The mid-Pacific trough (MPT), occurring in the upper troposphere during boreal summer, acts as an atmospheric bridge connecting the climate variations over Asia, the Pacific, and North America. The first (second) mode of empirical orthogonal function analysis of the MPT, which accounts for 20.3% (13.4%) of the total variance, reflects a change in its intensity on the southwestern (northeastern) portion of the trough. Both modes are significantly correlated with the variability of tropical Pacific sea surface temperature (SST). Moreover, the first mode is affected by Atlantic SST via planetary waves that originate from the North Atlantic and propagate eastward across the Eurasian continent, and the second mode is influenced by the Arctic sea ice near the Bering Strait by triggering an equatorward wave train over the northeast Pacific.
A stronger MPT shown in the first mode is significantly linked to drier and warmer conditions in the Yangtze River basin, southern Japan, and the northern United States and wetter conditions in South Asia and northern China, while a stronger MPT shown in the second mode is associated with a drier and warmer southwestern United States. In addition, an intensified MPT (no matter whether in the southwestern or the northeastern portion) corresponds to more tropical cyclones (TCs) over the western North Pacific (WNP) and fewer TCs over the eastern Pacific (EP) in summer, which is associated with the MPT-induced ascending and descending motions over the WNP and the EP, respectively.
The mid-Pacific trough (MPT) is one of the most prominent upper tropospheric circulation features in the Northern Hemisphere. It is also known as the tropical upper tropospheric trough (Krishnamurti 1971; Sadler 1976, 1978) and the midoceanic trough over the central Pacific (Colton 1973; Kousky and Gan 1981). The trough extends from the midlatitudes into the subtropics along a northeast–southwest direction (Sadler 1967) and peaks in the upper troposphere during boreal summer. Although it is believed that the MPT may be affected by many factors such as convection in the intertropical convergence zones (Webster 1972; Hoskins and Rodwell 1995) and radiative cooling over the North Pacific (Zhao et al. 2007), the exact mechanisms have never been fully understood.
The MPT can be decomposed into stationary and transient eddy parts (Ferreira and Schubert 1999). The stationary part features a climatological appearance of the trough, while the transient part refers to the synoptic-scale disturbances. Previous effort has been devoted to investigations of the interaction between the MPT and tropical cyclones (TCs) (Kelley and Mock 1982; Price and Vaughan 1992; Whitfield and Lyons 1992; Patla et al. 2009). On the one hand, the MPT affects the origin and development of TCs by modifying the vertical wind shear in which TCs are embedded (Gray 1968; Pfeffer and Challa 1992; Montgomery and Farrell 1993; Hu et al. 2017). On the other hand, the MPT disturbances provide the initiation to TCs by extending the trough into the middle and lower troposphere in some cases (Hodanish and Gray 1993). In turn, TCs could also affect the MPT through Rossby wave energy dispersion associated with the TCs (Ferreira and Schubert 1999). Patla et al. (2009) proposed a conceptual model about how MPT influences TC tracks, offering a useful operational guidance for weather forecasters. Recently, Wu et al. (2015) found that the MPT experienced a pronounced westward shift during the past three decades in all of the available reanalysis datasets, suppressing TC genesis in the eastern portion (east of 145°E) of the western North Pacific due to the enhanced vertical wind shear associated with the MPT shift.
Moreover, the MPT can act as an atmospheric bridge connecting the eastern and western hemispheres as well as tropical and extratropical regions (Lu et al. 2017). Recent studies have suggested possible associations of the MPT with various atmospheric teleconnections. For example, Zhang et al. (2005) noted that a stronger South Asian high (SAH) was accompanied by a stronger and more extensive western North Pacific (WNP) subtropical high (WNPSH). The enhancement and expansion of WNPSH could then push the MPT eastward, which subsequently results in a stronger Mexican high. Thus, the MPT bridges the climate variations in Asian–Pacific–North American regions. In addition, as the MPT tilts along a southwest–northeast direction over the subtropics, it could promote exchanges of heat and momentum between different latitudes (Magaña and Yanai 1991; Waugh and Polvani 2000). Indeed, the MPT can sometimes extend so far south into the tropics that the equatorial winds over the south of MPT are switched from the trade easterlies to westerlies, forming the so-called equatorial westerly duct that allows atmospheric disturbances to propagate from one hemisphere into the other (e.g., Webster and Holton 1982; Webster and Yang 1989; Tomas and Webster 1994; Yang et al. 2015).
