Variations in tropical sea surface temperature patterns and the phasing relative to climate change in higher-latitudes provide insight into the mechanisms of climate change on both orbital and shorter time-scales. Here, we present well-dated, high-resolution records of planktonic foraminiferal δ18O and Mg/Ca-based SST spanning the last deglaciation from the Sulu Sea, located in the western equatorial Pacific. The results indicate that the last glacial maximum was 2.3 ± 0.5°C cooler than present in the Sulu Sea with a concomitant decrease in sea surface salinity. The similarity between variations in surface salinity in the Sulu Sea, the western and eastern equatorial Pacific, and the Greenland ice-core record suggests that the observed changes in salinity reflect large-scale rearrangement of atmospheric patterns, which were coherent and synchronous throughout the Northern Hemisphere. The results suggest that the glacial equatorial Pacific climate was strongly influenced by both tropical, and extra-tropical forcing, although it is not clear whether interannual (ENSO) variability is a good analogue of glacial-interglacial climate change.
 Under current conditions, interannual variations in the distribution of equatorial sea surface temperature (SST) related to the El Niño-Southern Oscillation (ENSO) exert a major effect on atmospheric circulation and global climate. Longer-term changes of SST patterns in the equatorial Pacific might also have had large consequences for extra-tropical climate during the last glacial maximum (LGM). Model simulations suggest that decreased seasonality associated with reduced summer insolation would favor a shift in the mean climate state of the equatorial Pacific to more El Niño-like conditions during the LGM relative to the modern climate state [Clement et al., 1999]. Other model studies have suggested a more uniform cooling of tropical SST in response to extra-tropical forcing [Andreasen et al., 2001]. The paleoceanographic evidence is, however, controversial. For example, Lea et al.  showed similar cooling during the LGM in cores from the eastern and western equatorial Pacific. Based on δ18Owater reconstructions that indicate glacial freshening of the Ontong Java Plateau region in the western equatorial Pacific (WEP) relative to the eastern equatorial Pacific (EEP), Lea et al.  argued in favor of more La Niña-like conditions. Subsequent studies, however, countered that the tropical Pacific was not more La Niña-like on broader spatial scales [Andreasen et al., 2001; Oppo et al., 2003], whereas others propose that the glacial tropics were more El Niño-like than today. This argument was made based on glacial reconstructions of weaker zonal and meridional SST gradients in the equatorial Pacific [Koutavas et al., 2002] and relatively salty conditions in the Mindanao Sea in the WEP [Stott et al., 2002]. Here we assess the amplitude and phasing of climate change in a high-resolution sediment record from the Sulu Sea, a semi-enclosed shallow-silled basin located between South China Sea and the western equatorial Pacific (Figure 1), and discuss the implications to glacial tropical climatology.
 Surface and subsurface waters from the South China Sea enter the Sulu Sea from the northwest side through the ∼420 m deep Mindoro Strait. Shallower passages connect the Sulu Sea to the Sulawesi Sea (∼250 m) and through the Philippine archipelago (110 and 65 m) to the WEP. The modern climatology of the Sulu Sea is strongly influenced by the combined effects of the East Asian monsoon and ENSO. Mean annual SST is 28.7°C, varying seasonally by less than 2°C. Sea surface salinity (SSS) in the Sulu Sea reflects its intermediate position between the relatively salty WEP and the low salinity South China Sea. During low sea level stands, however, the relative contribution of surface water from South China Sea was greatly enhanced due to the hydraulic constraints on the influx of Pacific surface water.
 A record of temporal variations in δ18O and Mg/Ca of the mixed-layer planktonic foraminifer Globigerinoides ruber, was generated from Sulu Sea IMAGES core MD972141 (8.8°N, 121.3°E, 3633 m water depth) located near the site of ODP769A [Linsley, 1996]. Here we discuss the high resolution results (2-cm sample interval representing ≤200 years/sample) from the uppermost part of the core spanning 4.5 to 22 ky. The analytical methods and other aspects of core MD972141 have been discussed elsewhere [Dannenmann et al., 2003; Oppo et al., 2003].
2. Results and Discussion
 The planktonic δ18Oruber record from the Sulu Sea shows a 1‰ LGM to Holocene negative shift (Figure 2; LGM is defined as the interval of 18–24 ky B.P.), which is similar to the global ice volume effect [Schrag et al., 1996]. The LGM-Holocene Δδ18Oruber amplitude is consistent with earlier results from ODP site 769A [Linsley, 1996]. Concomitantly, the Mg/Ca ratio in G. ruber shells shows an LGM-Holocene increase of about 0.9 mmol mol−1, suggesting that Sulu Sea SST was significantly colder during the LGM than at present. Carbonate preservation in the Sulu Sea is much better than in the open Pacific Ocean. The high planktonic foraminiferal preservation in the basin results in relatively high Mg/Ca. This yields unreasonably high SST values when using open-ocean calibrations, because the latter were calibrated for more dissolved foraminifera [Lea et al., 2000] (e.g. estimated core top SST = 30°C and LGM SST = 27.6°C). Instead, we apply the preservation-dependent calibration of Rosenthal and Lohmann ; using the average, core-top shell weight (212–300 μm size fraction) of 6.74 μg, the Mg-temperature relationship for the Sulu Sea is, (Mg/Ca)ruber = 0.28Exp(0.095SST). Applying this equation to the average core top value of Mg/Ca = 4.32 mmol mol−1 we calculate SST of 28.9°C, in agreement with modern mean annual value of 28.7°C at the Sulu Sea (note however that the core top age is ∼4.5 ky). Applying the equation to the LGM data we estimate glacial cooling of 2.3 ± 0.5°C in the Sulu Sea.
