By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
 High temperature extremes (HTEs) have received increasing attention due to their rising impacts on human mortality, regional economies, and ecosystems. In this study, HTEs are defined as days with daily maximum temperature above the 90th percentile derived from a specific climatological period. Occurrences of summer HTEs in Southeast China (south of 35°N and east of 105°E) associated with atmospheric anomalies are investigated. Two key domains in the upper level that are associated with HTE variation, the “exit” and the “tail” of the East Asian Jet Stream (EAJS), are identified. Poleward displacement of the exit is associated with warming tropospheric temperatures over East Asia and tends to be linked with high HTE frequency, while enhancement of the tail is associated with cooling tropospheric temperatures in the northern Pacific and tends to be linked with low HTE frequency. A possible reflection effect of the stratosphere on cool summers in Southeast China is proposed. Furthermore, these two domains are in essence two sectors of the phase-locked circumglobal teleconnection (CGT) pattern in the Northern Hemisphere. Linkages are found between HTEs in Southeast China and precipitation anomalies in the Indian summer monsoon region, and also in extratropical regions such as northeastern Europe. These teleconnections are set up through the CGT pattern associated with the westerly jet in the midlatitudes. These findings may be a source of variability and predictability of HTEs in Southeast China.
 High temperature extremes (HTEs), referring to extreme events with temperatures over specific thresholds, have been occurring with increasing frequency, consistent with expectations that such events will become more common under the background of global warming. In China, many previous studies have focused on the trends of extreme indices based on observational records [Ding et al., 2010; Ding and Qian, 2011; Qian and Lin, 2004; Qian et al., 2011; Su et al., 2006; Wei and Chen, 2009; Yan et al., 2002; Yan et al., 2011; You et al., 2008; You et al., 2010; Zhai and Pan, 2003; Zhang et al., 2008]. These studies have recorded different trends, with uncertainties in their exact patterns, in different areas of China. For instance, Zhai and Pan  suggested that the number of hot days showed a slightly decreasing trend for China as a whole. Zhang et al.  found a significant upward trend in the frequency and intensity of HTEs in the western and northern regions of the Yellow River basin, while trends in eastern China were generally not significant.
 Some studies have revealed that strengthening anticyclonic circulations and positive anomalies of geopotential height centered over the Yellow Sea and Korean Peninsula may result in rapid warming over Eurasia, which will further contribute to increasing HTEs in China [Ding et al., 2010; Gong et al., 2004; You et al., 2010; Yan et al., 2011]. They have further suggested that the western North Pacific Subtropical High (WNPSH) contributes to this anticyclonic anomaly [Ding et al., 2010; Gong et al., 2004]. In practice, the East Asian Summer Monsoon (EASM) interacts with the El Niño-Southern Oscillation (ENSO) on an interannual time scale, and this interaction has a strong contribution to the climate in East Asia [e.g., Huang et al., 2004; Chan and Zhou, 2005; Zhou and Chan, 2007; Zhou et al., 2006; Wang et al., 2008; Yuan et al., 2008a; Li et al., 2011; Feng et al., 2011; Wang et al., 2011; Li and Zhou, 2012; Chen et al., 2012; Yuan et al., 2012]. Recent studies have revealed a “capacitor effect” of the Indian Ocean [e.g., Yang et al., 2007; Yuan et al., 2008b, 2008c; Xie et al., 2009]. The Indian Ocean warming induced by El Niño can persist into the summer following El Niño through internal air-sea interaction [Du et al., 2009], triggering an anticyclone anomaly in the northwestern Pacific and having an impact on near-surface temperature [Hu et al., 2011], and hence HTEs [Hu et al., 2012], in China.