Therefore, understanding the variations and characteristics of the MPT is helpful for untangling the interconnected weather events and climate anomalies in the Asian–Pacific–North American regions. Although the synoptic features of the MPT and its influence on TCs have been studied previously, the interannual and interdecadal variations of the MPT and their climate impact are poorly understood so far. In fact, the MPT is often regarded as a passive circulation feature induced from the upstream SAH, with its impact on the climate in the remote regions being considered negligible. Given this belief, few studies have been conducted about the multiscale variations of the MPT, leaving an unanswered question about the interannual variability of the MPT and its climate effect.
Given that the MPT connects the Asian and North American climate, as well as the tropics and the midlatitudes, the present study focuses on the interannual variation of MPT, its possible effect on the Asian–Pacific–North American climate, and the responsible physical mechanism. Although ENSO is the dominant mode of the climate variability in this region, ENSO alone cannot fully explain the variability of the MPT (Wang and Wu 2016). It is important to seek and analyze other climate impacting factors for possible improvement over the MPT-related climate forecasts.
The rest of the paper is organized as follows. Section 2 describes the datasets and analysis methods applied. Section 3 discusses the climatological and interannual features of the MPT. Sections 4 and 5 present the associated atmospheric circulation anomalies and discuss the possible factors underlying the MPT variability. The effects of MPT on Asian–Pacific–North American climate are illustrated in section 6, followed by a summary and a further discussion in section 7.
2. Data and method
This study uses the monthly mean sea surface temperature (SST) data from the National Oceanic and Atmospheric Administration (NOAA) extended reconstructed SST version 4 (Huang et al. 2015). The SST dataset has a horizontal resolution of 2° × 2° and is available from 1854 to the present. The monthly precipitation data are obtained from the Global Precipitation Climatology Center (GPCC; Schneider et al. 2011), with a horizontal resolution of 1° × 1° for the period of 1901 to the present. Surface air temperature anomalies are acquired from the Goddard Institute for Space Studies (GISS/NASA), with a horizontal resolution of 2° × 2° for the period of 1880 to the present. The monthly mean sea ice concentration data provided by the Japanese Reanalysis (JRA; Hirahara et al. 2014), with a horizontal resolution of 1° × 1° for the period of 1850 to the present, are available from the website at https://www.esrl.noaa.gov/psd/data/gridded/data.cobe2.html. The JRA sea ice data merged the satellite observations from the Nimbus-5 Scanning Multichannel Microwave Radiometer (SMMR) from October 1978 to July 1987 and the Special Sensor Microwave Imager (SSM/I) and the Special Sensor Microwave Imager/Sounder from August 1987 to December 2010. The JRA sea ice data for the Arctic Ocean before the satellite era from January 1870 to November 1978 were taken from Walsh and Chapman (2001), who used various regional datasets based primarily on ship reports and aerial reconnaissance. To explore the impact of the MPT on tropical cyclones, we also analyze the tropical cyclone datasets from the Joint Typhoon Warning Center (JTWC) and the Tropical Prediction Center (TPC) best-track reanalysis from the website at http://weather.unisys.com/hurricane/. In this study, we examine the linear relationships between the dominant modes of MPT and the tropical cyclone genesis numbers over the WNP and the east Pacific (EP) during the midsummer.
The atmospheric data, including geopotential height and three-dimensional velocity at multiple levels, are obtained from the NCEP–NCAR Reanalysis (Kalnay et al. 1996). The geopotential height and horizontal wind obtained from ERA-Interim for 1979 to the present are also used to compare with the NCEP–NCAR data. The Niño-3.4 index is downloaded from the NOAA Environmental Systems Research Laboratory’s (ESRL’s) website at http://www.esrl.noaa.gov/psd/data/climateindices. The analysis period of this study is 1948–2014 for NCEP–NCAR and 1979–2014 for ERA-Interim.
