The leading mode of boreal winter precipitation variability over the tropical Pacific for the period 1980–2010 shows a west–east dipole pattern with one center over the western North Pacific (WNP) and Maritime Continent and the other center over the equatorial central Pacific (CP). Observational evidence shows that the variability of the East Asian upper-tropospheric subtropical westerly jet (EAJ) has a significant correlation with precipitation anomalies over the WNP and CP and that tropical precipitation anomalies over WNP and CP have a distinct influence on the variation of the EAJ. A series of numerical experiments based on a linear baroclinic model are performed to confirm the influence of the heating anomalies associated with precipitation perturbations over the WNP and CP on the EAJ. The results of numerical experiments indicate that a heat source over the WNP can excite a northward-propagating Rossby wave train in the upper troposphere over East Asia and facilitate a poleward eddy momentum flux. It results in the acceleration of the westerlies between 30° and 45°N, which favors a northward displacement of the EAJ. The response induced by a heat sink over the CP features a zonal easterly band between 25° and 40°N, suggesting that the response to heat sink associated with negative precipitation anomalies over the CP may weaken the EAJ. A strengthened relationship was found between tropical Pacific precipitation and the EAJ since 1979. The modeling results suggest that the shift of mean states might be responsible for the strengthened EAJ–rainfall relationship after 1979.
Based on observational and theoretical analyses over the past decades, it has become clear that the diabatic heating anomaly associated with tropical convection has pronounced influences not only on the tropics but also on the midlatitudes (Hoskins and Karoly 1981; Jin and Hoskins 1995; Kummerow et al. 2000; McBride et al. 2003; Wu et al.2012). Many investigations have been undertaken in order to understand the role of tropical heating with regard to the variability of midlatitude circulations (e.g., Wallace and Gutzler 1981; Nitta 1987; Zheng et al. 2013). Observational studies have found that anomalous tropical precipitation has a significant impact on extratropical atmospheric circulation changes through Rossby wave propagation (e.g., Hoskins and Ambrizzi 1993; Webster and Chang 1998; Yoo et al. 2012), which has been documented as an important linkage between the circulations at midlatitude and tropical convection. Some efforts have been made to understand the relationship between the circulation change at the middle-to-high latitudes and anomalous convection over the tropical Indian and Pacific Oceans (Zheng et al. 2013), the Maritime Continent (Lau and Chan 1983a,b), the western North Pacific (WNP; Lu 2001; Wang 2002; Zheng et al. 2013; Park and An 2014; Jia et al. 2015), central Pacific (CP; Wallace and Gutzler 1981), and subtropical East Asia (Kwon et al. 2007; Lu and Lin 2009). In particular, convection over the WNP has been claimed to hold the key for the variability of atmospheric circulation over East Asia in some previous studies. For instance, Park and An (2014) found that the tropical WNP convection significantly influences the North Pacific atmospheric circulation in boreal winter by modulating the jet stream.
Numerical models have also been applied to understand whether anomalous precipitation can significantly influence the midlatitude circulations in various time scales. On the intraseasonal time scale, Huang and Lu (1989) demonstrated that the enhanced convective activities over the WNP could lead to the intensified variation of subtropical high over East Asia by using an atmospheric general circulation model (GCM). On the interannual time scale, Lu and Lin (2009) indicated that tropical precipitation anomalies, compared with subtropical precipitation anomalies, may be relatively trivial for the variation of large-scale meridional teleconnection over East Asia via a simple linear model. On the decadal time scale, Zheng et al. (2013) examined the main mode of rainfall variations over the Indo-western Pacific. They suggested that the stronger rainfall mode after the climate regime shift of the 1970s may lead to the decadal shift of atmospheric circulation over East Asia based on the results obtained from a simple two-level model.
Extensive investigations have been undertaken in order to figure out the interaction between anomalous precipitation and upper-tropospheric circulations over the midlatitudes. According to current knowledge, anomalous precipitation can induce poleward-propagating Rossby waves (Hoskins and Karoly 1981; Chen and Trenberth 1988; Lee et al. 2011a,b). Lau et al. (2000) suggested that the westerly jet over East Asia plays a quite important role in linking tropical heating to extratropical circulation systems. Thus, it is requisite to devote attention to the interaction between tropical precipitation and the East Asian upper-tropospheric subtropical westerly jet (EAJ). Many studies have focus on the interaction between the EAJ and precipitation over tropical and subtropical regions (e.g., Yang and Webster 1990; Li and Zhang 2014). On the one hand, the change of EAJ has a pronounced impact on the variation of precipitation. The variations of the EAJ could significantly contribute to rainfall anomalies over East Asia via influencing the intensity of the South Asian high in summer on interannual time scales (Liao et al. 2004). On the other hand, some studies have investigated the impact of precipitation variation on the westerly jet. For instance, Yang and Webster (1990) noticed that the heating field in the adjacent hemisphere is important in determining the location and magnitude of the EAJ in boreal summer. On decadal time scales, Kwon et al. (2007) showed that the decadal decrease of the zonal wind speed over East Asia could be understood as a barotropic response to the heating associated with increased precipitation in the southeastern part of China after the mid-1990s in summer.