 The low-level anticyclonic anomaly excited by the Indian Ocean warming develops in the northwest Pacific, propagating a wave train to the midlatitudes of East Asia and forming a cyclonic anomaly there, which has been called the East Asia-Pacific pattern [Huang and Lu, 1989; Huang and Yan, 1999] or the Pacific-Japan pattern [Nitta, 1987]. This meridional teleconnection pattern connects variability in the tropics to the midlatitude climate system, such as the East Asian Jet Stream (EAJS). Many efforts have been made to investigate the variability of the EAJS in boreal summer and its connections to the EASM [e.g., Lau et al., 2000; Liang et al., 2001; Lu, 2004; Yang et al., 2004]. Meanwhile, the effects of the coupling of the WNPSH and the EAJS on the climate of China, especially on the precipitation over eastern China [e.g., Tao et al., 1958; Zhang and Guo, 2005; Zhu et al., 2011; Lin et al., 2010], have long been recognized. Other studies have investigated the influences of the summer EAJS on near-surface air temperature [Wang et al., 2013]. For example, Yu et al.  suggested that the surface-cooling trend downstream of the Tibetan Plateau in recent decades has been accompanied by southward shifts in the EAJS and a weakening EASM. Sun et al.  indicated that geopotential height anomalies at middle and high levels in EAJS regions link the summer North Atlantic Oscillation (NAO) with summer near-surface temperatures in China. However, how HTEs in China are associated with the EAJS has not been well documented in previous studies.
 The westerly jet stream travels around the Northern Hemisphere (NH) in midlatitudes in the tropopause. The circumglobal teleconnection (CGT) pattern was first revealed along with its Rossby waveguides in boreal winter [Hoskins and Ambrizzi, 1993; Hsu and Lin, 1992]. This theory was then extended to boreal summer, and a relatively weaker circumglobal waveguide belt was found [Ambrizzi et al., 1995]. The summer CGT pattern has received increasing attention in recent decades [Lu et al., 2002; Ding and Wang, 2005; Ding et al., 2011; Wang et al., 2012]. The CGT tends to couple with Indian summer monsoon (ISM) rainfall anomalies and is more likely to occur in summers preceding the peak phases of the ENSO cycle [Ding et al., 2011]. The ISM-EASM teleconnection [Wang et al., 2001] has been addressed as one portion of this global teleconnection. More importantly, climate anomalies (rainfall and near-surface temperature anomalies) are associated with the CGT pattern along its route [Ding and Wang, 2005], including climate in China [Huang et al., 2011].
 The above review suggests that most previous studies have focused on the long-term trend of extreme indices. Some studies have investigated the interannual variability of HTEs and their linkage to the atmospheric system, but the issue is still not fully understood. The purpose of this study is to investigate atmospheric linkages from the extratropics to the interannual variability of HTEs in east-central and southeast China. A brief description of the data and analysis methods used in this study is given in the next section. Section 3 discusses the main atmospheric factors associated with HTEs in the study domain. Linkages of the EAJS with HTEs are identified in section 4, and impacts of the CGT on HTEs are investigated in section 5. The final section summarizes our main findings along with some discussion.
2 Data and Definition
 In the study of climatic extremes, homogenization of observation data is important. In this study we use the China homogenized historical temperature (CHHT) data set [Li and Yan, 2009], which includes homogenized daily temperature series from 549 stations in China spanning from 1960 to 2008. The stations are densely distributed in eastern China but are scarce on the Tibetan Plateau and in the Tarim Basin in western China, as shown in Figure 1a. This homogenized data set has been widely used in temperature-extreme studies in China in recent years [Ding et al., 2010; Hu et al., 2011; Qian et al., 2011; Yan et al., 2011]. The study domain, south of 35°N and east of 105°E, is also shown in the box in Figure 1a. This domain covers Southeast China and the region downstream of the Yangtze River valley and has 204 gauge stations (black dots in Figure 1a).
 Extreme events are easy to recognize but difficult to define [Diaz and Murnane, 2008]. The high threshold used to define extreme events can be chosen in many different ways. In this study, the daily maximum temperature (DMT) from the CHHT, with seasonally varying thresholds rather than arbitrary fixed thresholds, is used to define HTEs. The 90th percentiles of DMT derived from a climatological period of 1960–1999 are taken as the thresholds of hot days. For percentile transformation, we adopt the empirical formula method introduced by Beard  and described in detail by Jenkinson . A binomial filter is applied to the series of percentiles to remove synoptic turbulence.
