(2006) 202–212www.elsevier.com/locate/yqres

Quaternary Research 66

Devils Hole, Nevada, δ18O record extended to the mid-Holocene

Isaac J. Winograd a,⁎, Jurate M. Landwehr a, Tyler B. Coplen a, Warren D. Sharp b, Alan C. Riggs c,Kenneth R. Ludwig b, Peter T. Kolesar d

a U.S. Geological Survey, 432 National Center, Reston, VA 20192, USAb Berkeley Geochronology Center, 2455 Ridge Road, Berkeley, CA 94709, USA

c U.S. Geological Survey, Denver Federal Center, MS 413, Lakewood, CO 80225, USAd Department of Geology, Utah State University, Logan, UT 84322, USA

Received 20 May 2005Available online 2 August 2006

Abstract

The mid-to-late Pleistocene Devils Hole δ18O record has been extended from 60,000 to 4500 yr ago. The new δ18O time series, in conjunctionwith the one previously published, is shown to be a proxy of Pacific Ocean sea surface temperature (SST) off the coast of California. Duringmarine oxygen isotope stages (MIS) 2 and 6, the Devil Hole and SST time series exhibit a steady warming that began 5000 to >10,000 yr prior tothe last and penultimate deglaciations. Several possible proximate causes for this early warming are evaluated. The magnitude of the peak δ18O orSST during the last interglacial (LIG) is significantly greater (1 per mill and 2 to 3°C, respectively) than the peak value of these parameters for theHolocene; in contrast, benthic δ18O records of ice volume show only a few tenths per mill difference in the peak value for these interglacials.Statistical analysis provides an estimate of the large shared information (variation) between the Devils Hole and Eastern Pacific SST time seriesfrom ∼41 to ∼2°N and enforces the concept of a common forcing among all of these records. The extended Devils Hole record adds to evidenceof the importance of uplands bordering the eastern Pacific as a source of archives for reconstructing Pacific climate variability.© 2006 University of Washington. All rights reserved.

Keywords: Late Pleistocene; Holocene; Southwestern U.S.; Nevada; Devils Hole; Paleoclimate; SST; Eastern Pacific; Tropical Pacific; Atmospheric teleconnection

Introduction

In earlier studies (Winograd et al., 1992; Ludwig et al., 1992),we presented a continuous 500,000-yr time series of Pleistoceneclimate recorded in vein calcite from Devils Hole cave insouthern Nevada (Fig. 1). Those reports initiated a decade-longdiscussion regarding the validity of the Milankovitch hypothesisof the Pleistocene ice ages (Broecker, 1992; Imbrie et al., 1993;Kerr, 1997; Karner and Muller, 2000; Henderson and Slowey,2000). They also spurred renewed interest in cave carbonatedeposits as archives of Quaternary paleoclimate (McDermott,2004).

Herein, we present new data that bring the Devils Holeoxygen-18 (δ18O) time series forward from 60,000 to 4500 yrago, extending our original record (which covered the period568,000 to 60,000 yr ago) into the mid-Holocene. Having this

⁎ Corresponding author. Fax: +1 703 648 5832.E-mail address: ijwinogr@usgs.gov (I.J. Winograd).

0033-5894/$ - see front matter © 2006 University of Washington. All rights reservdoi:10.1016/j.yqres.2006.06.003

new mid-Wisconsin to mid-Holocene record enables us tocompare the Devils Hole record with marine and continentalrecords during times when dating uncertainties of the variousclimate proxies are relatively minor, thereby enhancingrecognition of similarities and differences in their phases. Forexample, we can determine if the much discussed phase offsetbetween the Devils Hole δ18O time series and the penultimatedeglaciation – as recorded by foraminiferal δ18O – was aunique event, or if it also occurred during the last deglaciation.Additionally, we address to what extent the Devils Hole δ18Otime series is a proxy of temperature versus a mixture of severalvariables (i.e., ocean water δ18O, moisture source, seasonalityof recharge, etc.). We also examine the question whether theDevils Hole δ18O time series is predominantly a localpaleoclimate record (Imbrie, 1992; Imbrie et al., 1993), aregional record (Herbert et al., 2001), or a proxy of interhemi-spheric relevance (Landwehr and Winograd, 2001). Finally, the230Th uranium-series dating of the Devils Hole calcite (Ludwiget al., 1992) was replicated in 1997 using both 230Th and 231Pa

ed.

Figure 1. Western United States and adjacent northeastern Pacific Ocean showing the location of Devils Hole and marine cores discussed in text.

203I.J. Winograd et al. / Quaternary Research 66 (2006) 202–212

(Edwards et al., 1997) and is now unquestioned. However, thecorrection that must be added to the ages to account forgroundwater transit/travel time through the aquifer tributary toDevils Hole was not resolved in our earlier work. We now cansay that the correction is <2000 yr.

Extending the Devils Hole record to the mid-Holocene

Devils Hole core DH-11, which yielded our 500,000-yrrecord, was recovered from a depth of ∼30 m below the watertable (∼45 m beneath the ground surface). Cessation of veincalcite growth ∼60,000 yr ago was perplexing, given theabsence of any hiatuses in the growth of this fracture coatingduring the preceding ∼500,000 yr (Winograd et al., 1992) andthe likelihood that the groundwater from which the calcite wasprecipitated has always been slightly supersaturated withrespect to calcite (Riggs et al., 1994). We speculated that at∼60,000 yr ago the fissure that comprises Devils Hole openedto the surface. Thereafter, calcite precipitation ceased due toland surface inputs of organic and mineral matter andsubaquatic growth that poisoned calcite precipitation from theslightly supersaturated groundwater (saturation index, SI=0.2;Plummer et al., 2000).

This line of reasoning suggested that portions of the cavernup the hydraulic gradient (i.e., upstream in the aquifer) from thenear-surface DH-11 core site might yield calcite that grewuninhibited by contaminants. Several divers, including two ofthe authors (ACR and PTK), searched Devils Hole Cave no. 2, asimilar cave ∼200 m north of Devils Hole, for recentlyprecipitated calcite coatings. The first dives in 1994 retrievedseveral samples one of which, DHC2–3, was precipitated from80,000 to 19,000 yr ago, thus providing 41,000 yr of new recordplus a 20,000-yr overlap with core DH-11 (Winograd et al.,1996). Additional dives in 1996 yielded sample DHC2–8,which grew between 30,000 and ∼4500 yr ago, and whichoverlaps DHC2–3 by 11,000 yr.

Thin-section petrography – to check for hiatuses in calcitegrowth, and for calcite recrystallization or dissolution –preceded sampling of the calcite for stable isotopic analysesand uranium-series dating. No such events were noted. Startingat the free (i.e., outer or wetted) face of each sample, andproceeding inward at right angles to the growth laminae, wemicro-milled 209 samples, each 0.25 mm thick, for δ18Oanalyses. The stable isotope data were plotted against theirdistance inward from the free face of the samples, and theresulting curves were used to select intervals for uranium-series

Table 1Mass spectrometric 230Th/U ages of DHC2–8 samples versus distance of samplecenter from free face of vein

Distance(mm)

Observed230Th/232Thact.

Age (103 yr)detritus-corrected

Age (103 yr)corrected fordetritus andexcess 230Th

Initial234U/238Uact.