We apply an empirical orthogonal function (EOF) analysis (vector EOF) on the mean horizontal winds to capture the dominant modes of the MPT during the midsummer (July and August) when the MPT reach the maximum intensity (see Fig. 1), with an equal area weighting at each grid point by multiplying the square root of the cosine of latitude due to the decrease in grid area toward the pole (North et al. 1982). This vector EOF analysis enables a more thorough analysis of wind components than scalar analyses (Legler 1983). Linear regression or correlation analysis is performed by regressing or correlating the anomalous atmospheric circulation, precipitation, and other physical quantities on/with the principal components (PCs) corresponding to the vector EOF. This study focuses on the midsummer [i.e., July–August (JA)], and so each set of successive 2-month season is considered with physical fields leading or lagging the PCs [e.g., May–June (MJ) and June–July (JJ) lead the PC in JA by two months and one month, respectively, whereas August–September (AS) lags the PC in JA by one month].
Considering the dominant correlation between ENSO and other variables, in most of the correlation/regression analyses the ENSO signal captured by the Niño-3.4 index is removed via a simple linear regression method. It should be noted that it is impossible to remove the influence of ENSO completely by only the Niño-3.4 index (Hu et al. 2016a,b), as the ENSO signal is not always linear and stationary in time (Capotondi 2015). To investigate the interannual variations of MPT and its climate effects, we also remove the linear trend of MPT before conducting the correlation/regression analysis. We used the Student’s t test to assess the statistical significance of results from correlation analysis. A value of correlation above 0.33 (0.24), for a 36-yr (65-yr) length of 1979–2014 (1948–2014), is used to estimate the 95% confidence level. Here the correlation (r) is defined as
in which and are the sample mean of and from i = 1 to i = n [n is the sample size, the same as in Hu et al. (2016c)]. The threshold of correlation coefficient |r0| for the 95% confidence level is defined as , where can be acquired by checking tables about the critical values of the Student’s t distribution with n − 2 degrees of freedom (von Storch and Zwiers 2002).
3. Climatological and interannual features of MPT
Figure 1 shows the monthly climatological patterns of 200-hPa geopotential height (H200) and its departures from the annual mean for all 12 months. Within the annual cycle, the largest changes occur over East Asia and the North Pacific, accompanied by a shift in the phase of planetary-scale stationary waves. In winter (from December to February), a distinct trough and ridge are seen over East Asia and the west of North America, respectively, due to strong land–ocean thermal contrast (Smagorinsky 1953; Wang and Ting 1999). From late spring to early summer (May and June), the atmospheric circulations undergo a dramatic change when the East Asian trough and the Alaskan ridge weaken rapidly, while the Asian summer monsoon and the SAH are quickly established (Yeh et al. 1959; Webster and Yang 1992). In northern midsummer (JA), the SAH, the MPT, and the North American ridge dominate the northern subtropics, due mainly to monsoon heating (Wang and Ting 1999). In particular, the SAH and the MPT evolve approximately in tandem, and both reach their maximum intensities in July and August. After September, the SAH and the MPT weaken quickly and winter patterns appear again.
Figure 2 shows more details about the climatological characteristics of the MPT and its surrounding atmospheric circulation in midsummer (JA), when the MPT reaches peak intensity. As seen in Fig. 2a, the upper-level SAH and MPT correspond to the lower-level Asian monsoon low and WNPSH, respectively, implying that the atmospheric circulations in the northern subtropics are completely out of phase in the vertical direction. Figure 2b shows the vertical cross section of zonally asymmetric height and vertical velocity along 30°N. It indicates that vigorous upward motions occur under the western side of the MPT due to the monsoon-related convection over East Asia and the WNP. In addition, strong descending motions are found under the eastern side of the MPT, which seem to be linked to the Pacific subtropical high.
Figure 3a shows the standardized deviations of upper-tropospheric horizontal winds, in which the largest values appear over the midlatitudes and the equatorial Pacific regions, which could be attributed to the intrinsic large variability of the westerly jet streams (Joseph and Sijikumar 2004; Schiemann et al. 2009) and the large variability associated with ENSO (Yang and Hoskins 2013; Kim et al. 2014). There is, however, substantial variability over the subtropical Pacific, coinciding with the MPT area. Overall, one expects the subtropics to be stable with small variability due to the relatively weaker coupling between the upper troposphere and lower troposphere (Charney 1963). Thus, it is interesting to explore why large variability exists in the MPT region and whether such variability is forced locally or remotely.
Figure 3b further shows the frequency distribution of MPT centers during the period of 1948–2014. The MPT centers are identified as the locations of minimum zonally asymmetric geopotential height at 200 hPa. In general, the MPT centers are distributed along a northeast–southwest direction, concentrating around 30°N of latitude, with a maximum frequency of 9 years located at 32.5°N and 152.5°W. In extreme cases, MPT centers can shift as far northeast as to 40°N and 130°W. It can also be seen that the MPT centers tend to shift more frequently in the east–west direction than in the south–north direction.
