1. Introduction

The East Asian summer rainfall exhibits significant interannual variability. It has been known for many years that the rainfall anomaly in central-eastern China and Japan tends to be out of phase with the rainfall anomalies in northern and southern China (e.g., Lau 1992; Tian and Yasunari 1992; Shen and Lau 1995; Nigam 1994; Weng et al. 1999; Hsu and Liu 2003). This out-of-phase relationship reflects the tripole structure in the East Asian summer rainfall distribution. Studies have been carried out to document the characteristics of this tripole pattern and to propose mechanisms responsible for the existence of the pattern. In addition to the notable relationship with the El Niño (e.g., Tian and Yasunari 1992; Weng et al. 1999; Lau and Weng 2001), one of the frequently cited mechanisms is the influence of the Pacific–Japan (PJ) pattern (Nitta 1987; Lau 1992; Huang and Sun 1992; Nitta and Hu 1996; Kosaka and Nakamura 2006), which is forced by the anomalous heating over the anomalously warm sea surface temperature (SST) in the Philippine Sea. Statistically significant correlation is also found between the vertically integrated heating above the Tibetan Plateau and the tripole pattern (Liu et al. 2002; Hsu and Liu 2003). The stronger (weaker) than normal heating over the Tibetan Plateau is associated with the above (below) normal rainfall in central-eastern China. It demonstrates the potential heating effect of the Tibetan Plateau on the interannual variability of East Asian summer rainfall.

While the heating anomaly over the Philippine Sea was considered the major forcing for the PJ pattern, a study by Enomoto et al. (2003) pointed out the existence of other forcing, based on numerical simulation results. The Bonin high, which significantly affects the summer climate in Japan, is part of the Rossby wave–like perturbation forced by the diabatic heating anomaly in the Eurasian continent, instead of the anomalous heating in the Philippine Sea. They proposed that the Rossby wave energy disperses along the waveguide near the jet stream across the Eurasian continent to East Asia and results in the formation of the Bonin high. This wavelike perturbation, named the Silk Road pattern, is consistent with the study by Ambrizzi et al. (1995) on the Rossby wave energy propagation during the northern summer. Guan and Yamagata (2003) related the 1994 severe drought in Japan, which was characterized by an anomalous anticyclonic circulation over the Okhotsk Sea, to the Indian Ocean dipole. They suggested that the anomalous convection associated with the Indian Ocean dipole initiates perturbations near the jet stream, which in turn triggers the downstream Rossby wave energy propagation and results in the anticyclonic anomaly over the Okhotsk Sea and the 1994 dry/hot summer in Japan.

These cited studies can be roughly classified into two types: oceanic/tropical and continental/extratropical origins. Apparently, the tripole pattern is related to the wavelike structure that can be originated from either the oceanic forcing in the Philippine Sea or forcing over the Eurasian continent and the Tibetan Plateau. There is a possibility that forcing of both origins can initiate the tripole pattern. In view of the above information, the mechanism and origin of the tripole pattern is not completely understood. It is therefore worthwhile to further explore the nature of the tripole pattern, which has a significant influence on the East Asian summer climate.

In the course of this study, we identified the distinct characteristics in the positive and negative phases of the tripole pattern, which was unnoticed in previous studies. The arrangement of this article is as follows. The second section describes the data and methodology. The third and fourth sections present the major characteristics of the tripole pattern and the corresponding wave activity flux in both positive and negative phases. The associated sea surface temperature anomaly (SSTA) in both phases and the related vertical overturning circulation are shown in section 5. Conclusions and discussion are given in section 6.

2. Data and methodology

The data used in this study include the monthly data from 1) the 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40; Uppala et al. 2005), 2.5° × 2.5° resolution; 2) precipitation from the Climatic Research Unit (CRU; Mitchell et al. 2004), 0.5° × 0.5° resolution; and 3) the Global Sea Ice and Sea Surface Temperature dataset (GISST; Rayner et al. 1996), 1° × 1° resolution. To be consistent with the ERA-40, only summer (June–August) data from 1958 to 2001 were used.