0.38 12.0 5.21±0.16 4.83±0.24 2.771±0.0081.27 18.0 8.49±0.12 7.23±1.01 2.807±0.0092.29 76.3 17.55±0.51 17.28±0.56 2.786±0.0063.30 234.1 19.84±0.23 19.56±0.32 2.786±0.0084.32 286.3 21.94±0.13 21.63±0.28 2.798±0.0055.33 219.3 24.06±0.24 23.75±0.35 2.811±0.0056.35 192.6 25.98±0.31 25.63±0.42 2.805±0.0067.37 174.0 27.87±0.24 27.53±0.36 2.812±0.0058.38 162.7 29.99±0.26 29.67±0.37 2.811±0.0059.40 252.5 30.70±0.28 30.38±0.38 2.809±0.005

First sample was 0.762 mm thick; all others 1.016 mm thick. Calculated usinghalf lives for 230Th of 75,690±230 and 234U of 245,250±490 yr (2σ; Cheng etal., 2000). U content of samples ranged from 0.35 to 0.48 ppm, 232Th was 2.2–15 ppb (median=3.8). Analytical errors propagated at 95% C.I. Detrituscorrection assumes typical silicate with atomic Th/U=3.8±1.9 and 234U/238U,230Th/238U near secular equilibrium (±10%). Detritus corrections range from390 to 55 yr (median=105). U–Th ratios used for adsorption-corrected ages arecorrected for a flux of 2.7×10−20 moles of unsupported 230Th per cm2 per year,calculated from analyses of two samples scraped from the free face andextrapolated age of surface. Initial 234U/238U calculated from adsorption-corrected age.

Figure 2. The Devils Hole δ18O time series, 160,000 to 4500 yr ago. Blackdots, δ18O analyses; stars mark the tie points between calcite samples DH-11(160,000 to 78,700 yr ago), DHC2–3 (78,500 to 24,700 yr ago), and DHC2–8 (24,500 to 4500 yr ago). The location and 2-sigma errors for 13 TIMSuranium-series ages of samples DHC2–3 and DHC2–8 are shown above thetime series. (Three additional ages were redundant given the overlap betweenDHC2–3 and DHC2–8, but all uranium-series data are given in Tables 1 and2). Also shown, between ∼60,000 and ∼156,000 yr ago, are 8 previouslypublished TIMS ages for DH-11 (Ludwig et al., 1992; Winograd et al., 1997).Calcite growth rates given in the lower panel. The precision of the δ18Omeasurements is ±0.07 per mill.

204 I.J. Winograd et al. / Quaternary Research 66 (2006) 202–212

dating via the same thermal ionization mass-spectrometric(TIMS) methodology used to date DH-11 (Ludwig et al., 1992).The 230Th ages are summarized in Tables 1 and 2. The 230Thages were used, in turn, to interpolate the age of each of thesamples analyzed for δ18O. The overlapping DH-11, DHC2–3and DHC2–8 time series were spliced together at 78,500 yr and24,500 yr. These points were chosen to minimize any smalldifferences in the overlapping time series. The resulting δ18Otime series for the period 160,000 to 4,500 yr ago is shown inFigure 2. The precision of the δ18O measurements is ±0.07 permill.

Table 2Mass spectrometric 230Th/U ages of DHC2–3 samples versus distance of samplecenter from free face of vein

Distance(mm)

Observed230Th/232Thact

Age (103 yr)detritus-corrected

Age (103 yr) correctedfor detritus and excess230Th

Initial234U/238Uact

1 110 20.83±0.46 20.62±0.49 2.802±0.0087 204 31.24±0.17 30.92±0.28 2.812±0.00813 204 40.07±0.20 39.64±0.35 2.827±0.00823 289 56.76±0.30 56.11±0.51 2.815±0.00731 681 67.24±0.41 66.44±0.64 2.805±0.00741 227 81.45±0.65 80.43±0.88 2.786±0.008

Samples were 2 mm thick. Calculated using half lives for 230Th of 75,690±230and 234U of 245,250±490 years (2σ; Cheng et al., 2000). Median Ucontent=0.66 ppm. Analytical errors propagated at 95% C.I. Detritus correctionassumes typical silicate with atomic Th/U=3.8±1.9 and 234U/238U, 230Th/238Unear secular equilibrium (±30%). U–Th ratios used for adsorption-correctedages are corrected for a flux of 4.1×10−19 moles of unsupported 230Th per cm2

per year, calculated from analyses of two samples scraped from the free face andextrapolated age of surface. Initial 234U/238U calculated from adsorption-corrected age.

Five features of the new record (Fig. 2) merit discussion.First, the time series displays several abrupt isotopic shifts, themost prominent being the 0.6 per mill increase between 19,000and 17,000 yr ago. Note also the sharp shifts from ∼74,000 to∼72,000 yr ago and from∼65,000 to∼63,000 yr ago; these areonly 0.2 per mill but are anchored by two or moremeasurements at their upper and lower bounds and thus donot reflect measurement error. These abrupt shifts, occurringwell within 2000 yr, constitute one line of evidence that thecorrection of the Devils Hole chronology for groundwatertravel time is <2000 years. That is, the age of any feature ofinterest in the time series could be up to 2000 years older thanshown in Figure 2. (An extended discussion of the travel-timecorrection is given in Appendix A of the SupplementalMaterials.)

Second, the interval ∼75,000 to ∼34,000 yr ago, appro-ximately concurrent with marine oxygen isotope stages (MIS)3 and 4, does not display the Dansgaard–Oeschger (DO) cycles(i.e., millennial-scale high amplitude climatic shifts) that areprominent in the GRIP and GISP2 Greenland ice cores and thatapparently are also recorded by sediments from Ocean DrillingProgram (ODP) Sites 893A and 1017, respectively, about 400and 500 km southwest of Devils Hole (Hendy and Kennett,2000; Seki et al., 2002; Fig. 1). Millennial-scale oscillations inwestern Great Basin lake levels have also been inferred from

205I.J. Winograd et al. / Quaternary Research 66 (2006) 202–212

sediments of MIS 3 age (Benson, 2004). We do not believe thatthe absence of millennial-scale cycles reflects lack of resolutionof such cycles in the Devils Hole record, due for example, todispersion (i.e., mixing) within the aquifer (see Appendix A,Supplemental Materials). Some of the millennial-scale cyclesrepresented at Site 893A–for example, those believed to becorrelative with interstadials nos. 11,12, 14 and 16/17 inGISP2–are of several thousand year duration (see Fig. 2 inHendy and Kennett, 2000). Given the step-like isotopic shiftsin Devils Hole (within 2000 yr) and the ∼400-yr sampling stepin the interval encompassing MIS 4 and 3, these cycles shouldbe visible in our record, but they are not.

Third, about 32,000 yr ago, the δ18O in the Devils Holerecord begins a 1.1 per mill shift toward more positive values(i.e., warmer, see below) eventually peaking at ∼17,000 yr ago,after which it slowly declines about 0.2 per mill until the recordterminates about 4500 yr ago. This warming begins about12,000 yr before the initiation of global deglaciation, asrecorded by minimum sea level, at about 20,000 yr ago(Lambeck et al., 2002). Similar early warming occurred prior tothe penultimate deglaciation (Winograd et al., 1992; Herbert etal., 2001), but the magnitude of the phase offset betweenwarming seen in Devils Hole δ18O and the initiation of globalice-sheet retreat for that deglaciation has been problematic dueto the difficulty of obtaining measured rather than inferred datesfor the marine records between 145,000 and 120,000 yr ago, asdiscussed below.

Fourth, the prominent Younger Dryas (YD) cooling eventseen in numerous paleoclimate records between 13,000 and11,000 yr ago, including at ODP Sites 1017 and 1019 (Seki etal., 2002; Barron et al., 2003) off the California and Oregoncoasts (Fig. 1) and in a speleothem in the Klamath Mountains ofOregon (Vacco et al., 2005), is not seen in Devils Hole. Itsabsence may reflect the low temporal resolution of the DevilsHole record between 17,000 and 7000 yr ago, an interval whereeach sample interval of 0.25 mm represents ∼2500 yr, owing tovery slow calcite precipitation (Fig. 2). Alternatively, we notethat the YD is not recorded in all high resolution (≥15 cm/yr)marine records off the southern Oregon and California coasts(Sancetta et al., 1992; Sabin and Pisias, 1996, Fig. 14; Doose etal., 1997, Fig. 3); it is also absent from records of SierraNevadan and Mount Rainier glaciation in the western U.S.(Clark, 2003; Licciardi et al., 2004).

