In order to know possible phase propagation features, we also examine amplitudes and phases of the wintertime relative MWF SLPs again with respect to the phase at 50｡? 165｡? (Fig. 3). In the North Pacific after 1930, the phases exhibited continuous increase from Pacific North West to Kamchatka peninsula across through the amplitude maxima in the central North Pacific. The phase differences between acific North West to Kamchatka peninsula is about 90 degrees. Another notable feature is that the apparent Atlantic dipole pattern seen in Fig. 1 turns out that it is, in fact, a monopole structure with phase lags in the first few decades in the 20th century.
The validity of the phase propagation found in Fig. 3 in the Pacific sector is examined by calculating coherency and phase spectra between 45°N, 140°W and 60°N, 180 (Fig. 4). The coherency exhibits a significant peak on the bidecadal timescale, as well as the other three peaks on interannual timescales. The phase difference at the bidecadal peak is 93 degree, and is significantly different from zero and from 180°. Consequently, one can concluded that the propagation from Pacific North West to Kamchatka peninsula is statistically meaningful.
In order to examine closely the southward migration of the BO in the North Pacific sector, we plot a latitude-time section of the SLPs averaged zonally from 160｡? to 140｡? by using MWF and a conventional band-pass filter with the half-power points at 10 and 30 year periods (Fig. 5). In the first few decades, significant amplitudes are confined north of 60｡? then migrated toward the south from 1930 to 1950 with the center of the variability about 45｡? in 1950. From 1950 to the present, the latitudes of the amplitude maxima shifted slowly further toward the south, and the center of the variability after 1980 locates about 40｡?. Qualitatively the same features are captured by the conventional band-pass filter, but the band-pass filtered amplitudes are larger than the MWF amplitudes. The smaller amplitudes by the MWF are resulted from the narrower pass-band of it.
In the North Atlantic sector the comparison between the MWF SLPs and 10-30 band-pass filtered SLPs give substantially different visual impressions (Fig. 6). The band-pass filtered data pose much more complicated structures than the MWF data, especially from 1930-1960. This indicates that only a certain portion of the Atlantic variations on the time scales from 10-30 years varied coherently with the BO in the North Pacific.
3.2. SST and LAT
It is interesting to examine whether physical parameters of the atmosphere and ocean other than the SLP support or not the century-scale migration of the BO over the North Pacific. The parameter first to be examined should be SSTs, since most of the migration occurred over the central North Pacific. Unfortunately, the coverage of the available grids of the SST data is relatively poor in the early 20th century as shown in Fig. 7. Thus, we must carefully choose regions where SSTs are examined from the point of views of good coverage of available data and physically meaningful locations. In order to know what regions are suitable from the aspect of the physical meanings, we examine the first EQE of the cold- season SST after 1950, during which available SST data in the North Pacific have almost complete coverage.
