Simple Metrics for Representing East Asian Winter Monsoon Variability: Urals Blocking and Western Pacific Teleconnection Patterns

The intensity of the East Asian winter monsoon (EAWM) is closely related to severe conditions in densely populated regions in East Asia, including China, Japan, and Korea. A simple and predictable EAWM index (EAWMI) is important for seasonal forecasting and for studying the long-term variations of the EAWM. Based on large-scale circulation features over the EAWM region (Fig. 1), many previous works have constructed different EAWMIs using: (1) the mean SLP (MSLP) gradient between East Asia and the Pacific Ocean; (2) the lower tropospheric wind along the coastal region of East Asia; (3) the East Asian trough; and (4) the upper tropospheric zonal wind shear (Wang et al., 2010a). Yet, the EAWM is a complicated system involving the air-sea interaction between the Asian continent and the Pacific Ocean, and the topographic forcing exerted by the Tibetan Plateau. The interannual variations of surface air temperature (SAT) in the northern and southern parts of the EAWM region are uncorrelated (Wang et al., 2010a). Therefore, it is challenging to construct a unified index that can successfully capture the SAT variations in the entire EAWM region.

Figure 1.  Large-scale circulation features of the EAWM: (a) MSLP (shading; units: hPa) and the 1000-hPa wind (vectors; units: m s$^-1$); and (b) U$_250$ (shading; units: m s$^-1$) and Z$_500$ (contour interval: 60 gpm). In (a), the Tibetan Plateau is colored black, while A, B and C denote the Siberian high, Aleutian low and near-surface northeasterly flow over the East Asian coast, respectively. In (b), D and E denote the East Asian jet stream and East Asian trough, respectively.

The evolution of the EAWM system is related to middle and upper tropospheric remote signals from both upstream and downstream (e.g., Chang and Lau, 1980; Joung and Hitchman, 1982; Lau and Lau, 1984; Hsu and Wallace, 1985; Takaya and Nakamura, 2005, 2013; Song et al., 2014). A typical severe cold-air outbreak in the EAWM region is characterized by an inverted-omega geopotential height pattern. This is composed of two blocking ridges over the Urals-Siberia region and the North Pacific (Yeh et al., 1962; Ding, 1994). A recent study by (Cheung and Zhou, 2015) showed that the long-term variation in the number of cold days (not the year-to-year SAT variability) in Hong Kong, a coastal city in Southeast China, is closely related to the frequency of the Urals blocking pattern index (UBI) and the western Pacific teleconnection pattern index (WPI). In this paper, we will further show how both the UBI and WPI constitute one of the most robust signals accounting for the year-to-year SAT variability in the EAWM region. Unlike most previous works, which tended to focus on the relationship between blocking/teleconnections and EAWM variability (Gong et al., 2001; Lee and Jhun, 2006; Liu et al., 2014; Lim and Kim, 2015), our analysis begins with the SAT in the EAWM region.

Following the definition of (Wang et al., 2010a), the SAT variation in the EAWM region will also be described by the two major temperature modes. The northern temperature mode (NTM) captures the temperature variation north of 40^°N well, whereas the southern temperature mode (STM) successfully depicts the temperature variation south of 40^°N. Compared to 19 EAWMIs and other well-known teleconnection indices, we will demonstrate that the UBI shows the strongest correlation with the NTM, which is poorly captured by most of the EAWMIs. The WPI shows an equally strong correlation with the STM with four EAWMIs. Because the UBI-NTM and WPI-STM correlations exceed 0.7 in magnitude on both interannual and interdecadal timescales, we suggest that Urals blocking (UB) and western Pacific (WP) teleconnection are crucial for EAWM studies. Based on the findings of this study and other recent studies, we will also discuss the key to defining a representative EAWMI.

