4.1. Hydrographic Changes on the California Margin During the Last Glacial Episode
 The discovery of millennial scale variability in Santa Barbara Basin sediments, and its striking similarity to the Greenland ice core temperature record, led previous authors to suggest that the California Margin recorded Northern Hemisphere-wide variability, and that this signal was transmitted rapidly via atmospheric teleconnections [Behl and Kennett, 1996; Hendy and Kennett, 1999, 2000]. However, the assumptions involved in quantifying temperature from δ18O left uncertainty as to the magnitude of the millennial-scale temperature changes on the California margin. The Mg/Ca record from Site 1017 indicates that millennial-scale stadial and interstadial events of the last glacial episode identified from δ18O are synchronous with large, 4.5 ± 1.5°C changes in SSTs that occurred abruptly within 100 to 200 years (Figure 3). Both Mg/Ca and δ18O of planktonic foraminifera reflect the temperature of calcification; however, δ18O-calcite also bears the signal of the δ18O-water. In marginal settings, δ18O-water can be strongly affected by freshwater run-off and evaporation/precipitation, as well as by salinity differences in water masses and changes in continental ice volume. Using the G. bulloides Mg/Ca-based temperatures and the published δ18O of the coeval samples [Hendy et al., 2004], we have calculated the δ18O-water variations over the past 60 kyr at Site 1017. Following previous protocol, we used the Bemis et al.  Orbulina LL paleotemperature equation (T = 16.5 − 4.80(δ18O-calcite − δ18O-water)) and a regional δ18O/salinity relationship of the Southern California Bight of δw = (0.39S) − 13.23 [Spero and Lea, 1996]. Bemis et al.  indicate that the Orbulina LL equation yields the most realistic temperatures for glacial age G. bulloides in nearby Santa Barbara Basin. To remove the effect of continental ice volume changes on δ18O-water, we calculated a global mean δ18O-water record from the centennial-scale sea level record of Siddall et al. , assuming a δ18O-water change of 0.1‰ for every 10 m of sea level change [Lea et al., 2002]. The high-resolution Siddall et al.  record covers the interval from 60–22 kyr; for the interval from 10 to 22 kyr, we used the low-resolution global mean δ18O-water record of Waelbroeck et al. . We interpolated both the Waelbroeck et al.  and the Siddall et al.  records to 200 years to match the sample resolution of the 1017 record. The resulting global mean δ18O-water record was subtracted from the 1017E δ18O-water record to obtain the Δδ18O-water record (Figure 4).
 The millennial-scale stadial and interstadial temperature oscillations of the last glacial episode are clearly apparent in the δ18O-water record, with stadials corresponding to more negative δ18O-water and interstadials corresponding to more positive δ18O-water values (Figure 4). Similarly, the Younger Dryas interval is apparent as a sharp decrease of 0.4 ± 0.2‰ in δ18O-water at 12.6 kyr. The inverse relationship between δ18O-water and temperature reflects the strong influence of both temperature and salinity on the isotopic record, with warm interstadials associated with relatively high salinity surface waters. Two different mechanisms may account for abrupt changes in SST on the California Margin during the last glacial: changes in upwelling intensity and changes in relative strength of the California Current and Southern California Countercurrent [Mortyn et al., 1996; Mortyn and Thunell, 1997; Herbert et al., 2001; Hendy et al., 2004; Hendy, 2010]. At present, Site 1017 lies beneath an active upwelling cell; an increase in past upwelling intensity could decrease SSTs by up to 10°C [Locarnini et al., 2006]. Previous work indicates that coastal California regions experienced greater productivity during the last glacial relative to interglacials, likely due to enhanced upwelling [Mortyn and Thunell, 1997; Hendy et al., 2004]. However, on the millennial scale, there is no clear relationship between stadials and enhanced productivity at Site 1017; in fact, interstadial events were generally more productive than stadials [Hendy et al., 2004]. Furthermore, our δ18O-water record indicates that sea surface salinity decreased during stadial events, whereas increased upwelling would likely increase surface salinity. We favor millennial scale changes in the relative strengths of the California Current and subtropical Countercurrent as the primary explanation for our observed stadial-interstadial changes in SST and salinity, although we recognize that increased glacial upwelling could enhance the magnitude of the observed SST changes.