Since there exists a significant trend in the tripole pattern and the interannual variability is the major focus in this study, the linear trend at each grid point was derived based on linear regression and then subtracted. The 9-yr running means were then computed and subtracted from the detrended data to exclude the decadal–interdecadal signals and mostly retain the interannual variability. At the end of this preprocess, only the summer-mean data from 1962 to 1997 were retained for further analysis.

3. Structure of the tripole pattern and the corresponding circulation

The empirical orthogonal analysis (EOF) was applied to the CRU precipitation data to retrieve the most recurrent pattern in East Asia. The first EOF (Fig. 1a), which explains 17.6% of total variance, clearly exhibits the well-known tripole structure as reported in previous studies (e.g., Weng et al. 1999; Chang et al. 2000a, b; Hsu and Liu 2003), with positive polarity in central-eastern China, South Korea, and Japan, and negative polarity in southern China–Taiwan and northern China. The positive and negative anomalies are elongated in the east–west direction and are relatively narrow in the meridional direction. This banded structure is a distinctive characteristic of the East Asian monsoon rainfall in the boreal summer. The corresponding time series (i.e., principal component; Fig. 1b) fluctuates from year to year, indicating significant interannual variability.

For the compositing purpose to reveal the characteristics of this tripole rainfall pattern, years with PC1 greater than 1 and less than −1 standard deviation were identified as positive and negative phases, respectively. Six summers (1963, 1965, 1969, 1980, 1991, and 1993) and seven summers (1966, 1967, 1973, 1978, 1990, 1992, and 1994) were chosen for the positive and negative phase, respectively. Composites of various variables were constructed and evaluated using Student’s t-test significance, based on the null hypothesis, to identify the statistically significant features at the 0.01, 0.05, and 0.1 levels. To further examine the reliability of the composite results, plots for each individual case were examined and similarity was found among all cases. Composites were also made by lowering the case-selection criterion to 0.8 standard deviation, yielding nine cases for both positive and negative phases. Great similarity was found between the composites based on 1 and 0.8 standard deviation. The results of these two further tests, which are not shown here, are consistent with the results presented in the following sections.

The rainfall and 850-hPa wind anomalies in the positive and negative phase are presented in Figs. 2a,b, respectively. During the positive phase when the rainfall is significantly above normal in central-eastern China, South Korea, and Japan, an anticyclonic anomaly is evident in the western North Pacific between 20° and 35°N. This anomalous circulation indicates stronger-than-normal southwesterly and westerly flow in southern China and the region south of Japan, respectively. More moisture is transported to central-eastern China and Japan (figures not shown), and results in larger rainfall. This feature also corresponds to a southwestward shift of the Pacific subtropical anticyclone and a weak trough to the north of this anticyclone. The former corresponds to the negative rainfall anomaly in southern China and Taiwan, while the latter corresponds to the positive rainfall anomaly in central eastern China and Japan.

In the negative phase, when a negative rainfall anomaly appears in central-eastern China and Japan, a cyclonic anomaly is observed in the western North Pacific between 20° and 35°N, corresponding to the weaker southwesterly and westerly in southern China and to the south of Japan, respectively. This indicates less moisture transport to central-eastern China, South Korea, and Japan, and therefore less rainfall (figures not shown). This feature corresponds to a northward shift of the Pacific subtropical anticyclone. Conversely, more rainfall occurs in southern China and Taiwan, corresponding to a stronger-than-normal moisture convergence with the monsoon trough extending from the South China Sea to the Philippine Sea. A comparison between the rainfall anomalies in the positive and negative phase reveals an interesting feature: a better organized tripole structure in the positive phase than in the negative phase. This distinct contrast is also observed in the corresponding heating and circulation anomalies shown below.

