Diversity of the Coupling Wheels in the East Asian Summer Monsoon on the Interannual Time Scale: Challenge of Summer Rainfall Forecasting in China

The East Asian summer monsoon (EASM) is a three-dimensional circulation system in the tropospheric wind fields and is characterized by lower and upper-level circulations and a subtropical front rainfall belt over East Asia during boreal summer (e.g., Lau and Li, 1984; Tao and Chen, 1987; Ding and Chan, 2005; Zhu et al., 2005, 2011; He and Liu, 2016). The low-level southwesterly flow is associated with the western Pacific anticyclone (WPA), whereas the upper-level northeasterly wind is connected with the South Asian high (SAH). The vertical shear in the meridional flow of the EASM and the diabatic heating released by the monsoonal rainfall conform to the Sverdrup vorticity balance (e.g., Rodwell and Hoskins, 2001; Liu et al., 2004).

The vertical coupling between the WPA and the SAH is essential for the onset of the summer monsoon in the South China Sea and determines the start of the EASM rainy season in May (e.g., Kueh and Lin, 2010; Liu and Zhu, 2016). The westward movement of the WPA and the eastward movement of the SAH enhance the meridional and vertical circulations of EASM and regulate the northward transport of warm air and the moisture flux, which induces anomalous summer precipitation over inland China from June to August (Dao and Chu, 1964; Zhang et al., 2002, 2005; Ren et al., 2007; Liu and Zhu, 2016). The Mongolian cyclone (MC), a cold vortex in the troposphere over Lake Baikal, is most crucial for the transport of the cold air modulating the EASM (e.g., Lau and Li, 1984; He et al., 2007; Zhu et al., 2012; Chen et al., 2017; Liu and Zhu, 2019). The distinct interactions of the MC with the SAH and the WPA determine the subseasonal variability of the EASM and rainfall over East Asia (e.g., Song et al., 2016).

The interannual variability of the WPA is regarded as the dominant factor determining the seasonal anomalies in the EASM due to its physical linkage with the external tropical forcing of El Niño–Southern Oscillation (ENSO) (Huang and Wu, 1989; Zhang et al., 1996; Wang et al., 2000, 2013; Wu and Zhou, 2008; Xie et al., 2009, 2016). However, increasing evidence has shown that the relation between the WPA and ENSO cannot fully explain the anomalous rainfall patterns of the EASM. Recent studies suggest that the EASM is also influenced by an anomalous SAH (Zhang and Wu, 2001; Wei et al., 2015), a meridional shift in the upper-level westerly jet over North Asia (Yang et al., 2002; Zhang et al., 2006; Chiang et al., 2017; Liu et al., 2018), and the Silk Road teleconnection pattern in the mid to high latitudes (Lu et al., 2002; Enomoto et al., 2003; Kosaka et al., 2009; Chen and Huang, 2012; Hong and Lu, 2016). Also, the anomalous year-to-year pattern of summer rainfall in China is determined by a distinct coupling circulation regime (Pang et al., 2019).

In the present study, we seek to understand more comprehensively the effects of the interactions of the WAP, SAH, and MC on the interannual anomalies in the three-dimensional circulation, to investigate the diverse coupling modes of the EASM, and discuss the possible challenges in the seasonal forecasting of rainfall over China based on the dominant modes of SST anomalies.

The atmospheric circulation fields (horizontal and vertical winds, geopotential height, etc.) are from the NCEP–DOE AMIP-II reanalysis, with a horizontal resolution of 2.5° × 2.5° in 17 vertical layers (Kanamitsu et al., 2002). The quality-controlled rainfall in monthly records for 160 in-situ stations is provided by the National Climate Center of the China Meteorological Administration (http://cmdp.ncc-cma.net/). The sea surface temperature (SST) is from ERSST.v4 (Huang et al., 2015), and all SST indices are calculated based on this dataset. The climatology is defined as the arithmetic average during the period 1981–2010. We firstly remove each field’s linear trend during the period 1979–2015, to avoid the impact of global warming (Zhu et al., 2012). Then, we calculate the standardized anomalies of the winds at 850 and 200 hPa, and in-situ rainfall anomalies, over mainland China. Finally, following Wang et al. (2008), we conduct a multivariant empirical orthogonal function (MV-EOF) analysis on this combined field of wind and rainfall anomalies during the boreal summer (June–August).

The rainfall occurs near the subtropical front of the EASM, along with substantial adjustment of the atmospheric static stability. The change in the atmospheric thermal structure follows the latent heat released by the EASM rainfall (Liu et al., 2004). An appropriate indicator of the EASM should include not only the wind fields but also the diabatic heating of the atmosphere. Previous research has shown that the potential vorticity is a good description of the low-level monsoonal flows (Yang and Krishnamurti, 1981), WPA (Rodwell and Hoskins, 2001; Liu et al., 2004), SAH (Wu and Liu, 2003; Ren et al., 2015), and midlatitude cyclone (Hoskins et al., 1985). Here, we apply the vertical component of the isobaric potential vorticity $ \mathrm{P}\mathrm{V}\cong -g(f+\mathrm{\xi })\partial \theta /\partial p $ to indicate the intensity of the WPA, SAH, and MC in the coupling modes of the EASM (Hoskins et al., 1985), where g = 9.8 m s⁻² is the gravitational acceleration, ($ f+\mathrm{\xi } $) in units of s⁻¹ is the absolute vorticity, and $ \partial \theta /\partial p $ in units of K m kg⁻¹ s^–2 denotes the static stability of the atmosphere. Finally, the unit of 10⁻⁶ m² s⁻¹ K kg⁻¹ is convenient in practice and we call it a “PVU” for short. The WPA index (WPAI) anomaly is defined as the average of the potential vorticity anomalies over the region (20°–30°N, 110°–130°E) at 850 hPa, and the SAH index (SAHI) and MC index (MCI) are defined as the average of the potential vorticity anomalies over the regions (25°–40°N, 90°–120°E) and (45°–60°N, 90°–120°E) at 200 hPa, respectively.

