Variations in the Pacific Decadal Oscillation over the past millenniumGlen M. MacDonald and Roslyn A. CaseDepartment of Geography, University of California, Los Angeles, California, USAReceived 19 January 2005; revised 25 February 2005; accepted 9 March 2005; published 20 April 2005.[1] Hydrologically sensitive tree-ring chronologies fromPinus flexilis in California and Alberta were used to producean AD 993 –1996 reconstruction of the Pacific DecadalOscillation (PDO) and to assess long-term variability in thePDO’s strength and periodicity. The reconstruction indicatesthat a 50 to 70 year periodicity in the PDO is typical forthe past 200 years but, was only intermittently a strongmode of variability prior to that. Between AD 1600 and1800 there is a general absence of significant variabilitywithin the 50 to 100 year frequency range. Significantvariability within in the frequency range of 50 to 100 yearsreemerges between AD 1500 and 1300 and AD 1200 to1000. A prolonged period of strongly negative PDO valuesbetween AD 993 and 1300 is contemporaneous with asevere medieval megadrought that is apparent in manyproxy hydrologic records for the western United States andCanada.Citation: MacDonald, G. M., and R. A. Case (2005),Variations in the Pacific Decadal Oscillation over the pastmillennium, Geophys. Res. Lett., 32, L08703, doi:10.1029/2005GL022478.1. Introduction[2] The Pacific Decadal Oscillation (PDO) is a leadingmode of multi-decadal variability in sea surface temperatures(SST’s) in the extratropical North Pacific [Mantua et al.,1997; Nigam et al., 1999; Minobe, 2000; Mantua and Hare,2002]. The PDO index is derived from an EOF analysis ofSST’s and positive phases of the PDO are typified by warmSST’s in the northeastern Pacific (Figure 1). The climatic andenvironmental impacts of positive and negative PDO phasesare of major importance to fisheries and water resources[Latif and Barnett, 1993; Cayan et al., 1998; Nigam et al. ,1999; Minobe, 2000; Barlow et al., 2001; Mantua and Hare,2002; Chavez et al., 2003]. In western North America,positive phases of the PDO are associated with climaticconditions similar to El Nin˜o – although weaker in expres-sion. These conditions include decreased winter precipita-tion, snowpac k and streamflow in the northwest and higherprecipitation in the southwest. Conditions reverse duringnegative PDO phases. The PDO has been shown to modulateclimatic teleconnections between North American climateand the equatorial Pacific during El Nin˜o and La Nin˜a events[Gershunov and Barnett, 1998; Brown and Comrie, 2004].[3] A positive PDO regime existed between 1977 and1997. The index now appears to be moving toward anegative or more variable state. The causes of shifts in thePDO remain uncertain [Mantua and Hare, 2002], butpossibly lie in the tropical Pacific [Latif and Barnett,1993; Zhang et al., 1998; Cane and Evans, 2000; Linsleyet al., 2000; Evans et al., 2001]. Shifts between states of thePDO may reflect non-linear dynam ics of the ocean-atmosphere system [Overland et al., 2000] and/or be forcedby factors such as strong El Nin˜o and La Nin˜a events[Biondi et al., 2001; Newman et al., 2003]. Fluctuations inthe strength of both positive and negative PDO phases occurin the E l Nin˜o/Southern Oscillation (ENSO) frequencybands of 2 to 7 years [Minobe, 2000].[4] Dominant periodicities in the 50 to 70 year andbidecadal year bands have been proposed for the PDO[Minobe, 1999]. Testing the persistence of dominant perio-dicities in the PDO is of key concern in predicting futureclimate variations in western North America. Unfortunately,instrumental records of SST’s are too short in duration toconfidently gauge the long-term persistence and multi-decadal variability of the PDO. Instrumental records ofNorth Pacific SST’s a re relatively sparse prior to the1940’s [Woodruff et al., 1987] and calculations of thePDO can be inconsistent with one another prior to that time[Biondi et al., 2001]. Here we present a >1000-year recon-struction of the PDO derived from tree-ring records andexamine the long-term multi-decadal variability of the PDO.We also demon strate that a prolonged depression in the PDOcorrelates with an episode of aridity apparent throughoutmuch of western North America between AD 900 and 1300.2. Study Sites and Methods[5] Southern California and western Canada lie at oppo-site ends of the PDO precipitation dipole (Figure 1) andhydrological variations related to the PDO should be ofopposite sign in the two regions. Both regions support Pinusflexilis (James) trees that are useful in producing dendro-climtological records of precipitation and streamflow [Caseand MacDonald, 1995, 2003; Hidalgo et al., 2001]. Chro-nologies from trees growing at the two ends of the PDOprecipitation dipole should maximize the probability ofproducing a robust representative reconstruction of thePDO. Chronologies were developed from living and deadPinus flexilis trees occurring in open groves on rockysubstrates. The California site is located near Mount SanGorgonio (34°040116°290) while the Alberta site lies on theflanks of the Rocky Mountains at Whirlpool Point nearNordegg, Alberta (52°000N116°270W). We used standardtechniques [Stokes and Smiley, 1968; Fritts, 1976; Cook andKairiukstis, 1990] to develop millennial length tree-ringchronologies for both sites. Minimum sample depth was sixradii and mean sensitivity ranged from 0.168 at SanGorgonio to 0.397 at Whirlpool Point. Both chronologiesdisplayed significant correlations (p 0.05) with monthlyprecipitation records and selected monthly PDO indexvalues. Using standard chronologies and the 1940– 1998PDO index from Mantua (http//:jisao.washington.edu/pdo/GEOPHYSICAL RESEARCH LETTERS, VOL. 32, L08703, doi:10.1029/2005GL022478, 2005Copyright 2005 by the American Geophysical Union.0094-8276/05/2005GL022478$05.00L08703 1of4pdo.latest) for calibration and verification purposes [Fritts,1976; Cook and Kairiukstis, 1990] we produced a multipleregression based reconstruction of annual PDO (Jan –Dec)that extends from AD 993 to 1996 (Figure 2 and Table 1).The 1940 to 1996 PDO record was used for calibrationbecause there is a high density of SST measurements and agood correspondence between different PDO indices overthis period. Verification (Figure 2 and Table 1) was doneusing sequence splitting (1940– 1967 and 1968 –1996). Asexpected, the regression coefficients for the Alberta chro-nology were consistently