 Near-surface temperature (a2m), tropospheric temperature, geopotential height, and wind velocity in pressure levels from the National Centers for Environmental Prediction-National Center for Atmospheric Research (NCEP/NCAR) [Kalnay et al., 1996; Kistler et al., 2001] are used for analysis of atmospheric circulation anomalies associated with HTEs. Sea surface temperature (SST) data is from the National Oceanic and Atmospheric Administration (NOAA) extended reconstructed SST (ERSST) [Smith et al., 2008], which was constructed using the most recently available International Comprehensive Ocean-Atmosphere Data Set (ICOADS) SST data. NOAA-ERSST version 3 data is provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA, from their Web site at http://www.esrl.noaa.gov/psd/.
 The present study focuses on interannual variability. The time series of HTE and all atmospheric fields involved in the analysis are averaged over the 3 months of June, July, and August (JJA), so pronounced seasonal cycles are not included. Furthermore, linear trends in all fields are removed to isolate the impacts of circulation anomalies from the effects of global warming.
3 Atmospheric Factors Associated With HTEs
 The spatially averaged time series of HTE frequency in JJA in the study domain is shown in Figure 1b. We first examine some common atmospheric factors that may contribute to HTEs. Correlations of HTE frequency with these factors are shown in Figure 2. HTEs are highly associated with near-surface air temperature anomalies, especially when a relative definition of “extreme” is used. Beyond local effects, teleconnections are also found over the Pacific in the near-surface temperature (Figure 2e). Anomalies over the tropical ocean may be associated with SST anomalies induced by tropical oscillations. Anomalies in the lower troposphere (Figure 2c) are similar to those near the surface. In the upper troposphere (Figure 2a), correlation in the low latitudes disappears, while correlation in the midlatitudes is enhanced, and a significant correlation belt from northern Africa to the northwestern Pacific is found, which implies that HTEs may be associated with atmospheric anomalies in the upper levels of the midlatitudes. This is also evident in the geopotential height fields (Figures 2b and 2d).
 Except for temperature convections and circulation anomalies, atmospheric sinking motions lead to the warming of air by adiabatic heating, and negative precipitation anomalies may also result in increasing HTEs. However, negative correlation between local precipitation and HTE frequency is weak (Figure 2f) and is significant only in a small area over the Yangtze River Delta and the southeast coast, while a large domain of significant positive correlation is found in the northwestern Pacific, east-central Asia, and the northwestern Indian Ocean, indicating teleconnections between diabatic heating related to increasing precipitation in these regions with temperature anomalies in eastern China. Local correlation between downward motion (omega) and HTEs is also not significant (figure not shown). In addition, the upstream signal over Eurasia is evident in Figures 2a and 2b, which will be further investigated in section 5.
 To further examine the connection between HTEs and temperature and circulation anomalies, composite analyses are conducted. As the mean of HTE frequency is 10 days/year, positive and negative anomaly years are selected as having more than 14 days/year and fewer than 6 days/year, respectively, as shown by the dashed lines in Figure 1b. Selected years are listed in Table 1.
Table 1. Positive and Negative Anomaly Years of HTE Frequency for Composite Analysis
 Composite temperature anomalies at three levels (upper troposphere composite at 600–300 hPa, lower troposphere at 925–700 hPa, and the near-surface) in high- and low-frequency HTE years are shown in Figure 3. It is especially noteworthy that in contrast to the correlation results, composite temperature anomalies in the upper level are larger and more significant than those near the surface, implying that upper-level warming (cooling) may have a greater contribution to the increasing (decreasing) occurrence of HTEs. The composite result also shows that, in high-frequency years, the anomaly signal is stronger over east-central Asia, while in low-frequency years, it is stronger over the northern Pacific. The composite geopotential height anomalies (Figure 4) are consistent with temperature. These results suggest that a tropospheric cooling, associated with a cyclonic anomaly in the tropopause above the northern Pacific near the Bering Sea, is connected with a low frequency of HTEs. Connections between HTEs and temperature and circulation anomalies over the midlatitudes of East Asia and the northern Pacific will be further investigated in the next section.