Since the rainfall data are available only over land, the apparent heat source (Q1) and apparent moisture sink (Q2) were calculated, based on Yanai et al. (1973), to represent the fluctuation of diabatic heating over both ocean and land. Overall, the Q1 and Q2 exhibit similar distributions in East Asia and the western North Pacific, indicating the dominance of latent heat release in the diabatic heating. Composites of vertically integrated Q1 in the positive and negative phase are shown in Fig. 3. A comparison between the Q1 and rainfall anomaly patterns indicates a great similarity between the two fields, reflecting the dominance of the convection fluctuation in the region. The tripole pattern is also evident in Q1. For example, in the positive phase (Fig. 3a), the elongated belt of the positive Q1 anomaly crossing central China, the East China Sea, South Korea, and Japan corresponds to the positive rainfall anomaly in the region. The negative Q1 anomalies in northern China and in the region covering southern China, Taiwan, and the Philippine Sea correspond to the negative rainfall anomalies. Similar correspondence can also be seen during the negative phase (Fig. 3b). The contrast in Q1 between the positive and negative phase is also evident, as in the rainfall. The Q1 in the positive phase exhibits a stronger and more clearly defined north–south wavelike structure than in the negative phase. For example, the negative Q1 anomalies in the northern and southern China in the positive phase are better defined than the positive anomalies in the same areas during the negative phase. There are also significant Q1 anomalies in the eastern Pacific, which will be discussed later.

To reveal its association with the large-scale circulation pattern, the corresponding composites of the 850- and 200-hPa relative vorticity anomalies are shown in Fig. 4. In the positive phase, a north–south wavelike pattern, with each element elongated zonally, is evident at 850 hPa from the equator to the Okhotsk Sea (Fig. 4a). A similar pattern is also evident at 200 hPa, except with a much larger east–west scale south of 30°N (Fig. 4b). Note the resemblance of this structure to the PJ pattern. The 200-hPa anomalies along the East Asian coast are located slightly to the north of their counterpart at 850 hPa, consistent with Kosaka and Nakamura (2006). These results indicate that the cyclonic and anticyclonic anomalies seen in Fig. 2 are simply one element of the large-scale wavelike pattern, which resembles the PJ pattern. The other wavelike pattern, which is more evident at 200 than at 850 hPa, exists over the Eurasian continent.

In the negative phase, a weaker wavelike pattern at 850 hPa is observed along the East Asian coast (Fig. 4c). Its north–south extension, restricted between 10° and 45°N, appears much smaller than its counterpart in the positive phase. A north–south wavelike pattern, which is weaker and less coherent than its counterpart in the positive phase, is also found at 200 hPa (Fig. 4d). On the other hand, the east–west wavelike pattern at 200 hPa, crossing the Eurasian continent between 30° and 50°N, is better organized and stronger in the negative phase than in the positive phase.

4. Wave activity

The comparison between the positive and negative phase presented above reveals an interesting contrast between the two phases. The north–south wavelike pattern along the East Asian coast is more active in the positive phase, while the east–west wavelike pattern over the Eurasian continent is more active in the negative phase. To reveal this contrast more clearly, the wave activity flux (WAF) based on Takaya and Nakamura (1997, 2001) was calculated for both the positive and negative phases. The WAF is designed in the quasigeostrophic framework to represent the energy propagation of Rossby wave–like perturbation, and has been used as a powerful diagnostic tool to identify the origin and energy propagation of the observed large-scale circulation perturbation (e.g., Honda et al. 1999). It has been shown in previous studies that the quasigeostrophic flow is still a dominant component in the tropical large-scale perturbation (e.g., Matsuno 1966; Webster 1972; Gill 1980). Their findings suggest that, although the WAF is derived based on the quasigeostrophic framework, it is still a useful diagnostic tool to examine the wave activity associated with the low-frequency and large-scale wavelike patterns shown in Fig. 4.

The 850-hPa WAF in the positive phase, presented in Fig. 5a, clearly shows the northward-pointing vectors from the equator to 40°N within the 120°–150°E longitudinal band, where the north–south wavelike perturbation is most evident. There is also an indication of the WAF pointing northeastward in the extratropical western North Pacific. This WAF distribution suggests a Rossby wave–like energy propagation from the tropical western Pacific to the extratropical western North Pacific along the East Asian coast, a signature reported in Nitta (1987). The counterpart in the negative phase is presented in Fig. 5c. Although there are also northward-pointing arrows along the East Asian coast, these arrows are confined in a much narrower region between 30° and 40°N. The WAF south of 20°N is either very weak or not pointing consistently northward, and is therefore not shown because of the plotting scheme that was designed to ignore small WAF and to emphasize the more significant features. This contrast between the WAF in the positive and negative phase confirms the above observation that the northward-propagating wave activity at 850 hPa along the East Asian coast is much coherent and stronger in the positive phase. It is also clear that the WAF originates in the Tropics in the positive phase, but in the subtropical western North Pacific in the negative phase. This suggests the stronger tropical influence on extratropical East Asia in the positive phase.