Although some efforts have been made to understand the influence of precipitation variation over the tropics on the westerly jet over the midlatitudes, relatively few studies have focused on the impact of tropical precipitation change on the EAJ in boreal winter, as well as the relative contributions of tropical rainfall over different regions to the EAJ variation. Thus, our focus here is on how anomalous heating associated with tropical precipitation anomalies over different regions affects the EAJ during boreal winter. We also devote our attention to the investigation of the relative contribution of tropical precipitation over the WNP and CP to the midlatitude zonal circulation change at upper levels. The possible mechanism responsible for the influences of winter precipitation variability over different tropical regions on the variation of the EAJ is one of the questions we attempt to address here. It has long been known that tropical rainfall exhibits a very strong interannual variation associated with El Niño–Southern Oscillation (ENSO) (e.g., Chou et al. 2003; Chou et al. 2009). Considerable attention has been focused on extratropical responses to the forcing of sea surface temperature (SST) rather than the associated convection (e.g., Webster 1982; Hoerling et al. 1997; Wu et al. 2009; Chen et al. 2014). However, the SST–convection relation varies with the seasons and regions (Lau et al. 1997; Wu and Kirtman 2007; Sabin et al. 2013). Consequently, it is necessary to directly focus on the effects of precipitation rather than SST on the circulations in the midlatitudes.
It is not easy to diagnose the causality between precipitation and circulation changes through observational analyses. Numerical model simulation may be a helpful tool to study the role of tropical precipitation anomaly in the variation of extratropical circulations over East Asia and effectively isolate the impact of tropical precipitation anomalies over different regions. Thus, a dry numerical model is used in this study to examine whether precipitation variability can significantly affect atmospheric circulations.
The rest of the paper is organized as follows. Datasets and model are described in section 2. In section 3, we present the spatial and temporal characteristics of precipitation variation over the tropical Pacific during boreal winter as background information. The relationship between the variation of the EAJ and precipitation anomalies over the WNP and CP is investigated here. And we also discuss the individual impact of tropical precipitation changes over the WNP and CP on the EAJ. Section 4 describes the results of the linear baroclinic model experiments. Section 5 gives a summary and discussion.
2. Datasets and model
The datasets used in the present study consist of 1) monthly variables derived from the National Centers for Environmental Prediction (NCEP)–U.S. Department of Energy (DOE) Reanalysis II with a horizontal resolution of 2.5° in both zonal and meridional directions and 17 vertical levels (Kanamitsu et al. 2002); 2) monthly precipitation data from the Global Precipitation Climatology Project, version 2.2 (GPCPv2.2; Adler et al. 2003), which combines observations and satellite precipitation data onto 2.5° × 2.5° global grids; 3) the monthly NOAA Precipitation Reconstruction (PREC) dataset for 1948–2010 (Chen et al. 2002); 4) monthly NCEP–National Center for Atmospheric Research (NCAR) reanalysis data from 1948 to 2010 (Kalnay et al. 1996); and 5) monthly data from the 40-yr European Centre for Medium-Range Weather Forecasts Re-Analysis (ERA-40) for 1957 through mid-2002 (Uppala et al. 2005). The datasets of PREC, ERA-40, and NCEP–NCAR reanalysis are only used to discuss the interdecadal changes of the relationship between the EAJ and tropical precipitation variations in section 4. Before performing our analysis, the climatological mean for the period 1980–2010 was removed, and the winter-mean was calculated as average of December to the following February [December–February (DJF)]. The winter of 1980 refers to the 1979/80 winter.
A simple dry model named linear baroclinic model (LBM) was used in our present study to examine the influence of tropical rainfall anomaly on the subtropical westerly jet. It consists of the primitive equations linearized about the observed winter climatology obtained from the NCEP–DOE reanalysis dataset for the period 1980–2010. This model was constructed based on a linearized atmospheric general circulation model (AGCM) developed at the Center for Climate System Research (CCSR), University of Tokyo, Japan, and the National Institute for Environmental Studies (NIES), Japan (Watanabe and Kimoto 2000). The model adopts a horizontal resolution using a T42 spectral truncation, roughly equivalent to 2.8° latitude × 2.8° longitude, with 20 vertical levels using a sigma coordinate system, and includes a horizontal (vertical) diffusion, Rayleigh friction, and Newtonian damping. The damping time scale was set at 0.5 day for the three lowest levels and two topmost levels, and 30 days for the other levels. We use a time integration method and integration is continued up to 30 days. The results of the 10-day average from day 20 to day 30 are shown as the steady response to a prescribed diabatic heating.
3. Relationship between tropical rainfall and EAJ
a. Leading mode of precipitation variability in winter
For the purpose of obtaining the spatiotemporal precipitation variability in tropics in boreal winter, the empirical orthogonal function (EOF) analysis is applied to GPCP winter precipitation anomalies. As shown in Fig. 1, the leading mode, which is distinguished from the other modes based on the method proposed by North et al. (1982), explains about 38.6% of the total variance for the period 1980–2010. A well-organized zonal dipole structure with a negative center over the WNP and Maritime Continent and a positive center over the CP is seen over the tropical Pacific. Based on the power spectrum analysis, the prevailing period of the corresponding principal component (PC1) is 2.5 yr. This suggests that the leading mode of tropical precipitation variability in winter is governed by interannual variations. As shown in the PC1 time series, the extreme events of the first mode usually appear in ENSO years, such as 1983 and 1998. The correlation coefficient between PC1 and winter-mean Niño-3.4 (5°S–5°N, 170°–120°W) index is 0.92, which is statistically significant at the 99% confidence level according to the Student’s t test. The linkage between the dipole pattern of tropical Pacific rainfall variability and ENSO has been previously recognized (e.g., Lau 1992; Webster and Yang 1992; Zhang et al. 1996; Xie and Arkin 1997; Cai et al. 2011). These previous studies found the precipitation anomalies exhibit a striking dipole pattern with a dry polarity over the WNP and Maritime Continent and a wet polarity over the equatorial eastern-central Pacific when El Niño matures in boreal winter. These studies also point out that ENSO exerts tremendous impacts on the dipole pattern of precipitation variations and circulation anomalies in the equatorial Pacific Ocean (Wang et al. 2000; Wang and Zhang 2002). However, the present study, as described above, will focus on the influence of tropical precipitation anomalies on atmospheric circulations in the midlatitudes, rather than the factors being responsible for the rainfall pattern.