4 Changes in HTEs Associated With the EAJS
4.1 Key Domains in the EAJS
 The westerly jet stream in the tropopause is one of the most important atmospheric systems in the subtropics. HTE frequency is highly correlated with temperature and circulation anomalies in the upper levels (Figures 2-4), which leads to our consideration of the relationship between the EAJS and HTEs. Correlation and composite analysis investigating the linkage between HTEs and jet streams are shown in Figure 5. Figure 5a presents correlation, while Figures 5b and 5c present the composite of 200 hPa zonal wind against HTE frequency. Shading indicates 95% significance levels (r test for correlation and student's t test for composite). A dipole structure of 200 hPa zonal wind anomalies over east-central Asia, positive in the north and negative in the south and divided around 40°N, is found in both the correlation and high-frequency years of the composite (Figures 5a and 5b). The negative anomaly belt along 30°N extends from the northern Indian Ocean to the East China Sea. The counterpart positive signal in the north is comparable to the negative one. This dipole structure of zonal wind in the tropopause is associated with the meridional displacement of the EAJS noted previously [e.g., Lin and Lu, 2005]. However, this dipole structure of zonal wind anomalies over east-central Asia nearly disappears in the composite of low-frequency HTE years (Figure 5c). Meanwhile, the teleconnection signal from the tropopause of the northern Pacific associated with cool summers in eastern China is enhanced compared to hot summers.
 The EAJS varies in both intensity and location on seasonal and interannual time scales [e.g., Yang et al., 2002]. Here we focus only on the interannual time scale. Zonal wind at 200 hPa is used to represent the interannual variability of the jet stream. Figure 6 shows the climatological location (thick black lines) and standard deviations (red lines) of the subtropical jet stream in the NH in boreal summer. To identify the linkage between the EAJS and HTEs over eastern China, we examine the variability of the EAJS in two domains (blue-dashed boxes in Figure 6) based on the correlation and composite results above. The first domain, over east-central Asia, enclosing 90°–140°E and 30°–50°N, covers the eastern region of the EAJS core and is traditionally called the “exit.” The second domain, over the northern Pacific, enclosing 160°E–160°W and 30°–50°N, is called the “tail” of the EAJS. High standard deviation areas are also taken into consideration in the selection of these two domains. To distinguish the spatial-temporal features of jet stream behaviors, empirical orthogonal function (EOF) analysis is applied.
 The leading EOF patterns of the exit and tail domains of the EAJS are shown in Figure 7. For the exit, the first three leading modes account for 42.4%, 22.6%, and 12.7% of the total variance, respectively. The first EOF mode (Figure 7a) exhibits a meridional out-of-phase structure, with the zero line at around 40ºN, where the climatological JJA axis of the EAJS is located (Figure 6). The loading with positive (negative) values is to the north (south) of 40ºN. This implies that the summer EAJS undergoes meridional displacement toward the pole. The second mode (Figure 7b) shows an anomaly center loading northwest of Bohai Bay. It appears that this pattern of the 200 hPa zonal wind exhibits the intensity changes of the summer EAJS. The third mode (Figure 7c) exhibits a tilted pattern with two positive values orienting in the southeast-northwest direction and two negative values centering in the northeast-southwest direction. This pattern may represent the interannual variation in the zonal location of the summer EAJS.
 For the tail domain, the first three leading modes account for 50.5%, 22.3%, and 8.6% of the total variance, respectively. In contrast to the exit domain, the first mode here represents the intensity, while the second mode stands for the meridional displacement of the jet stream. The first two leading patterns change with the meridional location of the EOF domain, depending on the coverage of the domain by the variability region of the jet. In addition, only the first two leading EOF modes stand out from the others in both the exit and tail domains, according to North et al. . The first two principal components (PCs) of the exit and the tail are referred to as the four EAJS indices: Exit1, Exit2, Tail1, and Tail2, hereafter. The following discussion will focus on these four PCs and their linkage with summer temperature.