Fifth, the newly extended and radiometrically dated DevilsHole δ18O record was examined for significant signalfrequencies using the Blackman–Tukey spectral analysisprocedure in AnalySeries 1.2 (Paillard et al., 1996). The fullrecord, ranging from 4500 to 567,700 yr ago, sampled at 1500-yr intervals, was found to contain signal with significant powerat periods of 93,700, 42,000, 30,000, and 17,600 yr. Theseperiods are listed in order of decreasing power, such that thevariance explained at 42,000 yr is 30% of that at 93,700 yr; at30,000 yr, 7%; and at 17,600 yr, 1%. We note that the periodexhibiting largest power is close to an orbital frequency foreccentricity (94,900 yr) and the period with second largestpower is close to the primary period for obliquity (41,000 yr).Band-pass filtering reveals a low-power, but significant, signal

at ∼3000 yr. The new portion of the record (i.e., from 4500 to80,000 yr ago, Fig. 2) was sampled at 400-yr intervals andfound to contain a dominant signal at 42,000 yr. Band-passfiltering confirmed that there are statistically significant, butmuch lower power, signals at 30,000, 21,000, 3000, and1500 yr, consistent with other paleoclimatic records.

Devils Hole δ18O as a proxy of SST off California

Given the strong mid- to high-latitude relation between δ18Oin precipitation and mean annual surface air temperature(Yurtsever and Gat, 1981), we assumed in our initial DevilsHole papers (Winograd et al., 1988, 1992) that the highestisotopic values displayed by the Devils Hole δ18O time seriesreflect relatively warm climate and the lowest values reflectrelatively cold climate. This assumption was supported by thestrong linear correlation (r=0.85 to 0.92) of Devils Hole δ18Owith the Vostok, Antarctica, ice core δ2H time series generallyaccepted to be a proxy of paleotemperature (Winograd et al.,1997, Fig. 8). However, similar to other proxy records, aquantitative evaluation of the numerous factors potentiallycontrolling the variations of δ18O in the Devils Hole veincalcite, and their relative importance, is not feasible. Difficultiesin extracting paleotemperature information from continentalδ18O archives (ice cores, speleothems, lake beds), especiallythose from low to mid-latitudes, have been discussed by Frickeand O'Neil (1999), Hendricks et al. (2000), and Alley andCuffey (2001). Accordingly, herein, we utilize an empiricalapproach to evaluating the paleoclimatic information embeddedwithin the Devils Hole δ18O time series.

That the Devils Hole δ18O time series closely tracks sea-surface temperature (SST) in selected northeastern Pacificmarine cores has been previously noted by Winograd et al.(1996, 1997), Lea et al. (2000), and Herbert et al. (2001). Toevaluate to what extent the most recent portion of the DevilsHole δ18O record also covaries with sea-surface temperature(SST) we performed two parallel analyses as shown below. Weobtained marine core records at five Eastern Pacific sites off theCalifornia coast ranging from the California–Oregon boundary(42°N) to the U.S.–Mexican boundary (∼32°N) that hadavailable both sea-surface temperature (SST) and benthic δ18Odata representing global ice volume. Focusing on the period160,000 to 4500 yr ago, we first obtain the correlations amongthese SST records and with the newly available Devils Holeδ18O. Because the marine cores generally cannot be dateddirectly beyond the range of carbon-14 dating, published ageshave been assigned by importing a SPECMAP-based chronol-ogy (Imbrie et al., 1984; Martinson et al., 1987). However, suchchronologies have been under discussion in the literature, inparticular for dates older than MIS 5d, that is prior to 120,000 yrago. Consequently, using more recently published studies forthe timing of MIS 6/5 boundary, we amended these portions ofthese five chronologies accordingly and repeated the afore-mentioned analyses for the period 160,000 to 4500 yr ago.

SST variations in the California Current during the past500,000 yr have been documented by analyzing the alkenonecontent of sediments from Ocean Drilling Program (ODP) Leg

Figure 3. Comparison of ODP Site 1012 SST time series (in red) with those forSite 1012 benthic δ18O (in black) and Devils Hole δ18O (in blue) for the period160,000 to 4500 yr ago. SST and benthic δ18O data from Herbert et al. (2001).Dashed red line, SST time series for the period 160,000 to 120,000 yr ago usingan amended chronology, as explained in text. The linear correlation coefficient(r) between SST and Devils Hole δ18O is 0.83 on both the published andamended chronology (Table 3).

206 I.J. Winograd et al. / Quaternary Research 66 (2006) 202–212

167 cores (Herbert et al., 2001; Yamamoto et al., 2004). (Use ofalkenones as a paleotemperature proxy has been questioned forselected oceanic environments, but this concern is not relevantto these cores as discussed in Appendix B of the SupplementalMaterials). Core site 1012 is located∼500 km south–southwestof Devils Hole, in the California Borderland in the SouthernCalifornia Bight, and is influenced seasonally by the northwardinvasion of the Southern California Countercurrent; and core

Table 3Pearson correlation (r) between the Devils Hole δ18O record and the SST time seinterpolated on a common 1700-yr time step

Core DH ODP 1020 ODP 893A ODP 1

PublishedODP1020 0.66 1.00ODP 893A 0.70 0.72 1.00ODP 1014 0.70 0.79 0.66 1.00ODP 1012 0.83 0.79 0.69 0.78EW9504 0.70 0.90 0.65 0.81LPAZ 0.66 0.81 0.62 0.68TR163-19 0.68 0.86 0.72 0.76

AmendedODP1020 0.84 1.00ODP 893A 0.77 0.76 1.00ODP 1014 0.83 0.81 0.72 1.00ODP 1012 0.83 0.82 0.72 0.83EW9504 0.85 0.91 0.69 0.83LPAZ 0.77 0.85 0.66 0.69TR163-19 0.85 0.88 0.75 0.76

The upper matrix gives results with SST chronologies as published by Herbert et al. (chronologies as amended per this paper. All correlations are significant at p<0.01.

1020 is located ∼1000 km northwest of Devils Hole, wellwithin the California Current (Fig. 1). For the period 420,000 to60,000 yr ago, Herbert et al. (2001) reported linear correlations(r) of 0.78 and 0.70, respectively, between the SST of thesecores and the Devils Hole δ18O time series; using the newlyextended Devils Hole record to 4500 yr ago, we obtaineffectively the same linear correlations of 0.80 and 0.68,respectively. Focussing on the period through MIS 6, we showin Figure 3 the ODP Site 1012 SST and the Devils Hole δ18Otime series, each on its own chronology from 160,000 to4500 yr ago, with a linear correlation of 0.83 between theserecords; for this period, the correlation with ODP site 1020 is0.66 (Table 3). Had we elected to compare these records for theperiod 110,000 to 4500 yr ago – thereby avoiding the portion ofthe marine record for which the chronology has been debated inthe literature – the correlations are substantively larger, namely0.88 and 0.76 between the Devils Hole record and cores 1012and 1020, respectively.

There is strong evidence (Henderson and Slowey, 2000;Johnson, 2001; Bard et al., 2002; Gallup et al., 2002; Speed andCheng, 2004) that the SPECMAP marine δ18O chronology(Imbrie et al., 1984), and its modification by Martinson et al.(1987), are up to 8000 yr too young during the periodencompassing the penultimate deglaciation (i.e., the transitionfromMIS6 toMIS5). Friddel et al. (2002,Table 1) suggested thatan age of 144,300±2900 yr ago should be assigned to marineisotope event 6.2 (the penultimate full glacial), 135,200±2500 yrago to event 6.0 (termination II), and 120,800±5800 yr ago toevent 5.51 (the youngest portion of the warm peak of substage5e). Additionally, Stirling et al. (1998) suggested an age of128,000±1000 yr ago for the oldest portion of the substage 5e sealevel high stand (marine isotope event 5.53).We followFriddel etal. (2002) and Stirling et al. (1998) in amending the SPECMAPages used for the period 145,000 to 120,000 yr ago in the benthicδ18O chronology of Herbert et al. (2001) for ODP Site 1012 by

ries from each of seven marine cores for the period 160,000 to 4500 yr ago,

014 ODP 1012 EW9504 LPAZ TR163-19

1.000.88 1.000.72 0.82 1.000.72 0.85 0.82 1.00

1.000.90 1.000.74 0.85 1.000.74 0.88 0.87 1.00

2001), Yamamoto et al. (2004) and Lea et al. (2000); the lower matrix, with SST(Self-correlation values of 1 for SST are shown to assist reading.)