According to the leading mode of winter precipitation anomalies, it is clear that precipitation over both the CP and WNP varies tremendously. Hence, investigating the variability of precipitation over these two regions is of central importance for understanding the precipitation fluctuation over the tropical Pacific during boreal winter. Consequently, we calculate the area-averaged rainfall anomaly over the WNP (5°–15°N, 105°–135°E) and CP (7.5°S–2.5°N, 165°E–165°W) as the western Pacific precipitation anomaly index (WPPA) and central Pacific precipitation anomaly index (CPPA), respectfully, for a better understanding of the impact of precipitation change in the two key regions. Figure 1c displays the time series of normalized WPPA and CPPA for the period 1980–2010. The correlation coefficient between CPPA and WPPA is −0.87, significant at the 99% confidence level. It suggests that the variation in precipitation over the WNP is tightly coupled with that over the CP. Figure 2 shows the correlation coefficients between precipitation anomalies and the WPPA/CPPA. Both the WPPA and the CPPA are significantly associated with a dipole pattern of precipitation anomalies in the tropical Pacific, with the opposite polarity because of the inverse relationship between the WPPA and the CPPA.
b. Relationship between tropical rainfall and EAJ
To detect the relationship between tropical precipitation anomalies and EAJ, we perform a composite analysis of atmospheric circulation anomalies of the winter mean in this subsection. The years during which the normalized WPPA and CPPA values are greater than 1.0 or less than −1.0 standard deviation are selected to make the composite. According to the criteria, six strong rainfall years (1983, 1987, 1992, 1998, 2003, and 2010) and eight weak rainfall years (1984, 1989, 1996, 1999, 2000, 2001, 2008, and 2009) are identified for the CPPA for the period 1980–2010 (Fig. 1c). For the case of WPPA, six strong rainfall years (1984, 1996, 1999, 2000, 2006, and 2009) and six weak rainfall years (1983, 1987, 1990, 1991, 1992, and 1998) are selected. Compared with the strong and weak convection years used in the composite in Park and An (2014; see their Table 1), the selected years are very different. Only two strong WNP rainfall years (1999 and 2000) one weak WNP rainfall year (1998) are the same.
During the strong rainfall years in the WNP, positive rainfall anomalies are located to the east of the Philippines with a maximum of 5–6 mm day−1 (Fig. 3a). During the strong rainfall years in the CP, a zonal band of positive rainfall anomalies appears over the equatorial eastern-central Pacific (Fig. 3b) with the maximum value of about 7–8 mm day−1. As shown in Figs. 3c and 3e, the composite zonal wind and streamfunction anomalies at 200 hPa feature an anomalous anticyclone over East Asia during the strong WNP rainfall years. There are westerly anomalies to the north of climatological location of the EAJ and easterly anomalies to its south. With the enhanced precipitation over the CP, the pattern of wind anomalies is similar to that corresponding to enhanced precipitation over the WNP, but with opposite signs (Figs. 3d and 3f). The pattern correlation of 200-hPa zonal wind between Figs. 3c and 3d is −0.77 and that of 200-hPa streamfunction between Figs. 3e and 3f is −0.92. It is inferred that the EAJ migrates northward (southward) with increased precipitation over the WNP (CP). Our results are consistent with those of Lau and Boyle (1987), who pointed out that the changes in heating over the western and central Pacific were accompanied by significant changes in the EAJ.
To determine whether the composite pattern is the primary mode of upper atmosphere circulation, we perform an EOF analysis of 200-hPa zonal wind anomalies in the domain of 0°–50°N, 90°–140°E during boreal winter. The leading EOF (EOF1) accounts for 42% of the total variance. The first mode features a pattern of alternating positive and negative centers over East Asia (not shown), which represents the north–south migration of the EAJ. It resembles the pattern shown in Figs. 3c and 3d, suggesting that the meridional pattern of zonal wind anomalies associated with positive (negative) rainfall anomalies over the WNP (CP) represents the dominant mode of the variability of the East Asian upper-tropospheric circulation during winter.
To further unravel the connection between precipitation change in the key regions and the upper-tropospheric circulation, we built two indices to depict the variations of the westerly jet, including its location and intensity. Based on the pattern with opposite sign anomalies of 200-hPa zonal wind in the meridional direction (Figs. 3c and 3d) and the leading mode of 200-hPa zonal wind anomalies in the domain of 0°–50°N, 90°–140°E during boreal winter (figure not shown), a winter EAJ index (WEAJI) measuring its meridional displacement is defined by using the difference between the 200-hPa zonal winds averaged over 22.5°–32.5°N, 100°–130°E and 32.5°–42.5°N, 100°–130°E. A positive (negative) WEAJI indicates a southward (northward) displacement of the EAJ. In addition, an index measuring the intensity of the EAJ (WEAJS) is defined using the area-averaged 200-hPa zonal winds over 27.5°–37.5°N, 100°–130°E. Table 1 depicts the correlation coefficients between precipitation and circulation indices. The correlation coefficient between the WPPA (CPPA) and WEAJI is −0.75 (0.67). The correlation coefficient between PC1 and WEAJI is 0.79. As expected, the EAJ migrates northward (southward) corresponding to increased precipitation over the WNP (CP). The precipitation anomalies in the key areas are also significantly correlated with the intensity of the EAJ. The correlation coefficient between the WPPA (CPPA) and WEAJS is 0.69 (−0.61). The correlation coefficient between PC1 and WEAJS is −0.65. This result shows that the westerly jet tends to be stronger (weaker) corresponding to positive precipitation anomalies over the WNP (CP).