 Linear correlation of HTE frequency with Exit1, Exit2, Tail1, and Tail2 is 0.45 (exceeds the 99% confidence level), 0.15, −0.33 (exceeds the 95% confidence level), and 0.10, respectively (Table 2). This implies that only the first leading PCs of the exit and tail of the EAJS have potential linkages with HTEs over eastern China. To further investigate the teleconnection of the EAJS and HTEs over Southeast China, a correlation map of each station in the study domain with Exit1 and Tail1 is given in Figures 8a and 8b, respectively. Green triangles represent stations above the 90% confidence level. There are 142 and 99 stations above this confidence level for Exit1 and Tail1, respectively. Figure 8a shows that significant positive correlation of Exit1 and HTEs tends to occur in the northwest. Along the southeast coast, regions of no correlation or even negative correlation appear. Meanwhile, Figure 8b shows that significant negative correlation of Tail1 and HTEs tends to occur south of the Yangtze River. If EOF analysis is applied to HTEs in our study domain (figure not shown), we find that the first two leading EOF modes are well separated from the others. These two leading EOFs account for 45.7% and 13.9% of the total variance, respectively. The first leading EOF shows a homogeneous pattern in the whole domain, while the second leading EOF shows a northwest-southeast opposing pattern. This suggests that in the southeast coastal region, near-surface temperature variation involves more complicated driving mechanisms due to interaction between the tropics and extratropics. But that is outside the scope of this study.
Table 2. Linear Correlation Coefficients of HTE Frequency With EAJS Indices, CGTI, and MFI
4.2 Coupled Thermodynamics of the EAJS and Tropospheric Temperature
 Regressions of 200 hPa geopotential height against the EAJS indices are shown in Figure 9 to demonstrate the circulation anomalies associated with the jet stream. Exit1 represents meridional displacement of the jet exit, associated with positive anomalies of geopotential height centered over the Yellow Sea and Korean Peninsula along the jet axis of 40°N. Previous studies have identified this anticyclonic anomaly center as an important contributor to HTEs over eastern China [Ding et al., 2010; Gong et al., 2004; You et al., 2010]. This positive anomaly of geopotential height in the upper level seems more correlated with the EAJS than with the WNPSH. Negative anomalies are located to the north and south of the positive anomaly center. It is noteworthy that large significant negative anomaly areas appear in the south, which implies important contributions from the tropics in the meridional displacement of the EAJS. Regression against Tail2 (Figure 9d) shows a similar sandwich-like structure of a trough-ridge system, but with a smaller anomaly area.
 Exit2 represents the intensity of the EAJS. It induces a dipole structure of anomalies, with cyclonic anomalies north of the jet axis and anticyclonic south of it. It is known that as air enters the jet core from the west, it is accelerated by the Coriolis force, resulting in a convergence to the north and a divergence to the south. In the jet exit region, the opposite happens, with divergence to the north and convergence to the south. As a result, air rises in the northern jet exit and sinks in the south [Uccellini and Koch, 1987]. Results in Figures 9b and 9c are consistent with this understanding. Regression patterns at 500 hPa (not shown) are similar to Figure 9. Note that cyclonic anomalies in the north are stronger than the counterpart anticyclonic anomalies in the south in both Figures 9b and 9c. This may be due to a stronger Coriolis effect at higher latitudes. In addition, the northeastward tilt of the climatological jet tail, as can be seen in Figure 6, may lead to much larger anomalies in the north compared to the south in this region.
 The thermo-structure associated with air motion in the exit region of the jet is shown in Figure 10, which presents details of temperature anomalies associated with Exit1 and Tail1 in both the upper and lower troposphere. Generally, positive (negative) temperature anomalies are associated with anticyclonic (cyclonic) circulation anomalies. This is also consistent with the thermodynamics of the jet stream in the exit region, warming and sinking to the south, while cooling and rising to the north. More importantly, temperature anomaly patterns associated with Exit1 and Tail1 resemble the correlation patterns in Figures 2a and 2c, suggesting that the exit and tail are two key domains that are connected with summer near-surface temperature in Southeast China. Increasing frequency of HTEs is linked to pole ward displacement of the exit associated with anticyclonic anomalies and tropospheric warming over the Yellow Sea and the Korean Peninsula. Decreasing frequency of HTEs is connected with enhancement of the tail intensity associated with cyclonic anomalies and tropospheric cooling from the northern Pacific to the midlatitudes of East Asia.
