports warm, high-salinity waters to Chatham Rise, where it is deflected eastward to continue along the STF [Heath, 1985]. East of the ECC lies the permanent Wairarapa Eddy, an anticyclonic surface feature that is centered at 41°00′S, 178°40′E [Roemmich and Sutton, 1998]. At present, there is no lateral particle transport in the upper 1000 m on either side of the Chatham Rise [King and Howard, 2001; Sikes et al., 2005], suggesting that the sediment record is not affected by surface circulation and particle advection from other areas to the core sites.
 The STF marks the rapid transition from subantarctic waters in the south to subtropical waters in the north. The site of core MD97-2120 south of the STF is therefore influenced by the cold, low-salinity waters of the Subantarctic Zone (summer SST: 15.1°C, salinity 34.4 [Uddstrom and Oien, 1999; Chiswell, 2002]), while core MD97-2121 lies in warmer, more saline subtropical waters (summer SST: 18°C, salinity 35 [Uddstrom and Oien, 1999]).
 Core TN057-21-PC2 (41°08′S, 7°49′E; 4981 m water depth) was retrieved from a drift site in the southeast Atlantic just north of the Agulhas Ridge that marks the southern boundary of the Cape Basin. In order to obtain a continuous 0- to 160-kyr-long record we extended the previously published 110-kyr-long SST record of core TN057-21-PC2 [Sachs and Anderson, 2003] using samples from ODP Site 1089 (40°56.18′S, 9°53.64′E, 4621 m water depth) that is located on the same drift 176 km northeast of TN057-21-PC2 (combined record referred to in the following as TN057-21/1089). The temperature records show very similar SST values in the overlapping interval (Figure 3 in section 2.2.2).
 Vigorous cyclonic bottom currents in the Cape Basin result in erosion in the northern and northeastern part of the basin and sediment deposition at the core sites in the southern Cape Basin [Tucholke and Embley, 1984]. Consequently, although the core site is located in subantarctic waters, the paleo-records along the core document climatic conditions in the subtropical part of the Cape Basin [Sachs and Anderson, 2003]. This was inferred from core-top alkenone-derived SST estimates of core TN057-21-PC2 that are some 6°C warmer than the overlying surface waters at the core site [Sachs et al., 2001]. Downcore SST variability shows no correlation with sediment focusing factors derived from unsupported 230Th along the core, and closely covaries with SST changes at nearby nondrift sites [Sachs and Anderson, 2003]. This has been suggested to indicate that the SST variations along cores TN057-21-PC2/1089 consistently document subtropical surface conditions and are not an artifact of changes in sediment advection [Sachs and Anderson, 2003]. We will therefore interpret the SST record in terms of temperature changes in the subtropical part of the Cape Basin.
2.2.1. Alkenone Analysis
 Sample preparation and alkenone analysis followed standard protocols described by Sachs and Lehman . Briefly, freeze-dried sediment samples (4–5 g) were weighed into stainless-steel cells with an equivalent amount of sodium sulfate (dispersing and drying agent). Two micrograms of a recovery standard (Ethyl triacontanoate in 20% acetone in hexane) were added and the samples were extracted with dichloromethane at 150°C and 2000 psi on a Dionex ASE-200 pressurized fluid extractor. The total lipid extracts were dried under a N2 stream, transferred into 2-mL autosampler vials and re-dissolved in 400 μL toluene containing 2 μg of n-hexatriacontane (n-C36, a quantitation standard). Samples were derivatized with bis(trimethylsilyl)trifluoroacetamide for 1 hour at 60°C and alkenones (C37 methyl ketones) were quantified by capillary gas chromatography on an Agilent 6890N with a Chrompack CP-Sil-5 column (60 m by 0.32 mm inner diameter, 0.25-μm film thickness), a programmable temperature vaporization inlet in solvent-vent mode, and flame-ionization detector. The oven was programmed from 110°C to 270°C at 40°C/min followed by a temperature increase of 2°C/min to 320°C, and an 18-min isothermal period.