207I.J. Winograd et al. / Quaternary Research 66 (2006) 202–212

accepting these time markers and then interpolating ages forbenthic δ18O data falling between them. To avoid a square wavetransition at 144,000 yr ago, we tapered the difference (betweenSPECMAP ages and more recent data for event 6.2) to zero at160,000yr ago.The published chronology for Site 1012SSTdatais based on the benthic δ18O time scale (Herbert et al., 2001) andso we obtained amended ages for the SST data in a consistentmanner. The dashed line on Figure 3 shows the portion of the SSTrecord whose chronology has been amended. For the period160,000 to 4500 yr ago the linear correlation (r) of the amendedSST time series for site 1012 with Devils Hole remains 0.83.Amending the chronology for site 1020 in a comparable manneryields effectively the same correlation, 0.84 (Table 3). In a similarfashion we compared the Devils Hole δ18O with SST at threemore sites from within the California Bight, namely ODP 893A,EW9504-03PC (EW03 on Fig. 1) and ODP 1014 (Fig. 1), forwhich both alkenone SST and benthic δ18O data are available(Herbert et al., 2001; Yamamoto et al., 2004). All showsubstantive and statistically significant linear correlations, butwhich are enhanced using the amended chronology (Table 3).The data for ODP Site 1014 are shown in Figure 4.

The Devils Hole δ18O and the five ODP Leg 167 SST timeseries discussed above differ from foraminiferal δ18O time seriesrepresenting global ice volume in two important respects,illustrated by Figures 3 and 4. First, the magnitude of the peakδ18O or SST during the last interglacial (LIG) is significantlygreater than the peak value of these parameters for the Holoceneby 1 per mill and 2 to 3°C, respectively (Figs. 3 and 4). In marineδ18O records, whether using the SPECMAP (Imbrie et al., 1984)stack of planktonic δ18O records or the benthic δ18O stack ofKarner et al. (2002), there are only a few tenths per mill

Figure 4. Comparison of ODP Site 1014 SST time series (in red) with those forSite 1014 benthic δ18O (in black) and Devils Hole δ18O (in blue) for the period160,000 to 4500 yr ago. SST and benthic δ18O data from Yamamoto et al.(2004). Dashed red line, SST time series for the period 160,000 to 120,000 yrago using an amended chronology, as explained in text. The linear correlationcoefficient (r) between SST and Devils Hole δ18O is 0.70 and 0.83 on thepublished and amended chronology, respectively (Table 3).

difference between the LIG and the Holocene. Second, asanticipated byWinograd et al. (1996) on the basis of vein sampleDHC2–3, both the Devils Hole and the SST records beginwarming ∼5000 to >10,000 yr prior to the start of the last twodeglaciations, which began at about 20,000 yr ago (Lambeck etal., 2002) and 135,000 yr ago (Martinson et al., 1987) or144,000 yr ago on the amended chronology (Friddel et al.,2002). That is, the warming was underway before and during thelast and the penultimate full glacials when the ice sheets were attheir maximum (Figs. 3 and 4).

Note that at three other Leg 167 drill sites (1017, 1018,and 1019; Fig. 1) these patterns are modified or absent. AtSites 1017 and 1018 an early SST warming occurred prior tothe penultimate, but not the last deglaciation (Mangelsdorf etal., 2002, Fig. 2). Cannariato and Kennett (2005) havedismissed the early SST warming and increase in Devils Holeδ18O before the penultimate deglaciation as artifacts of lowresolution records. We consider this notion to be unsuppor-table as discussed in Appendix C of the SupplementalMaterials. And, at Site 1019, which differs from other Leg167 cores in several respects (Lyle et al., 2000, Figs. 8–11),SST warming did not precede either of the last twodeglaciations, nor was the last interglacial peak temperaturewarmer than that for the Holocene (Mangelsdorf et al., 2002,Fig. 2). We tentatively attribute these differences to thecomplex physical oceanography of the California Current (seediscussion below).

The above described statistical and qualitative similaritiesbetween Devils Hole δ18O and these alkenone-derived SSTs offthe coast of California lead us to conclude that the Devils Holeδ18O time series (for the period 160,000 to 4500 yr ago) is a proxyfor northeastern Pacific SST. This conclusion does not requirecausation for the Devils Hole δ18O by these SST time series.Rather, both theSSTs andDevilsHole δ18Omaybe responding tothe same master forcing through an atmospheric teleconnection.

Possible causes for SST warming during MIS 6 and 2

Herbert et al. (2001) observed that SST in the CaliforniaCurrent began warming about 10,000–15,000 yr prior to thestart of the past 5 deglaciations and attributed this warming to a“collapse” (i.e., a substantial weakening) of the CaliforniaCurrent triggered by growth of the Laurentide Ice Sheet (LIS).We evaluate this and three other possible mechanisms for theearly SSTwarming: a period of sustained El Niño, a response toearly warming at tropical and subtropical latitudes, and aninterhemispheric causation.

1. Weakening of the California Current in response to growthof the LIS. Building on GCMmodeling by Kutzbach andWright(1985), which suggested that strong easterly winds weregenerated as a high pressure cell developed over the growingLIS, Herbert et al. (2001) proposed that these easterliesweakened the North Pacific High, and, in turn, the southward-flowing California Current. Aweakened current, they suggested,would lead to reduced coastal upwelling, which allowed thesubsequent eastward migration of warm water from the centralNorth Pacific gyre into the California margin and borderland.

Figure 5. Devils Hole δ18O (in blue) and SST (in red) time series from marinecores TR 163-19 and ODP 1020 for the period 360,000 yr ago to the present. Alltime series use their published chronology. SST data for TR163-19 from Lea etal. (2000); that for ODP 1020 from Herbert et al. (2001). Correlation coefficient(r) is for the period 360,000 to 4500 yr ago.

208 I.J. Winograd et al. / Quaternary Research 66 (2006) 202–212

Kutzbach andWright's (1985) analysis also suggested that at theLast Glacial Maximum (LGM), the northern jet stream split intotwo branches with the southern branch of the jet, andaccompanying storm tracks, purportedly displaced 20o latitudesouthward (to 30°N), bringing increased precipitation to theGreat Basin and filling Lakes Lahontan and Bonneville(Thompson et al., 1993; Benson, 2004). If the jet andaccompanying storm tracks were displaced southward duringthe period 30,000 to 17,000 yr ago, then the early SSTwarmingmight coincide with a migration of relatively warm central NorthPacific gyre waters eastward to the California coast.

Though intriguing, this explanation is problematic in threeways. First, the behavior of the California current is not clearlyknown. Several earlier studies have also proposed that theCalifornia Current weakened during the LGM (Lyle et al., 1992;Sancetta et al., 1992; Doose et al., 1997; Ortiz et al., 1997;Ganeshram and Pedersen, 1998), while others (Molina-Cruz,1988; Thunell and Mortyn, 1995; Mortyn et al., 1996; Gardneret al., 1997; Hendy and Kennett, 2000) have argued that theCalifornia Current increased in the strength during the LGM.Several of the cited studies explicitly noted latitudinal and/ordistance-from-shore variations in the proxies used to gage thepaleo-California Current. Given the physical oceanographiccomplexity of the modern California Current (Huyer, 1983;Lynn and Simpson, 1987), it is possible that the differingconclusions in the above cited paleoceanographic studies reflectcontemporaneous spatial differences within this eastern Pacificboundary current system. For example, as mentioned above, thealkenone SST time series for ODP Site 1017 (Seki et al., 2002)does not show the early warming prior to the last deglaciationseen in nearby core ODP 893A, nor does the SST time series forSite 1019 (Mangelsdorf et al., 2002, Fig. 2) resemble that ofnearby Site 1020 (Fig. 5).

