In this section, we examine the factors affecting the changes in WP variability. We first examine the role of tropical Pacific SST anomalies. Then, we discuss the role of North Pacific storm track activity.
a. Role of tropical Pacific SST anomalies
Studies have suggested that the tropical SST anomalies that are related to ENSO significantly contribute to WP variability by triggering an extratropical atmospheric teleconnection (Horel and Wallace 1981; Barnston and Livezey 1987; Koide and Kodera 1999; Wang et al. 2007; Furtado et al. 2012; He et al. 2013; Park et al. 2018; Aru et al. 2021). Figure 5 presents the lead-lag correlation coefficients of the wintertime WP index (D(-1)JF(0)) with the Niño3.4 index from the preceding summer (JJA(-1)) to the following summer (JJA(0)). Here, the time notations “-1” and “0” denote the order of the years. A remarkable difference was found in the WP-ENSO correlation between the two epochs (Fig. 5). During 1950–1974, the wintertime WP index had a significant positive correlation with the Niño3.4 index from the previous summer (JJA(-1)) to the following spring (MAM(0)). In contrast, the correlation of the winter WP index with the Niño3.4 index was weak during the 1989–2013 period. Hence, ENSO has a strong (weak) relationship with the WP during periods of strong (weak) WP variability. The above results imply that the observed changes in WP variability may be partly related to changes in the ENSO-WP relation.
To confirm the above speculation, we examined the 25-yr moving standard deviations of the winter WP index and the 25-yr running correlation coefficients of the winter WP index with the Niño3.4 index (Fig. 6a). As is clear in Fig. 6a, the changes in the winter WP-ENSO relation have very good agreement with the changes in WP variability. Figure 6b presents a scatter plot of the 25-yr moving standard deviations of the winter WP index versus the 25-yr running correlations of the winter WP index with the Niño3.4 index. The two quantities shown in Fig. 6b are highly correlated with each other, with correlation coefficients as high as 0.94. Therefore, the above evidence indicates that changes in the WP-ENSO relationship are closely related to variations in WP variability. Specifically, during period with strong (weak) WP-ENSO connections, the variability in the WP is strong (weak).
Figure 7 depicts maps of the atmospheric and SST anomalies in winter regressed onto the simultaneously calculated winter Niño3.4 index during 1950–1974 and 1989–2013. During 1950–1974, the geopotential height anomalies in the upper troposphere (200 hPa) featured marked negative anomalies over the Russian Far East extending eastward to the west of Alaska, accompanied by large-scale positive anomalies over the tropics (Fig. 7a). The Z500, 850-hPa wind and SLP anomalies (Figs. 7b-7c) were similar in this period to those related to the WP (Fig. 1a), with positive Z500, SLP and anticyclonic anomalies observed over the subtropical WNP and anomalies with opposite signs observed over the mid-high-latitude WNP. Hence, during this period, the ENSO-related SST anomalies in the tropics strongly impacted the WP and thus contributed to stronger WP variability. The atmospheric anomalies related to the Niño3.4 index over the North Pacific that occurred during 1989–2013 (Figs. 7e-g) show notable differences from those recorded during 1950–1974 (Figs. 7a-c). In particular, the atmospheric anomalies related to the Niño3.4 index exhibited a PNA-like pattern, with strong negative geopotential height, SLP and cyclonic anomalies over the mid-latitude North Pacific and the southeastern United States, along with anomalies with opposite signs over the subtropical northeastern Pacific and northern Canada (Figs. 7e-g). The above evidence indicates that the atmospheric anomalies generated by tropical ENSO projected more onto the PNA pattern during 1989–2013, in sharp contrast to the WP pattern during 1950–1974. Therefore, the weak contribution of the tropical SST anomalies related to ENSO to the WP may partly explain the weak WP variability recorded during 1989–2013.
The above evidence indicates that ENSO-related atmospheric anomalies showed notable differences between the two periods. A question naturally arose: what leads to changes in the spatial structure of atmospheric anomalies that are related to ENSO? Figs. 7d and 7h depict the regression patterns of winter SST anomalies upon the winter Niño3.4 index. The spatial patterns of the SST anomalies in the tropics appeared similar during the two periods, with significant SST warming occurring in the tropical central and eastern Pacific and in the tropical Indian Ocean, accompanied by SST cooling in the tropical western Pacific. In addition, the amplitudes of the SST anomalies in the tropical central and eastern Pacific were similar between the two periods. In particular, we examined a scatter plot between the 25-yr running correlations of the winter WP index and Niño3.4 index against the 25-yr running standard deviations of the Niño3.4 index (not shown). The correlation coefficient between the above two variables was weak, indicating that a change in the ENSO intensity could not explain changes in the spatial structure of ENSO-induced extratropical atmospheric anomalies and thus could not explain changes in the ENSO-WP relation.