2.2.2. Alkenone Paleothermometry
 Long-chain alkenones, synthesized by haptophyte algae (notably Emiliania huxleyi and Gephyrocapsa oceanica) [Volkman et al., 1980; Marlowe et al., 1984; Volkman et al., 1995], have become a useful tool in paleoceanography for the reconstruction of past sea surface temperatures and productivity. This application is based on the linear relation between the ratio of di- and tri-unsaturated C37-alkenones (U37K′ = C37:2/(C37:2 + C37:3)) and ambient water temperature [Brassell et al., 1986; Prahl and Wakeham, 1987; Prahl et al., 1988]. The most widely used empirical equation for SST estimation from the U37K′-index was proposed by Prahl et al.  on the basis of cultured E. huxleyi (U37K′ = 0.034T + 0.039, r2 = 0.994). A global compilation of alkenone analyses of surface sediments has confirmed this relation [Müller et al., 1998] (U37K′ = 0.033T + 0.044, r2 = 0.958) and suggests that the U37K′ index is best correlated with mean annual SST overlying the core sites. For the Antarctic and sub-Antarctic, however, Sikes and Volkman  demonstrated highest correlation of U37K′ in suspended particles with summer SSTs (U37K′ = 0.0414T − 0.156, r2 = 0.958), consistent with highest alkenone fluxes in this region during spring-summer [Ternois et al., 1998; Sikes et al., 2005]. This calibration has subsequently been confirmed to provide reliable estimates of past summer SSTs at subantarctic to Antarctic latitudes [Sikes et al., 1997; Pichon et al., 1998; Sikes et al., 2002, 2005].
 In most other oceanic areas, sediment U37K′ values have been suggested to integrate the annual temperature signal and best correlate with mean annual surface temperatures despite considerable seasonal variations in alkenone production [e.g., Müller et al., 1998; Prahl et al., 2005]. It may be argued that the increase in benthic grazing in response to increased primary production and organic matter flux to the seafloor [e.g., Herguera and Berger, 1991] annuls the dominance of alkenones produced during the seasonal flux maxima, resulting in an integrated U37K′ signal in the sediment that more uniformly reflects all seasons and hence correlates with mean annual surface temperatures [Sachs et al., 2000].
 Moreover, regional temperature calibrations are based on limited temperature ranges and short timescales and the benefits and detriments of such calibrations compared to global calibrations are still under debate (for a review, see Herbert ). We therefore report the SST estimates for subantarctic core MD97-2120 from the Southwest Pacific on two different SST scales using (1) the global calibration of Prahl et al.  reflecting mean annual SSTs (outer axis in Figures 2– 5) and (2) the regional Southern Ocean calibration of Sikes and Volkman  that yields summer temperature estimates (inner axis in Figures 2–5). We use Prahl et al.'s  calibration rather than the core-top calibration of Müller et al.  for historical reasons and consistency with other SST reconstructions that are mainly based on the Prahl et al.  calibration, though they result in nearly identical SSTs. For SST estimates along cores MD97-2121 and TN057-21/1089 that are both under subtropical influence (see above and Figure 1) we applied the calibration of Prahl et al. .
 Alkenone fluxes in the southwest Pacific vary seasonally and differ in subtropical and subantarctic waters. North of the Chatham Rise (Subtropical Zone), flux rates are highest during midsummer and U37K′-SSTs show a cold bias relative to satellite-derived SSTs. South of the rise (Subantarctic Zone), a broad flux maximum persists from late winter to early summer, peaking in late spring, with U37K′-SSTs closely tracking measured surface temperatures [Sikes et al., 2005]. The cold bias in U37K′-SSTs north of the Rise was attributed to nutrient stress and a possible contribution of alkenones produced at the base of the surface mixed layer [Sikes et al., 2005]. The latter has also been invoked to explain low U37K′-SSTs in the North Pacific [Prahl et al., 1993; Ohkouchi et al., 1999].
2.2.3. Age Models
 The chronology of core TN057-21-PC2, adapted from Ninnemann and Charles , is based on visual alignment of features in the benthic δ18O record and the SPECMAP benthic δ18O stack [Imbrie et al., 1993], and graphical tuning of the planktonic δ18O record to that of well-dated nearby core RC11-83 [Charles et al., 1996] (Figure 3). The age model for ODP Site 1089 is derived from visual alignment of stable oxygen isotope records to those of RC11-83 and through graphical tuning of its paleointensity record to similar records from the North Atlantic [Hodell et al., 2003; Stoner et al., 2003]. The stacked record TN057-21/1089 covers the past 160 kyr at a mean temporal resolution of 190 years. Mean sedimentation rates along cores TN057-21-PC2 and 1089 are 16.4 cm/kyr [Ninnemann et al., 1999] and 14 cm/kyr [Hodell et al., 2003], respectively.