Second, there is conflicting geologic evidence regardingpaleowind directions generated by the LIS during the LGM.Muhs and Bettis (2000) concluded that aeolian sediments in themid-continent North America do not support the existence ofstrong easterly winds during the LGM, as proposed by the GCMmodels cited by Herbert et al. (2001). In contrast, Sweeney andBusaca (2004) present evidence from Washington State thatappears to support strengthened easterly winds. Third, if growthof the LIS led to early SSTwarming, should not retreat of the icesheet have been accompanied by SST cooling, one not evidentin the expanded-scale SST time series for the past 30,000 yrpresented by Herbert et al. (2001, their Fig. 3)? The above threeconcerns do not disprove Herbert et al.'s (2001) regionally-limiting explanation for the early SST warming, but theyprompted us to explore other possibilities.

2. A period of sustained El Niño. There are major differencesof opinion in both the proxy-based and the atmospheric–oceanic modeling literature whether the mean state of the LGMwas more like an extended El Niño or La Niña event (Rosenthaland Broccoli, 2004). In view of this literature, might thewidespread early warming off California, recorded by the citedSST and Devils Hole time series between ∼30,000 and∼17,000 and ∼155,000 and ∼135,000 yr ago, perhaps reflectperiods during which El Niño events occurred with greater

frequency than today and possibly also with increasingintensity? We note that Bull et al. (2000) documented El Niñoperiodicities in Site 893A sediments during the past 160,000 yr,while Hinrichs et al. (1997) suggested, on the basis of alkenone-derived SST, that the peak last interglacial SST at ODP Site893A – which exceeded that for the Holocene by severaldegrees – may have been driven by a “prolonged El Niño-Southern Oscillation (ENSO) event” which intensified therelatively warm north-flowing Davidson Current and reducedthe cold south-flowing California Current.

An initial difficulty with the extended El Niño notion comesfrom core Site TR163-19, which is located in the easternequatorial Pacific at 2°N, 91°W (Lea et al., 2000). This site,perfectly situated to record an extended El Niño event, displaysearly SSTwarming prior to the penultimate deglaciation (whichbegan ∼135,000 yr ago on its published chronology and144,000 yr ago on the amended one), but none prior to the lastone which started ∼20,000 yr ago (Fig. 5).

Furthermore, a different proxy from the Devils Hole cave thatrecorded the water-table elevation in Brown's Room, a subaerialroom in the cave (see discussion in Appendix A of theSupplemental Materials and Szabo et al., 1994), indicates arapid lowering of the water table (to modern levels) starting at20,000±1400 yr ago, or 3000 yr prior to the attainment ofmaximum warmth (calcite δ18O) at 17,000±300 yr ago (Fig. 2).Apparently, the rapid increase in δ18O between ∼20,000 and17,000 yr ago in Devils Hole vein calcite (Fig. 2) and the SSTrise off California during MIS 2 were not accompanied byincreasing moisture transport into the Great Basin, as would beexpected during a period of El Niño events of increasingfrequency.

3. A response to early warming at tropical and subtropicallatitudes. At our suggestion, Lea et al. (2000) had compared

Figure 6. First principal component (PC1) for SST and Devils Hole δ18Onormalized time series using published (amended) chronologies against age. ThePC1 explains 78% (83%) of the total variability in the paleotemperature timeseries.

209I.J. Winograd et al. / Quaternary Research 66 (2006) 202–212

SST for core TR163-19 with Devils Hole δ18O for the period360,000 to 60,000 yr ago and obtained a correlation coefficient(r) of 0.73. With the Devils Hole record newly extended up to4500 yr ago (Fig. 5), we obtained effectively the same value,r=0.71. For the same interval, the correlation between DevilsHole δ18O and ODP 1020 SST is 0.69, effectively the same aswith TR163-19. Between the two cores, one equatorial (∼2°N)and the second off the northernmost coast of California(∼41°N), the correlation between the SST time series is 0.74.This indicates a strong teleconnection over 39 degrees oflatitude. Although consistent in timing and relative variability,we note that the magnitude of the warm SST excursions isdependent on latitude (Fig. 5).

In addition to SST series obtained at the five sites offCalifornia and the equatorial site TR163-19 for which both SSTand foraminiferal δ18O were available, we obtained data for asubtropical core, LPAZ, off southernmost Baja California (Fig.1) at ∼23°N (Herbert et al., 2001). Table 3 shows the linearPearson correlation (r) between all seven SST records for theperiod 160,000 to 4500 yr ago, using both published andamended chronologies; the correlations span 0.62–0.90 and0.66–0.91, respectively. Similarly, the correlations between theDevils Hole record and these SST records span 0.66–0.83 and0.77–0.85, respectively.

The high co-linearity among these records indicates a greatdeal of shared information, which we assessed using principalcomponents analysis (PCA). Each time series is treated as avariable measuring contemporaneous paleotemperaturerecorded at different locations. Because the Devils Hole recordis expressed as isotopic ratios and there are latitudinaldifferences in SST temperature range, we used normalizedtime series. We performed two PCA analyses, first with thepublished SST chronologies and then with the amended series.In both cases, the first principal component explained almostall of the system variation, 78% and 83%, respectively. In bothcases, the coefficients for the first principal component foreach of the 8 records were positive and almost equal in value(0.323 to 0.375; 0.323 to 0.371), indicating that all records,including the tropical and subtropical series, contributedequally to PC1, which strongly indicates a common forcingfactor. All of the other principal components were <0.1, so thatthey need not be considered as contributing significantly to thesystem variation. (We examined the second principal compo-nent although it was <0.1 for both the original and amendedchronologies, just to note that the values of the eightcoefficients indicated that the Devils Hole record has a closerrelationship with SST from the northern California Borderlandsites.) Figure 6 shows a plot of PC1 using the published andamended chronologies, respectively. The age at which theminimum SST occurs at MIS 2 is 30,000 yr ago; and at MIS 6,150,700 and 155,800 years ago, respectively, comparable tothose seen in the Devils Hole and SST records. We concludethat the paleotemperatures recorded at Devils Hole and in theseSST proxies covary substantially and are consistent with atropical Pacific origin.

4. An interhemispheric origin for the early SST warming.The warming during MIS 2 and 6 is not restricted to the

northeastern Pacific Ocean off California and Baja California orto equatorial site TR163-19. Two decades ago, the CLIMAPProject Members (1984) demonstrated that SST warmingpreceded the penultimate deglaciation by several thousandyears in low to mid-latitude Southern Hemisphere marine cores.Recently, Levine et al. (2001) demonstrated SST warmingpreceding the penultimate deglaciation by as much as14,000 years at 15 locations between 50°N and 5°S in theIndian, Pacific, and Atlantic Oceans. Winograd (2001) showedthat SST also preceded the last deglaciation by 5000 to15,000 years at a minimum of 13 sites located between 40°Nand 30°S in the Pacific and Atlantic Oceans. Thus, the SSTbehavior discussed above for many ODP Leg 167 sites andother cores is not unique, occurring in the low to mid-latitudesof all oceans. The above cited papers also show, however, thatthe SST warming prior to deglaciation does not occur every-where within the cited latitudes (see Fig. 2 in Winograd, 2001).Whether this interhemispheric early SSTwarming, in the low tomid-latitudes of the Pacific, Atlantic, and Indian Oceans prior tothe last two deglaciations, is of a common global causation isworthy of further study (Winograd, 2001). By association, then,the Devils Hole δ18O record may be indicative of the timing ofmajor shifts in SSTof a much larger region than the northeasternPacific Ocean.