As mentioned above, the precipitation anomalies over the WNP and CP are strongly coupled with each other. It is also noted that the negative precipitation anomaly over the CP is apparently dominant during the strong rainfall years of WNP (Fig. 3a). Thus, the composite horizontal wind anomalies based on the WPPA (Fig. 3c) may contain the impact of anomalous precipitation over the CP. The individual impact of anomalous convection over the WNP and CP cannot be seen in the preceding composite analyses. Subsequently, we should attempt to distinguish the influences of anomalous precipitation over the WNP and CP by partial correlation and numerical experiments.
c. Individual impact of anomalous precipitation on the EAJ
To isolate impacts of precipitation anomalies over the WNP and CP, partial correlation is applied. An anomalous circulation pattern associated with the WPPA (CPPA) after removing the effects of the CPPA (WPPA) is described by the partial correlation coefficients, denoted as WP|CP (CP|WP). The significance of the partial correlation is estimated based on the Student’s t test. The circulation anomalies associated with rainfall variability over the WNP and CP are assessed with correlation coefficient map of streamfunction and horizontal winds at 200 hPa onto the WP|CP and CP|WP. Some striking discrepancies have been identified, which will be discussed subsequently.
Figure 4 shows the partial correlations of upper-tropospheric streamfunction and horizontal wind anomalies with WPPA and CPPA for the period 1980–2010. After intentionally excluding the impact of precipitation variations over the CP, positive correlations appear over the subtropical Pacific and East Asia. It indicates that a significant anticyclonic anomaly appears over there with enhanced precipitation over the WNP and eastern China (Fig. 4a). At the upper levels, the pattern of wind anomalies is characterized by westerly anomalies to the north of 30°N and southeasterly anomalies to the south (Fig. 4c). After removing the influence of precipitation in the WNP, the streamfunction correlations show a pair of anticyclones straddling the equator and a cyclone over the North Pacific with positive precipitation anomalies over the CP (Fig. 4b). This pattern over the North Pacific is similar to the Pacific–North American (PNA) teleconnection (Wallace and Gutzler 1981). There is a wide zonal band of westerly anomalies between 135°E and 120°W located to the north of the subtropical anticyclone over the CP (Fig. 4d). Comparison with Fig. 4c reveals that the westerly jet over the south of Japan and the North Pacific tends to intensify because of significant enhancement of westerlies associated with positive precipitation anomalies over the CP, while the range of the East Asian subtropical jet might expand northward due to the anomalous anticyclone over eastern China associated with positive precipitation anomalies over the WNP.
According to the results of partial correlation, we speculate that the impact of precipitation anomalies over the WNP and CP on the variability of the subtropical westerly jet might be different. The result obtained from the partial correlation also implies that the meridional displacement of the EAJ between 90° and 130°E is relevant to precipitation anomalies over the WNP, while precipitation anomalies over the CP may affect the intensity of the subtropical jet over the North Pacific between 135°E and 120°W. Note that the precipitation variation over the CP seemingly does not strongly link to the variation of the EAJ between 100° and 130°E (Fig. 4d). The relationship between precipitation anomalies over the CP and zonal wind anomalies over the EAJ’s entrance region will be further discussed in section 4.
We have noted that the variation of precipitation in the two key regions is significantly related with the upper-tropospheric circulation and the results of partial correlations may give a hint of the individual impact of precipitation anomalies over the WNP and CP. Hence, a subsequent question is how anomalous precipitation over the tropical Pacific influences the atmospheric circulation, especially the EAJ during boreal winter. This part is investigated by virtue of partial correlations of wave activity flux defined and formulated by Takaya and Nakamura (2001) with the WP|CP and CP|WP. The definition of this wave activity flux (hereinafter T-N flux) can be written as
where f0 = 2Ω sinφ is the Coriolis parameter and φ and λ are latitude and longitude, respectively; z = −H lnp, where H is a constant scale height and p is the pressure divided by 1000 hPa; and N2 = (RapκH−1)(∂θ/∂z) is the buoyancy frequency squared where θ denotes potential temperature, Ra is the gas constant of dry air, and κ is defined as Ra normalized by the specific heat of air at constant pressure. Also, U and V denote a steady basic flow in the zonal and meridional directions, respectively; ψ is the streamfunction; and a is Earth’s radius. The three-dimensional perturbations deviating from the time mean are denoted by primes. The T-N flux considers a steady zonally inhomogeneous basic flow and could be applicable to small-amplitude quasigeostrophic disturbances, either stationary or migratory. It is independent of wave phase and parallel to the local group velocity of Rossby waves. So, it is a powerful diagnostic tool to identify the energy propagation of the observed perturbation (Takaya and Nakamura 2001; Zheng et al. 2013).