 The age scale for core MD97-2120 is based on 13 14C AMS dates, a tephrochronological marker event, and visual alignment of benthic δ18O and SST records to benthic δ18O of North Atlantic core MD95-2042 [Shackleton et al., 2000] and the Antarctic Vostok deuterium record [Petit et al., 1999], respectively [Pahnke et al., 2003; Pahnke and Zahn, 2005]. The mean temporal resolution along the 156-kyr-long record is 215 years.
 Core MD97-2121 was dated using a set of 9 well-dated tephra layers and 19 foraminiferal 14C AMS dates [Carter et al., 2002; Carter and Manighetti, 2006]. In the interval older than 42 kyr, the benthic δ18O record was graphically aligned to the SPECMAP benthic δ18O stack [Imbrie et al., 1993]. The record covers the past 135 kyr at a mean temporal resolution of 196 years. Close covariation of the benthic δ18O records of all three cores (Figure 3) demonstrates robustness of the individual age scales relative to each other and warrants direct comparison of their SST records.
3.1. Core TN057-21-PC2/ODP 1089
U37K′-derived SSTs along southeast Atlantic cores TN057-21/1089 range between 11.8° and 24.1°C over the past 160 kyr. Maximum SSTs occur during MIS5e, with temperatures on average 3.5°C warmer than during the early Holocene. Minimum SST values are reached at 47 and 43 kyr B.P., after which SSTs increase to 16.3°C at 25.5 kyr B.P. and then cool toward the LGM (∼20 kyr B.P.). SSTs during MIS6a of 16°C are some 2°C warmer than LGM temperatures. The mean temperature changes at the penultimate and last glacial-interglacial terminations in TN057-21/1089 are 6.5° and 5.6°C, respectively.
3.2. Core MD97-2121
 In core MD97-2121 from the subtropical southwest Pacific, U37K′-derived SST estimates yield values between 11.7° and 21.9°C over the past 135 kyr. The record shows a gradual decrease from an early Holocene maximum (18.2°C) toward 16.8°C at the core top, dated to 3 kyr B.P. The core-top estimate deviates from satellite mean annual temperatures (∼16°C) [Uddstrom and Oien, 1999], which may be due to the lack of recent sediments in the core.
 Maximum SSTs along the record occur during MIS5e with values on average 2.4°C higher than during the Holocene. Lowest SSTs are reached during the LGM. Similar to the record of TN057-21/1089, core MD97-2121 exhibits a prominent warming between 47 and 23 kyr with maximum temperatures of 16.1°C at 36.6 kyr B.P. The SST minimum of 14.4°C at 134 kyr near the base of the core is warmer by 1.1°C than the mean LGM SST of 13.3°C.
3.3. Core MD97-2120
 Subantarctic core MD97-2120 shows SST changes (using the calibration of Sikes and Volkman  and of Prahl et al.  (in parentheses)) from 9.5° to 17.8°C (6° to 16°C), with minimum and maximum temperatures at 24.12 kyr and during MIS5e, respectively. Similar to the subtropical SST records from north of Chatham Rise and from the southeast Atlantic, temperatures significantly decrease toward 47 kyr. The following temperature evolution shows no marked changes. Temperatures slightly warm toward 29–26 kyr B.P. starting at 36 kyr, and show an early temperature minimum of 9.5°C (6°C) at 24.1 kyr. Temperatures during MIS6a are some 1°C warmer than average LGM (19–22 kyr) SSTs. The core top temperature estimate using the calibration of Sikes and Volkman  of 13.1°C underestimates present-day satellite-derived summer SSTs of 14.6°C by some 1.5°C [Uddstrom and Oien, 1999]. Using the calibration of Prahl et al.  that yields mean annual temperatures, the core top SST of 10.2°C differs from satellite mean annual temperatures (11.3°C [Uddstrom and Oien, 1999]) by 1.1°C. The latter offset is slightly smaller, but since the last 2 kyr of the core are missing, we can not unambiguously determine which calibration provides better SST estimates.
4.1. Orbital-Scale SST Variability Over the Past 160 kyr: Global Comparison
 The temperature pattern in our subtropical records (MC97-2121, TN057-21/1089) differs from the subantarctic (MD97-2120) SST evolution by exhibiting strong cooling at 47 kyr, warming between 47 and 23 kyr and warmer temperatures during MIS6a than during the LGM. These patterns are also in contrast to changes in global ice volume recorded by benthic foraminiferal δ18O (Figure 3, bottom), that indicate gradual ice volume increase toward 20 kyr B.P., which implies global climate cooling and lowest temperatures during the LGM.