 To further investigate the thermodynamics of the westerly jet stream variation, vertical composite anomalies of temperature and geopotential height associated with HTEs are given in Figures 11-13. Figure 11 is a composite of a latitude-height cross section averaged over the longitudes between 90°E and 130°E in high-frequency HTE years. Over the midlatitudes of East Asia along the axis of the EAJS (around 40°N), warming in the troposphere is associated with cooling in the stratosphere, resulting in an increase in the maximum temperature contrast (Figure 11a) and pressure (Figure 11b) in this portion of the tropopause. This pressure increase along the normal jet axis strengthens the pressure gradient to the north of it, and hence a maximum increase in westerlies occurs north of 40°N through geostrophic balance.
 Figure 12 differs from Figure 11 not only because it is a composite of low HTE years but also because it is a longitude-height cross section averaged over a latitude range (40°N–60°N). This is for the convenience of viewing the intensification (not displacement) of the jet associated with temperature changes. In cool summers (Table 1), significant negative temperature anomalies occur in the troposphere, from the sea surface of the north-central Pacific to the tropopause, as shown in Figure 12a. The strongest temperature contrast (Figure 12a) and geopotential height anomaly (Figure 12b) are centered on the tropopause of the north-central Pacific, implying a maximum enhancement of wind speed here. Figure 13 is the same as Figure 11 but averaged over the longitudes between 160°E and 160°W in low-frequency HTE years. Together with Figure 12, a three-dimensional view of temperature and circulation anomalies corresponding to a low frequency of HTEs is presented. Tropospheric cooling over the north-central Pacific is obviously to the north of the subtropical jet axis. Meridional displacements of the tail may also occur (pressure gradient changes in Figure 13b) but are not significant.
 Significant negative geopotential height (Figure 12b) and temperature (Figure 13a) anomalies are found in the stratosphere, implying a possible linkage between the stratosphere and tropospheric cooling. Decreases (cyclonic anomalies) in geopotential height and increases in static stability (cooling in the troposphere and warming in the stratosphere) as shown in Figures 12 and 13 indicate increases in potential vorticity and a falling region in the midlatitudes of the tropopause [Yu et al., 2004], where downward flows from the stratosphere to the troposphere exist [Holton et al., 1995]. The troposphere affects the stratosphere through upward-propagating atmospheric waves, and changes in the stratosphere can feed back to the troposphere by creating fluctuations in the strength of the polar vortex formed by high-latitude winds [Baldwin et al., 2003]. Meanwhile, lower boundary forcing from the northern Pacific may also contribute to cool summers (Figure 14). Negative SST anomalies appear near the Bering Sea, a transition region between the cold Arctic air mass to the north and the warm maritime air mass to the south.
5 Circumglobal Teleconnection in the Northern Hemisphere
 Two key domains in the tropopause that are associated with the EAJS and are highly connected to summer near-surface temperature over eastern China were identified in the previous section. The question that immediately arises is as follows: Are there any other teleconnection signals in the extratropics that show linkage with HTEs over eastern China? Figure 15a shows the correlation map of HTE frequency with 200 hPa geopotential height, the same as in Figure 2b but for the entire NH. The correlation pattern shows an evident wave-like structure over the NH. Except for the previously identified two key domains over East Asia and the northern Pacific, positive correlations also appear over North Africa, northeastern Eurasia, and North America. This wave train pattern is close to the circumglobal teleconnection (CGT) revealed in previous studies [e.g., Hoskins and Ambrizzi, 1993]. The CGT pattern and stationary Rossby waves along with the upper-level westerly jet in the subtropics of the NH are notable in boreal winter. This teleconnection pattern is weaker in boreal summer but still significant and highly associated with climate variability in the NH [e.g., Ambrizzi et al., 1995; Lu et al., 2002].