The present study investigates 66 winters, from 1948/49 to 2013/14, using the NCEP-NCAR reanalysis datasets. These include the daily field of geopotential height (Z), the monthly field of SAT, MSLP, and the zonal and meridional components of wind (U and V). Following this introduction, the teleconnection indices (UBI and WPI) are defined in section 2. Section 3 presents the relationship between these teleconnection indices and the temperature variations in the EAWM region. Section 4 highlights the strength of these teleconnection indices to represent the EAWM variability. The results are summarized and discussed in section 5.

The prominent large-scale circulation features related to the SAT in the EAWM region are depicted by performing correlation analysis for the area-averaged SAT in the EAWM region (20^°-50^°N and 100^°-140^°E; green box in Fig. 2a) with the 500-hPa geopotential height (Z₅₀₀; shading in Fig. 2a). We use the following three steps to highlight the key regions responsible for the year-to-year variability of the SAT in the EAWM region (white contours in Fig. 2a):

• First, we repeat the correlation analysis for each of the grids in the EAWM region (221 grids in total). The output in the analysis is r₁(Λ,φ,i), where Λ [0, 357.5]^°E, \(\phi\in[0,90.0]^\circ\)N and i [1,221], and r₁ is the linear correlation coefficient between Z₅₀₀ and the SAT grid in the EAWM region.

• Second, to identify the regions accounting for a significant fraction of the SAT variability in the EAWM region, we set a threshold of 0.574 for r₁ (in magnitude; corresponding to 33.3% or one-third of the total variance). The threshold (0.574) is much greater than commonly used confidence levels (95% or 99%) with 65 degrees of freedom (where the length of study period is 66 years). The output in the analysis is N₁(Λ,φ,i) (1 if exceeding the threshold and 0 otherwise).

• Third, at each Z₅₀₀ grid point (Λ,φ) we count the percentage of SAT grids in the EAWM region with r₁ exceeding the threshold, where N₁(Λ,φ)=[∑_i=1²²¹N₁(Λ,φ,i)]/(221× 100) (%; white contours in Fig. 2a).

Figure 2.  Correlation between the area-averaged SAT in the EAWM region (green box) and (a) the Z$_500$ grids (shading), (b) the blocking frequency in the NH in the December-January-February period, where the percentage of SAT grids inside the box showing a linear correlation coefficient larger than 0.574 in magnitude with each of the Z$_500$ grids is denoted by the white contours in (a) , and that with blocking frequency is shown in (c).(see text for description)

Figure 3.  Eigenvector of the (a) UB and (b) WP teleconnection pattern. Linear correlation coefficients between SAT and the principal component of (c) UB and (d) WP. In (c, d), only the regions significant at the 95% confidence level are shaded, and the values larger than 0.574 in magnitude (equivalent to one-third of the explained variance) are enclosed by thick contours.

As shown in Fig. 2a, the year-to-year variability of the SAT over the EAWM region is teleconnected with both upstream and downstream signals over the extratropical region. Upstream of the East Asian continent is a tripole pattern, where a low region is centered over the European continent and the Mediterranean Sea (∼ 15^°E), while a high region is centered near the Ural Mountains (∼ 60^°E) and another low region can be found over western Siberia (∼ 90^°-100^°E). This is analogous to the dominant UB pattern (Wang et al., 2010b; Cheung et al., 2012). A significant negative correlation is also found over the Arabian Sea (∼ 20^°N, 60^°E), but further analysis reveals that this signal is significantly correlated with UB (figure not shown). On the other hand, downstream of the East Asian continent is a dipole pattern over the western North Pacific. This resembles the WP teleconnection pattern (Wallace and Gutzler, 1981). The above results suggest that UB and WP account for a significant fraction of the SAT variability in the EAWM region.

The Z₅₀₀ anomaly pattern resembles an inverted omega pattern (Yeh et al., 1962; Ding, 1994), suggesting that lower SAT in the EAWM region is related to the more frequent occurrence of blocking. The area-averaged SAT of the EAWM region is significantly correlated with the blocking frequency over 45^°-90^°E (the Urals sector) and 105^°-130^°E (the WP sector; Fig. 2b). Note that the blocking frequency is defined in the same way as the algorithm listed in (Cheung and Zhou, 2015), which identifies the regions showing a reversal of Z₅₀₀ gradient over the midlatitudes for at least four consecutive days.

