Interannual variability of the Asian subtropical westerly jet in boreal summer and associated with circulation and SST anomalies

Clim Dyn (2016) 46:2673 2688 DOI 10.1007/s00382-015-2723-x Interannual variability of the Asian subtropical westerly jet in boreal summer and associated with circulation and SST anomalies Yin Du 1,2 Tim Li 1,2 Zhiqing Xie 3 Zhiwei Zhu 1,2 Received: 20 May 2014 / Accepted: 20 June 2015 / Published online: 9 July 2015 Springer-Verlag Berlin Heidelberg 2015 Abstract The interannual variability of the Asian Subtropical Westerly Jet (ASWJ) in boreal summer is investigated through the diagnosis of 54-year (1960 2013) NCEP/NCAR reanalysis data. The main characteristics of two leading empirical orthogonal function patterns of 200 hpa zonal wind anomalies are the meridional displacement and southwest northeast tilting of ASWJ. The first leading mode has significant periods of 4.9 years, whereas the second mode has significant periods of 3.6 and 7.7 years, respectively. The two modes exhibit an equivalent barotropic structure, and are associated with a distinctive north south and east west dipole rainfall pattern in China, respectively. The positive phase of the first leading mode appears during El Nino developing phase, whereas the positive phase of the second mode occurs during La Nina decaying phase. A mechanism is put forth based on observational analysis and AGCM sensitivity experiments. The positive phase of the first mode is primarily driven by the combined effect of a cold SST anomaly (SSTA) in mid-latitude North Pacific and a warm SSTA in tropical Indian Ocean and Pacific. In response to the SSTA forcing, * Tim Li timli@hawaii.edu Yin Du duyinxie@163.com 1 2 3 Climate Dynamics Research Center and Earth System Modeling Center, International Laboratory on Climate and Environment Change, Nanjing University of Information Science and Technology, Nanjing, China Department of Atmospheric Sciences, International Pacific Research Center, School of Ocean and Earth Science and Technology, University of Hawaii at Manoa, Honolulu, HI, USA Jiangsu Climate Center, Nanjing, China a zonally oriented north south tropospheric temperature dipole is induced. While the tropospheric warming in the tropics arises from El Nino like heating, the tropospheric cooling in the mid-latitudes arises possibly from the local SSTA forcing. For the positive phase of the second mode, the upper-tropospheric anticyclonic vorticity anomaly in the east pole arises from local SSTA forcing in North Pacific, whereas the cyclonic anomaly in the west pole results from southeastward Rossby wave energy emanation from North Atlantic to East Asia. Keywords Asian Subtropical Westerly Jet Meridional temperature gradient Sea surface temperature anomaly AGCM experiments 1 Introduction The Asian Subtropical Westerly Jet (ASWJ) is one of the important circulation systems that affect weather and climate in East Asia. Its seasonal migration, with a northward jump in boreal summer and a southward retreat in boreal winter, is the most pronounced signal of seasonal transition of upper tropospheric circulation in the region (Ye et al. 1958). From climatological point of view, the meridional movement of ASWJ is associated with the onset and withdraw of the Asian summer monsoon, signifying by the beginning and ending of the main rainy season in East Asia (Li and Wang 2005; Zhang et al. 2006; Reinhard et al. 2009). The land-sea thermal contrast and the monsoon heating play an important role in the formation of ASWJ (Krishnamurti 1979; Zhang 1980; Wu et al. 2008). Dong et al. (1999) pointed out that the seasonal variation of ASWJ center is closely connected to the seasonal change

2674 Y. Du et al. of tropical monsoonal heating. Convective activity in one hemisphere could affect the location and intensity of the subtropical westerly jet in the other hemisphere (Yang and Webster 1990). The thermodynamic forcing of the large terrain in Tibet Plateau (TP) also plays an important role (Wu et al. 2009a, b, 2010; Duan and Wu 2009). Global model simulation results showed that heat sources over TP affect the intensity of the subtropical westerly jet. In early summer, the terrain thermal effect over TP accelerates the northward jump of ASWJ. The northward shift of ASWJ corresponds to a temperature gradient reversal in the middle and upper troposphere over the Asian Continent (Li et al. 2004). In addition to a pronounced annual cycle, ASWJ also experiences a marked interannual variation. The location and intensity changes of ASWJ are often associated with the year-to-year variation of rainfall in the middle and lower reaches of the Yangtze River (MLYR). The shift of ASWJ latitudinal location may affect the onset of the East Asian summer monsoon and the region of heavy rain in China (Zhou et al. 2003). Typically, when the jet shifts southward (northward) during the period of Meiyu, more (less) precipitation appears in the Yangtze River (Dong et al. 1987). The inter-decadal variability of East Asian summer rainfall is also related to the meridional shift of ASWJ (Liao et al. 2004; Kuang and Zhang 2006; Lin and Lu 2005). Lu et al. (2011) showed that zonal wind intensity in the upper troposphere in boreal summer has been significantly weakened over Northeast Asia and the subtropical western North Pacific since the mid-1990s. Zhu et al. (2010) also revealed the weakening of the ASWJ and related changes in the rainfall pattern in East China after late 1990s. When the westerly jet core at the upper troposphere is located over the western Pacific and accompanied with a low-level southwesterly jet at 850 hpa, heavy precipitation occurs in the lower reaches of the Yangtze River Valley. Associated with the heavy rainfall are strong low-level convergence and water vapor supply (Du et al. 2009). Wang et al. (2008) suggested that a warm temperature trend over the TP would favor a strengthened East Asian monsoon and thus the northward shift of ASWJ. Sampe and Xie (2010) showed that when the westerly jet tilted poleward with height during the rainy season, northward moisture transport was strengthened in the East Asian monsoon region. Sea surface temperature (SST) change in the tropical Pacific may also affect the decadal variability of ASWJ (Liao et al. 2006; Xuan et al. 2011). For example, Xuan et al. (2011) found that before 1980, the East Asian westerly jet was mainly influenced by the SST in North Pacific, whereas after 1980 it was primarily influenced by the SST in tropical western Pacific. The objective of this study is two folds. Firstly, we aim to reveal the dominant modes of ASWJ on the interannual timescale in boreal summer and its associated circulation and SST patterns. Secondly, we attempt to understand the cause of the ASWJ interannual variability. In addition, we will reveal the connection of the ASWJ modes to precipitation anomalies in China. The remaining part of this paper is organized as follows. In Sect. 2, we introduce the data, analysis methods and numerical model experiments. The dominant ASWJ patterns are revealed by use of an empirical orthogonal function (EOF) analysis in Sect. 3. In Sect. 4, we discuss the relation between the ASWJ modes and anomalous atmospheric circulation, SST and precipitation patterns. In Sect. 5, we discuss possible mechanisms through which the SST anomaly influences the dominant ASWJ modes. A conclusion and discussion are given in the last section. 2 Data, methods and numerical model experiments The primary data used for an observational analysis are (1) multi-level (from 1000 to 100 hpa) horizontal wind components, vertical velocity (ω, in pressure coordinate), geopotential height and temperature fields, precipitation, upward longwave radiations flux at the surface from the National Centers Environmental Prediction and National Center for Atmospheric Research (NCEP/NCAR) reanalysis for the period of 1960 2013, which incorporates satellite observations and in situ data and has a horizontal resolution of 2.5 (Kalnay et al. 1996), (2) monthly Precipitation data of 740 stations in China provided by the National Climate Center in China, (3) a global gridded observational SST dataset taken from the National Oceanic and Atmospheric Administration Extended Reconstructed SST data set (ERSST) at a 2 2 global grid (Smith and Reynolds 2004), and (4) the Nino3 index obtained from NOAA website http://www.esrl.noaa.gov/psd/gcos_wgsp/ Timeseries/Nino3/. Our analysis will focus on the period of 1960 2013. The analysis methods applied include the Empirical Orthogonal Function (EOF) analysis, the Power Spectrum Analysis (PSA), and the Singular Value Decomposition (SVD) analysis (Van Storch and Zwiers 1999). The EOF is applied for identifying the dominant modes of interannual variability of ASWJ. The PSA is used to identify dominant periodicity of the EOF principal components. The SVD is applied to confirm the EOF analysis result by revealing covariance coupled patterns of upper-tropospheric wind and SST. Before the above analysis, the data are standardized, with a linear trend removed. Both composite analysis