The primary impact of precipitation variations over the WNP on the EAJ is confirmed by a careful examination of the partial correlation between the T-N flux and WPPA (CPPA) when the effects of precipitation variations over CP (WP) are removed. When our attention is devoted to the relationship between the acceleration/deceleration of the zonal-mean zonal wind and anomalous perturbations, a simplified T-N flux was formed here. In light of our focus on variations of the zonal westerly jet during boreal winter where the meridional flow is relatively trivial, we assume that the basic states are dominant by zonal flows (V = 0). Then, the T-N flux becomes the traditional EP flux (Eliassen and Palm 1961) after the zonal average is taken in Eq. (1). The divergence of simplified T-N flux is closely correlated with the variability of zonal-mean flow. It indicates that the original T-N flux could help us explore the horizontal propagation of Rossby wave energy excited by the heating in tropics (shown in Fig. 5), while the simplified T-N flux could favor a better understanding of the impact of heat-induced distributions on the EAJ (shown in Fig. 6).
At the upper levels, correlations of the T-N flux with WPPA are mostly poleward to the north of 20°N over East Asia (Fig. 5a). Consistent with previous studies, the poleward-propagating Rossby wave trains excited by tropical convection can reach middle-to-high latitudes, modulate extratropical circulations, and even affect polar amplification (Hoskins and Karoly 1981; Schneider et al. 2012; Yoo et al. 2011; Yoo et al. 2012). In Fig. 6a, we note that there are positive correlations at the upper levels between 20° and 40°N, indicating a flux divergence to the north of 20°N. Thus, there is increased wave absorption in the vicinity of 30°N. The poleward and upward propagation of simplified T-N flux is dominant in the midlatitudes. The horizontal component of simplified T-N flux is proportional to the product of eddy momentum flux and cosφ. Thus, it suggests that there is eddy momentum flux from the tropics to the midlatitudes associated with anomalous precipitation over the WNP. The enhanced poleward propagation of Rossby wave activity associated with precipitation variations over the WNP results in an increased eddy momentum flux divergence in the midlatitudes, leading to an acceleration of zonal winds at the upper levels. When the effects of precipitation variations over the WNP are removed (Fig. 5b), the poleward propagation of T-N flux over East Asia apparently weakens and significant correlations appear over the subtropical Pacific. By the vicinity of the date line, the correlations are mostly equatorward. For the vertical component of simplified T-N flux (Fig. 6b), the anomalous center of the flux divergence at upper levels is located between 25° and 55°N. It implies an acceleration of zonal winds at the upper levels corresponding to enhanced precipitation over the CP.
Based on Fig. 6, it seems that the variation of zonal mean flow is more relevant to the variability of anomalous precipitation over CP than that over WNP. However, from Fig. 5b, the associated T-N flux varies largely in the zonal direction. When the zonal average is done, significant signals over the North Pacific account for a large proportion of the variations shown in Fig. 6b. Focusing on East Asia, the precipitation change over WNP plays a relatively more important role in the variability of the zonal flow (Fig. 5a), compared with that over CP (Fig. 5b). Overall, the results based on the partial correlations suggest that the poleward-propagating Rossby waves excited by tropical thermal forcing are of considerable importance to the variation of the basic flow over East Asia.
We note that significant signals of T-N flux appear over the central Pacific around 20°N between 170° and 120°W after the effects of CPPA are removed (Fig. 5a). Actually, the precipitation and horizontal wind anomalies regressed onto the WP|CP were also examined (figures not shown). When the effects of WPPA are removed, significant negative precipitation anomalies still appear over the central South Pacific and North Pacific. An anticyclone was located over the central Pacific between 0° and 30°N in the middle and upper troposphere (not shown). The anticyclone anomaly over the central North Pacific could also be seen in Fig. 4c. Hence, we speculate that the negative precipitation anomalies over the CP may result in significant signals there (Fig. 5a) even if we removed the effects of CPPA. Thus, the impacts of precipitation change over the WNP and CP cannot be completely isolated through simple statistical methods because the precipitation variations over the WNP is highly correlated with that over the CP (r = −0.87). The results obtained from the partial correlations may be partly credible and the further numerical investigations are required. Accordingly, a series of model experiments with a dry model named LBM are conducted to investigate the relative role of tropical precipitation anomalies over the WNP and CP in atmospheric circulation change over the midlatitudes in the next section.
4. Numerical results by LBM
To determine whether the variation in precipitation over the WNP and CP results in the variation of the EAJ, several numerical experiments based on LBM were carried out. As the LBM experiments in this study include only the linear response, the model results to be presented may be understood qualitatively, not quantitatively.
Figure 7 shows the heating over the tropical Pacific prescribed in the model in three numerical experiments. In the first experiment, to investigate the impact of precipitation variation over the WNP, an elliptic heat source placed onto the WNP (centered at 10°N, 118°E) was sustained for the integration period (Fig. 7a). Its maximum at 400 hPa (σ = 0.45) is 3.7 K day−1 (Fig. 7b). It is approximately equivalent to heating associated with a precipitation of 4–5 mm day−1 and consistent with the composite results (Fig. 3a). In the second experiment, the heat perturbation is centered at 5°S, 170°W and takes an elliptical form in the horizontal (Fig. 7c). The vertical profile of the heat sink peaks at 400 hPa with a vertically averaged heating rate at the center of −4.5 K day−1 (Fig. 7d). This absolute value of heating rate is roughly equivalent to 6–7 mm day−1 of anomalous precipitation (Fig. 3b). With the aim of investigating the joint impact of the anomalous heating over the WNP and CP in the third experiment, the heat source over the WNP and the heat sink over the CP are identical to the heating imposed in the first two experiments, respectively (Figs. 7e,f). Since the model we used in the present study is linear, the responses of the joint impact experiment are actually the results adding the first two numerical experiments together. However, to further explore the relatively contribution to the variation of the EAJ clearly, the results of the joint impact experiment are still shown in the present study.
a. Impact of heat source over the WNP
We now consider the impact of anomalous heating associated with precipitation anomalies over the WNP on the westerly jet at the upper levels. To emphasize the steady response to a prescribed diabatic heating, the results of 10-day average response from day 20 to day 30 are shown in Fig. 8.