In summary, of these four possible causes for the SSTwarming prior to the last and penultimate deglaciations (i.e.,during MIS 2 and 6), the last two – a tropical origin or aninterhemispheric one – appear most plausible.

Comparisons with present day observations

Numerous studies for the southwestern U.S. (Ely et al., 1994;Kahya and Dracup, 1994; Cayan et al., 1999; Higgins et al.,2000; McCabe and Dettinger, 2002; Pool, 2005) consistently

210 I.J. Winograd et al. / Quaternary Research 66 (2006) 202–212

document that occurrence of snowpack, extreme precipitationevents, and high stream flow accord with El Niño years. Otherstudies (Pandey et al., 1999; Friedman et al., 2002a) stress asouthwestern source of the moisture entering the Great Basin.The primary recharge area of the aquifer carrying groundwaterto Devils Hole is the Spring Mountains, the highest mountainrange in southern Nevada (∼3630 m), located ∼60 km east–southeast of the cave. Winograd et al. (1998) have shown thatmodern groundwater recharge in this upland is also stronglyinfluenced by El Niños and that recharge occurs principallyfrom winter–spring precipitation, a finding subsequentlysupported by Friedman et al. (2002b). Collectively, thesestudies point to a tropical or subtropical Pacific origin for themodern δ18O signal in Spring Mountains' winter–springprecipitation, a signal subsequently to be recorded as variationsin δ18O in the vein calcite coating the walls of Devils Holecavern. The two strongest El Niños of the 20th century, those of1982/83 and 1997/98, were accompanied by exceptionallystrong Aleutian Lows ( Simpson, 1992; Schwing et al., 2002)that favored advection of relatively warm north Pacific gyrewater (and presumably also storm tracks) to the California coast,providing yet another potentially important source of precipita-tion bound for the southern Great Basin.

That tropical and subtropical sources of inland-boundprecipitation also were prevalent during the past 160,000years is suggested by the presence of ENSO periodicities insediments deposited during the last 160,000 years at ODP Site893A (Bull, 2000) and during the past 30,000 years in core 480in the Gulf of California (Fig. 1; Schaaf and Thurow, 1998). Andcomparing the Devils Hole δ18O with SST from eastern tropicalPacific core TR163-19, Lea et al. (2000) observed that “SSTchanges in eastern Pacific are transmitted via moisture carriedby the westerlies to the western sector of North America.”

We are not the first to suggest a connection between easterntropical Pacific SST and the δ18O of winter precipitation onmajor uplands near or bordering the eastern Pacific Ocean.Bradley et al. (2003) and Vuille et al. (2003) demonstrated astrong positive correlation between eastern tropical andsubtropical Pacific SSTs (for November–February) and δ18Oin Andean ice cores from Bolivia and Peru for the period 1961–1997, with the SSTs themselves being strongly influenced byENSO. Vuille et al. (2003) concluded that such isotopic records“provide an important archive for reconstructing Pacific climatevariability.” We agree and suggest that Devils Hole δ18O timeseries represents an exceptionally well-dated proxy of mid-Pleistocene to mid-Holocene variations of SST off theCalifornia, and perhaps as well in the eastern tropical Pacific.

Conclusions

The Devils Hole, Nevada, mid-to-late Pleistocene (568,000–60,000 yr ago) δ18O record has been extended forward to themid-Holocene (∼4500 yr ago). This new record closelyresembles sea surface temperature (SST) time series frommarine sediments cored off the California coast. For the periodfrom 160,000 to 4500 yr ago, representing the best-datedsegment of these marine records, about 80% of the variation in

SST is shared with the Devils Hole record. During MIS 2 and 6,both the Devils Hole and SST time series record a steadywarming that began 5000 to >10,000 yr prior to the last andpenultimate deglaciations. Four possible causes for this earlywarming are examined. Though a continental archive, DevilsHole shares via an atmospheric teleconnection the δ18Osignature in tropical and subtropical moisture.

Acknowledgments

We are indebted to the late Scheck Exley and to RayHoffman, Paul Deloach, Guy Bryant, and Paul Smith for theirinvaluable experience and help during the 1992, 1994, and 1996SCUBA exploration of, and calcite sampling in, Devils Holeand Devils Hole cave #2. Professor Masanobu Yamamotokindly shared his SST data for ODP Site 1014. We thankMichael C. Ierardi for preparation of Fig. 1. The manuscriptbenefited measurably from review comments by Dan Muhs,Rick Forester, and QR's anonymous reviewers. Authorcontributions to this paper are as follows: Winograd andLandwehr jointly interpreted the data and wrote the paper;Coplen provided the analysis of the vein calcite samples forδ18O, while Sharp and Ludwig provided TIMS dates, generat-ing the U-series data in Tables 1 and 2; Riggs organized and ledthe SCUBA expeditions that found the vein calcites; andKolesar examined the calcites petrographically and was a keymember of the SCUBA team.

Appendix A. Supplementary data

Supplementary data associated with this article can be foundin the online version at doi:10.1016/j.yqres.2006.06.003.

References

Alley, R.B., Cuffey, K.M., 2001. Oxygen and hydrogen isotopic ratios of waterin precipitation: beyond paleothermometry. In: Valley, J.W., Cole, D.R.(Eds.), Stable Isotope Geochemistry. Reviews in Mineralogy andGeochemistry, vol. 43.Mineralogical Society of America,Washington, D.C.,pp. 527–553.

Bard, E., Antonioli, F., Silenzi, S., 2002. Sea-level during the penultimateinterglacial period based on a submerged stalagmite from Argentarola Cave(Italy). Earth and Planetary Science Letters 196, 135–146.

Barron, J.A., Heusser, L., Herbert, T., Lyle, M., 2003. High resolution climaticevolution of coastal north California during the past 16,000 years.Paleoceanography 18(1) (20), 1–19.

Benson, L.V., 2004.Western lakes. In: Gillespie, A.R., Porter, S.C., Atwater, B.F.(Eds.), The Quaternary Period in the United States. Elsevier, New York,pp. 185–204.

Bradley, R.S., Vuille, M., Hardy, D., Thompson, L.G., 2003. Low latitude icecores record Pacific sea surface temperatures. Geophysical Research Letters30 ((4)23), 1–4.

Broecker, W.S., 1992. Upset for Milankovitch theory. Nature 359, 779–780.Bull, D., Kemp, A.E.S., Weedon, G.P., 2000. A 160-K.y.-old record of El Nino-

Southern Oscillation in marine production and coastal runoff from SantaBarbara Basin, California, USA. Geology 28, 1007–1010.

Cayan, D.R., Redmond, K.T., Riddle, L.G., 1999. ENSO and hydrologicextremes in the Western United States. Journal Climate 12, 2881–2893.

Cheng, H., Edwards, R.L., Hoff, J., Gallup, C.D., Richards, D.A., Asmerom, Y.,2000. The half-lives of uranium-234 and Thorium-230. Chemical Geology169, 17–33.

211I.J. Winograd et al. / Quaternary Research 66 (2006) 202–212

Clark, D.H., 2003. Complex timing and patterns of glaciation in the AmericanCordillera during Termination I. XVI INQUA Congress Program withAbstracts, Paper 88-4, p. 231.

CLIMAP Project Members, 1984. The Last Interglacial Ocean. QuaternaryResearch 21, 123–225.

Doose, H., Prahl, F.G., Lyle, M.W., 1997. Biomarker temperature estimates formodern and the last glacial surface waters of the California Current Systembetween 33° and 42°N. Paleoceanography 12, 615–622.