 Some U37K′-SST records from nearby core sites (GeoB 1028-5, GeoB 1710), however, lack the prominent warm anomaly at 47–23 kyr documented in SST profiles described above but exhibit higher temperatures during MIS6a than during the LGM and cooling toward ∼47 kyr B.P. [Schneider et al., 1995, 1999]. U37K′-SST values in some records from the tropical and subtropical Atlantic and Pacific [Villanueva et al., 1998; Herbert et al., 2001; Liu and Herbert, 2004] are higher during MIS6a than during the LGM but decrease toward the LGM without a marked warming during MIS3 (Table 2). A sea surface temperature record from 17°S in the southeast Pacific (TG7) likewise shows temperatures during MIS6a that exceed those of the LGM, but is devoid of substantial changes on timescales <100 kyr [Calvo et al., 2001a].
 We note that a large part of the low- and mid-latitude SST records shown in Figures 4 and 6are from upwelling areas. This could introduce local temperature variability and obscure the regional climatic imprint on SST [e.g., Pichevin et al., 2005]. However, several of the records mentioned above, including those from the Caribbean [Schmidt et al., 2004], the central South Atlantic [Schneider et al., 1999], the Indian Ocean [Rostek et al., 1997] and the new records presented in this study, are from areas not directly influenced by upwelling, making us confident that the consistent SST patterns are not primarily caused by upwelling dynamics.
 Our subantarctic SST record (MD97-2120), on the other hand, shows more similarities with the benthic δ18O record and several U37K′-SST records from high latitudes in the South Indian Ocean (core RC11-120 [Mashiotta et al., 1999]), the South Pacific (core E11-2 [Mashiotta et al., 1999]; cores U939, U938 [Sikes et al., 2002]), northeast Pacific (ODP Site 1020 [Herbert et al., 2001]), the North Atlantic (core MD95-2040 [Pailler and Bard, 2002]) and the climate records from Greenland and Antarctic ice cores (e.g., GISP2, Vostok, EPICA [Grootes and Stuiver, 1997; Petit et al., 1999; EPICA Community Members, 2004]) (see also Table 2). The long-term gradual cooling from ∼60 kyr B.P. to the LGM, recorded in these records, as well as similar temperatures during MIS6a and MIS2 are features of the global ice volume record documented by benthic δ18O (Figure 5). They are, however, distinctly different from the temperature evolution in our subtropical cores and other mid- to low-latitude SST records described above.
 In the North Atlantic, surface ocean records often show a dominant millennial-scale pattern associated with the episodic discharge of icebergs from the Northern Hemisphere ice sheets that complicate the identification of longer-term features [e.g., Chapman et al., 1996; Madureira et al., 1997; Sachs and Lehman, 1999; Bard et al., 2000; Bard, 2002]. Other high-latitude SST records from the Sea of Okhotsk [Seki et al., 2004] and the Sea of Japan [Ishiwatari et al., 2001] show anomalously warm temperatures during the LGM, which has been ascribed to the isolation of these areas from the open ocean during times of lower sea level; that is, these records likely reflect very localized SST changes.
 The “anomalous” glacial temperature pattern in our subtropical cores is observed in many alkenone SST records from low and mid latitudes (Figures 4 and 6 and Table 2), implying a wide geographic footprint. The glacial SST pattern we observe in our subantarctic core south of the Chatham Rise tracks global ice volume changes as recorded by benthic foraminiferal δ18O and appears to be characteristic of SST records from subpolar and polar latitudes of both hemispheres (Figures 5 and 6 and Table 2). Disappearance of the “low-/mid-latitude temperature signal” in core MD97-2120 just 5° in latitude south of core MD97-2121 may be due to the strong hydrographic front of the Subtropical Front that marks the transition from subtropical to subantarctic waters and therefore amplifies the separation of the low-/mid- and high-latitude temperature patterns (Figure 1).