To identify the key blocking region accounting for a significant fraction of the SAT variability in the EAWM region, we follow the three steps that constructed the white contours in Fig. 2a. First, we correlate the blocking frequency at each longitude with each of the 221 SAT grids inside the EAWM region. The output of this correlation analysis is r₂(Λ,i), where Λ [0,357.5]^°E and i [1, 221]. Second, we use N₂(Λ,i) to record whether the correlation exceeds the threshold (0.574 in magnitude), where N₂(Λ,i)=1 if exceeding this threshold and N₂(Λ,i)=0 otherwise. Third, for each longitude, we count the percentage of SAT grids in the EAWM region with r₂ exceeding the threshold, where N₂(Λ)=[∑_i=1²²¹N₂(Λ,i)]/(221× 100) (Fig. 2c).

As shown in Fig. 2c, more than 30% of the SAT grids are found to show a linear correlation coefficient larger than 0.574 in magnitude with the UB sector, but much less than 10% of the SAT grids have such a strong correlation with the WP blocking. Therefore, the Urals sector is the major blocking sector strongly linked to SAT variability in the EAWM region.

Accordingly, we deduce that a blocking pattern centered near the Urals and a north-south-oriented dipole over the western North Pacific are two large-scale atmospheric signals related to SAT variability in the EAWM region:

• Signal (1) can be represented by the UBI. This is defined as the first leading EOF pattern obtained from the covariance matrix of Z₅₀₀ enclosing the region (30^°-80^°N, 0^°-120^°E) (Fig. 3a). Each grid of the covariance matrix is weighed by cosφ. The definition is similar to that in (Cheung et al., 2012), but a larger domain is chosen here in order to capture the tripole pattern in Fig. 2a. Unlike in (Cheung and Zhou, 2015), the area-averaged blocking frequency is not adopted here because the EOF pattern can be easily reproduced, and this EOF pattern is closely related to the UB frequency.

Figure 4.  (a, b) Spatial patterns of the first two leading EOF modes (EOF1 and EOF2) of SAT over the EAWM region (0$^\circ$-60$^\circ$N, 100$^\circ$-140$^\circ$E) using the ERA-40 dataset for the period 1957/58-2000/01: (a) NTM; (b) STM (units: $^\circ$C) [redrawn from Wang et al. (2010a, Figs. 3a and c)]. (c, d) Standardized principal component of the NTM/STM (black line) and the time series of the UBI (blue line)/$-$WPI (red line). The linear correlation coefficient between the two time series is shown in the top-right.

• Signal (2) is captured well by the western Pacific index (WPI). This is defined as the first leading EOF obtained from the correlation matrix of the Z₅₀₀ anomalies enclosing the region of (20^°-80^°N, 90^°E-120^°W), which resembles the WP pattern (Wallace and Gutzler, 1981; Cheung and Zhou, 2015). Because the positive sign of the WPI represents a lower Z₅₀₀ over the high-latitude region of the North Pacific, its sign is reversed in Fig. 3b.

After defining the two teleconnection patterns in Figs. 3a and b, their standardized principal component is correlated with the SAT field in order to illustrate their relationship with the EAWM temperatures. As shown in Figs. 3c and d, the two patterns are associated with two distinct temperature patterns in Asia. The UBI is significantly negatively correlated with the SAT over the midlatitude region in Asia (including Siberia, central and northern parts of China, Korea, and Japan), with the strongest negative correlation southeast of Lake Baikal (Fig. 3c). The WPI, on the other hand, is accompanied by a temperature dipole anomaly pattern over the Asia-Pacific region. In particular, the negative phase of the WPI is significantly negatively correlated with the SAT over the coastal region in East Asia and Southeast Asia, including southeastern China, South Korea, southern Japan, and Vietnam (Fig. 3d). In short, the UBI forms a very strong linkage with the SAT north of 40^°N, whereas the WPI is strongly linked to the SAT south of 40^°N in the EAWM region.