Interannual variability of the Asian subtropical westerly jet in boreal summer and associated 2675 Table 1 Description of the control and sensitivity experiments using the AGCM Experiment CTRL EXP_M1a EXP_M1b EXP_M1c EXP_M1d EXP_M2a EXP_M2b EXP_M2c EXP_M2d Description Forced by climatological monthly SST Forced by climatological monthly SST in CTRL plus a SSTA field over the tropical Pacific and IO (20 S 20 N, 40 E 80 W) derived based on the composite SSTA pattern from Fig. 5c in JJA Same as EXP_M1a, except with the SSTA over mid-latitude north Pacific Ocean (20 N 60 N, 120 E 80 W) Same as EXP_M1a, except with the SSTA over tropical Pacific and IO and mid-latitude north Pacific (20 S 60 N, 40 E 80 W) Same as EXP_M1a, except with the SSTA over tropical and North Atlantic (20 S 60 N, 80 W 0 W) Forced by the climatological monthly SST in CTRL plus a SSTA field over tropical Pacific and IO (20 S 20 N, 40 E 80 W) derived based on the composite SSTA pattern from Fig. 5f in JJA Same as EXP_M2a, except with the SSTA over mid-latitude north Pacific Ocean (20 N 60 N, 120 E 80 W) Same as EXP_M2a, except with the SSTA over tropical Pacific and IO and mid-latitude north Pacific (20 S 60 N, 40 E 80 W) Same as EXP_M2a, except with the SSTA over tropical and North Atlantic (20 S 60 N, 80 W 0 W) and linear regression methods are performed to validate against each other, to document the statistically robust relationship between upper-tropospheric zonal wind anomalies and other atmospheric and oceanic variables including vertically integrated temperature, vorticity, geopotential height, low-level flow, precipitation and SST. A local t test is applied to examine the statistical significance of the composite differences, and its formula is x ȳ t = ( ) 1/2 S 1m1 + 1 m2 ( where S 2 = m1 1 m1 m 1 i=1 (x i x) 2 + m 2 1 m2 m 2 i=1 (y i / ȳ)2) (m 1 + m 2 2), m 1 and m 2 are the sample numbers for the positive and negative cases, and x and ȳ are the means for the positive and negative cases, respectively. To examine how the anomalous SST pattern influences the interannual ASWJ variability, an atmospheric general circulation model (AGCM) was employed. The AGCM used in the study is ECHAM4.6 developed by the Max Planck Institute for Meteorology (MPI; Reockner et al. 1996). The AGCM has a horizontal resolution of 2.8 2.8 (T42) and 19 vertical levels extending from the surface to 10 hpa. The AGCM can better simulate the Asian summer monsoon (Cherchi and Navarra 2003; Fu et al. 2002; Jiang et al. 2005). In order to investigate physical processes through which the SST anomaly influences the interannual variability of ASWJ in boreal summer, we designed the control and sensitivity experiments listed in Table 1. In the control experiment (CTRL), the model is integrated for 15 years with specified the climatological monthly mean SST field. Parallel to the CTRL, in the first set of the sensitivity experiment we specify a SST pattern that is the sum of the climatological monthly mean SST field and a SSTA pattern regressed against the first dominant ASWJ mode. In the second set of sensitivity (1) Fig. 1 The climatology (shaded, U 25 m s 1 ) and standard deviation (contour, unit: m s 1, interval: 0.4) of the 200 hpa zonal wind in summer (June August) from 1960 to 2013. Thick black line is the axis of the westerly jet experiment, the SSTA pattern is derived from the regressed field of the second dominant ASWJ mode. In both sets of sensitivity experiments, various regional SSTA patterns based on the aforementioned regressed fields are specified in order to understand the effect of the regional SSTA forcing on the formation of the dominant ASWJ modes. 3 Characteristics of interannual variability of ASWJ Figure 1 shows the climatology and standard deviation of 200 hpa zonal wind field averaged in June July August (JJA). Climatological jet stream axis is approximately zonally oriented along 40 N. There are three cores along 40 N, located Caspian, north of Tibet Plateau, and the northwest Pacific Ocean, respectively, from west to east. Among the three cores, the strongest appears over north of the TP (between 80 E and 100 E). The standard deviation field shows four activity centers in the region, two of which residing at north and the other two at south of the