 A number of SST records from low to mid latitudes do not show the temperature pattern described above (see Table 2) and instead more closely follow the “high-latitude” pattern. Different paleo-SST proxies (e.g., planktonic foraminiferal Mg/Ca, faunal assemblage structure, U37K′) often show considerable differences in both absolute values and variability along records [e.g., Sikes and Keigwin, 1994; Chapman et al., 1996; Sikes and Keigwin, 1996; Weaver et al., 1999; Sikes et al., 2002]. This can also be seen in the South Atlantic core studied here (TN057-21/1089), where the U37K′-derived SST record differs from a temperature record based on radiolarian assemblages [Cortese and Abelmann, 2002]. Such differences are usually explained by invoking differences in water column and/or seasonal temperatures reflected by the proxies, or differences in sensitivity to other environmental factors (e.g., carbonate dissolution affects Mg/Ca and faunal SST estimates, while alkenone unsaturation ratios are unaffected [Thunell and Honjo, 1981; Brown and Elderfield, 1996]). Faunal assemblage structures are further prone to variations in thermocline depth and nutrient concentrations that can mask actual temperature variations [Watkins and Mix, 1998]. Nutrient or light limitation has been shown to cause U37K′-SST biases toward lower and higher temperatures, respectively [Prahl et al., 1993; Epstein et al., 1998]. Moreover, alkenones attach to fine particles and are therefore more sensitive to sediment transport than sand-sized foraminifera, which can result in temporal and absolute temperature offsets between U37K′- and foraminifera-derived estimates [Bard, 2001; Ohkouchi et al., 2002; Mollenhauer et al., 2003]. Despite these potential effects on alkenone SST reconstructions, the similarities in low-/mid-latitude SST patterns at core sites influenced by different oceanic and sedimentary conditions suggest that our U37K′-SST records and those shown in Figures 4 and 5 and listed in Table 2, primarily reflect changes in sea surface temperature. Moreover, given the above-mentioned difficulties with different SST proxies, the correlation of some SST records from tropical and subtropical latitudes with high-latitude temperatures does not compromise our interpretation that the “anomalous” SST pattern in our subtropical cores is a prevailing low- to mid-latitude feature that is not found at latitudes poleward of ∼43°N/S.
4.2. Mid-/Low- Versus High-Latitude SST Variability Over the Past 160 kyr
 The new mid-southern latitude SST records from the southwest Pacific and southeast Atlantic presented here indicate significant deviations from changes in global ice volume and the temperature evolution at high latitudes on orbital timescales. This is unexpected with respect to the principal notion that the Late Quaternary temperature and climate pattern follows global ice volume variations as recorded in benthic foraminiferal δ18O. However, the clear recurrence of this deviation in other SST records reveals this “anomaly” as common feature of low to mid latitudes, suggesting a differential temperature response at low/mid and high latitudes.
 In considering possible causes for these different SST trends at high and low/mid latitudes we focus on insolation as possible external climate driving force because it (1) varies meridionally and (2) has a timescale of variations similar to the SST trends we observe.
4.2.1. Temperature Changes During Late MIS3: Potential Driving Forces
 The late-MIS3 warm anomaly in our subtropical South Pacific and South Atlantic SST records that is also present in other SST records from low and mid latitudes, occurred in the middle of the last glacial period when polar ice sheets were still growing [Imbrie et al., 1993; Schrag et al., 2002] and high-latitude temperatures decreasing toward that LGM temperature minimum [e.g., Grootes and Stuiver, 1997; Mashiotta et al., 1999; Petit et al., 1999]. The classic Milankovitch climate driving force, i.e., summer insolation at 65°N, was relatively high throughout MIS3, with only a small drop at 47 kyr B.P., and gradually decreased during the time of the low-/mid-latitude SST warm anomaly toward a minimum at ∼20 kyr B.P. (Figure 4). The temperature pattern observed in tropical to temperate oceanic regions can therefore not be explained by a direct response of the surface ocean to Northern Hemisphere summer insolation. Instead, the observed cooling at 47 kyr B.P. and following SST warm anomaly coincide with obliquity and precession maxima and minima, respectively (Figure 4). Obliquity causes in-phase mean annual insolation changes between ∼43°N and S and opposite phasing poleward of ∼43°N/S. Low obliquity therefore results in increased insolation equatorward of 43°N/S and decreased insolation poleward of 43°N/S, and vice versa [Loutre et al., 2004]. This 180°-phase shift in obliquity-driven mean annual insolation changes near 43°N/S and a response of ocean temperatures to mean annual insolation forcing would be consistent with the opposing SST trends observed here between records from low/mid (<∼43°N/S) and high latitudes (>∼43°N/S).
