In record time (!), here is the summary of our discussion this morning, courtesy of Keith H.
The Pacific Decadal Oscillation is a multi-decadal oscillation in North Pacific sea surface temperatures (SSTs) that affects precipitation and temperature patterns during the winter season in North America. When the PDO is in its warm phase, cool SSTs and anomalously low sea level pressure exist in the Northern Pacific Ocean. The PDO Index quantifies the strength of the Pacific Decadal Oscillation by taking the difference between Pacific SST anomalies north of 20N and global SSTs. Typical phases in the PDO last approximately 20-30 years, with observed abrupt changes in sign of the PDO Index in 1946 and 1977. Since the instrumental record for SSTs only goes back until approximately 1900, only three clear PDO phases have been observed. Therefore, tree ring records in regions that are significantly affected by changes in the PDO are used as a proxy for the PDO Index for the years before instrumental records are available. These records are chosen by finding sites where winter precipitation in the 20th century is highly correlated to the PDO Index and can therefore be used as a proxy for the PDO Index. If winter precipitation is highly correlated with the PDO Index, then it is assumed that trees in these areas are highly sensitive to the amount of winter precipitation that falls, and the winter precipitation mostly varies based on the Pacific Decadal Oscillation. A physical mechanism for the Pacific Decadal Oscillation has not been identified.
Biondi et al. (2001) gathered six tree ring chronologies from Southern California and Baja California to represent the PDO Index back to ~1650. Although the winter precipitation in this region does not appear to be correlated with the PDO Index (Figure 1, Mantua and Hare 2001), the Tree Ring Index matches the PDO Index remarkably well from about 1920-1995. Therefore, the Tree Ring Index from these trees is a suitable proxy for the PDO Index before 1900. The decadal-scale variability of the PDO that is observed in the 20th century is also observed in these tree ring chronologies. The reconstructed PDO Index from Biondi et al. (2001) has a correlation of 0.57 with the reconstructed PDO Index from MacDonald and Case (2005).
Mantua and Hare (2001) assembled various studies that used different proxies to represent the PDO Index back to around 1600. They observe that while the dendrochronologies are in sync during some periods, many periods exist where they show "little, if any correspondence with each other." This indicates the likelihood that other influences besides the PDO were affecting precipitation and/or the growth of trees in areas where tree rings were sampled.
MacDonald and Case (2005) used two tree ring chronologies (one from southern California and one from western Canada) as a proxy for the PDO Index back to approximately AD 900. These locations were chosen because they represent a dipole of precipitation from the PDO, where increases in precipitation one location due to the PDO is coincident with decreases in precipitation at the other location. The tree ring chronologies were significantly correlated to the PDO Index during the period of observed SSTs, but displayed rather low correlations with other PDO proxies (r = 0.19 with Gedalof and Smith (2001)). The low sample size of tree ring chronologies may not be truly representative of regional climatic variations that are caused by the PDO, but coincident local climatic variations instead. Perhaps the addition of more tree ring chronologies would provide more confidence about the structure of the PDO before the presence of instrumental records.