2676 Y. Du et al. (d) (e) (c) (f) Fig. 2 The leading EOF mode patterns (a, d) and corresponding principal components (b, e) of 200 hpa zonal wind (m s 1 ) in JJA from 1960 to 2013. The bottom panel c, f shows the power spectrum of the two principal components. The dashed lines in b, e denote 0.9 of standard deviation. The dashed red (blue) line in c, f denotes a 90 % (95 %) confidence level climatological jet axis. Along the zonal direction, the activity centers may be divided into the East Asia section and the West Asia section. To reveal the dominant modes of interannual variability of ASWJ in boreal summer, we performed an EOF analysis of JJA 200 hpa zonal wind field for a 54-year (1960 2013) period. Figure 2 presents the first two EOF patterns of the 200 hpa zonal wind anomaly and corresponding principal components (PCs). The first leading EOF pattern (hereafter M1), which accounts for 24 % of the total variance, is dominated by a meridional dipole structure, with positive (negative) zonal wind anomalies to the south (north) of the climatological jet axis (40 N). Thus, M1 is closely related to the meridional displacement of the zonal mean ASWJ. The second leading EOF pattern (hereafter M2), which explains 14.9 % of total variance, is characterized by a quarter-pole structure, with positive anomalies appearing over Iranian Plateau and Mongolian Plateau and negative anomalies over eastern Europe and East China. Thus, M2 reflects the northeastward tilting of ASWJ. According to the rule by North et al. (1982), the first mode is statistically separated from the rest of the eigenvectors. The second mode, albeit not totally separated from the higher modes, is still a large fraction of the total variance. The power spectrum analysis of time series of the two leading EOF modes reveals that they have distinctive peaks (Fig. 2c, f). The first principal component (PC1) has a significant peak at the period of 4.9 years. After being removed its linear trend, the second principal component (PC2) shows two significant periods at 3.6 and 7.7 years, respectively. The PC1 and PC2 peaks exceed a 95 % confidence level. The result implies that the uppertropospheric zonal wind exhibits significant interannual variability. In addition, M2 also shows a marked interdecadal variation.

Interannual variability of the Asian subtropical westerly jet in boreal summer and associated 2677 Table 2 List of selected positive and negative years for two leading EOF modes (σ denotes the standard deviation of the two principal components) PC 0.9 σ PC 0.9 σ PC1 1965, 1974, 1982, 1987, 1991, 1993, 1998, 2002, 2009 1961, 1971, 1973, 1975, 1978, 1984, 1990, 1994, 2006, 2013 PC2 1963, 1972, 1981, 1988, 1999, 2000, 2004, 2010, 2012 1973, 1976, 1980, 1983, 1984, 1986, 1990, 1993, 2003 (e) (f) (c) (g) (d) (h) Fig. 3 Horizontal maps of zonal wind at 200 hpa (contour, top, m s 1 ), geopotential height at 200 and 500 hpa (contour, middle, gpm), and temperature averaged from 500 to 200 hpa (contour, bottom, K) for M1 and M2, respectively. The 90 % (95 %) confidence levels are light (dark) shaded 4 Atmospheric and oceanic anomalies associated with the ASWJ modes Next we examine atmospheric circulation and SST fields associated with the two leading ASWJ modes, using both the regression and composite analysis methods. For the regression analysis, all fields are regressed against PC1 and PC2 time series. For the composite analysis, we select cases based on the following criterion: the standardized PC1/PC2 time series is greater than 0.9 or less than 0.9. Table 2 lists all cases selected for the composite analysis. Nineteen years are selected for M1, including nine positive years 1965, 1974, 1982, 1987, 1991, 1993, 1998, 2002 and 2009, and the ten negative years 1961, 1971, 1973, 1975, 1978, 1984, 1990, 1994, 2006 and 2013. Eighteen years are selected for M2, including positive years 1963, 1972, 1981, 1988, 1999, 2000, 2004, 2010 and 2012, and negative years 1973, 1976, 1980, 1983, 1984,1986, 1990, 1993 and 2003.