In the lower troposphere (Fig. 8a), an anomalous cyclone response appears over the South China Sea and Philippines. An anomalous westerly is observed to the west of the heat source placed onto the WNP with a relatively weak easterly to the east. At the upper levels (Fig. 8b), the heat source induces a wave train pattern with a northeast–southwest tilt structure, which features a cyclone anomaly over the North Pacific and an anticyclonic anomaly over eastern China and South Asia. Thus, there are westerly anomalies to the north of the East Asian anticyclone and easterly anomalies to its south (Fig. 8c). It represents a northward displacement of the EAJ. Note that the pattern of streamfunction in Fig. 8c is, to some extent, similar to that of the partial correlations with the WP|CP (Fig. 4c). Consistent with the preceding results in section 3c, enhanced precipitation over the WNP significantly favors a northward displacement of the EAJ by inciting a Rossby wave train, which features a distribution of alternating anticyclone and cyclone over East Asia and the North Pacific. Many previous studies with numerical models have used forcing over the tropical western Pacific and the results show that the heat anomaly over the WNP generally excites a northward-propagating Rossby wave train (Hoskins and Ambrizzi. 1993; Lu and Lin 2009; Zheng et al. 2013), which significantly modulates the EAJ.
To further confirm that this pattern is a Rossby wave train excited by the tropical forcing, the associated T-N flux is shown in Fig. 8d. Consistent with the teleconnection pattern from East Asia to the North Pacific, the T-N flux emits from around 20°N toward Japan and is mostly poleward. Overall, the result indicates that the poleward Rossby wave propagation is related to the wave train pattern. A heat-induced Rossby wave train extending from East Asia to the North Pacific enhances the westerly to the north of climatological location of westerly jet.
b. Impact of heat sink over the CP
In the second experiment, we devote our attention to the individual effect of the heat sink related with negative precipitation anomalies over the CP. Figure 9 illustrates the model responses of streamfunction and horizontal winds in the case of CP heat sink. The response at the lower level features westerlies over the equatorial central-eastern Pacific and easterlies over the equatorial western Pacific (Fig. 9a). At the upper levels, Fig. 9b shows a pattern of alternating cyclone and anticyclone centers from the CP to the North Pacific. The wave train of streamfunction anomalies is similar to a teleconnection pattern known as the PNA proposed by Wallace and Gutzler (1981). As demonstrated by Hoskins and Karoly (1981) and Horel and Wallace (1981), a heating anomaly associated with El Niño can excite a stationary barotropic Rossby wave train, whose spatial pattern is somewhat similar to the PNA. Figure 9b also shows a cyclone to the northwest of the heat sink at upper levels, which is associated with the equatorial Rossby wave response. At the same time, to the east of the heat sink the Kelvin wave response with equatorial easterlies extends eastward along the equator to 90°E (Fig. 9a). The equatorial response is consistent with the typical Gill (1980) pattern.
A cyclone develops over the subtropical Pacific and an anticyclone over the North Pacific (Fig. 9b). Thus, there is a strong easterly anomaly band around 30°N (Fig. 9c). It implies that the heat sink over the CP may weaken the westerlies over East Asia and the North Pacific. The associated T-N flux does not exhibit any meaningful information about the propagation of Rossby wave (figure not shown).
Observational results in preceding section show a pattern with anomalous easterlies around 30°N between 135°E and 120°W, suggesting that the variability of the EAJ intensity is possibly relevant to anomalous precipitation over the CP (Fig. 4d). However, the variations in precipitation in these two regions are tightly correlated with each other (r = −0.87). The observational results with respect to the individual effect of rainfall anomalies over the WNP and CP might be partially inaccurate. In this subsection, the numerical experiments demonstrate that the heat sink over the equatorial CP possibly result in the weakening of the maximum westerlies over East Asia and the North Pacific, which is different from the observations (Figs. 3d and 4d).
c. Joint impact of the heat source over the WNP and heat sink over the CP
In this subsection, emphasis is placed on the joint impact of the precipitation anomalies over the WNP and CP. When precipitation over the WNP is enhanced and that over the CP is suppressed, a cyclone develops over the South China Sea at 850 hPa and a weak anticyclone appears over the North Pacific (Fig. 10a). At 200 hPa (Fig. 10b), anticyclonic anomalies over eastern China and South Asia build up and a PNA-like wave train extends from the equatorial CP to the North Pacific. Based on the response of streamfunction induced by the WNP and CP heating anomalies, there are northeasterly anomalies to the south of this anticyclone over eastern China and westerly anomalies to the north. The anticyclone over East Asia in Fig. 10b resembles that in Fig. 8b. It suggests that the heating anomaly over the WNP, compared to the heat sink over the CP, may be a dominant factor responsible for the meridional movement of the EAJ between 110° and 150°E. Note that the cyclonic anomalies over CP induced by anomalous heating of both the WNP and CP (Fig. 10) are similar to that induced by the heat sink over the CP (Fig. 9) and the easterly response to the north of this cyclone is dominant over the North Pacific around 30°N. Hence, the heat sink over the CP may be one of the possible factors that affect the strength of the subtropical jet over the North Pacific.