Figure 5.  As in Figs. 4c and d but for (a, b) the interannual component ($<$8 years) and (c, d) the interdecadal component ($>$8 years) of the time series.

During the observational period, the interannual and interdecadal variations of the SAT in the EAWM region can be described mainly by two dominant modes (Wang et al., 2010a). Applying an EOF analysis to the covariance matrix of SAT in the EAWM region (0^°-60^°N, 100^°-140^°E), (Wang et al., 2010a) called the first two EOFs the NTM and STM. The spatial patterns of the two temperature modes are given in Figs. 4a and b. Following their definitions, the two temperature modes were obtained using the ERA-40 datasets. Then, the principal component of the two SAT modes was obtained by projecting the two eigenvectors onto the NCEP data, where the climatological mean difference between the NCEP and ERA-40 data was subtracted before the projection. The two modes account for 72% of the total variance, where the NTM (STM) explains well the SAT variability north (south) of 40^°N. The region explained by the UBI (WPI) seems to coincide with that of the NTM (STM).

The UBI-NTM and WPI-STM relationships are revealed in Figs. 4c and d, where their year-to-year correlations are larger than 0.7 in magnitude and the explained variances are greater than 50%. Because the intensity of the EAWM has been found to have undergone a strong decadal variation in the late 1980s (e.g., Jhun and Lee, 2004; Wang et al., 2009b; Wang et al., 2010a), one might wonder if these relationships are still strong on interannual timescales. In other words, the relationship in Figs. 4c and d might be mainly due to the decadal variation. In this regard, we applied an 8-year high-pass (low-pass) Lanczos filter with 21 weights to extract the interannual (interdecadal) component of each time series in Figs. 4c and d (Fig. 5).

As can be seen in Figs. 5a and b, both the UBI-NTM and WPI-STM relationships still have a linear correlation coefficient of larger than 0.7 in magnitude on interannual timescales. On the other hand, NTM and STM undergo a decadal variation around 1986/87, where they change from a predominantly positive to negative phase (Fig. 5c-d). Correspondingly, the SAT in the EAWM region tends to be below normal before this period and above normal afterward (Lee et al., 2013). In the late 2000s, the two temperature modes become less negative due to some severe winters, which is consistent with the strengthening tendency of the EAWM after 2004/05 (Wang and Chen, 2014a). In comparison, both the UBI and WPI can capture the decadal change in the 1980s (dashed colored lines in Figs. 4c and d). The correlation between the interdecadal time series of the UBI and NTM is 0.887, and that between the WPI and STM is 0.883. Apparently, the UB and WP are two large-scale teleconnection patterns that can account for a significant fraction of SAT variability in different parts of the EAWM region on interannual and interdecadal timescales.

To measure how well the UBI and WPI can capture SAT variability in the EAWM region compared to other EAWMIs and teleconnection patterns [Arctic Oscillation (AO), ENSO Niño3 index; Eurasian pattern (EU), Scandinavian pattern (SCAN)], all indices are correlated with the NTM and STM in Figs. 6a-d. The definition of all EAWMIs is given in Table 1, whereas the definitions of the teleconnection patterns are the same as those of the CPC/NOAA (http://www.cpc.ncep.noaa.gov/data/teledoc/telecontents.shtml). Because all climate indices on the CPC/NOAA website are available for 1950 onwards, the correlation analysis is confined to the period 1950/51-2013/14.

Figure 6.  Correlation between EAWMIs and the (a) NTM and (b) STM, where the first column of each plot shows the correlation throughout the entire study period, and the second column shows the 31-year running correlation. (c, d) As in (a, b) but for the correlation with teleconnection indices.