4. Dominant modes of MPT and associated circulation anomalies
a. Vector EOF modes of MPT variability
To better classify MPT variability, an area-weighted vector EOF analysis is applied to the 200-hPa horizontal winds. The analysis domain is 15°–50°N and 160°E–120°W, as outlined by the blue box in Fig. 3a, covering the area of both maximum subtropical variability and climatic position of the MPT. The first two modes respectively explain 20.4% and 13.1% of the total variance (26.5% and 14.7% for the ERA-Interim dataset, 1979–2014), which are statistically distinguishable from the rest of the eigenvectors according to the rule by North et al. (1982).
The spatial patterns of the first two modes of MPT are presented in Figs. 4a and 4c, in terms of the correlation coefficients between H200 and the principal components (PC1 and PC2). The first mode is dominated by significant anomalies in the southwestern side of the MPT (Fig. 4a), which reflects a change in MPT intensity over the WNP. The second mode, however, is related to the significant anomalies in the northeastern side of the MPT (Fig. 4c), and it reflects the change in MPT intensity over the eastern North Pacific. The corresponding time series of the two modes are illustrated in Figs. 4b and 4d. The first mode exhibits strong interannual signals, combined with multidecadal variations with a shift from a predominantly negative phase to a positive phase around the 1990s. On the other hand, the second mode shows a significant decreasing trend for the period of 1948–2014. The time series calculated from the ERA-Interim dataset are also included in Figs. 4b and 4d, which are consistent with those from the NCEP–NCAR product.
We also repeat the vector EOF analysis for a larger domain covering most of the Northern Hemisphere (20°–80°N, 20°E–60°W). The spatial patterns related to the first mode of the MPT are very similar to those patterns from the analysis of the smaller domain (with a correlation coefficient of 0.77 between the two corresponding PC1s), although the variance explained drops to 10.5%. The previous second mode of MPT variability falls to the third mode for the extended domain (with a correlation coefficient of 0.52 between the corresponding PC2s). These results further demonstrate the robustness of the first two modes of the MPT.
b. Atmospheric circulation anomalies associated with the dominant modes of the MPT
Figure 5 shows the regressions of atmospheric circulation and Plumb wave activity fluxes (WAFs; Plumb 1985) onto the dominant modes of the MPT, in which the linear trends of the PCs and the field variables have been subtracted. The first mode corresponds to higher pressure across the entire midlatitude belts (Fig. 5a), covering North Africa, South and East Asia, the North Pacific, North America, and the Atlantic. Several isolated positive centers are identified along the entire midlatitude belts, with most prominent anomalies over East Asia, the North Pacific, North America, and the North Atlantic. Meanwhile, remarkable anomalies of alternative cyclonic and anticyclonic circulations are seen over the western North Pacific. In comparison, the second mode of MPT reveals predominant signals over the eastern North Pacific (Fig. 5b), which are oriented in a north–south direction, connecting the tropics and the Arctic region.
Figures 5a and 5b show the association of WAFs with the two dominant MPT modes. The first mode is associated with southward and eastward propagating wave trains, which diverge over East Asia and converge in the southwestern portion of the MPT, resulting in intensification of the trough at its first mode. Similarly, the second mode is related with an equatorward propagating wave train, which originates from the northeastern Pacific and converges in the northeastern portion of MPT, leading to strengthening of the MPT at its second mode. Besides, the second mode of MPT also corresponds to a relatively weak wave train originating from the central Pacific, which could be related to the central Pacific warming that is shown in Fig. 6b. The anomalous WAFs seem to appear at the exit of East Asian westerly jet, which may be related to the meridional displacements and the changes in intensity of the jet stream. As seen in Figs. 5c and 5d, an intensified MPT in the first mode is significantly related to the northward shift of the jet stream over East Asia and the western North Pacific, while the intensified MPT in the second mode corresponds to the weakening jet stream over the eastern North Pacific. These results suggest that the wave trains associated with MPT variability may be sensitive to the subtle westerly jet structure. Thus, we have also computed the WAF proposed by Takaya and Nakamura (2001, hereafter TN01) to understand the impact of uneven basic flow. The regressed patterns of TN01’s WAFs reveal a pattern similar to Plumb’s (1985) WAF (figure not shown), showing two splitting wave trains over the western (first mode) and the eastern (second mode) North Pacific. One propagates eastward along the westerly jet stream and the other propagates southeastward, which affects the variability of the MPT. Compared to Plumb’s WAFs, the equatorward propagating wave trains of TN01 WAFs associated with the MPT seem to be more prominent, implying the possible effect of zonally asymmetric westerlies on MPT variability.