Apart from the anomalous heating over the WNP and CP and the mean flow, other factors may also affect the location and intensity of the EAJ. In Fig. 10b, the anticyclonic anomaly appears over eastern China and South Asia, but differs appreciably from the diagnosed one (Fig. 3). A major difference between the simulated and diagnosed zonal wind anomalies is that the simulated centers of streamfunction anomaly are located to the north of Indo-China and the Indian subcontinent. A possible reason might be that the model used in this study is a dry version, and thus the heat source or heat sink imposed on the model cannot result in any anomalous tropical precipitation as in reality. For instance, the observational result shows that there are positive precipitation anomalies over the WNP, South Asia, and Maritime Continent (Fig. 3a). The anomalous heating with an elliptical form in the horizontal (Fig. 7a) in the numerical experiments, however, is quite different from the observations (Fig. 3a). This might be one of the factors influencing the pattern of the heat-induced Rossby wave train. Overall, the modeling results, to some extent, capture the major characteristics in spite of some appreciable discrepancies.
d. Nonlinear response to tropical heat forcing
The model used in the present study focuses on the linear response to the steady forcing and rules out the nonlinear interactions between forcing and basic states. The present model is linearized about the basic states in integration period. However, the direct response to tropical forcing can change the extratropical flow, which modifies the wave energy propagation from the tropics (Lin and Derome 2004). It was noted that the small variations in the basic flow can make important differences in the response (Ting and Held 1990; Hall 2000). Nevertheless, the numerical experiments by Hoerling et al. (1997) show that the extratropical response to the tropical Pacific SST anomalies does not depend on the zonally varying flow of El Niño or La Niña states. Hence, in this subsection, to investigate the dependence of the upper-tropospheric zonal wind response to the tropical forcing in the two cases of WNP and CP, two numerical experiments are designed.
In the present results, the response of linear integration when using the model climatology as a basic state and heat-induced Rossby wave train is clear (Figs. 8 and 9). Lin and Derome (2004) discussed the nonlinearity of extratropical response to El Niño/La Niña based on some linear numerical experiments. Following Lin and Derome (2004), as the heat-induced direct linear response, the anomalies of day 15 to the heating anomalies placed onto the WNP or CP are superposed on the model primary climatology, denoted as WNP or CP basic states, respectively. To assess whether a changed basic state can significantly alter the forced solution, the WNP and CP basic states are used as the basic states of the linear baroclinic model. To test the sensitivity of anomalies imposed into the model primary basic states, we have also performed the numerical atmospheric response with day 10 and day 20. It is found that the spatial patterns of the upper-tropospheric circulations are rather insensitive to the chosen anomalies.
Figures 11a and 11c illustrate the response to the heat source over the WNP with the WNP basic states. Comparison with Fig. 8 shows that the anticyclone over East Asia does not change a lot and the cyclone over the North Pacific significantly intensifies. The result indicates that nonlinear interactions between forcing and basic states might be trivial for the anticyclone response at the westerly jet over East Asia. As shown in Figs. 11b and 11d, the pattern of streamfunction induced by the heat sink over the CP resembles that in Fig. 9. The heat-induced cyclone over the subtropical CP intensifies, resulting in decelerating the maximum westerlies near the date line. The interactions of heat sink over the CP and basic flow may further enhance the heat-induced cyclone and tend to slow down the zonal westerly wind.
e. Sensitivity to the basic state
We have performed some other numerical experiments by imposing heating anomalies with different basic background flow, and found that the responses of circulations are sensitive to these differences. Before the sensitive experiments, we noted an interdecadal change in the relationship between tropical Pacific convection and the EAJ. We examined the changes in tropical Pacific precipitation and found some differences between the two periods before and after 1979. To avoid the influence of plausible unrealistic interdecadal variability in this study, the precipitation anomalies obtained from the NOAA PREC dataset and the zonal wind obtained from the NCEP–NCAR reanalysis and ERA-40 are all subjected to a harmonic analysis to obtain periods shorter than 8 yr. The leading EOF mode of winter-mean precipitation anomalies (obtained from the NOAA PREC dataset) over the tropical Pacific Ocean for the period 1949–2010 (figure not shown) is very similar to that for 1979–2010 (Fig. 1a). Consistent with the increased magnitude of ENSO (Wang et al. 2008), the PC1 amplitude has increased after the late 1970s. Based on the correlation analysis between the zonal wind anomalies (obtained from NCEP–NCAR reanalysis and ERA-40) in winter for 1949–2010 (ERA-40 for 1958–2002) and PC1 (Fig. 12), the relationship between precipitation variations and the EAJ is apparently different between the two periods. The results obtain from the ERA-40 are very similar to that from the NCEP–NCAR reanalysis, and the pattern correlation between Figs. 12a and 12c (Figs. 12b and 12d) is 0.88 (0.95).