It would be interesting to see whether such a contrast exists in the upper troposphere. Figure 5b presents the 200-hPa WAF in the positive phase. It is interesting to note that the northward propagation of wave activity is not seen along the East Asian coast. Instead, the southward-pointing arrows suggest southward energy propagation, which is connected to the wave activity flux pointing from the Asian continent to the Pacific. This southward WAF is likely due to the advective effect of the background northerly, which is included in the definition of WAF. However, the westward-pointing arrows near 40°N in the 80°–100°E region do not support the notion that the wave activity propagates across the Eurasian continent. It follows that these activities are originated in the Asian continent near the coast. On the contrary, the 200-hPa WAF in the negative phase (Fig. 5d) clearly indicates a more consistent eastward propagation of wave activity across the Eurasian continent to the East Asian coast and turns southeastward to the western North Pacific. The eastward-propagating wave activity in the negative phase is confined in a narrower region over the Eurasian continent and exhibits more coherent propagation tendency than the counterpart in the positive phase. This contrast between the 200-hPa WAF in the positive and negative phase are consistent with the observation in Fig. 4 that the east–west wavelike pattern is better organized in the negative phase than in the positive phase.

The above results reveal the marked contrasts not only between the two phases but also between 850 and 200 hPa. To reveal these contrasts more clearly, vertical cross sections of these wavelike perturbations averaged over 35°–45°N and 120°–150°E are examined. The 120°–150°E cross section of vorticity anomaly and WAF for the positive phase is shown in Fig. 6a. The vorticity anomaly exhibits a wavelike pattern in the whole meridional domain. At a brief glance, one tends to have an impression that the wavelike pattern tends to tilt slightly northward in the vertical, exhibiting an equivalent barotropic structure. However, a closer examination reveals the existence of two wavelike patterns, one in the upper troposphere and another one in the lower troposphere. For example, there are two separated vorticity maxima between 5° and 20°N: one in the middle and lower troposphere near 10°N and another one in the upper troposphere near 15°N. These two positive vorticity anomalies in the upper and lower troposphere seem decoupled. The negative vorticity anomaly between 20° and 35°N also indicates a secondary maximum in the lower troposphere, which has been confirmed by plotting normalized anomaly (not shown here). By examining the shading that indicates the statistically significant region, one can also identify two wavelike patterns, one in the lower troposphere from the equator to 60°N and another one in the upper troposphere between 10° and 60°N. This observation suggests the coexistence of these two wavelike patterns in the lower and upper troposphere along the East Asian coast. This interpretation is supported by the WAF distribution. In the upper troposphere, the southward-pointing WAF indicates a southward propagation of wave activity from extratropical to subtropical East Asia. The northward-pointing WAF in the lower troposphere suggests the wave activity propagation from the Tropics to the midlatitude along the East Asian coast. This result is consistent with recent finding by Kosaka and Nakamura (2006) and implies the inadequacy to explain the PJ-like pattern in terms of the barotropic Rossby wave propagation forced by the tropical heating (e.g., Nitta 1987; Huang and Sun 1992; Kurihara and Tsuyuki 1987).

The cross section of Q1 and the p-coordinate vertical velocity are shown in Fig. 6b. The major regions of upward and downward motion coincide with the major heating and cooling anomalies, reflecting the dominance of the convective latent heating. This interpretation is supported by the fact that the Q2 profile tends to be located slightly lower than the Q1 profile (not shown; Yanai et al. 1973). A comparison between the upper-tropospheric vorticity and Q1 pattern reveals an interesting phase relationship between the two fields. In the region north of 15°N, the Q1 and vorticity tend to be in quadrature with each other in such a manner that the cooling (heating) anomaly tends to be located to the south of the negative (positive) vorticity anomaly. This phase relationship is consistent with a Rossby wave–like activity propagating from north to south, which will be elaborated on later. However, the Q1 near 10°N is located right over the positive vorticity anomaly in the middle/lower troposphere and under the near-zero or weak negative anomaly in the upper troposphere. This phase relationship, which is distinctively different from its counterpart in the higher latitude, resembles the local atmospheric response to a deep heating (e.g., vertical stretching resulting in positive vorticity anomaly in the lower troposphere and negative vorticity anomaly in the upper troposphere). Since the upper troposphere near 10°N is also under the influence of southward-propagating wave activity, the cancellation between the positive anomaly related to this wave activity and the negative anomaly associated with local heating likely leads to the near-zero vorticity anomaly in the upper troposphere near 10°N.