 At present, the core site is affected by the seasonal interplay between the relatively cold, fresh California Current and the warmer, more saline, Southern California Countercurrent. During the summer the northerly position of the North Pacific High results in a stronger California Current at the expense of the warm subtropical Countercurrent, whereas during the fall and winter, the North Pacific High migrates southward, and the less saline California Current weakens, resulting in a more northerly influence of more saline subtropical Countercurrent water along the California Margin [Huyer, 1983; Harms and Winant, 1998; Bograd et al., 2002]. The 1017 δ18O-water record implies that this scenario is a plausible mechanism for observed millennial-scale variability during the last glacial episode and the Younger Dryas. The peak-to-peak change in δ18O-water between stadial and interstadials is ∼0.5–1.0‰, implying salinity changes of ∼1 to 2.5 psu. The calculated salinity change is in part dependent on the choice of δ18O/salinity relationship used. Using a regional δ18O/salinity relationship for the Southern California Bight of 0.4‰ δ18O per 1 psu [Spero and Lea, 1996] implies a salinity change of 1.25 to 2.5 psu, whereas using a global mean δ18O/salinity relationship of 0.5‰ δ18O per 1 psu [Broecker, 1989] implies a smaller salinity change of 1 to 2 psu.
 Salinity variability in the California Current is primarily controlled by longshore flow, with equatorward advection freshening upper ocean waters, and poleward advection increasing the salinity [Schneider et al., 2005]. Site 1017E lies below the present-day high salinity gradient region of the California Current, where upper ocean salinity has ranged from 33.15 to 33.7 in the last 40 years [Schneider et al., 2005]. We note that our reconstructed salinity variations are larger than this modern range. However, the long-term mean salinity of California coastal surface waters ranges from <33.0 off Mendocino, near the core of the California Current, to >34.6 off Baja California, which is dominated by the subtropical Countercurrent [Lynn, 1967]. Large changes in advection could conceivably increase the range of salinity variability at Site 1017, as the boundary between the California Current and subtropical Countercurrent shifted. Furthermore, large (0.6) variations in salinity have been reported from recent observations in the Gulf of Alaska [Overland et al., 1999; Schneider et al., 2005]. We suggest that changes in the salinity of the source waters of the California Current, along with changes in the relative strength of the California Current and subtropical Countercurrent, may account for the range of salinity reconstructed from our δ18O-water record.
 Previous work supports the idea that cold events were accompanied by a major increase in the advection of subarctic waters along the California Margin via the California Current, which shifted the influence of the subtropical Countercurrent southward of Site 1017. Thunell and Mortyn  and Mortyn et al.  hypothesized that a southward shift of the North Pacific High during the last glacial contributed to a strengthening of the California Current and a decrease in SSTs along the California Margin, while Hendy and Kennett  and Hendy et al.  suggest a similar mechanism accounted for cold SSTs during millennial-scale stadial events in the Santa Barbara Basin during Marine Isotope Stage 3.
 Variations in the positions of North Pacific atmospheric highs and lows control not only surface currents but also regional patterns of evaporation and precipitation. Colder polar temperatures of the last glacial episode would likely have shifted the jet stream and resultant storm tracks southward, increasing precipitation in southwestern North America. Evidence for this effect is observed in New Mexico cave deposits [Asmerom et al., 2010], as well as in a high-resolution pollen record from Santa Barbara Basin [Heusser, 1998], both of which indicate relatively more arid conditions during interstadial events and wetter conditions during stadial events. Increased precipitation during stadial events could also account for some of the observed δ18O-water changes in the ODP 1017E record, by reducing salinity in coastal California waters. At present, coastal run-off due to precipitation is relatively low in Southern California; however, episodic flood layers are evident in Holocene Santa Barbara Basin cores [Schimmelmann et al., 2003]. Both the marine SST and the terrestrial precipitation records support the hypothesis that D/O events in Greenland ice cores were associated with a reorganization of Northern Hemisphere atmospheric circulation, and that the resultant repositioning of the North Pacific high and low pressure systems led to both SST and water mass changes in the Northeast Pacific. Oceanic conditions in the northeast Pacific appear to have responded abruptly despite the large distance from the direct effects of Northern Hemisphere ice sheets and North Atlantic circulation changes.
4.2. Amplitude and Phasing of Deglacial Sea Surface Temperature Changes on the California Margin
 The timing and amplitude of temperature changes on the California margin has strong implications for the mechanisms of climate change. Whereas foraminiferal δ18O and sediment records suggest that the California margin is tightly coupled to Northern Hemisphere climate changes via atmospheric teleconnections [Thunell and Mortyn, 1995; Behl and Kennett, 1996; Hendy and Kennett, 1999], other investigators have suggested that the North Pacific deglacial response may have been more regional, reflecting the sensitivity of the region to the movement of oceanographic fronts [Herbert et al., 2001; Kiefer and Kienast, 2005]. The 1017 Mg/Ca record provides an opportunity to examine phase relationships in a single core within a single faunal group, and to compare the Mg/Ca SST record with other proxy records of temperature from the same core.
