Isolating the Decadal Component of the Pacific Decadal Oscillation

Modes of low-frequency variability, such as the Pacific Decadal Oscillation (PDO), are traditionally defined with EOF analysis, where principal components are ranked by the amount of SST variance they explain. Under this definition, the PDO exhibits variability on a wide range of timescales, though it is typically associated with decadal timescales. We employ low-pass component analysis to rank components instead by the ratio of interdecadal to intradecadal SST variance in unforced preindustrial control runs of coupled climate models and in the NOAA Extended Reconstructed Sea Surface Temperature (ERSST) data set. This analysis identifies a PDO-like pattern of variability, the principal component of which operates predominantly on decadal timescales without much ``noise" on seasonal and interannual timescales. This decadal PDO index exhibits significant autocorrelation at 6 year time lags, but still explains nearly 50% of the temporal variance of the PDO as traditionally defined using EOFs. It is associated with temperature variability in the Kuroshio extension, Bering Sea, and Gulf of Alaska and has little manifestation in tropical SSTs. When there is a warm SST anomaly in the Kuroshio, there are positive surface upward energy fluxes, a reduced Aleutian low, and higher surface air temperatures over the United States and Russia. These temperature anomalies in the Kuroshio extension result from changes in strength and extent of the North Pacific subpolar gyre in response to wind-stress forcing, as illustrated by the significant lead-lag relationship between annual-mean North Pacific wind stress curl anomalies and the decadal PDO index. A cyclonic anomaly over the North Pacific leads to a cold Kuroshio SST anomaly after a lag of 2-5 years. This has implications for the predictability of North Pacific SSTs and the teleconnections to land-surface climate over North America and Asia.