The situation is different in the negative phase. The cross section presented in Fig. 6c reveals a wavelike structure that is largely confined to the north of 20°N with the maximum amplitude in the upper troposphere. In the region between 10° and 20°N, the perturbation appears only above 500 hPa. The maximum amplitude of this pattern in the upper troposphere is consistent with the southward-pointing WAF predominantly in the upper troposphere. In contrast to its positive phase counterpart, the vorticity anomaly in the lower troposphere to the south of 20°N is not observed. A plot of a normalized vorticity anomaly did not indicate the existence of local maxima in the lower troposphere, either (not shown). This characteristic is supported by the barely observed northward-pointing WAF in Fig. 6c and the lack of the Q1 anomalies in the Tropics in Fig. 6d. These results suggest that the tripole rainfall pattern in the negative phase is predominantly associated with the southward-propagating wave activity that has an extratropical and continental origin.

Distinct contrasts between the positive and negative phase are also observed over the Eurasian continent as seen in Figs. 4 and 5. To further explore these contrasts, we examine the vertical cross section of various variables averaged over 35°–45°N. Figure 7 presents the cross section of vorticity, WAF, vertical velocity, and Q1. In the positive phase, two positive vorticity anomalies with maximum amplitude in the upper troposphere are observed in 50°–90°E and over the East Asian coast, respectively. The former one is restricted mainly in the upper troposphere, while the latter one exhibits deep vertical structure with the amplitude increasing and the vertical axis titling westward with height in the whole troposphere. In contrast, the negative vorticity anomaly between the two positive anomalies is very weak. WAF, which is predominantly located in the upper troposphere, does not seem to indicate a consistent eastward-propagating wave activity. The corresponding Q1 and vertical velocity (Fig. 7e), which are much weaker than their counterparts in 120°–150°E, indicate that the positive vorticity anomalies tend to be associated with positive Q1 anomalies and anomalous upward motion.

In the negative phase, the wavelike perturbation in the upper troposphere is evident from 50° to 150°E (Fig. 7b). The feature is similar to its counterpart in the positive phase, except the much clearer evidence of the positive vorticity anomaly over the eastern Tibetan Plateau. Being consistent with this feature, WAF indicates a more consistent eastward-propagating wave activity than in the positive phase. The Q1 and vertical velocity shown in Fig. 7f are also better organized than in the positive phase. The positive (negative) Q1 anomaly and anomalously upward (downward) motion are found located to the east of the positive (negative) vorticity anomalies. This phase relationship is consistent with an eastward-propagating Rossby wave activity, which will be elaborated on later.

It is interesting to note the existence of the statistically significant negative (positive) vorticity anomaly near the surface of the eastern Tibetan Plateau in the positive (negative) phases. These features can be seen clearly in the normalized vorticity anomaly field shown in Figs. 7c,d. Note that the strength of these anomalies in the positive and negative phase is about the same, a characteristic different from the asymmetry in the strength of anomaly (between the positive and negative phase) in many other regions as discussed above. The negative and positive vorticity anomalies are situated under negative and positive Q1 anomalies (and anomalously downward and upward motion), respectively, in the positive and negative phase. This relationship is consistent with the notion that the vertical squashing and stretching result in the near-surface negative and positive vorticity anomalies, respectively. This result seems to suggest that the heating anomalies over the eastern Tibetan Plateau are able to trigger large-scale wavelike perturbation and contribute to the formation of the tripole pattern as suggested by previous studies (e.g., Wu et al. 2002, Hsu and Liu 2003). The observation of the westward tilting of the vorticity anomalies with height in the lower troposphere east of 100°E and the eastward-pointing WAF in 120°–140°E are in agreement with this interpretation.