Studies have demonstrated that the spatial distributions of atmospheric anomalies induced by tropical SST anomalies are highly sensitive to the zonal locations of SST anomalies in the tropics (Trenberth et al. 1998; Hoerling and Kumar 2002; Barsugli and Sardeshmukh 2002; Zhou et al. 2014; Jo et al. 2015; Soulard et al. 2019; Chen et al. 2021). In particular, the ENSO-induced atmospheric anomalies over the North Pacific shifted eastward (westward) when the ENSO-related SST anomalies in the tropical Pacific were located farther eastward (westward) (Jo et al. 2015; Soulard et al. 2019; Chen et al. 2021). Following Chen et al. (2021), a quantification of the zonal shift of the ENSO-related SST anomalies in the tropics was defined as the longitude of the 0°C isotherms of the SST anomalies in the equatorial western Pacific. For accuracy, we interpolated the SST data onto a higher-resolution grid of 0.5°×0.5°, taking the mean longitude of the 0°C isotherms between 5°S and 5°N as the index describing the east-west movement of the ENSO-related SST anomaly pattern, termed the “zero line”. Figure 8a presents a scatter plot of the zero line of SST anomalies in the tropical Pacific versus the 25-yr moving standard deviations of the WP index. Figure 8b shows a scatter plot of the zero line versus the 25-yr running correlation of the winter WP index with the Niño3.4 index. The correlation coefficients between the two quantities shown in Fig. 8a and Fig. 8b were − 0.50 and − 0.38, respectively, both of which are significant at the 99% confidence level according to two-tailed Student’s t-test. These results suggest that changes in the zonal location of tropical SST anomalies related to ENSO may partly contribute to changes in the spatial structure of the atmospheric anomalies induced by ENSO; this result is consistent with previous findings (Jo et al. 2015; Soulard et al. 2019; Chen et al. 2021).
b. Impact of ENSO on the WP variability in a Coupled Model Simulation
The above analyses indicate that the observed changes in the variability in the WP were partly attributed to changes in the ENSO-WP relationship. In this subsection, we use a long historical simulation of a coupled climate model over 1850–2014 to confirm this conclusion. This long simulation was conducted using the National Aeronautics and Space Administration (NASA) Goddard Institute for Space Studies (GISS) climate model GISS-E2.1-G, which is participating in CMIP6 (Kelley et al. 2020). A more comprehensive description of GISS-E2.1-G can be found in Kelley et al. (2020). The atmospheric variables obtained from the GISS-E2.1-G historical simulation were interpolated to a common horizontal resolution of 2.5°×2.5° to facilitate comparisons with the observations. The WP index in GISS-E2.1-G was obtained by projecting the model-simulated Z500 anomalies onto the observed EOF1 loading pattern.
We first examined the ability of the GISS-E2.1-G model to simulate the spatial patterns of the WP and ENSO. Figure 9a displays the winter 500-hPa geopotential height anomalies obtained by regression on the normalized WP index for the 1850–2014 period. This figure depicts a canonical meridional dipole pattern over the subtropical western North Pacific and mid- to high-latitude North Pacific and the Far East, bearing a close resemblance to the WP shown in the observations (Fig. 1a). In addition, ENSO-related SST anomalies are simulated well by the model, featuring prominent SST warming in the tropical central and eastern Pacific and Indian Ocean, together with SST cooling in the tropical western Pacific (Fig. 9b). Moreover, the GISS-E2.1-G model simulated the spatial distribution of the WP-related SAT anomalies over Eurasia and the North Pacific well, as has been reported in previous studies (Sung et al. 2019, 2020; Aru et al. 2021). Hence, it is reasonable to use the GISS-E2.1-G model to examine changes in the variability in the WP and changes in the WP-ENSO relation.
Figure 10a shows the 25-yr moving standard deviations of the WP index and the 25-yr running correlation coefficients between the winter WP index and Niño3.4 index obtained with the GISS-E2.1-G model. The change in the variability in the wintertime WP corresponded well with the variation in the WP-ENSO relation, except during the early 20th century (Fig. 10a). The correlation coefficient between the time series of the 25-yr moving standard deviation of the WP index and the 25-yr running correlation of the WP index with the Niño3.4 index reached 0.62, which was significant at the 99% confidence level (Fig. 10b). Hence, the GISS-E2.1-G-derived long-term simulations further confirmed that changes in the winter WP variability are closely related to changes in the ENSO-WP relation. In addition, a statistically significant correlation coefficient of 0.35 was found between the variation in the WP-ENSO relation and the variation in the zero line of the ENSO-related SST anomalies in the tropical Pacific (Fig. 10c). This finding confirmed the observed results in which changes in the wintertime WP-ENSO relation were determined to be partly related to the zonal shifts of ENSO-related SST anomalies in the tropics.