Among the 19 EAWMIs listed in Table 1, only five show a correlation of 0.50 or above in magnitude with the NTM in the entire study period (Fig. 6a). Two of them (WC14 and H15) are new indices, following the review of (Wang et al., 2010a). They consider the meridional MSLP or Z₅₀₀ gradient over East Asia instead of solely the zonal gradient. The remaining three EAWMIs describe either the intensity of the Siberian high (G01) or the East Asian trough (S97 and CS99b). In contrast, only three EAWMIs have a correlation smaller than 0.50 in magnitude with the STM (Fig. 6b). These results are consistent with those of previous studies (e.g., Wang et al., 2010a; Wang and Chen, 2014a), where most of the EAWMIs were found to show a strong correlation with the STM but not the NTM.

As mentioned in (Wang et al., 2010a), a cold NTM is characterized by a northwestward shift in the East Asian trough toward Lake Baikal (∼50^°N, 110^°E) and a northwestward shift in the Siberian high [which is located at (40^°-65^°N, 80^°-120^°E) in the climatology]. A cold STM, on the other hand, is related to a deepening of the East Asian trough and a stronger surface MSLP gradient between the Siberian high and the subtropical western North Pacific. Most of the EAWMIs listed in Table 1 belong to groups 1 and 2 (13 out of 19), where group 1 is defined by the zonal MSLP gradient between the Asian continent and the Pacific Ocean, and group 2 is defined by the lower tropospheric V over the coastal region in East Asia. These EAWMIs describe mainly the large-scale circulation features south of 40^°N. Accordingly, they tend to show a strong correlation with the STM that represents the SAT variability south of 40^°N (Fig. 4b).

Compared to the EAWMIs and other well-known teleconnection indices, the UBI has the strongest correlation with the NTM, whereas the WPI has a correlation equally as strong (∼0.7) as that of four of the EAWMIs with the STM. Such strong year-to-year relationships are stable throughout the study period, as evidenced by slight changes in the 31-year running correlation. The above conclusion is still valid if we repeat the correlation analysis for the interannual and interdecadal component of the indices (same as Fig. 5), except that more EAWMIs show a correlation with the STM greater than 0.7 on the interannual timescales, and H15 shows a stronger correlation with the NTM (0.913) than the UBI (0.888) on the interdecadal timescales (figures not shown). Therefore, both the UBI and WPI are representative for measuring EAWM temperatures. Because many EAWMIs can represent the STM well, the remaining part of this section focuses on the NTM.

The UBI-NTM relationship is related to the linkage between UB and both the Siberian high and the East Asian trough. As can be inferred from Fig. 3a, the positive phase of the UBI corresponds to stronger northerly geostrophic wind from the polar region towards western Siberia. This enhances the cold air intensity (i.e., negative SAT anomaly) over the climatological Siberian high region. Moreover, the stronger meridional-type circulation during the positive phase of the UBI tends to intensify the cyclonic flow over the East Asian continent, which can be deduced by the negative Z₅₀₀ anomaly over East Asia (Fig. 3a). This negative Z₅₀₀ anomaly pattern is analogous to the northwestward shift of the East Asian trough from the south of Japan towards the East Asian continent.

Compared to the UBI, SCAN shows a weaker correlation with the NTM throughout the study period. However, the 31-year moving correlation suggests that the SCAN-NTM relationship has become stronger in recent decades, which is comparable to the UBI-NTM relationship. Previous studies have also identified its impact on the EAWM (Bueh and Nakamura, 2007; Bueh et al., 2011; Sohn et al., 2011; Liu et al., 2014). The difference between SCAN and the UBI arises from their centers of action over Eurasia; those of SCAN are located at a higher latitude. The above results suggest that the large-scale circulations associated with the NTM are characterized by a quasi-stationary Rossby wave-train pattern recurring over Eurasia. This is also related to the location of UB. In future work, we will look to further explain the dynamic mechanisms responsible for the recurring position of blocking and the quasi-stationary Rossby wave train over Eurasia.