As discussed above, the wavelike anomaly pattern may be interpreted as stationary Rossby wave, which are likely the result of the balance between the perturbation vorticity advection by mean flow, the beta effect, and the vertical stretching due to heating and cooling. These three are recognized as the dominant terms in the vorticity equation for the large-scale circulation (e.g., Hoskins and Karoly 1981). Detailed vorticity budget calculation is needed to understand the exact way of balance. Since it is not the major focus of the present study, such a quantitative calculation is not performed in this study. Nevertheless, an intuitive and qualitative interpretation seems useful and is proposed as follows.

Over the extratropical Eurasian continent, the prevailing flow in the upper troposphere is westerly. The anticyclonic vorticity tendency in the upper troposphere, produced by the heating over the eastern Tibetan Plateau (i.e., 100°–120°E in Figs. 7b,f), is likely balanced by the eastward advection of positive vorticity anomaly by the mean westerly, and the northward advection of smaller planetary vorticity by the anomalous southerly. A balance in the upper troposphere near 80°E can also be understood in the similar way.

The north–south wavelike pattern along the East Asian coast can be understood as follows. In the upper troposphere, the mean zonal flow is westerly to the north of 30°N and the mean meridional flow is northerly to the south of 40°N. The anticyclonic vorticity tendency in the upper troposphere, produced by the heating between 30°–40°N (e.g., Figs. 6a,b), is likely balanced by the advection of positive vorticity anomaly by the mean westerly and northerly flow, because the positive vorticity anomaly is located to the northwest of the positive Q1 (Figs. 3a and 4b). In the region near 20°N where the prevailing flow is northeasterly, the southward advection of the negative vorticity anomaly by the mean northerly mean flow is likely balanced by the stretching effect due to the cooling (or subsidence). The beta effect is likely small due to the zonal elongation of the vorticity anomalies.

5. Sea surface temperature and circulation

The results presented above suggest the existence of two wavelike patterns of different origins. It would be interesting to understand their relationship with the SSTA, especially in the positive phase when the wavelike pattern of tropical origin along the East Asian coast is most active. Shown in Fig. 8 are the SSTA patterns from the preceding spring to the following winter of the chosen summers. It is clearly shown that the positive phase of the tripole pattern is associated with the development of the positive SSTA in the equatorial eastern Pacific, which appears in the preceding spring and expands westward in the following seasons until the following winter (Figs. 8a,c,e,g). On the contrary, weaker and less persistent negative SSTA are observed in the equatorial eastern Pacific in the negative phase (Figs. 8b,d,f,h). Note that the negative SSTA becomes very weak and statistically insignificant in the following winter.

In contrast, there are no obvious SSTAs in the tropical western Pacific in both positive and negative phases. However, an investigation of Q1 and Q2 reveals significant anomalies in the tropical western Pacific as shown in Fig. 3. In the positive phase (Fig. 3a), a wavelike pattern of Q1 with each center elongated in the east–west direction is clearly evident in the western North Pacific. This north–south pattern corresponds well to the perturbation shown in Figs. 6a,b. In the negative phase (Fig. 3b), a wavelike pattern also appears in the same region, but with weaker amplitude and less coherent structure. This contrast between the positive and negative phases is particularly evident in the tropical region. For example, a pair of positive and negative Q1 anomalies straddling the equator in the western Pacific is observed in the positive phase, while no significant anomalies are observed in the negative phase. The lack of significant SSTA in the tropical western Pacific, where significant Q1 anomalies are observed in the positive phase, suggests that the corresponding wavelike pattern is not directly forced by the anomalous sea surface temperature in the tropical western Pacific.

Conversely, a north–south pair of positive and negative Q1 anomalies, which are zonally elongated and meridionally narrow, is clearly evident in the tropical eastern Pacific in both positive and negative phases. In the positive phase, the positive Q1 anomaly near the equator is located over the positive SSTA while the negative Q1 anomaly is located to the north of it. This feature seems to reflect the southward shift of the intertropical convergence zone (ITCZ) in the eastern Pacific, which often occurs with a positive SSTA in the equatorial eastern Pacific. A reversed situation (e.g., negative Q1 anomaly over negative SSTA and positive Q1 anomaly to the north of it) occurs in the negative phase. This relationship reveals the effect of SST on tropical convection as in the El Niño and La Niña conditions (Rasmusson and Carpenter 1982), although the chosen cases are not necessarily classified as El Niño or La Niña.