 Several other influences of obliquity on climate are conceivable. First, changes in winter insolation at high latitudes are dominated by obliquity as the tilt angle of the Earth's axis determines whether high latitudes receive insolation during winter or not. However, it is evident that insolation close to zero (winter insolation at high latitudes) will have a very limited effect on temperature changes and is thus unlikely to have significantly contributed to past climatic changes. Second, the insolation gradient between different latitudes closely varies with obliquity and has been suggested to exert a major influence on moisture transport to high latitudes and hence snow accumulation at the poles [Raymo and Nisancioglu, 2003]. The resulting ice albedo feedback would influence global climates, providing a possible explanation for the dominant 41-kyr period in climate records prior to ∼900 kyr B.P. [Raymo and Nisancioglu, 2003]. However, an almost exclusively obliquity-driven variation in insolation gradients is only achieved between narrow latitude bands, for example, between 65°N and 30°N June insolation. Moreover, while insolation gradients may account for changes in the source area of moisture that is transported to the poles and the vigor of this transport (see below and Vimeux et al. ), they cannot explain the differential temperature patterns observed at low/mid and high latitudes. Finally, as discussed above, mean annual insolation at all latitudes exclusively varies with obliquity and is symmetric about the equator with a phase shift at ∼43°N/S. These insolation changes are small at low to mid latitudes with amplitudes of 0.6 W/m2 (40°N/S) to 3.6 W/m2 (0°). However, as the upper ocean integrates the insolation signal over several years, mean annual insolation changes may have a significant impact on variations in sea surface temperature. Moreover, low- to mid-latitude warming may have been reinforced by local greenhouse effects caused by increased water vapor and cloud formation over warm ocean areas [e.g., Hallberg and Inamdar, 1992; Gupta et al., 1996]. Pierrehumbert  suggested that this mechanism may be active despite the negative feedback of radiative cooling caused by increased cloud cover, because it is compensated for by additional heating through adiabatic compression due to subsidence in the subtropics and latent heat release in convective areas. Such feedback mechanisms suggest that mean annual insolation changes may constitute an important contributing driving force to the observed temperature evolution at the end of MIS3.
 In contrast to mean annual insolation changes that are exclusively driven by obliquity variations, daily insolation in the tropics is dominated by orbital precession and is therefore of opposite phasing about the equator [Berger et al., 1993]. As mentioned above and shown in Figure 4, the late-MIS3 warm anomaly in tropical and subtropical SST records coincides with a minimum in precession and hence with a maximum in Northern Hemisphere summer insolation, but a minimum in Southern Hemisphere summer insolation. The consistency of this pattern in both hemispheres suggests that it is unlikely to be a response to local daily insolation forcing and rather caused by the interaction of relatively weak mean annual insolation forcing with global, i.e., 65°N summer insolation forcing. The eccentricity minimum at this time may have been instrumental in increasing the contribution of obliquity-driven mean annual insolation to climate forcing by dampening the precessional amplitude (see Figure 4). Small changes in precession implicate weak seasonal modulation of insolation and hence a reduced effect of those changes on climate. These conditions may have enhanced the response of sea surface temperatures to variations in mean annual insolation at low/mid latitudes.
 The most striking examples for obliquity signals in climate records from the Late Quaternary are the deuterium excess records from Antarctica [Vimeux et al., 1999, 2001, 2002] and Greenland [Masson-Delmotte et al., 2005]. Deuterium excess (d) reflects moisture source temperatures, suggesting the d variability in polar ice mirrors SST changes in the main moisture source areas, the low to mid latitudes, or it is an expression of spatial changes in moisture source and subsequent transport to high latitudes in response to changing meridional temperature gradients. Another example is surface temperature and salinity changes in the North Atlantic that show opposing trends north and south of ∼43°N during MIS5, in line with a dominant response to obliquity-driven mean annual insolation changes [Cortijo et al., 1999]. A recent analysis suggests that the timing of Pleistocene deglaciations is consistent with forcing by every second or third obliquity maximum [Huybers and Wunsch, 2005]. The authors propose that obliquity maxima, by increasing high-latitude insolation, may increase basal ice sheet temperatures and lead to melting of thick ice sheets. Alternatively, as suggested by Johnson  for the penultimate deglaciation, the lowered meridional insolation gradient during obliquity maxima may have led to a reduction in poleward moisture transport and hence starved high-latitude ice sheets of precipitation. These studies demonstrate the potentially crucial impact of obliquity variations on climate.