The mean state of the upper-tropospheric circulation was examined and the results show that the background flow experiences an interdecadal change since the late 1970s (figure not shown). Hence, it is hypothesized that the shift of the mean state before and after 1979 may give rise to the interdecadal change in the enhanced relationship between tropical Pacific convection and the EAJ. Subsequently, some numerical experiments are performed to examine the sensitivity to the background flow. The basic states imposed into the model in this subsection are the observed winter climatology obtained from the NCEP–NCAR reanalysis dataset for the period 1949–79. The heat source over the WNP and heat sink over the CP are identical to the heating imposed in sections 4a and 4b, respectively. Figures 12e and 12f show the responses of 200-hPa streamfunction and winds to the heating after the basic flow has been altered. Compared with Figs. 8b and 8c, the anticyclone and the anomalous easterly to the south of 30°N in Fig. 12e apparently weaken. Furthermore, the response of the anomalous easterly to the heat sink over the CP with the basic states of 1949–79 also exhibits relatively weak signals over East Asia (Fig. 12f), compared with Fig. 9c. The feeble anomalous westerly and easterly in the vicinity of the EAJ might be evidence for the subdued correlation between the Pacific precipitation and the EAJ before 1979.
5. Summary and discussion
The impact of the precipitation anomalies over the tropical Pacific on the upper-tropospheric subtropical westerly jet over East Asia during boreal winter has been investigated by diagnosis and numerical experiments. In boreal winter, the leading mode of precipitation anomalies over the tropical oceans for the period 1980–2010 is characterized by a dipole pattern with a positive anomalous center over the CP and a negative one over the WNP. The precipitation variations over the CP and WNP are significantly correlated with the meridional displacement of the EAJ. It is found that when precipitation over the WNP (CP) is enhanced, EAJ tends to migrate northward (southward). To investigate the individual impact of anomalous precipitation over the WNP and CP, partial correlation is performed. The partial correlations of streamfunction and wind anomalies with WP|CP shows an anticyclone centered around 30°N over East Asia with anomalous westerlies to its north and easterlies to its south. The partial correlations with CP|WP show a strong westerly band over the central North Pacific. There is a hint of different impact of precipitation anomalies over the WNP and CP on the upper-tropospheric westerly jet.
To examine the relative contribution of precipitation variability over the CP and WNP to the variation of the EAJ, a series of numerical experiments based on a simple dry baroclinic model have been performed. A heat source over the WNP, which is equivalent to positive precipitation anomalies over the western Pacific, excites a northward-propagating Rossby wave train in the upper troposphere over East Asia. A strong westerly anomaly appears to the north of the anticyclone over eastern China and an easterly anomaly to its south, which enhances the westerly flow and favors a northward shift of the EAJ. The response induced by a heat sink over the CP is quite different from that induced by the WNP heat source. A zonal easterly band is built up to the north of the heat-induced cyclonic circulation over the subtropical central Pacific, suggesting that the precipitation change over the CP may influence the intensity of the EAJ, instead of its location. The response of atmospheric circulations in the midlatitudes is similar to the diagnostic results when the joint impact of precipitation anomalies over these two regions is taken into account. Therefore, it could be concluded that positive precipitation anomalies as heat source anomalies over the WNP play a relatively dominant role in the meridional displacement of the EAJ and negative precipitation anomalies over the CP may be a possible factor that changes the intensity of the EAJ. In addition, many studies have suggested that the transient eddy feedback reinforces the extratropical responses to the thermal forcing (Lin and Derome 1997, 2004). The lack of eddy forcing in this linear model may lead to the differences of circulations between the observations and numerical experiments in the midlatitudes.
Furthermore, the interactions of tropical forcing and basic flow are discussed in the present study. The numerical results show that the direct response to the heat sink over the CP changes the atmospheric circulations. Then, the changed background flow intensifies the response circulations. The response to heat source over the WNP does not seem to depend on the nonlinear interactions between forcing and basic states over the westerly jet region.
Many studies demonstrated that the ENSO amplitude has increased and its periodicity has become longer after the late 1970s (e.g., Gu and Philander 1997; An and Wang 2000; Wang et al. 2008). Hence, we investigated the interdecadal enhanced relationship between the tropical Pacific precipitation anomalies and the EAJ since 1979. The sensitive experiments to the background flow in section 4e give a hint of the considerable importance of the changed mean states before and after 1979 in the interdecadal strengthened relationship between the precipitation anomalies and the EAJ. However, more diagnosis and numerical experiments need to be undertaken to further investigate the role of the mean flow and what authentically gives rise to interdecadal changes in the enhanced relationship between tropical Pacific convection and the EAJ since the late 1970s.
It is worth mentioning that the obtained relationship between the precipitation anomalies over the WNP and the meridional displacement of the EAJ is in agreement with Park and An (2014). However, our work compared precipitation changes over WNP and CP and their relative contributions to the variability of the EAJ. The results show that the individual effects of tropical precipitation changes over the WNP and CP on the EAJ are quite different, which was not discussed in Park and An (2014). Another major difference between our work and Park and An (2014) is that the role of mean flow in the interdecadal enhancement of the relationship between the precipitation anomalies over the WNP and CP and the EAJ was discussed in our paper. In addition, the possible mechanism responsible for the impact of tropical precipitation variation on the EAJ, proposed in our work from the perspective of momentum transportation, enhances the understanding of the impact of tropical heating on the midlatitude circulation variability.
The authors thank Prof. M. Watanabe for his permission to use the LBM. This research was jointly supported by National Key Basic Research and Development Projects of China (2014CB953901), National Natural Science Foundation of China (41175076 and 412111046), and State Key Laboratory of Severe Weather opening project. RW acknowledges the support of a National Natural Science Foundation of China grant (41275081). YG acknowledges the support of the high-performance grid computing platform of Sun Yat-sen University.