Based on recent advances in EAWM studies (e.g., Lee and Jhun, 2006; Wang et al., 2010a; Takaya and Nakamura, 2013; Kim et al., 2014; Liu et al., 2014; Hu et al., 2015; Leung and Zhou, 2015) and the results of this study, an EAWMI should be able to represent the following anomalous large-scale circulation features in order to capture the SAT variability in the majority of the EAWM region:

• The strength of the Siberian high, which is the cold-core system related to cold air activity over the EAWM region;

• The meridional-type circulation (or zonal index) over East Asia, which is related to the meridional displacement of the East Asian trough and the cold-air pathway in the EAWM region.

The former feature is captured well by any index enclosing the climatological Siberian high region (e.g., Hu et al., 2015). This is closely related to the teleconnection pattern centered near the Ural Mountains, which can be described by the UBI [Urals blocking pattern (Cheung et al., 2012)]. The latter feature is closely related to the deepening of the East Asian trough (Wang et al., 2009a; Leung and Zhou, 2015). This resembles the large-scale circulation features associated with the WPI (Lim and Kim, 2013; Takaya and Nakamura, 2013; Wang and Chen, 2014b). Therefore, variability in the EAWM can be explained well by the large-scale circulation features upstream and downstream of the EAWM region.

However, the SAT variability north and south of 40^°N in the EAWM region (NTM and STM) is quite different on interannual timescales. As a result, very few EAWMIs show an equally strong correlation with both the NTM and STM. Among the EAWMIs considered in this study, only CS99b (a trough index) shows a correlation above 0.60 with both the NTM and STM. We have also tried to use a linear combination of the UBI and WPI to represent the NTM and STM, and the optimization is 3× UBI-2× WPI. The correlation between this linear combination and the NTM (STM) is 0.614 (0.626), which is close to that of CS99b. In other words, a unified and representative EAWMI (say, with an explained variance larger than 50% in the entire EAWM region) is not easy to establish. One possible way is to consider the latitude and longitude dependence of SAT on the two important EAWM circulation features mentioned above (i.e., EAWMI is a function of latitude and longitude).

To improve the seasonal forecasting of temperature anomalies in the EAWM and to project future changes in the EAWM, more in-depth analyses should be carried out to explore the underlying dynamics regarding the large-scale teleconnections related to the EAWM, such as the UB and WP. The WP is related to ENSO, which has been discussed extensively in previous studies (e.g., Li, 1990; Wang et al., 2000; Zhou et al., 2007; Wang et al., 2008; Takaya and Nakamura, 2013; Chen et al., 2014). On the other hand, UB has been studied less extensively.

On synoptic timescales, UB is an important dynamic precursor of severe cold-air outbreaks in the EAWM. The occurrence of UB strengthens cold-air advection from the polar region to western Siberia. This enhances the Siberian high and subsequent cold-air outbreaks in the EAWM (Tao, 1957; Takaya and Nakamura, 2005; Lu and Chang, 2009; Zhou et al., 2009; Cheung et al., 2013). On seasonal timescales, when UB occurs frequently in a particular winter, the Siberian high intensity tends to be higher than normal. Because of the close linkage between the Siberian high intensity and the NTM, the temperature tends to be lower in the northern part of the EAWM (Wang et al., 2010b; Chang and Lu, 2012; Cheung et al., 2012). Moreover, the occurrence of UB is also characterized by a stronger meridional-type flow over East Asia. The more frequent occurrence of UB partly enhances the southward intrusion of cold air into the southern part of the EAWM. This potentially causes more frequent cold extremes in the EAWM, as in Hong Kong (Cheung and Zhou, 2015). Therefore, the dynamics of UB should be investigated in depth in order to improve the understanding of the EAWM on different timescales. In addition to the major dynamic factors that lead to the occurrence of UB, we should understand the factors related to the recurring location of UB.