 Dominant precession signals have been reported from several low-latitude climate records and suggested to reflect a response to the precession-driven part of Northern Hemisphere summer insolation [e.g., Perks et al., 2002] or local daily insolation [Baker et al., 2001; Bush et al., 2002; Cruz et al., 2005]. Orbital configurations similar to that during MIS3 with minima in eccentricity, obliquity and precession also occurred during MIS6 (∼150 kyr B.P.) and 8 (∼260 kyr B.P.). But eccentricity values were lower around 45 kyr B.P. than during the two earlier glacial periods. The paleoclimatic evidence for these periods is considerably more ambiguous and sparser, as fewer records cover these periods. Records showing a clear and pronounced warming at the end of MIS6 include our record from south of Chatham Rise (Figure 4) and two records from the North Atlantic at 40°N (MD95-2042, MD95-2040 [Pailler and Bard, 2002]) and 43.5°N [Villanueva et al., 1998] and the southeast Atlantic [23°S, Kirst et al., 1999]. Several records from the low/mid latitude Atlantic (GeoB1016-3, 1105, 1710 [Schneider et al., 1999]; GeoB112 [Nürnberg et al., 2000] (and the tropical Pacific (TR163-10, ODP Site 806 [Lea et al., 2000]) show a small positive temperature departure, while other SST records, for example, from the Atlantic (GeoB1008 [Schneider et al., 1995]) and the North Indian Ocean (MD90-963 [Rostek et al., 1997]), exhibit no discernable warm anomaly at this time. Only a few records exist that extend to or beyond MIS8. A clear sea surface warming is documented in the Mg/Ca-derived SST record from subantarctic Pacific core MD97-2120 [Pahnke et al., 2003], and in the alkenone-SST record from tropical Atlantic core GeoB1105 [Schneider et al., 1999]. Alkenone-SST records from the northeast Pacific (La Paz 21P and ODP Site 1012 [Herbert et al., 2001]) and Mg/Ca-derived SST records from the tropical Pacific (TR163-19 and ODP Site 806 [Lea et al., 2000]), on the other hand, do not exhibit a temperature anomaly.
 Similar to the warm excursion during MIS3, the MIS6 SST anomaly would be in line with mean annual insolation forcing presumably in combination with a surface ocean response to the precession-driven increase in Northern Hemisphere summer insolation (Figure 4). The presence of this temperature excursion in core MD97-2120 south of Chatham Rise may indicate the transitional position of this site between mid- and high-latitude climate impacts.
4.2.2. Long-Term SST Evolution Over the Past 160 kyr
 Higher temperatures during MIS6a than during the LGM and the long-term temperature evolution seen in many paleo-SST records from low/mid latitudes show similar timing and amplitudes as Earth's orbital eccentricity trend. The 100-kyr period of eccentricity is modulated by the 400-kyr cycle of eccentricity, resulting in lowest values since the past ∼350 kyr at the time of the 47-kyr SST minimum and subsequent warm anomaly. While the effect of eccentricity changes on insolation is negligible, eccentricity modulates the amplitude of precession and hence the seasonal insolation contrast. With precession being the dominant orbital parameter influencing daily insolation changes at low latitudes, eccentricity may have contributed to low-/mid-latitude climate changes through its impact on the amplitude of precession. In contrast, high-latitude insolation changes are less affected by precession, which may explain why high-latitude temperatures during MIS6 were similar to those during the previous and following glacials and hence relatively colder than at low and mid latitudes. Eccentricity imprints on SST changes have previously been described from tropical and subtropical Atlantic sites [Schneider et al., 1999] and east Pacific paleo-records [Calvo et al., 2001a].
4.3. Implications for Meridional Temperature Gradients and Global Climate Variations
 The distribution of paleo-SST records with contrasting temperature patterns indicates differential heating between low/mid and high latitudes (Figures 4, 5, and 6) and suggests an increase in the meridional thermal gradient at 47–23 kyr B.P. For such conditions, the thermal wind balance predicts stronger winds [Peixoto and Oort, 1992] and thus enhanced atmospheric transport of heat and moisture from low to high latitudes. Momentum conservation requires an increase in westerly wind strength and zonal surface ocean circulation in response to increased poleward atmospheric and ocean transport. A resulting sharpening of thermal oceanic fronts at the subtropical-subpolar boundary has been shown to have a positive feedback on atmospheric temperature gradients [Cessi, 2000] and may have reinforced the disparate SST pattern at the end of MIS3.