2678 Y. Du et al. Our calculations show that the results from the regression and composite analysis methods are quite similar. Figure 3a, e show the horizontal distributions of the composite zonal wind fields at 200 hpa associated with the two modes. As expected, the main characteristics of the upper-tropospheric wind anomalies are consistent with the EOF patterns shown in Fig. 2. That is, for M1 two positive centers are located south of the climatological jet axis, one being over Iranian Plateau and the other over central China extended to South of Japan, whereas two negative centers are located north of the climatological jet axis, one being over Balkhash Lake and the other over Southeast of Lake Baikal; for M2, a quarter-pole pattern appears, with activity centers being located over the north of Caspian Sea, south of Aral Sea, north of northeast China, and central China, respectively. The distinctive zonal wind patterns between M1 and M2 motivate us to further examine the vertical structure of the zonal wind, geopotential height and temperature fields. It is found that the zonal wind anomaly field in lower levels (say, at 500 and 850 hpa) exhibits a similar pattern with reduced amplitude. This indicates that the vertical shear pattern is same as the upper-tropospheric zonal wind anomaly pattern. According to the thermal wind relationship, the vertical shear of zonal wind is balanced by the meridional gradient of layer-mean temperature u T = u g (p 1 ) u g (p 0 ) = R ( ) ( ) T p0 ln (2) f ȳ p p 1 where p 0 = 500 hpa, p 1 = 200 hpa, u g (p 0 ) and u g (p 1 ) are the geostrophic wind, T is the average temperature between p 0 and p 1, and f is the Coriolis parameter. Equation (2) states that the change of zonal wind is proportional to the meridional temperature gradient averaged between the two pressure levels. In other words, a strengthened westerly wind at 200 hpa corresponds to an enhanced layer-mean meridional temperature gradient. The equivalent barotropic vertical structure of the zonal wind anomalies may be inferred from the distribution of geopotential height anomalies. Composite JJA geopotential height field at 200 and 500 hpa and the layer-mean temperature anomalies for M1 and M2 are depicted at the middle and lower panel of Fig. 3. Note that the height fields exhibit a clear barotropic vertical structure. For example, two significant negative height centers appear along the climatological jet axis (40 N) at both 200 and 500 hpa (Fig. 3b, c). Associated with the two low pressure centers are cyclonic flows. The spatial distributions of 500 hpa height anomalies are similar to those at 200 hpa, with their amplitude at 500 hpa being weaker. The high (low) anomalies in the 200 hpa geopotential height fields at middle panel of Fig. 3 correspond well the negative (positive) vorticity fields. The 500 200 hpa layer-mean temperature field shows two significant negative anomaly centers along 40 N (Fig. 3d). One center is located over north of the Iranian Plateau, and the other over Bohai Bay. Based on the hydrostatic relationship, the cold temperature anomaly is consistent with the increase of amplitude of the negative height anomaly with altitude. Meanwhile, this temperature pattern is consistent with the thermal wind relationship, that is, a cooling along 40 N causes a strengthened (weakened) meridional temperature gradient to its south (north); as a consequence, the westerly wind strengthens (weakens) to its south (north). A further examination of the temperature vertical profile shows that a consistent cooling happens throughout the middle and upper troposphere, with a maximum cooling occurring at 300 hpa. In contrast to the zonally symmetric distribution in M1, an east west dipole pattern appears in the height anomaly field along 40 N in M2 (Fig. 3f, g). The layer-mean temperature pattern resembles the geopotential height at 200 hpa (Fig. 3h). The temperature anomaly in the east pole is greater than that in the west pole, so are the geopotential height and vorticity fields. Therefore, the formation of anomalous jet patterns is essentially related to processes through which the anomalous temperature patterns formed. We will investigate the possible cause using an AGCM in Sect. 5. The interannual variability of ASWJ closely links to precipitation anomalies in China. Figure 4 shows JJA rainfall anomalies and corresponding 850 hpa wind fields. For M1, significantly positive rainfall anomalies appear along the Yangtze River (30 N) east of 90 E, whereas negative rainfall anomalies occur over North China around 40 N (Fig. 4a). Associated with the precipitation pattern are two cyclonic centers at low level along 30 40 N (Fig. 4b). The cyclonic center to the east is much stronger. The northerly flow associated with the center prevents northward water vapor transport from the tropical oceans; as a result, a drought anomaly appears in North China. The anomalous wind converges over south of the Yangtze River, resulting in strong upward motion and rainfall anomalies there. The strengthened rainfall is also consistent with enhanced baroclinicity at 30 N as the upper-tropospheric westerly increases over the same latitudinal band. For M2, the precipitation anomaly in China shows an east west dipole pattern, with negative rainfall anomalies appear over central and eastern China and positive anomalies over western China (Fig. 4c). The dipole rainfall pattern is consistent with the low-level flow pattern (Fig. 4d), which exhibits an anticyclonic center to the east and a cyclonic center to the west. The low-level flow patterns in both M1 and M2 resemble in general those in upper troposphere.

Interannual variability of the Asian subtropical westerly jet in boreal summer and associated 2679 (c) (d) Fig. 4 Composite precipitation percentage in JJA (a, c, %) and wind fields at 850 hpa (b, d, m s 1 ) for M1 and M2. Contour interval is 10. The 90 % (95 %) confidence levels are light (dark) shaded. A and C denote anticyclone and cyclone, respectively Figure 5 illustrates the SSTA patterns associated with the two modes. It appears that M1 occurs during El Nino developing summers. Note that in the preceding winter, a large-scale SSTA pattern appeared in the eastern equatorial Pacific (Fig. 5a), and this SSTA increased as the season progressed from the preceding winter to the concurrent summer (Fig. 5b, c). A northeast southwest oriented cold SSTA appeared in the extratropical western North Pacific (WNP) in preceding winter, and it persisted from the winter to the concurrent summer. During the concurrent summer, there is marked negative SSTA over the mid-latitude WNP. Meanwhile, there is a basin-wide warming over the tropical Indian Ocean (IO). As shown by our numerical modeling result in Sect. 5, this combined tropical warming mid-latitude cooling pattern is responsible for southward shift of ASWJ over East Asia. The SSTA evolution associated with M2 resembles a La Nina decaying phase. The composite SSTA pattern shows a decaying pattern from the preceding winter to the concurrent summer, with cold anomalies appearing in the eastern tropical Pacific and positive anomalies over mid-latitude WNP, tropical IO and mid-latitude western South Pacific (Fig. 5d f). To confirm that the SST anomalies associated with M1 and M2 are associated with El Nino developing and La Nina decaying phases, we examine the longer time evolution characteristics of the Nino3 index. Figure 6 indicates that the composite SSTA associated with M1 does increase slowly from the preceding autumn to the concurrent winter, and after reaching to a peak in the concurrent winter, it decreased rapidly and developed into a cold episode prior to the following winter. In contrast, the cold SSTA associated with M2 decayed from the preceding winter to the next winter. Furthermore, based on the spatial distribution of the SSTA, the frequencies on the various phases in El Nino and La Nina are calculated between the preceding winter to concurrent winter. The result showed that most of positive (negative) phase cases of the first and second modes are concurrent with El Nino (La Nina) developing and La Nina (El Nino) decaying phases, respectively. To examine how well the ASWJ and SST are connected, an SVD analysis was further performed, in which we derived the coherent patterns between JJA 200 hpa zonal wind and SST anomaly fields. Figure 7 shows the SVD analysis result. The dominant SVD mode structures are similar to the previous EOF and composite analyses shown in Figs. 2a and 5c. The variance of the zonal wind explained by SVD1 is 19 %. Therefore, the first leading ASWJ mode in boreal summer is closely associated with the positive SSTA in the tropic IO and tropical eastern Pacific and the cold SSTA in WNP. In contrast, the second SVD mode (figure not shown) does not correspond to the