To further explore this indirect effect, vertical cross sections of Q1 and vertical velocity averaged between 5°N and 10°S, a latitudinal belt over the positive and negative SSTA in the positive and negative phases, are shown in Fig. 9. In the positive phase (Fig. 9a), a positive Q1 anomaly and anomalously upward motion are observed in the eastern Pacific (e.g., 150°–120°W) where the positive SSTA exists, while negative Q1 anomaly and anomalously downward motion are found in the western Pacific between 100° and 140°E, where no significant SSTA is observed. A zonally oriented overturning circulation (e.g., Walker-type circulation), with upward motion in the east and downward motion in the west, seems to exist in this narrow latitudinal band. These results suggest that the negative Q1 and downward vertical velocity anomalies in the equatorial western Pacific could be induced by the zonally overturning circulation, which are driven by the positive Q1 anomalies above the positive SSTA in the equatorial eastern Pacific. The anomalies in the tropical western Pacific in turn induce the meridionally overturning circulation and the north–south wavelike pattern in the western North Pacific, which emanates poleward along the East Asian coast.

One may suspect whether a zonally overturning circulation anomaly (but in the reversed direction) also exits, associated with the negative Q1 anomaly near 10°N in the eastern Pacific. An examination of the cross section along the latitudinal band between 7.5° and 15°N (e.g., crossing the negative Q1 in the eastern Pacific and the positive Q1 in the western Pacific) does not clearly reveal a zonally overturning circulation anomaly in the reversed direction (not shown). This suggests that the negative Q1 anomaly does not seem to have a similar impact on the western Pacific like the positive Q1 anomaly near the equator.

Similar features in the negative phase appear to be less significant and coherent. As shown in Fig. 9b, negative Q1 anomaly and anomalously downward motion are observed in the eastern Pacific, and positive Q1 anomaly and anomalously upward motion are found in the western Pacific. However, the positive Q1 anomalies in the western Pacific are not statistically significant and appear in three individual centers separated by negative Q1 anomalies. A zonally overturning circulation connecting the equatorial eastern and western Pacific is not well organized, either. This result seems to suggest that the heating anomalies in the eastern Pacific have a much weaker impact on the tropical western Pacific to induce a meridionally aligned wavelike pattern in the negative phase.

The observed positive-correlation relationship between the Q1 and SSTA in the tropical eastern Pacific does not exist in the extratropical western North Pacific. Instead, a negative-correlation relationship is observed, that is positive (negative) Q1 anomaly over negative (positive) SSTA. Furthermore, the tripole rainfall pattern in the positive (negative) phase tends to lead the negative (positive) SSTA in the extratropical western North Pacific. The positive and negative phases of the tripole pattern correspond to the positive and negative heating anomalies (implying more and less rainfall, respectively), respectively, in central-eastern China, Japan, and over the ocean to the east of Japan. The SST anomalies are very weak and appear in limited areas in the previous spring, and reach the maximum strength in the autumn. It is speculated that the out-of-phase relationship and the temporal phase lag is the result of atmospheric influence on the ocean surface. This interesting phenomenon deserves further exploration but is beyond the scope of the present study.

6. Conclusions and discussion

This study investigates the characteristics of the tripole rainfall pattern in East Asia during the northern summer. The positive phase corresponds to the positive rainfall anomaly in central-eastern China, Korea, and Japan, and the negative rainfall anomalies to the north (e.g., northern China) and south (southern China and Taiwan). The negative phase exhibits the same pattern but in opposite signs. This anomalous rainfall pattern exhibits a banded structure (i.e., zonally elongated in relatively narrow latitudinal bands). Two wavelike circulation patterns are found associated with the tripole pattern. One aligns meridionally in the western North Pacific along the East Asian coast, resembling the PJ pattern (Nitta 1987). The other one aligns zonally over the Eurasian continent in the extratropical region between 30° and 50°N, resembling the Silk Road pattern (i.e., Ambrizzi et al. 1995; Enomoto et al. 2003). These wavelike patterns exhibit the equivalent barotropic vertical structure titling slightly either northward with height over the western North Pacific or westward over the Eurasian continent, with amplitude increasing with height.