 The small LGM-Holocene Δδ18Owater amplitude (∼0.5‰) in the Sulu Sea, reconstructed from paired δ18Oruber and Mg/Ca-derived SST, indicates that the glacial cooling in the Sulu Sea was associated with significant surface water freshening. As discussed elsewhere [Oppo et al., 2003] and elaborated on below, we believe that both hydraulic and climatic factors influenced the SSS of the Sulu Sea. To gain further insight into the spatial significance of these changes, we compare δ18Oruber, SST, and δ18Owater records from the Sulu Sea and other equatorial sites including the southern South China Sea (core 18287-3 at 5°39′N 110°39′E; [Kienast et al., 2001]), Mindanao Sea off the southeastern coast of the Philippines (MD98-2181 at 6°3′N 125°83′E; [Stott et al., 2002], Ontong Java Plateau (ODP 806B at 0°19′N 159°22′E), and two cores from the eastern equatorial Pacific, TR163-19 (2°16′N 90°57′W [Lea et al., 2000]) and V21-30 in the cold tongue (1°13′S 89°41′W [Koutavas et al., 2002]). We also compare the Sulu Sea record with a continental record of East Asian monsoon variability based on the U-Th dated stalagmite δ18O record of the Hulu Cave (33.5°N) in China [Wang et al., 2001].
 The δ18Owater values extracted from sediment records at two western equatorial Pacific sites, the Sulu Sea and the Ontong Java Plateau near the core of today's western Pacific warm pool, become progressively heavier between 22 and 16.5 ky, suggesting increasing SSS just as sea level began to rise (Figure 3). The increasing trend in SSS in the Sulu Sea may partly be attributable to the decrease in summer/winter precipitation ratio in the East Asian monsoon as suggested from the Hulu Cave record [Wang et al., 2001]. However, similar trends in the Ontong Java Plateau and both EEP sites indicate that a post-glacial increase in SSS is a broad spatial feature of the equatorial Pacific, thereby suggesting a large-scale rearrangement of atmospheric patterns. The similarity in surface water δ18Owater composition at the Ontong Java Plateau and EEP sites during the LGM suggests a reduction in the zonal salinity gradient of the equatorial Pacific relative to the present, primarily due to glacial freshening of the WEP [Lea et al., 2000]. Unlike today, the records also show that the δ18Owater composition of the Sulu Sea was virtually identical to that of the South China Sea during the LGM and significantly different than observed in the Ontong Java Plateau. These observations suggest that in addition to climatological changes, hydraulic controls due to lower sea level limited the inflow of saltier equatorial Pacific surface water relative to inflow of fresher South China Sea waters, which enhanced the freshening of the glacial Sulu Sea. The Mindanao Sea record [Stott et al., 2002] is the only exception to this pattern: of the five sites with records extending into the LGM, it is the only one that does not show the post-glacial increase in SSS suggesting that it was more saline than the other sites during the LGM and early stage of deglaciation.
 The glacial cooling of 2.3 ± 0.5°C in the Sulu Sea is matched by similar cooling of 2-3°C observed in western (Ontong Java Plateau, Mindanao) and northeastern (TR163-19) equatorial Pacific records. In contrast, V21-30 exhibits a much smaller glacial cooling of ∼1.2°C [Koutavas et al., 2002]. Because V21-30 records conditions in the cold tongue of the eastern equatorial Pacific, the new records suggest both reduced zonal equatorial SST gradient as well as lower meridional SST gradient within the eastern equatorial Pacific during the LGM relative to the present. Koutavas et al.  proposed that the LGM reduction in the equatorial Pacific east-west SST gradient and contraction of the cold tongue are consistent with a persistent El Niño-like pattern in the tropical Pacific, a corollary consistent with model studies [Clement et al., 1999]. Similarly, based on evidence of relatively salty conditions in the Mindanao Sea, Stott et al.  proposed super-El Niño conditions for the LGM, and argued that this is consistent with the reduced zonal salinity gradient inferred from δ18Owater records. It is important to note, however, that the observed decreases in equatorial Pacific SST gradients during the LGM are significantly smaller than the anomalies between the eastern and western Pacific SST observed during modern El Niño events. For example, at the LGM the zonal SST gradient was ∼1°C smaller than the Holocene gradient, compared to the 3–4° anomalies during the historical El Niño events. In fact, such small variations occur seasonally, with the warming of the cold tongue during the winter months of February and March. The analogy of El Niño-like conditions also implies weaker Walker circulation and trade winds, and therefore reduced equatorial upwelling in the EEP during the LGM. Recent evidence suggesting a decrease in productivity and nutrient concentrations in the south equatorial current, is consistent with reduced upwelling along the Peru margin [Loubere, 2002]. In contrast, there is evidence for higher glacial productivity further north and south of the equator, which is interpreted to reflect a general enrichment in surface water nutrient concentrations [Loubere, 2002]. The paleo-records also suggest that the change in tropical Pacific SST patterns during the LGM was superimposed on an overall cooling of tropical SST, likely in response to extra-tropical forcing [Andreasen et al., 2001].