 To explore the CGT pattern over the NH, EOF analysis is applied to the 200 hPa geopotential height. Only the first two leading EOF modes are well separated from the others, according to North et al. . These two leading EOF patterns explain 25.5% and 12.5% of the total variance, respectively, and are shown in Figures 15b and 15c. The first leading EOF pattern shows a homogeneous positive pattern over the whole NH, while the second leading EOF pattern presents positive signals along the midlatitudes. This positive route starts from North Africa; crosses the Eurasian continent, with a maximum center over East Asia and the northern Pacific; crosses North America; and extends to the Atlantic. A comparison of Figures 15a and 15c immediately shows that the wave-like correlations in Figure 15a are entirely within the positive route of EOF2 in Figure 15c. More importantly, wave structures in the correlation map are totally consistent with the high variation centers of the second leading EOF pattern, which represents the CGT pattern [Ding and Wang, 2005]. Interdecadal change in the boreal summer CGT has been recently revealed, and remarkable intensity change has been found over the North America-Atlantic-Europe sector [Wang et al., 2012]. But the five-sector pattern of the phase-locked CGT remains unchanged.
 The EOF1 at the 200 hPa geopotential height represents the summertime large-scale background mean flow over the NH, which is supposed to include more signals of global warming. We define a summertime mean flow index (MFI) using the normalized PC1 at the 200 hPa geopotential height. The explained variance of this PC1 is 34.5% and has a significant correlation of 0.33 with HTEs before linear trends are removed. This correlation drops to 0.28 and is not significant at the 95% confidence level after linear trends are removed (Table 2). This reflects the impact of global warming on near-surface HTEs, and it can be isolated with linear trends removed in all data fields. We define the normalized PC2 at the 200 hPa geopotential height in the NH as the CGT index (CGTI). Linear correlations of the CGTI with HTE frequency and the EAJS indices are given in Table 2. The CGTI shows a significant correlation with most of the EAJS indices, as expected. These correlations are a consequence of the results in Figures 9 and 15c. More importantly, a significant (at the 95% confidence level) correlation of 0.33 is obtained between the CGTI and HTE frequency.
 A map of the correlation between the CGTI and HTE frequency at each station is given in Figure 8c. It shows that significant positive effects of the CGT pattern on HTEs are confined between the Yangtze River and the South China Sea, in a latitude range of 20°N to 30°N, similar to the pattern in Figure 8b but with opposite values. Also of note is that 101 of the 204 stations are statistically significant in this correlation map (green triangles).
 Figure 16a is the same as Figure 2f, showing the correlation of HTEs with JJA precipitation but for the whole NH. Negative correlation between HTEs and local precipitation is weak. Positive correlations track circumglobally in the NH, close to the path of the phase-locked CGT pattern. However, these correlations are below the 95% confidence level in most regions in the composite analysis (figure not shown). In the composite of high-frequency HTE years, only a region of anomalies over northeastern India and the Indian Ocean remains significant.
 Tropical precipitation is a measure of tropical forcing [Wallace et al., 1992]. Previous studies have revealed that precipitation in the ISM region maybe the key latent heat source of the CGT. For example, Rodwell and Hoskins  pointed out that increasing ISM rainfall tends to excite baroclinic Rossby wave response to the northwest. Joseph and Srinivasan  also proposed a similar argument. Yasui and Watanabe  noted in a modeling study that the heating anomaly most responsible for the CGT pattern is located over the eastern Mediterranean region. Lin  suggested that ISM convective activity is associated with a global pattern that has a far-reaching connection in both hemispheres. The ISM has global linkages with the CGT because the Asian high, one of the key action centers of the CGT in subtropical Asia, is tightly coupled with ISM rainfall [Ding and Wang, 2005; Wang et al., 2012].
 Based on the correlation in Figure 16a, an ISM region is chosen (dashed box). Correlations of precipitation anomalies in this region at the 200 hPa geopotential height are shown in Figure 16b. This figure further reveals that the ISM has global linkages with the CGT. Here we see a linkage between the HTE over Southeast China and diabatic heating by precipitation in the ISM region. They are connected by an atmospheric bridge in the upper level of the midlatitudes along with the wave train of the westerly jet: the CGT pattern. However, in comparing Figure 16b to Figures 15a or 15c, we find that the maximum difference appears in northern Eurasia and extends northward to the Arctic. We argue that this difference may be linked to precipitation anomalies in northeastern Europe, for which we can find evidence from Figure 16c. This implies that local diabatic heating in the extratropics may also contribute to CGT patterns.