The variabilities of MPT seem to largely originate in extratropical regions. Indeed, when ENSO signals were removed (figure not shown), the features of regression patterns remained almost unchanged. For the first mode, the negative height anomalies over the WNP and sub-Arctic regions experience a slight decrease, and the positive height anomalies over the midlatitudes remain the same and even become somehow stronger, particularly over North America and the Atlantic. For the second mode, the negative height anomalies over the Arctic region become even more significant. These results imply that the dominant modes of MPT variability may be also affected by other forcing sources beyond ENSO.
The regressions of SST anomalies (SSTAs) on PCs of MPT are shown in Fig. 6. Both modes are significantly correlated with the Pacific SST. The first mode is significantly related with a La Niña–like pattern, with negative SSTAs in the central-eastern Pacific. This mode also exhibits a close relationship with the Atlantic SST. The correlation coefficient between PC1 (PC2) and the Niño-3.4 index is −0.465 (0.23), which exceeds the 99% (90%) confidence level, implying that both the first and second modes of MPT are affected by ENSO, to different extents. It should also be noted that the regressed SST patterns in Figs. 6a and 6b seem to represent the eastern-Pacific (EP) type and the central-Pacific (CP) types of El Niño (e.g., Ashok et al. 2007; Kug et al. 2009), respectively. The EP type of El Niño corresponds to northward propagating waves and anomalous high pressure on the southwestern portion of MPT (Wang and Wu 2016), implying weakening of the MPT at its first mode. However, the CP type of El Niño is correlated with northeastward propagating waves (Fig. 5b), which lead to anomalous low pressure on the northeast portion of MPT, suggesting intensification of the MPT at its second mode. Zhang et al. (2012) have investigated the differences in atmospheric teleconnections associated with different types of El Niño, indicating that the EP type of El Niño was associated with wave trains that tilted more in the north–south direction, while the CP type of El Niño was significantly correlated with wave trains that tilted more in the northeast–southwest direction. Their results, although focused on the boreal autumn, were similar to the features shown in Figs. 5a and 5b.
5. Factors and physical mechanisms for the dominant modes of MPT
The maps of lead–lag correlation between global SST and PC1 (also PC2) are shown in Fig. 7, where the linear signal of ENSO is removed by regressing out the Niño-3.4 index from the PCs. The left panels of Fig. 7 indicate that the first mode of MPT is significantly correlated with SSTAs in the equatorial and North Atlantic regions at all lags. In particular, the Atlantic SSTs significantly lead the first mode by more than 4 months. This result implies that the Atlantic SST may be a driver to the variations of the first mode of MPT. In the North Pacific (NP), positive SSTAs are found in the simultaneous correlation. Although the NP SSTAs may also contribute to the variability of MPT by altering the NP storm tracks and triggering planetary waves, we tend to recognize the North Pacific SSTAs as a response to the atmospheric, rather than a cause, because the NP SSTAs associated with the PCs seem to only appear and strengthen during the periods when SST lags MPT.
Moreover, there are significant warmings in the northern Indian Ocean (NIO) and the western Pacific (WP), which tend to lead the MPT by as long as 4 months, evolving gradually from the NIO/WP to the North Pacific. Xie et al. (2009) indicated that the NIO/WP warming could persist through the boreal summer by initiating a series of air–sea interaction, which may also be a source of the North Pacific variability. Indeed, the WP warming is significantly correlated to the so-called Pacific–Japan teleconnection during the boreal summer (e.g., Nitta 1987; Kosaka and Nakamura 2006), which bridges the climate between the Indo-Pacific and North Pacific regions. However, the current study is mainly focused on the Atlantic SST emphasizing its relatively robust signals in intensity and locations.