2680 Y. Du et al. (d) (e) (c) (f) Fig. 5 Composite Sea surface temperature anomaly (K) from preceding winter (December February; top) and spring (March May; middle) to current summer (June August; bottom) with PC1 and PC2, respectively. The 90 % (95 %) confidence levels are light (dark) shaded second EOF mode. This suggests that the quarter-pole pattern shown in Fig. 2d may be less coherence with tropical SSTA. 5 Mechanism through which SSTA influences dominant ASWJ modes As discussed in the previous section, the interannual variability of ASWJ is closely related to global SSTA patterns (Fig. 5). In this section we further examine how the global SSTA patterns induce upper-tropospheric zonal wind anomalies in East Asia, through a set of idealized AGCM experiments. Before conducting AGCM experiments, we first examine the zonal mean circulation feature in a meridionalvertical cross section associated with M1 and M2. Different from a zonally uniform feature in M1, the geopotential height anomaly associated with M2 exhibits a zonal dipole pattern along 40 N (Fig. 3f). This implies that the dynamic origin of the geopotential height anomalies in the two poles differs. To the east of 80 E, a zonally oriented positive height anomaly appears, and it expands from 80 E to 160 E and beyond. This positive height anomaly overlies a zonally oriented warm SSTA along the same latitude in the mid-latitude North Pacific (Fig. 5f). To the west of 80 E, a wave train like pattern emerges in the geopotential height field (Fig. 3f). Figure 8 shows the latitude-height cross section of composite zonal wind, temperature and geopotential height averaged at 40 E 160 E for M1 and 80 E 160 E for M2. For M1, the most remarkable feature in the zonal wind field is a positive center south of 40 N and a negative center north of 40 N. This north south dipole is consistent with the zonally oriented wind pattern shown in Figs. 2 and 3. Both the zonal mean positive and negative zonal wind anomalies have a maximum peak near 200 hpa, consistent with a typical equivalent barotropic

Interannual variability of the Asian subtropical westerly jet in boreal summer and associated 2681 Fig. 6 The time series of the composite Nino3 index (K) from before 2-year winter (December February, DJF( 2)) to the next year winter (DJF(1)) for a M1 and b M2, respectively. Red bar represents the current summer 1.6 1.2 0.8 0.4 0-0.4 DJF(-2) MAM(-1) JJA(-1) SON(-1) DJF(-1) MAM(0) JJA(0) SON(0) DJF(0) MAM(1) JJA(1) SON(1) DJF(1) -0.8 0.3 0.0 DJF(-2) MAM(-1) JJA(-1) SON(-1) DJF(-1) MAM(0) JJA(0) SON(0) DJF(0) MAM(1) JJA(1) SON(1) DJF(1) -0.3-0.6-0.9 Fig. 7 The leading first SVD mode patterns between U200 (a, m s 1 ) and SSTA (b, K) in JJA. The 90 % (95 %) confidence levels are light (dark) shaded structure in the mid-latitudes (Fig. 8a). Note that the amplitude of the positive zonal wind anomaly south of 40 N is greater than that of the negative zonal wind anomaly north of 40 N. Such a feature is consistent with the greater strength of the meridional temperature gradient anomaly south than north of 40 N, according to the thermal wind relation. In the temperature field (Fig. 8b), a zonal mean cold anomaly center appears at 40 N and a zonal mean warm anomaly appears in the tropics. This leads to a greater increase of meridional temperature gradient south of 40 N and a decrease but with less strength of meridional temperature gradient north of 40 N. Therefore, a question of how the M1 ASWJ pattern forms is transferred into a question of how the vertically integrated meridional temperature gradient anomalies, or more specifically how the tropical warming and mid-latitude cooling anomalies, are induced. Figure 8b shows a tropospheric cooling at 40 N. Consistent with this tropospheric cooling is a local negative geopotential height anomaly (Fig. 8c). This negative height anomaly has a minimum in 200 hpa so that its vertical profile satisfies the hydrostatic relationship with the temperature. The negative height center at 200 hpa is in phase with the positive vorticity center in situ, confirming the quasi-geostrophic nature. For M2, a positive (negative) zonal wind anomaly appears north (south) of 45 N, implying a negative vorticity anomaly in situ. The negative vorticity center is co-located with a tropospheric warming anomaly below 200 hpa and a positive geopotential height anomaly that increases with altitude below 200 hpa.