 The G. bulloides Mg/Ca record at Site 1017 indicates that sea surface temperatures on the California margin experienced a large, rapid warming during the deglaciation (Figure 3). The initial rise in SST occurred at 18.1 ± 0.5 kyr, ∼1.5 kyr prior to the initial decrease in δ18O that marks the beginning of the deglaciation. This initial warming marks the onset of a 2.6 ± 1.2°C oscillatory event that lasted 1.5 kyr; it was followed by a rapid warming of 5.8 ± 1.6°C that began at 16.5 ± 0.5 kyr and that lasted ∼1.1 kyr (Figure 3). The entire deglacial warming was 7.4 ± 0.8°C, from an initial low of 8.5 ± 0.4°C at 18.1 kyr to 15.8 ± 0.4°C in the Bølling at 14.6 kyr. The magnitude of deglacial warming was approximately equal to the modern seasonal range of SSTs in the region. Site 1017 represents the first Mg/Ca SST record from the northeast Pacific. Few other Mg/Ca records from the greater North Pacific region exist, however, two Mg/Ca SST records from the subarctic northwest Pacific indicate slightly smaller deglacial warming of 4–6°C [Sarnthein et al., 2006; Kiefer et al., 2001]. Several factors may account for the smaller subarctic warming, including choice of species for analysis and oceanographic setting. Analyses of subarctic N. pachyderma may reflect subsurface habitat or stronger seasonality than the midlatitude G. bulloides at Site 1017. Alternatively, enhanced glacial upwelling at Site 1017 could potentially amplify the deglacial SST change.
 At Site 1017, previously published work affords us the opportunity to directly compare SST records based on different proxies, including U37k′ [Seki et al., 2002], foraminiferal faunal transfer functions [Hendy and Kennett, 1999] and Mg/Ca (this study). Because all of the temperature proxies were completed on the same core, the proxies can be directly compared on the same age model, and without effects of spatial or temporal differences in SST (Figure 5). As expected, all of the temperature proxies indicate colder temperatures during the last glacial episode than during the Holocene. D/O- type millennial-scale variability during Marine Isotope Stage 3 (MIS3) is clear in each of the proxy temperature records, with D/O oscillations represented from IS4 to IS17. Although all of the temperature proxies indicate similar timing of MIS3 oscillations, the size of the MIS3 peaks differs. The foraminiferal-based proxies show larger temperature oscillations throughout the record than do alkenone-based SSTs. In addition, U37k′ temperatures are generally warmer at 1017E than those indicated by planktonic foraminiferal proxies. Of the three proxies, foraminiferal faunal temperatures show the greatest warming during the deglaciation: 12 ± 0.7°C from the last glacial maximum to the early Holocene. This is largely due to the cold glacial temperatures as determined by foraminiferal faunal assemblages; faunal temperatures and Mg/Ca temperatures are similar in the early Holocene. U37k′-based temperatures indicate a small, gradual warming of 2.8 ± 0.8°C between 18.7 and 11.6 ± 1 kyr (Figure 5). Mg/Ca indicates a warming of 7.4 ± 0.8°C between 18.1 and 14.6 kyr, about halfway between the two other proxy estimates. Similar discrepancies between U37k′ and foraminiferal based paleotemperature proxies have been reported for other middle and high latitude sites, and previous authors have suggested that alkenone-producing phytoplankton in these environments may be recording warm season SSTs rather than mean annual average temperatures [Bard, 2001]. A compilation of Holocene Mg/Ca and U37k′ SST records from the eastern equatorial Pacific also exhibits significant differences between the two proxies, which are likely due to differences in seasonal fluctuations in coccolithophorids and foraminifera [Leduc et al., 2010]. Coccolithophorids have a relatively narrow thermal tolerance and therefore U37k′ records may reflect bloom events during the season of highest productivity [Leduc et al., 2010]. Foraminifera have seasonal fluctuations but often remain abundant year-round, as is the case with G. bulloides in central California waters today [Kincaid et al., 2000].