Edwards, R.L., Cheng, H., Murrell, M.T., Goldstein, S.J., 1997. Protactinium-231 dating of carbonates by thermal ionization mass spectrometry:implications for Quaternary climate change. Science 276, 782–786.

Ely, L.L., Enzel, Y., Cayan, D.R., 1994. Anomalous North Pacific atmosphericcirculation and large winter floods in the Southwestern United States.Journal Climate 7, 977–987.

Fricke, H.C., O'Neil, J.R., 1999. The correlation between 18O/16O ratios of mete-oric water and surface temperature: its use in investigating terrestrial climatechange over geologic time. Earth Planetary Science Letters 170, 181–196.

Friddel, J.E., Thunell, R.C., Heusser, L.E., 2002. Direct comparison of marineand terrestrial climate variability during marine isotope stages 6 and 5:results from Santa Barbara Basin ODP Hole 893A. Paleoceanography 17((1)2), 1–12.

Friedman, I., Harris, J.M., Smith, G.I., Johnson, C.R., 2002a. Stable isotopecomposition of waters in the Great Basin, United States. 1. Air-masstrajectories. Journal Geophysical Research 107D ((19) ACL14), 1–14.

Friedman, I., Smith, G.I., Johnson, C.A., 2002b. Stable isotope compositions ofwaters in the Great Basin, United States. 2. Modern precipitation. JournalGeophysical Research 107D ((19) ACL15), 1–22.

Gallup, C.D., Cheng, H., Taylor, F.W., Edwards, R.L., 2002. Directdetermination of the timing of sea level change during Termination II.Science 295, 310–313.

Ganeshram, R.S., Pedersen, T.F., 1998. Glacial–interglacial variability inupwelling and bioproductivity off NW Mexico: implications for Quaternarypaleoclimate. Paleoceanography 13, 634–645.

Gardner, J.V., Dean, W.E., Dartnell, P., 1997. Biogenic sedimentation beneaththe California Current System for the past 30 Kyr and its paleoceanographicsignificance. Paleoceanography 12, 207–225.

Henderson, G.M., Slowey, N.C., 2000. Evidence from U–Th dating againstNorthern Hemisphere forcing of the penultimate deglaciation. Nature 404,61–66.

Hendricks, M.B., DePaolo, D.J., Cohen, R.C., 2000. Space and time variation ofδ 18O and δ D in precipitation: can temperature be estimated from ice cores?Global Biogeochemical Cycles 14, 851–861.

Hendy, I.L., Kennett, J.P., 2000. Dansgaard–Oeschger cycles and the CaliforniaCurrent system: Planktonic foraminiferal response to rapid climate change inSanta Barbara Basin, Ocean Drilling Program Hole 893A. Paleoceano-graphy 15, 30–42.

Herbert, T.D., Schuffert, J.D., Andreasen, D., Heusser, L., Lyle, M., Mix, A.,Ravelo, A.C., Stott, L.D., Herguera, J.C., 2001. Collapse of the CaliforniaCurrent during glacial maxima linked to climate change on land. Science293, 71–76.

Higgins, R.W., Schemm, J.-K.E., Shi, W., Leetmaan, A., 2000. Extremeprecipitation events in the western United States related to tropical forcing.Journal Climate 13, 793–820.

Hinrichs, K.-U., Rinna, J., Rullkotter, T., 1997. A 160-kyr record of alkenone-derived sea-surface temperature from Santa Barbara Basin sediments.Naturwissenschaften 84, 126–128.

Huyer, A., 1983. Coastal upwelling in the California Current System. Progressin Oceanography 12, 259–284.

Imbrie, J., 1992. A good year for Milankovitch. Paleoceanography 7,687–690.

Imbrie, J., Hays, J.D., Martinson, D.G., McIntyre, A., Mix, A.C., Morley, J.J.,Pisias, N.G., Prell, W.L., Shackleton, N.J., 1984. The orbital theory ofPleistocene climate: support from a revised chronology of the marine δ18Orecord. In: Berger, A., Imbrie, J., Hays, J., Kukla, G., Saltzmann, B. (Eds.),Milankovitch and Climate, Part 1. Reidel Publishing Company, Boston,pp. 269–305.

Imbrie, J., Mix, A.C., Martinson, D.G., 1993. Milankovitch theory viewed fromDevils Hole. Nature 363, 531–533.

Johnson, R.G., 2001. Last interglacial sea stands on Barbados and an earlyanomalous deglaciation by differential uplift. Journal Geophysical Research106C, 11543–11551.

Kahya, E., Dracup, J.A., 1994. The influence of Type 1 El Nino and La Ninaevents on streamflows in the Pacific southwest of the United States. JournalClimate 7, 965–976.

Karner, D.B., Muller, R.A., 2000. A causality problem for Milankovitch.Science 288, 2143–2144.

Karner, D.B., Levine, J., Medeiros, B.P., Muller, R.A., 2002. Constructing astacked benthic δ18O record. Paleoceanography 17 ((3)2), 1–16.

Kerr, R.A., 1997. Second clock supports orbital pacing of the ice ages. Science276, 680–681.

Kutzbach, J.E., Wright Jr., H.E., 1985. Simulation of the climate of 18,000 yearsB.P.: results for the North American/North Atlantic/European sector andcomparison with the geologic record of North America. Quaternary ScienceReviews 4, 147–187.

Lambeck, K., Yokoyama, Y., Purcell, T., 2002. Into and out of the last glacialmaximum: sea level change during oxygen isotope stages 3 and 2.Quaternary Science Reviews 21, 343–360.

Landwehr, J.M., Winograd, I.J., 2001. Dating the Vostok ice core record byimporting the Devils Hole chronology. Journal of Geophysical Research106D, 31853–31861.

Lea, D.W., Pak, D.K., Spero, H.J., 2000. Climate impact of late Quaternaryequatorial Pacific sea surface temperature variations. Science 289,1719–1724.

Levine, J., Karner, D.B., Muller, R.A., 2001. Warming at 140 Ka: causalityproblem for Milankovitch. EOS. Transactions of the American GeophysicalUnion 82S, F3.

Licciardi, J.M., Clark, P.U., Brook, E.J., Elmore, D., Shama, P., 2004. Variableresponses of Western U.S. glaciers during the last deglaciation. Geology 32,81–84.

Ludwig, K.R., Simmons, K.R., Szabo, B.J., Winograd, I.J., Landwehr, J.M.,Riggs, A.C., Hoffman, R.J., 1992. Mass-spectrometric 230Th–234U–238Udating of the Devils Hole calcite vein. Science 258, 284–287.

Lyle, M., Zahn, R., Prahl, F., Dymond, J., Collier, R., Pisias, N., Suess, E., 1992.Paleoproductivity and carbon burial across the California Current: themultitracers transect, 42°N. Paleoceanography 7, 251–272.

Lyle, M., Koizumi, I., Delaney, M.L., Barron, J.A., 2000. Sedimentary record ofthe California Current System: middle Miocene to Holocene: a synthesis of167 results. Proceedings of the Ocean Drilling Program. Scientific Results,vol. 167. Texas A and M University, pp. 341–376.

Lynn, R.J., Simpson, J.J., 1987. The California Current system: the seasonalvariability of its physical characteristics. Journal Geophysical Research 92C,12947–12966.

Mangelsdorf, K., Guntner, U., Rullkotter, J., 2002. Climatic and oceanographicvariations on the California continental margin during the last 160 kyr.Organic Geochemistry 31, 829–846.

Martinson, D.G., Pisias, N.G., Hays, J.D., Imbrie, J., Moore Jr., T.C.,Shackleton, N.J., 1987. Age dating and the orbital theory of the ice ages:development of a high-resolution 0 to 300,000-year chronostratigraphy.Quaternary Research 27, 1–29.

McCabe, G.J., Dettinger, M.D., 2002. Primary modes and predictability of year-to-year snowpack variations in the western United States from teleconnec-tions with Pacific Ocean climate. Journal Hydrometeorology 3, 13–25.