2682 Y. Du et al. Fig. 8 Latitude-height cross sections of composite zonal wind (a, d, m s 1 ), temperature (b, e, K) and geopotential height (c, f, gpm) averaged from 40 E to 160 E in JJA for M1 and 80 E 160 E for M2. The 90 % (95 %) confidence levels are light (dark) shaded (d) (e) (c) (f) It is also interesting to note that the mid-latitude cooling (warming) in the eastern part of domain associated with M1 (M2) overlaps with the underlying cold (warm) SSTA in North Pacific, suggesting a possible impact of local SSTA to the atmosphere. Alexander et al. (2002) suggested that atmosphere ocean coupling in the north Pacific could significantly modify atmospheric circulation anomalies in boreal summer. Lau and Nath (2006) pointed out that the climatological monsoon trough over the subtropical western Pacific might facilitate positive feedbacks between the atmosphere and ocean. Zhu and Sun (2006) found that the North Pacific SST anomaly could enhance the upper troposphere westerly jet over the western North Pacific. This local SSTA forcing process will be further examined in the subsequent numerical experiments in which the SSTA in the mid-latitude North Pacific is specified. As shown in Fig. 3f, M2 exhibits a zonal dipole pattern of the 200 hp geopotential height anomaly. To examine a possible upstream effect on the cyclonic vorticity anomaly associated with in the west pole of M2, we plot the horizontal distribution of composite meridional wind, relative vorticity and wave-activity flux fields (Fig. 9). For quantitatively estimating the energy propagation of a stationary Rossby wave, we calculate the wave activity flux formulated by Takaya and Nakamura (2001) W = 1 (3) 2 ū( ψ 2 x ψ ψ ) ( ) xx +ῡ ψ x ψ y ψ ψ xy ( ) ( ) Ū ū ψ x ψ y ψ ψ xy +ῡ ψ 2 y ψ ψ yy where Ū(ū, ῡ) is the climatological JJA mean wind field, and ψ is the anomalous stream-function field (deviation from the climatological JJA mean field). The observational analysis clearly illustrates that there is a Rossby wave train, which emanates from North Atlantic all the way southeastward to China (Fig. 9a). The typical wavelength of this stationary Rossby wave is 80 in longitude. This wave train forms an alternated cyclone-anticyclone-cyclone pattern; as a result, there is an anomalous

Interannual variability of the Asian subtropical westerly jet in boreal summer and associated 2683 Fig. 9 Horizontal distribution of composite meridional wind (a, m s 1 ) and relative vorticity (b, shading, 10 5 s 1 ) and wave-activity flux (vector, m 2 s 2 ) at 200 hpa for M2. The vector with the scales less than 0.5 m 2 s 2 is omitted. The 90 % (95 %) confidence levels are light (dark) shaded in cyclonic vorticity center at 60 E, 40 N, which corresponds well to the negative geopotential height center in the west pole shown in Fig. 3f. Figure 9b shows that there are pronounced eastward and southward wave activity fluxes associated with the wave train, indicating that the Rossby wave energy propagates southeastward in the region. Physically, the change of the upper-tropospheric zonal wind should link closely with the change of meridional temperature gradients in the region. To examine which part of the global SSTA (shown in Fig. 5c, f) is responsible for the formation of the anomalous meridional temperature gradient, we conduct idealized numerical experiments using ECHAM4.6, in which we specify a SSTA field in various domains. In a control experiment, the model is integrated for 15 years with specified climatologic monthly SST. In the subsequent four sensitivity experiments, the model is integrated for the same length, forced by the summer SSTA spatial distribution in (1) tropical Pacific and IO (20 S 20 N, 40 E 80 W, named EXP_M1a and EXP_M2a), (2) mid-latitude North Pacific (20 N 60 N, 120 E 80 W, named EXP_M1b and EXP_M2b), (3) tropical Pacific and IO and mid-latitude North Pacific (20 S 60 N, 40 E 80 W, named EXP_M1c and EXP_ M2c), and (4) tropical and North Atlantic (20 S 60 N, 80 W 0 W, named EXP_M1d and EXP_M2d) from Fig. 5c, f. The difference between the sensitivity and the control experiment reflects how the regional SSTA affects the ASWJ. Figure 10 shows the simulated 200 hpa zonal wind fields in all these sensitivity experiments. Just by eye, one may see that the combined tropical Pacific and IO and midlatitude North Pacific SSTA forcing (EXP_M1c and EXP_ M2c) can well capture the structures in observed M1 and M2 modes (Fig. 10c, g). To assess quantitatively the simulation results, we rely on the pattern correlation coefficients between the simulated and observed U200 anomalies for M1 and M2 over the region of 20 N 60 N, 40 E 160 E. As shown in Table 3, positive pattern correlation coefficients appear in EXP_M1a, EXP_M1c and EXP_M2c. To sum up, for M1 mode, a maximum pattern correlation (0.73) occurs in EXP_M1c. For M2 mode, a maximum correlation (0.26) appears in EXP_M2c. This indicates that the combined effect of the tropical and mid-latitude North Pacific SST anomalies is critical in forming the dominant upper-tropospheric subtropical jet patterns in boreal summer. The numerical results above suggest that the tropical Pacific and IO SSTA forcing only (EXP_M1a and EXP_M2a), the mid-latitude North Pacific SSTA forcing only (EXP_M1b and EXP_M2b), or the tropical and North Atlantic SSTA forcing only (EXP_M1d and EXP_ M2d) cannot generate the observed M1 and M2 structures (Fig. 10; Table 3). For example, in EXP_M1b and EXP_M2b, the east pole over the northwest Pacific is well reproduced, but zonal wind anomaly over the continent is not. As a result, a negative pattern correlation coefficient of U200 anomalies appears over the region of 20 N 60 N, 40 E 160 E for both the cases. On the other hand, the mid-latitude North Pacific SSTA plays an important role in influencing the local atmospheric circulation. Table 3 indicates that the best simulations of the ASWJ patterns result from EXP_M1c and EXP_M2c, in which a combined SSTA pattern over tropical Pacific and IO and mid-latitude North Pacific is specified (Fig. 11a, d). Figure 11 shows the spatial distribution of simulated 200-hPa zonal wind, precipitation and 500 200-hPa mean temperature difference fields between EXP_M1c and CTRL and between EXP_M2c and CTRL, respectively. In M1, with a specified warm SSTA in the tropical Pacific and IO and a cold SSTA in the mid-latitude WNP, the model is able to reproduce a 200-hPa zonal wind anomaly pattern similar to the observed (Fig. 3a). As expected, a positive precipitation anomaly appears over the tropical center eastern Pacific in response to El Nino like warming in the eastern equatorial Pacific (Fig. 11b). This El Nino induced heating excites

