Zhao and Li  indicated that atmospheric circulation anomalies display a significant WWR in the central North Pacific Ocean, closely related to the SSTA WWR near 40°N. Furthermore, the area average over the central North Pacific (CNP: 165°E–160°W, 35°N–47°N) is used by Zhao and Li  to measure the interannual variability of the WWR in the North Pacific. We start from the original definition of the WWR (the mean climatic characteristic) to determine whether the WWR exists in the CNP region each year. Previous studies used lag correlation analysis to define the WWR [e.g., Alexander et al., 1999]. The lag correlations between monthly anomalies for February and monthly anomalies for each subsequent month through February of the year after next year were calculated in the CNP region. The lag correlations have two significant characteristics: a significant decline during the following summer and an increase again during the following winter, which was called the WWR [e.g., Alexander and Deser, 1995]. To efficiently detect the WWR and non-WWR years, its two characteristics are quantitated for each year by using the following criteria: (1) for positive (negative) anomalies during the winter, winter anomalies are greater (less) than anomalies during the following summer; (2) the following winter anomalies are greater (less) than anomalies in the preceding summer and have the same sign as anomalies in the preceding winter. It is a positive (negative) WWR year if a year meets the two criteria and a positive (negative) non-WWR year if it does not. In this way,Zhao and Li identified 18 WWR years (positive cases: 1951, 1956, 1965, 1966, 1968, 1971, 1974; negative cases: 1959, 1960, 1977, 1978, 1983, 1985, 1986, 1994, 1996, 1998, 1999) and 36 non-WWR years (positive cases:1950, 1952, 1953, 1954, 1955, 1957, 1962,1963, 1967, 1969, 1972, 1976, 1982, 1989, 1990, 1991, 1993, 2000, 2002; negative cases:1958, 1961, 1964, 1970, 1973, 1975, 1979, 1980, 1981, 1984, 1987, 1988, 1992, 1995, 1997, 2001, 2003) in the CNP region during the period 1950–2003.
 Figure 1shows time-altitude profiles of composite geopotential height anomalies between 1000 and 70 hPa in the CNP region during the WWR and non-WWR years. Note that the climatological seasonal cycle of geopotential height has been removed from the monthly values prior to the calculations. For the positive cases of the WWR (Figure 1a), the seasonal evolution is characterized by two reversals in the sign of geopotential height anomalies in the CNP region. The geopotential height anomalies in the first winter are positive with a maximum in January–February; they change to negative in the following summer with a maximum in June–August; then they return to positive again in the second winter. The geopotential height field in the CNP region displays WWR from the lower layer to the upper layer during the WWR years, which shows an equivalent barotropic vertical structure and the centers of anomalies are located in the high troposphere (500–300 hPa). Therefore, unlike the reemergence mechanism of the SSTA WWR, the recurrent atmospheric circulation anomalies in the second winter do not come from those of the previous winter through anomalies in the intervening summer. It seems that mechanisms of the WWR are markedly different between the atmosphere and ocean. Seasonal evolution of the geopotential height anomalies is similar during the negative cases of the WWR (Figure 1b). But the evolution of the atmospheric circulation anomalies during the non-WWR years (Figures 1c and 1d) differs markedly from that during the WWR years. Atmospheric circulation anomalies in winter do not recur in the following winter. Based on these results, the possible causes of the atmospheric in the CNP are investigated in section 3.