As seen from the right panels of Fig. 7, the second mode of MPT is only weakly correlated with global SST, except for the North Pacific and Indian Ocean SSTs. It should be mentioned that the Indian Ocean SSTA may be related to the residual signals of El Niño (see Fig. 6b) as it is impossible to remove the nonlinear effect of ENSO completely by a regressed method. Significant signals are mainly observed in the correlations when the MPT leads the North Pacific SST, implying that the North Pacific SSTA is a response to the atmosphere rather than a cause. What factors may be responsible for the second mode of the MPT? There are several reasons for exploring the possible impact of the Arctic sea ice. First, it is shown in Fig. 5b that the geopotential height anomalies associated with the second MPT mode are strongly linked to Arctic signals and an equatorward wave train. Second, the decreasing trend in PC2 is reminiscent of the recent rapid loss of the Arctic sea ice during the boreal summer (Stroeve et al. 2007; Comiso et al. 2008). These reasons lead us to investigate the relationship between the MPT and Arctic sea ice.
Figure 8 presents the lead–lag correlations between the second mode of the MPT and the Arctic sea ice concentration. When the sea ice leads the second mode by 2 months, positive correlations are observed near the Bering Strait. The correlation patterns become more significant and more extensive when the sea ice leads the MPT by 1 and 0 months. It should be noted that both ENSO signals and linear trends have been removed from the PC2, implying that the Arctic sea ice near the Bering Strait plays a potential role in affecting the second mode of the MPT.
Figure 9 shows the lead–lag regression maps of H200 and WAF onto the first two PCs of the MPT. For the first mode of the MPT, significant anomalies of H200 occur over the North Atlantic, North Africa, and the Eurasian continent during the period when the physical fields lead the MPT by 2–3 months. By the period of 1-month lead, anomalous high pressure emerges over the North Pacific and strengthens substantially in JA and AS, which are accompanied by enhanced WAFs. In comparison, significant anomalies of H200 associated with the second mode of the MPT mainly appear over the Arctic regions, leading the PC2 by 2 months, which subsequently induce an equatorward wave train over the eastern North Pacific in JJ and JA. In other words, although significantly antecedent signals associated with the MPT can be found over the Atlantic and Arctic regions, the anomalous wave trains that affect the MPT emerge only one month in advance, implying that the seasonally varying basic flow influences the triggering of anomalous wave trains related to the MPT. Moreover, the intensification of the anomalies over the western (eastern) North Pacific suggests an interaction between the Atlantic (Arctic) forcing signals and the Pacific jet stream variability, or the internal atmospheric dynamics, which plays an important role.
To better explain how the Atlantic SST and the Arctic sea ice influence the MPT, we define two indices from the domain-averaged values: the Atlantic SST index (ASST), calculated as , where A and B represent the regions outlined by the black boxes in Fig. 7 and the Arctic sea ice concentration index (ASIC), calculated as , where C denotes the region near the Bering Strait outlined in Fig. 8. The selections of boxes A, B, and C are based on the lag–lead correlation relationships between the Atlantic SST/Arctic SIC and the PCs. We have tested with different regions such as the entire North Atlantic (0°–55°N, 80°–10°W) and obtained similar results (not shown).
Figures 10a and 10b show the ASST and ASIC indices. As seen in Fig. 10a, except for the warming trends, the ASST also shows profound interannual and multidecadal variations, switching from a predominant negative phase to a positive phase around the mid-1990s. The correlation coefficient between ASST (detrended) and the PC1 is 0.41, exceeding the 95% confidence level. Figures 10c and 10d show the regression maps of H200 and WAF onto the detrended ASST and the interannual component of ASST (by subtracting the 9-yr low filtered and detrended ASST), respectively. Although significant anomalies of H200 associated with the detrended ASST can be observed within the MPT domain (Fig. 10c), the physical mechanism is unclear due to the ambiguity of teleconnection pattern. In comparison, after removing the low-frequent component of ASST, an anomalous wave train is clearly seen over the Eurasian continent (Fig. 10d), which originates from the North Atlantic and eventually affects the southwestern portion of MPT. On the other hand, except for a sharply decreasing trend, the Arctic sea ice also shows considerable interannual variability. The correlation coefficient between ASIC (detrended) and the PC2 is 0.36, which exceeds the 95% confidence level. As seen in Fig. 10e, the ASIC shows significant correlations with H200 and WAF over the eastern North Pacific. A positive Arctic sea ice anomaly can lower tropospheric geopotential height over the Arctic regions due to surface cooling, which subsequently induces an equatorward propagating wave train that deepens the MPT on its northeastern portion.