 A large area of significant correlation is found along the northern Pacific coastal regions. In the composite of low-frequency HTE years (figure not shown), only anomalies in these regions remain above the 95% confidence level. This highlights the contribution of the key domain associated with cool summers that we identified in section 4 and suggests that these precipitation anomalies are associated with anomalies in the Pacific sector of the CGT.
6 Conclusion and Discussion
 This study investigates different atmospheric factors associated with the interannual variability of HTEs in Southeast China. It is evident that signals from the upper levels of the extratropics are linked with HTEs. We suppose these circulation and temperature anomalies in the upper troposphere to be associated with variations in the EAJS, including its meridional displacement and intensity. Meanwhile, on an interannual time scale, linkages of local precipitation and downward air motion with HTEs are relatively weak.
 Two key domains in the upper level that are highly associated with the HTE variation in Southeast China, the so-called exit and tail of the EAJS, are identified. The pole ward displacement of the EAJS in the exit domain is associated with warming tropospheric temperatures around 40°N (climatological axis of the westerly jet), while the intensification of the EAJS in the tail domain is associated with cooling tropospheric temperatures in the northern Pacific and midlatitudes of East Asia. In other words, northward displacement of the EAJS in the exit domain seems to have a greater contribution to high HTE years, and enhancement of the tail has a greater contribution to low HTE years. Variability of the EAJS and tropospheric temperature are highly interrelated, and the key domains of the EAJS identified in this study may be a source of variability and predictability of HTEs in Southeast China.
 On one hand, it is well known that the subtropical westerly jet in the tropopause is a result of a strong temperature gradient, and hence pressure gradient, in the narrow flow duct. This means that the jet stream is a “passive” atmospheric system of temperature change and Coriolis force deflection. On the other hand, it is notable that gravity-inertia waves are believed to originate downstream of the jet stream as part of a geostrophic adjustment process [Uccellini and Koch, 1987]. This geostrophic adjustment excites a zonally oriented stationary Rossby wave train along the upper-level jet, resulting in anomalous air temperatures [Sun et al., 2008]. It is easy to understand why the pole ward displacement of the EAJS in the exit is related to more HTEs: variation in the subtropical jet, its position in particular, is an important “indicator” of temperature advection, which is a crucial source of temperature [e.g., Yu et al., 2004], and hence temperature extreme, variability [e.g., Mahlstein et al., 2012]. We presented the three-dimensional structure of temperature and circulation anomalies associated with cool summers in section 4 (Figures 12 and 13) and raised the possibility that tropospheric cooling associated with the westerly jet intensifies in low-frequency HTE years and is linked to the stratosphere. Anomalies in the lower stratosphere have nonlocal dynamical effects that change winds in the troposphere [Haynes et al., 1991], and weak changes to these winds could be amplified by interactions with “synoptic-scale” waves [Baldwin et al., 2003]. The intensification of the EAJS in the tail over the north-central Pacific is probably the amplifier for the downward reflection of the stratosphere to the troposphere and the near-surface, as suggested by Figures 9 and 10. These two key domains in the upper level that are associated with near-surface temperature are in essence two sectors of the phase-locked CGT pattern in the NH. As the CGT pattern travels along with the waveguide of the subtropical jet stream, it influences precipitation and near-surface temperatures below it. This study reveals that the CGT is linked with HTEs between 20°N and 30°N in eastern China, i.e., mainly in the subtropical regions south of the EAJS axis. Previous studies have suggested that diabatic heating of precipitation in the ISM region may be an important heat source of the CGT [Ding and Wang, 2007; Joseph and Srinivasan, 1999; Lin, 2009; Yasui and Watanabe, 2010]. We reveal a linkage between HTEs in Southeast China and rainfall anomalies in the ISM region, and also in extratropical regions such as northeastern Europe. These teleconnections are set up through the CGT pattern associated with the westerly jet in the NH, which also provides potential predictability of HTEs.
 This research is supported by 973 Basic research program Grant 2009CB421401, the National Nature Science Foundation of China grant 41175079, and Macao Meteorological and Geophysical Bureau (SMG) project 9231048.