2684 Y. Du et al. (e) (f) (c) (g) (d) (h) Fig. 10 The 200-hPa Zonal wind differences (contour, m s 1 ) between sensitivity experiments and CTRL for M1 and M2, respectively. The 200-hPa zonal wind larger than 25 m s 1 is shaded (with an interval of 2.5) Table 3 Pattern correlation coefficients between the simulated and the observed 200 hpa zonal wind fields over the region of 20 N 60 N, 40 E 160 E for M1 and M2, with * (**) denoting 95 % (99 %) confidence level Tropical Pacific and IO SSTA forcing experiment Mid-latitude North Pacific SSTA forcing experiment Tropical Pacific and IO and mid-latitude North Pacific SSTA forcing experiment Tropical Atlantic and North Atlantic SSTA forcing experiment M1 0.29** 0.39** 0.73** 0.55** M2 0.22* 0.50** 0.26** 0.51** warm Kelvin waves eastward in the troposphere, resulting in a tropospheric warming in the tropical belt (Wallace and Kousky 1968; Gill 1980). This El Nino effect, along with the tropical IO warming effect, causes a zonally oriented tropical warming in the middle-to-upper troposphere, leading to an increase of meridional temperature gradients near 30 N (beyond 40 N; Fig. 11c). The thermal wind relation requires that an upper-tropospheric westerly (easterly) jet anomaly appears south (north) of 40 N. Thus, the numerical sensitivity experiments above point out the importance of both the warm SSTA in the tropics and the cold SSTA in mid-latitude North Pacific in determining the variation of the 200-hPa zonal wind anomaly. The simulated 200-hPa zonal wind field associated with M2 shows a northeastward tilting structure (Fig. 11d). To the west of 80 E, a positive (negative) center appears to south (north) of 45 N. Compared to the observation, the transition latitude shifts slightly northward, possibly due to the model bias in the climatologic jet axis location. To the east of 80 E, a positive (negative) center appears to north (south) of 40 N, consistent with the observed. The axis of the tilting zonal wind is dynamically consistent with the

Interannual variability of the Asian subtropical westerly jet in boreal summer and associated 2685 (d) (e) (c) (f) Fig. 11 Differences of 200-hPa zonal wind (a, d, m s 1 ), precipitation (b, e, mm day 1 ), temperature between 500 and 200 hpa (c, f, contour, K) and meridional temperature gradient (c, f, shaded, K (5 lat) 1 ) between EXP_M1c, EXP_M2c and CTRL, respectively. Blue boxes in c, f represent the same domain in a, d axis of tilting meridional temperature gradient maximum (red color in Fig. 11f). A warm (cold) temperature anomaly appears to the southeast (northwest) of the axis. It is likely that the middle and upper tropospheric warm temperature anomaly is attributed to large-scale positive precipitation anomalies that extend from tropical IO, South China Sea, to western Pacific at 30 N (Fig. 11e). A further study is needed to confirm this hypothesis. As shown in Fig. 9, the Rossby wave train emanated from North Atlantic may also contribute to the uppertropospheric zonal wind anomaly associated with M2 in the west pole. Figure 12 shows the horizontal distribution of the simulated 200-hPa meridional wind anomaly and associated wave activity flux in EXP_M2c and EXP_M2d. In EXP_M2c, an upper-tropospheric wave train pattern emanates from the North Atlantic, passing though the north of Europe and mid-latitude Asia. The southeastward wave train resembles the observed pattern shown in Fig. 9a. The wave activity flux field (Fig. 12b) confirms that the wave energy propagates southeastward from Europe to East Asia. In EXP_M2d, although an upper-tropospheric wave train emanating from the North

2686 Y. Du et al. (c) (d) Fig. 12 Same as Fig. 9, but for the simulated difference fields (left panel) between EXP_M2c and CTRL and (right panel) between EXP_M2d and CTRL (contour interval: 0.5, zero line omitted) Atlantic appears, but the wave train is mainly eastward and less southeastward. An interesting issue is how the tropical Pacific and IO heat sources excite the high-latitude wave train. A recent study by Ding et al. (2011) suggested a Global Circumglobal Teleconnection (CGT) pattern and a western North Pacific-North America (WPNA) pattern in Northern Hemisphere summer, in which tropical heating in Indian and western North Pacific monsoon regions can stimulate the two teleconnection patterns in 200 hpa geopotential height field. The CGT and WPNA patterns connect, through global jet stream waveguide, tropical monsoon heat sources and mid-latitude perturbation centers including North Pacific and Atlantic. Therefore, it is likely that the wave train pattern seen in EXP_M2c is triggered remotely by anomalous heat sources in the tropical WNP and IO. While the Atlantic SSTA can also contribute to the wave train development, its effect seems shifting to further north. The cause of this northward shift is currently under investigation. To sum up, the formation of the first leading ASWJ mode (i.e., the M1 mode) is likely a manifestation of zonally oriented extratropical atmospheric temperature gradient changes in response to both the warm SSTA in tropical Pacific and IO and the cold SSTA over mid-latitude western North Pacific. The formation of the second leading ASWJ mode (i.e., the M2 mode) is attributed to the combined effects of high-latitude Rossby wave train, tropical heating and mid-latitude North Pacific SSTA forcing. 6 Conclusion and discussion Using 54-year NCEP/NCAR reanalysis data, we investigated the dominant interannual modes of 200 hpa zonal wind anomalies in boreal summer. It is found that the first leading EOF pattern (M1) is characterized by a north south dipole structure, while the second leading mode (M2) is characterized by a quarter-pole structure over the Asian Continent. While M1 represents the meridional displacement of ASWJ, M2 represents the southwest northeast tilting of ASWJ. A power spectrum analysis reveals that M1 has a significant peak at 4.9-year periods, whereas M2 has a significant peak at 3.6- and 7.7-year periods. The atmospheric circulation associated with the two modes are characterized by an equivalent barotropic structure, that is, wind anomalies in the lower and mid-troposphere resemble those in the upper troposphere, with their amplitude increasing with altitude. As a result, the vertical shear of the zonal wind anomaly is primarily controlled by the upper tropospheric wind component. Atmospheric