6. Associated Asian–Pacific–North American climate anomalies
a. Precipitation and surface air temperature
Figure 11 shows the correlation of Asian and North American precipitations with the PCs of MPT, in which the linear trends have been removed from the PCs. As shown in Figs. 11a and 11b, more precipitation over South Asia, northern China, and the eastern United States and less precipitation over the Yangtze River basin, southern Japan, and northern United States are associated with the first mode of the MPT. The features associated with the second mode are quite different. Figures 11c and 11d show that correlations are dispersed and less significant over South Asia, and that precipitation is significantly light over southwestern United States in association with the second mode.
When ENSO signals are removed, the correlations between Asian precipitation and the PCs decrease substantially, especially over South Asia (figure not shown). Nevertheless, the correlation patterns over North America suffer little change. If the ASST (ASIC) signals are removed, however, the patterns of correlation with the PCs become insignificant over North America, but remain similar with the previous over South Asia, implying the important roles of the Atlantic SST and the Arctic sea ice. It is also possible that both precipitation and the MPT patterns may respond to ENSO separately and the link between the MPT and the precipitation over South and East Asia may be weak.
The relationships between surface air temperature (SAT) and the dominant modes of the MPT are shown in Fig. 12. Corresponding to the first mode (Figs. 12a,b), warmer SAT occurs across almost the entire middle latitudes, including the Middle East, East Asia, and the northern United States. Over South Asia, colder SAT is seen with a relation to more precipitation. Associated with the second mode, significantly warmer SAT appears over the southwestern United States, which coincides well with less precipitation over this region. Our analysis reveals a feature (figure not shown) that is identical to the previous: the patterns of correlation between Asian (North American) SAT and the PCs of MPT are strongly modulated by ENSO (Atlantic SSTAs and Arctic sea ice). Indeed, the anomalous atmospheric circulations related to the dominant modes of MPT show a barotropic structure. Under the control of high pressure, enhanced descending motions suppress convection and allow more incoming radiation, which results in less precipitation and warmer surface air temperature, and vice versa.
b. WNP and EP tropical cyclones
Climatologically, the MPT is associated with strong ascending and descending motions under its upstream and downstream portions (Fig. 2b), respectively. Figure 13 shows the correlations between outgoing longwave radiation and the first two PCs of the trough, where both ENSO signals and linear trends have been removed from the PCs based on the ERA-Interim data. An intensified MPT in the first mode is significantly correlated with enhanced convection over the WNP and suppressed convection over the CP. However, an intensified MPT in the second mode corresponds to stronger convection over the WP and the northeastern Pacific, accompanied with suppressed convection over the EP and Central America. Therefore, for both the modes, an intensified MPT tends to contribute upward motions over the WNP/WP and descending motions over the CP/EP, which affect the occurrence of WNP and EP TCs, given that the WNP (EP) TCs mainly occur and develop beneath the upstream (downstream) portions of the MPT.
Figure 14 shows the scatter diagrams of the TC genesis numbers over the WNP/EP and the PCs. It should be noted that the TC samples used in these figures only include the TCs for the period of 1979–2014 due to the relatively weaker relationship between the MPT and the TC genesis numbers for the earlier period, which could be related to the relatively poor data quality before 1979. Again, both ENSO signals and linear trends in the PCs have been excluded. As seen in Fig. 14a, an intensified MPT in the first mode tends to increase the WNP TC numbers, but their correlation is insignificant, with a coefficient of 0.17. However, the first mode of the MPT presents a much stronger relationship with the EP TC numbers and tends to decrease the EP TC numbers (Fig. 14c), with a correlation coefficient of −0.311.
In comparison, the second mode of MPT shows stronger relationships with the TC numbers over both the WNP and EP. Figures 14b and 14d indicate that a strengthened MPT in the second mode favors generation of the WNP TCs but suppresses the occurrences of the EP TCs. The coefficient of correlation between the WNP (EP) TC numbers and the PC2 is 0.45 (−0.33), statistically exceeding the 99% (95%) confidence level with an effective degree of freedom of 33 for a total of 35 seasons (1979–2014).