 The Sulu Sea record exhibits millennial-scale oscillations both in δ18O and SST superimposed on the deglacial trends. Most notably, the δ18O oscillations parallel climate oscillations in Greenland. However, the new high-resolution record from the Sulu Sea reveals the strong coherence with the Greenland ice core record in greater detail. For example, the heaviest δ18O values between 16 and 16.5 ky B.P. are most likely associated with the Heinrich event (H1). The rapid decrease in foraminiferal δ18O at about 14.8 ky B.P. coincides with the Bølling Transition (B-A) and the pronounced δ18O excursion between ∼12.8 and 11.0 ky B.P. correlates with the Younger Dryas (Y-D). The Sulu Sea δ18O record also shows a significant positive excursion centered at about 8.5 ky B.P., likely associated with the 8.2 ky cooling event, known from the Greenland ice core. In contrast with δ18O, the Mg/Ca-based SST record shows (within the error of our analyses of ±0.5°C) no discernible variations during these millennial events. The absence of significant SST change in the Sulu Sea during these events indicates that millennial-scale variability in the δ18Oruber record is primarily driven by changes in SSS. We interpret the strong correlation among δ18Owater records from the South China Sea, Sulu Sea, Mindananao Sea and the speleothem record of changes in East Asian monsoon intensity from the Hulu Cave (Figure 2) to indicate that suborbital variability in Sulu Sea hydrography is coupled to changes in the intensity of the East Asian monsoon, whereby relatively fresh surface waters in the Sulu Sea coincide with greater net precipitation associated with higher intensity of the summer monsoon. It appears that suborbital variations in δ18Owater are also coherent and synchronous with variations in the cold tongue (V21-30), suggesting that these changes were part of a wider tropical climatic response. It is also noteworthy that under present conditions a weaker boreal winter monsoon is often associated with El Niño events, possibly reflecting a teleconnection between climate conditions in East Asia and SST anomalies in the equatorial Pacific [Wang et al., 2000]. It is possible therefore, that the positive covariation in δ18Owater, and by inference salinity, between the western and eastern equatorial Pacific indicates that similar extratropical-tropical interactions also occurred on millennial time-scales.
 Studies of SST change in tropical regions since the LGM suggest a spatially inhomogeneous initiation of rapid deglacial warming. Whereas SST appears to rise synchronously with the Northern Hemisphere Bølling warming in the South China Sea [Kienast et al., 2001], the timing of warming in the open tropical Pacific is more ambiguous. Post LGM surface water warming in the Sulu Sea began at 20 ky B.P., about 3.5 ky before the abrupt deglacial shift in δ18Oruber but synchronous with the warming inferred from records from the Mindanao Sea, TR163-19 and core V21-30, in the western and eastern equatorial Pacific, respectively. The offset between Sulu Sea SST and δ18Oruber is consistent with observations from other equatorial Pacific sites and reflects changes in hydrology. Within dating uncertainty, the equatorial warming began synchronously with the beginning of global sea level rise [Hanebuth et al., 2000], and the gradual early warming of both Greenland and Antarctica [Grootes et al., 2001]. The Sulu Sea Mg/Ca-based SST record, however, displays none of the well-known North Atlantic millennial-scale oscillations like the Bølling warming or Younger Dryas cooling. In fact, Sulu Sea SST reaches modern values (∼29°C) during the Younger Dryas, which is accompanied by the positive shift in δ18Oruber, suggesting that deglacial changes in the WEP hydrology continued for ∼1000 years after surface water reached Holocene temperatures. The same pattern of warming is also found in other equatorial Pacific records. The difference in timing between the South China Sea site, on the one hand, and the open Pacific sites on the other hand, indicates that the marginal sea record of South China Sea is more strongly affected by continental climate than the open ocean sites and therefore shows greater similarity to the Greenland ice core record. The absence of Northern Hemisphere millennial variability in SST in open ocean equatorial Pacific sites may reflect the integrated effects of changes in heat transfer and the associated tropical SST patterns driven by tropical dynamics and background cooling driven by global climate change. These variations likely had major implications for global climate change.
 We thank the IMAGES Program for allowing access to core MD97-2141. This project has been funded by NSF Awards #OCE 9987060 to YR, #OCE 9710097 to DWO, and #OCE 9710156 to BKL. We also acknowledge the support of French MENRT, TAAF, CNRS/INSU and IFRTP to the Marion Dufresne and the IMAGES program. This is WHOI contribution number 10850.