 Although the Mg/Ca SST record at Site 1017 exhibits warming ∼1.5 kyr prior to the major deglacial decrease in δ18O, there is no evidence of the 10 kyr or more lead indicated by U37k′ SST records on the California Margin [Herbert et al., 2001; Seki et al., 2002]. Early deglacial warming inferred from U37k′ SSTs on the California margin has previously been used to infer a shut-down of the California Current during the last Glacial Maximum (LGM) in response to a displacement of wind systems due to the presence of the Laurentide Ice Sheet [Herbert et al., 2001]. At Site 1017E, U37k′ results indicate glacial SSTs occur episodically between 50 and 19 kyr, after which the U37k′ temperatures rise gradually to peak warm temperatures around 10 kyr [Seki et al., 2002]. Because both the U37k′ and Mg/Ca records are derived from the same core, there can be no offset due to age model disagreement. However, the reduced warming in the U37k′ temperature record makes it difficult to identify the initiation of the warming within the overall relatively stable temperatures of the LGM. Based on the 1017E U37k′ record, deglacial warming began at ∼19 kyr [Seki et al., 2002], 3.5 kyr prior to the abrupt δ18O decrease that marks the onset of the Bølling. The apparent SST lead implied by the U37k′ record may be due to the low magnitude of the U37k′ signal. Alternatively, the differences between the alkenone and foraminiferal-based proxies may reflect resuspension and post-depositional transport of fine-grained organic particles. Sand-sized foraminifera generally are deposited directly beneath their site of origin, whereas the finer-grained alkenone bearing sediments may be selectively advected long distances [Bard, 2001]. Resuspension and lateral advection of sediments has been indicated as a factor in asynchroneity of alkenone and foraminiferal ages from the continental slope off west Africa, and Site 1017E is in a similar region of upwelling and vigorous current flow [Mollenhauer et al., 2003].
4.3. Deglacial Atmospheric and Oceanic Reorganization in the North Pacific
 The 1017E Mg/Ca and δ18O records indicate that the extratropical Pacific experienced a large SST increase during the deglaciation, and that the temperature change was accompanied by a significant change in sea surface salinity. These observations provide evidence for a reorganization of oceanic/atmospheric circulation in the Northeast Pacific during the deglaciation that led to a shift in precipitation zones and/or the boundary between the California Current and the subtropical Countercurrent.
 At Site 1017E, SST increased ∼8°C between 18 and 15 kyr. Over the same time interval, planktonic foraminiferal δ18O decreased by ∼2.5‰. The estimated global seawater δ18O-water change due to ice volume change on the deglacial is −1.0 ± 0.1‰ [Schrag et al., 1996; Adkins et al., 2002; Schrag et al., 2002]. Removing the ice volume and temperature components of the δ18O record leaves a residual δ18O-water increase of ∼0.9‰ (Figure 4) and implies an increase in surface water salinity of ∼1.8–2.2 (using the regional calibration) between 18 and 15.5 kyr.
 The observed δ18O-water increase at Site 1017E during the deglaciation is similar in direction and amplitude to the millennial-scale δ18O-water oscillations observed during interstadial events of the last glacial episode, and this increase is consistent with an increase in the relative strength of warm, saline poleward flow of the subtropical Countercurrent during deglaciation. Previous investigators have suggested that the California Current weakened during the deglacial in response to a northward shift of the North Pacific High as the high latitudes warmed [Thunell and Mortyn, 1995; Hendy and Kennett, 1999]. Our observations are consistent with this interpretation.
 In addition, the increase in surface water δ18O on the California margin may have been amplified by a decrease in precipitation as atmospheric and oceanic circulation patterns shifted as a result of the deglaciation. In the modern climate regime, warming of the northern subarctic region results in a northward shift of the jet stream and resultant winter storm tracks, and leads to lower annual precipitation in the southwestern United States [McCabe, 1996]. Lake levels and speleothem δ18O in the southwestern U.S. both suggest drier conditions during the Holocene than the glacial [Menking et al., 2004; Rasmussen et al., 2006; Oster et al., 2009; Asmerom et al., 2010]. Similarly, pollen assemblages from nearby Santa Barbara Basin indicate that the deglaciation was characterized by a shift from cooler, wetter conditions of the glacial interval to the warm, more arid conditions of the Holocene [Heusser, 1998]. Asmerom et al.  postulate that the jet stream and Pacific Intertropical Convergence Zone shifted rapidly northward during millennial and orbital scale warm episodes of the past 60 kyr, leading to abrupt precipitation changes in southwestern North America, which are consistent with our observations of SST and salinity changes on the California Margin. Similarly, recent model results indicate that a reduction of the Atlantic-Pacific moisture transport during glacial weakening of Atlantic Meridional Overturning Circulation resulted in a southward shift of the Pacific Intertropical Convergence Zone [Okazaki et al., 2010] and movement of precipitation zones south, with the converse taking place during on the deglaciation. Overall, our data indicate that California margin SST and SSS responded similarly on orbital and millennial time scales, as ocean and atmospheric circulation responded to temperature changes in the high northern latitudes.