Therefore, for both the first mode and the second mode, an intensification of the MPT is associated with more TCs over the WNP but fewer TCs over the EP. As shown in Fig. 13, an intensified MPT compels stronger upward (downward) motions in its upstream (downstream) portion, providing favorable (unfavorable) conditions for TC genesis over the WNP (EP). The relationships of the two MPT modes with WNP/EP TCs are different in several aspects. The first mode is more strongly connected with the EP TCs, compared with the WNP TCs. However, the second mode shows a more robust relationship with the WNP TCs than with the EP TCs. These features should be related to the large-scale circulation conditions that are induced by the MPT.
7. Summary and discussion
This study has documented the climatological and interannual characteristics of the MPT including the dominant modes of the trough and their associated climate anomalies. The MPT acts as an atmospheric bridge that connects the climate over Asia and North America, with strong ascending (descending) motions on it western (eastern) side. Climatologically, the MPT peaks at the 200- and 150-hPa levels during midsummer and extends from the midlatitudes to the subtropics in a northeast–southwest direction. It varies in tandem with the SAH.
The MPT exhibits a large interannual variability. Its first (second) mode reflects a change in the intensity of the trough on its southwestern (northeastern) side. Both modes of the MPT are significantly correlated with the tropical Pacific SST. Moreover, the first MPT mode is significantly correlated to the Atlantic SST, and the second MPT mode is linked with the variation of Arctic sea ice near the Bering Strait. On the one hand, the Atlantic warming would increase the upper-level geopotential height, which triggers an eastward propagating wave train across the Eurasian continent, and intensifies the MPT on its southwestern portion. On the other hand, a positive Arctic sea ice anomaly can lower the upper tropospheric geopotential height near the Bering Strait due to surface cooling, which induces an equatorward propagating wave train that deepens the MPT on its northeastern portion.
Associated with an intensification of the MPT on its southwestern side, more (less) precipitation occurs over South Asia and northern China (the Yangtze Basin and southern Japan). Correspondingly, warmer (colder) SAT appears over East Asia (South Asia) due to the control of anomalous high pressure (monsoon convection). However, the second mode of the MPT shows sporadic and less significant relationships with Asian precipitation and SAT. The relationship between Asian climate and the MPT is strongly modulated by ENSO. When ENSO signals are removed, the patterns of correlation between the trough and Asian precipitation/SAT over South Asia and East Asia change remarkably.
The first (second) mode of MPT is also significantly related with the precipitation and SAT over the northern (southwestern) United States, attributed mainly to the Atlantic SST (the Arctic sea ice near the Bering Strait). The patterns of correlation between the MPT and North American climate suffer little influence from ENSO, but are strongly affected by the Atlantic SST and the Arctic sea ice near the Bering Strait.
The numbers of tropical cyclone genesis over the WNP and the EP are also closely related to the dominant modes of MPT. Both the first mode and the second mode of MPT are positively (negatively) correlated with the TC numbers over WNP (EP). An intensified MPT is accompanied by more TCs over the WNP and fewer TCs over the EP. However, the first (second) mode of MPT presents a stronger relationship with the TC numbers over the EP (WNP), compared to the WNP (EP). Overall, the second mode of MPT is more strongly connected with the WNP and EP TC numbers than the first mode.
It should be pointed out that although we have presented the relationship between the dominant modes of MPT and the numbers of Pacific tropical cyclones, the cause-and-effect aspect of this relationship deserves more investigation. The tropical cyclones may also influence the MPT through Rossby wave energy dispersion associated with the TCs (Ferreira and Schubert 1999). However, considering the relative spatial and time scales between the MPT and the TCs, it is more likely that the MPT modulates the TC genesis (Hu et al. 2017). In addition, although the dominant modes of MPT are significantly related to the Pacific and Atlantic SSTs and the Arctic sea ice, the impact from intrinsic atmospheric variability, such as the circumglobal teleconnection, the Pacific–Japan teleconnection, and the Arctic Oscillation, should not be excluded.
The authors wish to thank Profs. Ming Cai, Zhaohua Wu, and Jianhua Lu for a number of constructive discussions. The comments from three anonymous reviewers were helpful for improving the overall quality of this paper. This study was supported by the National Key Research and Development Program of China (2016YFA0602703), the National Key Scientific Research Plan of China (Grant 2014CB953900), the National Natural Science Foundation of China (Grants 41705050, 41690123, 41690120, 91637208, and 41661144019), the National Postdoctoral Program for Innovative Talents (Grant BX201600039), and the CMA Guangzhou Joint Research Center for Atmospheric Sciences.