Interannual variability of the Asian subtropical westerly jet in boreal summer and associated 2687 temperature anomalies are in a thermal wind relationship with the vertical shear anomalies. In M1, a zonally symmetric negative temperature anomaly appears along the climatologic jet axis near 40 N, whereas in M2 a zonal dipole temperature pattern occurs. M1 is associated with an increase (decrease) of precipitation in the Yangtze River (North China), whereas M2 is associated with an east west asymmetric rainfall pattern in China, with less (more) rainfall in the central and eastern (western) China. The SSTA pattern associated with M1 resembles that of the developing phase of El Nino, with warm SST anomalies in the tropical Pacific and IO and cold SST anomalies in midlatitude North Pacific. In contrast, M2 is associated with a La Nina decaying phase, with a cooling in tropical eastern Pacific but a warming in the tropical western Pacific and North Pacific. A latitude-height cross-section analysis shows that the anomalous cooling in the middle latitudes and the anomalous warming in the tropics associated with the M1 mode are primarily driven by the cooling effect in the mid-latitudes of the ocean-atmospheric interaction and convective heating in the tropics. A positive geopotential height (associated anticyclonic vorticity) anomaly in the middle latitude in east pole of the M2 mode is caused by anomalous tropospheric warming in situ, whereas a negative height (associated cyclonic vorticity) anomaly in upper troposphere in the west pole of the M2 mode is induced by southeastward emanation of a Rossby wave train in Eurasia Continent. Idealized ECHAM4.6 AGCM experiments are further conducted to understand the mechanism through which the SSTA influences the ASWJ patterns. It is found that the SSTA in both the tropical Pacific and IO and the midlatitude North Pacific is essential to form the M1 mode. On one hand, El Nino like warming induces the tropospheric warming around the tropical belt. On the other hand, a cold SSTA in the middle latitude causes the tropospheric cooling in situ. For M2, a zonal oriented tropical heating appears in the tropical western Pacific monsoon trough, South China Sea and tropical IO, in response to the decaying La Nina SSTA forcing. This heating, along with the warm SSTA in North Pacific, leads to a southwest northeast tilted meridional temperature gradient anomaly, which is responsible for the tilted structure of zonal wind anomaly in M2. The tropical heating, meanwhile, triggers a mid-latitude Rossby wave train emanating from North Europe to East Asia. In this study, based on idealized atmospheric model experiments, we demonstrate the relative roles of regional SST anomalies in causing anomalous temperature and zonal wind changes in mid-latitude Asia. It is worth mentioning that the model has a systematic bias in simulating the mid-latitude wave train associated with M2. The simulated wave train shifts slightly northward. As a result, the vorticity anomaly in the west pole is biased to the north. A direct diagnosis approach is to analyze the upper-tropospheric zonal momentum budget, in order to reveal specific processes that give rise to the zonal wind changes. We tried this direct approach but found that there was a large bias in the upper-tropospheric zonal momentum budget, possibly due to a large error in estimate pressure gradient forcing terms over large terrain regions and errors in estimate subscale mixing/diffusion terms in the strong jet region. Acknowledgments This study was supported by China National 973 Project 2015CB453200, NSFC Grants 41205063 and 41475084, Jiangsu Province Grant BK2012888 and BK2011831, and Jiangsu Shuang Chuang Team. The International Pacific Research Center (IPRC) is partially sponsored by the Japan Agency for Marine-Earth Science and Technology (JAMSTEC). This is SOEST contribution number 1127, IPRC contribution number 1127 and ESMC number 054. References Alexander MA, Blande I, Newman M et al (2002) The atmospheric bridge: the influence of ENSO teleconnections on air sea interaction over the global oceans. J Clim 15:2205 2231 Cherchi A, Navarra A (2003) Reproducibility and predictability of the Asian summer monsoon in the ECHAM4-GCM. Clim Dyn 20:365 379 Ding Q, Wang B, Wallace JM et al (2011) Tropical-extratropical teleconnections in boreal summer: observed interannual variability. J Clim 24:1878 1896 Dong M, Zhu WM, Wei FY (1987) The characteristics of zonal winds at 500 hpa in Eurasian region and its relation to the weather in China. J Appl Meterol Sci 2(2):166 173 Dong M, Yu JR, Gao ST (1999) A study on the variations of the westerly jet over East Asia and Its relation with the tropical convective heating. Chin J Atmos Sci (in Chinese) 23(1):62 70 Du Y, Zhang YC, Xie ZQ (2009) Impacts of the zonal position of the East Asian westerly jet core on precipitation distribution during Meiyu of China. Acta Meteorol Sin 23(4):506 516 Duan AM, Wu GX (2009) Weakening trend in the atmospheric heat source over the Tibetan Plateau during recent decades part II: connection with climate warming. J Clim 22:4197 4212 Fu XH, Wang B, Li T (2002) Impacts of air sea coupling on the simulation of mean Asian summer monsoon in the ECHAM4 model. Mon Weather Rev 130:2889 2904 Gill AE (1980) Some simple solutions for heat-induced tropical circulation. Q J R Meteorol Soc 106:447 462 Jiang DB, Wang HJ, Lang XM (2005) Evaluation of East Asian climatology as simulated by seven coupled models. Adv Atmos Sci 22(4):479 495 Kalnay E, Kanamitsu M, Kistler R et al (1996) The NCEP/NCAR 40-year reanalysis project. Bull Am Meteorol Soc 77:437 470 Krishnamurti TN (1979) Compendium of meteorology part 4: tropical meteorology. World Meteorology Organization, Geneva Kuang XY, Zhang YC (2006) The impacts of position abnormalities of the East Asia subtropical westerly jet on summer precipitation in the middle-lower reaches of the Yangtze River. Plateau Meteorol (in Chinese) 25(3):382 389 Lau NC, Nath MJ (2006) ENSO modulation of the interannual and intraseasonal variability of the East Asian Monsoon a model study. J Clim 19:4508 4529 Li T, Wang B (2005) A review on the western North Pacific monsoon: synoptic-to-interannual variabilities. Terr Atmos Ocean Sci 16:285 314