 Invigorated meridional circulation in both atmosphere and ocean may have resulted in increased moisture transport from the warm low/mid latitudes to the cold polar regions. The modern moisture sources for Antarctic and Greenland precipitation are the mid-southern latitude oceans [Delaygue et al., 2000; Reijmer et al., 2002] and the mid- to high-latitude North Atlantic and Pacific oceans [Charles et al., 1994], respectively. A model simulation [Delaygue et al., 2000] and the deuterium excess records from Antarctica [Vimeux et al., 1999, 2001] suggest that the low-latitude moisture contribution to Antarctica slightly increased during the LGM at the expense of the high-latitude source as a result of steeper meridional temperature gradients. This highlights the importance of the low/mid latitudes as moisture supplier for continental ice sheets today as well as during glacial times. Enhanced ice accumulation due to increased moisture supply may have increased high-latitude cooling through ice albedo and altitude feedbacks. In turn, lower high-latitude temperatures would have led to tundra expansion at the expense of forests, further increasing the albedo effect and amplifying cooling [e.g., Gallimore and Kutzbach, 1996]. A low- to mid-latitude moisture source and increased poleward moisture transport may therefore be the key to sustained ice accumulation at times of high-latitude cooling.
 Using an ocean-atmosphere coupled model, Khodri et al.  simulated climate conditions at the last glacial inception (115 kyr B.P.) and suggested that enhanced equator-to-pole thermal gradients induced by seasonal perturbations in insolation forcing led to an increase in northward heat and moisture transport. Similar dynamics may have been active just prior to the LGM owing to differential heating of high and low/mid latitudes. Simple box models of the climate system also suggest a critical role for changes in sea-ice extent, through its albedo and insulating effects, in regulating high-latitude precipitation and ice accumulation [Gildor and Tziperman, 2000, 2001]. With Milankovitch forcing incorporated into the model, midlatitude atmospheric and ocean temperatures directly respond to insolation changes, while high-latitude climate variables rather respond to changes in ice mass balance [Gildor and Tziperman, 2000]. Our SST observations would be in line with such scenario in that low- to mid-latitude temperatures follow obliquity-driven mean annual insolation changes and precession-modulated Northern Hemisphere summer insolation, while high-latitude temperatures decreased in concert with increasing global ice volume.
 Three alkenone records from the midlatitude Southern Hemisphere document SST variability over the past 160 kyr and thus the last interglacial period to the present. Two of these records that document subtropical temperature variability in the southwest Pacific and southeast Atlantic reveal coherent temperature patterns that deviate significantly from changes in global ice volume and the temperature evolution at high northern and southern latitudes. Most notably, subtropical temperatures were higher during MIS6a compared to the LGM, anomalously low at 47 kyr and warmed 47–23 kyr B.P., when high latitudes cooled and global ice volume increased. Comparison with other available paleo-SST records indicates that these patterns are a consistent feature of low to mid latitudes in both hemispheres. This “anomalous” temperature evolution is consistent with a response of sea surface temperature to mean annual insolation changes that are exclusively driven by obliquity and have the same phasing in this latitude band (<43°N/S). Temperature changes at high latitudes, on the other hand, more closely follow the global ice volume evolution as recorded in benthic foraminiferal δ18O. These differential temperature changes between low/mid and high latitudes imply an increase in the meridional thermal gradient during the last glacial that may have resulted in enhanced moisture transport to high latitudes. Increased moisture supply to the poles may have been instrumental in sustaining and promoting continued ice sheet growth during the last glacial period culminating in maximum ice extent and high latitude cooling ∼20 kyr B.P. This highlights the role of the low to mid latitudes in the global climate system and emphasizes their potential as important players in the ice-age cycles of the Late Quaternary.
 We are grateful to Barbara Manighetti and Lionel Carter (Wellington, New Zealand) for providing sediment samples, the age model, and the benthic oxygen isotope data from core MD97-2121. Ying Chang and Maria Schriver assisted with alkenone analyses. We thank Eva Calvo and an anonymous reviewer for helpful comments that improved the manuscript. Funding for this research was provided in part by the Gary Comer Science and Education Foundation (J. P. S.), the Jeptha H. and Emily V. Wade Award for Research (J. P. S.), and a Henry L. and Grace Doherty Professorship (J. P. S.).