Role of North Pacific storm track activity
Apart from tropical SSTs, studies have suggested that the interaction between the low-frequency mean flow and synoptic-scale eddies is a crucial factor affecting the formation and maintenance of atmospheric circulation patterns associated with the WP (Lau 1988; Chang and Fu 2002; Linkin and Nigam 2008; Tanaka et al. 2016; Wettstein and Wallace 2010). This, therefore, suggests that changes in the wave-mean flow interactions and associated synoptic-scale eddy feedbacks may also contribute to changes in the WP variability. Figures 11a and 11c show regression maps of storm track anomalies against the winter WP index for the 1950–1974 and 1989–2013 periods, respectively. During 1950–1974, marked increases in storm track activity occurred in the area of approximately 40°-60°N over the North Pacific (Fig. 11a), corresponding well with westerly anomalies in the same region (Fig. 3c). Studies have demonstrated that increased (decreased) storm track activity is associated with westerly (easterly) anomalies and is immediately accompanied by positive anticyclonic vorticity forcing to the south and cyclonic vorticity forcing to the north (Lau et al. 1988; Chen et al. 2014), helping maintain a WP-like meridional atmospheric dipole pattern (Fig. 3c). Positive storm track anomalies were also observed over the North Pacific during 1989–2013, but these anomalies were less statistically significant (Fig. 11c). We further calculated the geopotential height tendency to quantitatively examine changes in the feedback of the synoptic-scale eddy to the mean flow (e.g., Lau and Holopainen 1984; Lau 1988; Cai et al. 2007). Figures 11b and 11d show the regression maps of geopotential height tendency at 500 hPa onto the winter WP index during the two studied periods. During 1950–1974, the spatial structure of the geopotential height tendency anomalies (Fig. 11b) was similar to those of the atmospheric anomalies related to the WP index (Fig. 3b). In particular, notable negative geopotential height tendency anomalies occurred over the Russian Far East, extending eastward to the Aleutian region, and positive geopotential height tendency anomalies were found over East Asia, extending eastward to the central North Pacific (Fig. 11b). Hence, during 1950–1974, the feedback of synoptic eddies to the mean flow was conducive to the formation and maintenance of the WP. During 1989–2013, the geopotential height tendency anomalies also exhibited a meridional dipole pattern over the North Pacific but with much weaker amplitudes and less significance compared to those that occurred during 1950–1974. Hence, a stronger wave-mean flow interaction and stronger feedback of synoptic-scale eddies to the mean flow may also partly contribute to the stronger WP variability that occurred during 1950–1974.
What contributes to changes in the feedback strength of synoptic-scale eddies? Studies have indicated that the feedback strength of synoptic-scale eddies is related to changes in the mean-state circulation and amplitude of storm track activity (Lau 1988; Chang and Fu 2002; Jin et al. 2006a, b; Kug et al. 2009a; Jin 2010). We examined the difference in the mean flow between the 1950–1974 and 1989–2013 periods. The difference was weak over the North Pacific (not shown). This implies that the change in the feedback strength of the synoptic-scale eddies to the mean flow is not related to changes in the mean-state circulation. Then, we further examined the change in the amplitude of the storm track activity. Following previous studies (Lee et al., 2012; Wettstein and Wallace, 2010; Ma and Zhang 2018), we used EOF analysis to obtain the dominant mode of storm track activity over the North Pacific (20°-90°N, 120°E-100°W). The long-term trend was removed prior to the EOF analysis. Figure 12a depicts the EOF1 of the winter storm track activity over the North Pacific, and Fig. 12b shows the corresponding PC time series. The EOF1 of the North Pacific storm track activity accounted for 19.62% of the total variance and was characterized by a monopole pattern over the North Pacific. The center of EOF1 corresponded well with the maximum center of the mean winter storm track, as has been reported in previous studies (Lee et al., 2012; Wettstein and Wallace, 2010). This suggests that EOF1 represents variations in the wintertime North Pacific storm track intensity. Thus, the corresponding PC1 time series was used to represent the intensity index of the North Pacific storm track. The variability in the PC1 time series was stronger during 1950–1974 than during 1989–2013, indicating that the amplitude of the North Pacific storm track activity was stronger during 1950–1974 than during 1989–2013, corresponding to the stronger feedback of synoptic-scale eddies to the mean flow measured during the earlier period. Figure 13 shows a scatter plot of the 25-yr moving standard deviations of the North Pacific storm track intensity index against the 25-yr moving standard deviations of the winter WP index. The correlation coefficient between the two variables shown in Fig. 12 reached 0.69, which was significant at the 99.9% confidence level. This result indicates that changes in the intensity of the North Pacific storm track also play a role in modulating changes in the WP intensity.