McDermott, F., 2004. Paleo-climate reconstruction from stable isotopevariations in speleothems: a review. Quaternary Science Reviews 23,901–918.

Molina-Cruz, A., 1988. Late Quaternary oceanography of the mouth of the Gulfof California: the polycystine connection. Paleoceanography 3, 447–459.

Mortyn, P.G., Thunell, R.C., Anderson, D.M., Stott, L.D., Le, J., 1996. Seasurface temperature changes in the Southern California borderlands duringthe last glacial–interglacial cycle. Paleoceanography 11, 415–430.

Muhs, D.R., Bettis, E.A., 2000. Geochemical variations in Peoria loess ofwestern Iowa indicate paleowinds in midcontinental North America duringlast glaciation. Quaternary Research 53, 49–61.

Ortiz, J., Mix, A., Hostetler, S., Kashgarian, M., 1997. The California Current ofthe last glacial maximum: reconstruction at 42°N based on multiple proxies.Paleoceanography 12, 191–205.

212 I.J. Winograd et al. / Quaternary Research 66 (2006) 202–212

Paillard,D.,Labeyrie,L.,Yiou,P., 1996.Macintoshprogramperforms time-seriesanalysis. EOS, Transactions-American Geophysical Union 77, 379.

Pandey, G.R., Cayan, D.R., Georgakakos, K.P., 1999. Precipitation structure inthe Sierra Nevada of California during winter. Journal Geophysical Research104D, 12019–12030.

Plummer, L.N., Busenberg, E., Riggs, A.C., 2000. In-situ growth of calcite atDevils Hole, Nevada: comparison of field and laboratory rates to a 500,000year record of near-equilibrium calcite growth. Aquatic Chemistry 6,257–274.

Pool, D.R., 2005. Variations in climate and ephemeral channel recharge insoutheastern Arizona, United States. Water Resources Research 41(w11403), 25 (doi:10.1029/2004WR003255).

Riggs, A.C., Carr, W.J., Kolesar, P.T., Hoffman, R.J., 1994. Tectonicspeleogenisis of Devils Hole, Nevada, and implications for hydrogeologyand the development of long continuous paleoenvironmental records.Quaternary Research 42, 241–254.

Rosenthal, Y., Broccoli, A.J., 2004. In search of paleo-ENSO. Science 304,219–221.

Sabin, A.L., Pisias, N.G., 1996. Sea surface temperature changes in theNortheastern Pacific Ocean during the past 20,000 years and theirrelationship to climate change in Northwestern North America. QuaternaryResearch 46, 48–61.

Sancetta, C., Lyle, M., Heusser, L., Zahn, R., Bradbury, J.P., 1992. Late-glacialto Holocene changes in the winds, upwelling, and seasonal production of theNorthern California Current System. Quaternary Research 38, 359–370.

Schaaf, M., Thurow, J., 1998. Two 30,000 year high-resolution grey value timeseries from the Santa Barbara Basin and Guaymas Basin. In: Cramp, A.,MacLeod, C.J., Lee, S.V., Jones, E.J.W. (Eds.), Geologic Evolution of Oceanbasins: Results from Ocean Drilling Program. In: Special Publication, vol.131. Geological Society, London, pp. 101–110.

Schwing, F.B., Murphee, T., de Witt, L., Green, P.M., 2002. The evolution ofoceanic and atmospheric anomalies in the northeast Pacific during the El Niñoand La Niña events of 1995–2001. Progress in Oceanography 54, 459–491.

Seki, O., Ishiwatari, R., Matsumoto, K., 2002. Millennial climate oscillations inNE Pacific surface waters over the last 82 Kyr: New evidence fromalkenones. Geophysical Research Letters 29 ((23) 59), 1–4.

Simpson, J.J., 1992. Response of the Southern California Current system to themid-latitude North Pacific coastal warming events of 1982–1983 and1940–1941. Fisheries Oceanography 1, 57–79.

Speed, R.C., Cheng, H., 2004. Evolution of marine terraces and sea level in thelast interglacial, Cave Hill, Barbados. Bulletin of the Geologic Society ofAmerica 116, 219–232.

Stirling, C.H., Esat, T.M., Lambeck, K., McCulloch, M.T., 1998. Timing andduration of the last interglacial: evidence for a restricted interval ofwidespread coral reef growth. Earth and Planetary Science Letters 160,745–762.

Sweeney, M.R., Busacca, A.J., 2004. Glacial anticyclone recorded in Palouseloess of northwestern United States. Geology 32, 705–708.

Szabo, B.J., Kolesar, P.T., Riggs, A.C., Winograd, I.J., Ludwig, K.R., 1994.Paleoclimatic inferences from a 120,000-yr calcite record of water-tablefluctuations in Browns Room of Devils Hole, Nevada. Quaternary Research41, 59–69.

Thompson, R.S., Whitlock, C., Bartlein, P.J., Harrison, S.P., Spaulding, W.G.,1993. Climatic changes in the western United States since 18,000 yr B.P. In:Wright, H.E., et al. (Ed.), Global Climates Since the Last Glacial Maximum.Univ. Minnesota Press, pp. 468–513.

Thunell, R.C., Mortyn, P.G., 1995. Glacial climate instability in the NortheastPacific Ocean. Nature 376, 504–506.

Vacco, D.A., Clark, P.U., Mix, A.C., Cheng, H., Edwards, R.L., 2005. Aspeleothem record of Younger Dryas cooling, Klamath Mountains, Oregon,USA. Quaternary Research 64, 249–256.

Vuille, M., Bradley, R.S., Healy, K., Werner, M., Hardy, D.R., Thompson, L.G.,Keimig, F., 2003. Modeling δ18O in precipitation over the tropical Americas:2. Simulation of the stable isotope signal in Andean ice cores. JournalGeophysical Research 108D ((6), ACL 2), 1–17.

Winograd, I.J., 2001. The magnitude and proximate cause of ice-sheet growthsince 35,000 yr. B.P. Quaternary Research 56, 299–307.

Winograd, I.J., Szabo, B.J., Coplen, T.B., Riggs, A.C., 1988. A 250,000-yearclimate record from Great Basin vein calcite: implications for MilankovitchTheory. Science 242, 1275–1280.

Winograd, I.J., Coplen, T.B., Landwehr, J.M., Riggs, A.C., Ludwig, K.R.,Szabo, B.J., Kolesar, P.T., Revesz, K.M., 1992. Continuous 500,000-yearclimate record from vein calcite in Devils Hole, Nevada. Science 258,255–260.

Winograd, I.J., Coplen, T.B., Ludwig, K.R., Landwehr, J.M., Riggs, A.C., 1996.High resolution δ18O record from Devils Hole, Nevada, for the period 80 to19 Ka. EOS. Transactions of the American Geophysical Union 77 (17),S169.

Winograd, I.J., Landwehr, J.M., Ludwig, K.R., Coplen, T.B., Riggs, A.C., 1997.Duration and structure of the past four interglaciations. Quaternary Research48, 141–154.

Winograd, I.J., Riggs, A.C., Coplen, T.B., 1998. The relative contributions ofsummer and cool-season precipitation to groundwater recharge, SpringMountains, USA. Hydrogeology Journal 6, 77–93.

Yamamoto, M., Oba, T., Shimamune, J., Ueshima, T., 2004. Orbital-scale anti-phase variation of sea surface temperature in mid-latitude North Pacificmargins during the last 145,000 years. Geophysical Research Letters 31,L16311.

Yurtsever, V., Gat, J.R., 1981. Atmospheric waters. In: Gat, J.R., Gonfiantini,R. (Eds.), Stable Isotopic Hydrology: Deuterium and Oxygen-18 inthe Water Cycle, IAEA Technical Report Series 210, IAEA, Vienna,pp. 103–142.