The first mode (MV-EOF1) of the EASM accounts for 13.5% of the total covariance and displays a meridional dipole of rainfall anomalies, with more (less) rainfall to the south (north) of the Yangtze River in China. The anomalous winds feature a weakened WPA, SAH, and MC, connected by anomalous westerly winds in the lower and upper levels over East Asia (Fig. 1a). The suppressed WPA and MC at 850 hPa display a cyclonic and anticyclonic circulation prevailing over Southeast and Northeast China, respectively, accompanied by a weakened SAH and MC at 200 hPa over central China and Lake Baikal. The suppressed MC shows a barotropic structure with a vertical tilt to the northwest, which connects the weakened SAH via the easterly anomalies of the EASM around 40°N at 200 hPa. As a result, anomalies in the lower-level northeasterly and upper-level southwesterly emerge south of 40°N, leading to a weaker EASM to the east of 100°E.

Figure 1.  First two MV-EOF modes of the anomalous summer rainfall over mainland China and winds at 850 and 200 hPa during 1979–2015. Panels (a), (b) and (c) indicate the modes of rainfall anomalies and wind anomalies at 850 hPa and 200 hPa of MV-EOF1, respectively. Panels (d), (e) and (f) are similar, but for the MV-EOF2. The gray shading in the middle column indicates topography higher than 3000 m.

The second mode (MV-EOF2) accounts for 10.7% of the total covariance, presenting a typical mei-yu front with above-normal rainfall along the Yangtze River and Southwest China, together with strong interaction between the intensified WPA and the southeastward shift of the MC at 850 hPa. At 200 hPa, the enhanced SAH is coupled with the southeastward shift of the MC over Indochina and the Korean peninsula via the anomalous westerly around 30°N over the Yangtze River to Japan.

These two modes are significantly orthogonal according to the criterion of North et al. (1982). They show two different interactions among the EASM circulation. The first mode is ascribed to the EASM intensity, but the second is associated with the spatial displacement of the EASM. Note that the WPA, SAH and MC anomalies are interconnected and act as a cogwheel comprising the meridional monsoonal flow and the westerly jet between the upper and lower troposphere over East Asia. These two dominant modes represent the diverse coupling wheels of the EASM on an interannual time scale.

The spatial structure of the EASM coupling wheels are well organized in the potential vorticity fields associated with the principal components (PC1 and PC2), with a different vertical shear of the meridional flow along 115°E (Fig. 2). In the first mode, the MC and the WPA are weakened at 850 hPa, presenting negative and positive anomalies of potential vorticity over Northeast China and the western North Pacific, respectively (Fig. 2a). By contrast, the coupling between the MC and the SAH at 200 hPa exhibits a dipole of anomalous potential vorticity over Lake Baikal and central China (Fig. 2b). This coupling wheel accompanies a less stable atmosphere ($ \partial \theta /\partial p>0) $ over East China beneath the positive potential vorticity anomaly in the upper troposphere (Fig. 2c). The mid–lower troposphere thus becomes colder over the East Asian continent, which decreases the land–sea thermal contrast to weaken the meridional monsoonal flow and restrain the above-normal rainfall over southern China in summer (Fig. 1a).

Figure 2.  Correlation coefficients of the vertical component of the potential vorticity (shading; units: PVU), wind anomalies (vectors; units: m s⁻¹), and the vertical gradient of the potential temperature ($ \partial \theta /\partial p $; contours; units: K hPa⁻¹) with (a–c) PC1 and (d–f) PC2 during the period 1979–2015 at (a, d) 850 hPa, (b, e) 200 hPa and (c, f) 115°E of the height–latitude cross section. The solid and dashed contours in panels (c) and (f) indicate positive and negative correlations of $ \partial \theta /\partial p $ exceeding a significance level of 0.05.

In the second mode, a dipole of potential vorticity anomalies extends northeastward south and north of 25°N, suggesting an interaction between the southward shift of the WPA and the enhanced MC at 850 hPa (Fig. 2d). The enhanced MC manifests westward tilting of the positive potential vorticity anomaly in the vertical direction, along with a southward shift in the SAH over Indochina at 200 hPa (Fig. 2e). In this way, the negative anomaly of the upper-level potential vorticity is above the Yangtze River, where the anomaly of the low-level potential vorticity is negative, leading to a more stable troposphere ($ \partial \theta /\partial p<0) $ in situ (Fig. 2f). Consequently, the potential temperature should rise to strengthen the land–sea thermal contrast over the East Asian continent, where the meridional monsoonal circulation is enhanced to support the above-normal rainband along the Yangtze River (Fig. 1d).

The interannual variability of the EASM illustrates the diversity of the coupling wheels in circulation associated with the summer rainfall anomalies in China. To verify the relationship between the anomalous rainfall and the circulation, we apply EOF and MV-EOF analysis to the anomalous rainfall and wind fields at 850 and 200 hPa, respectively (figures not shown). The amplitude of the anomalous rainfall becomes weaker after excluding the wind fields. However, the first two modes of anomalous rainfall are similar to those in the MV-EOF analysis in Fig. 1. The first two principal components (PC1 and PC2) of the rainfall anomalies are significantly positive. The correlations with the anomalous wind counterparts are +0.64 and +0.60, respectively, suggesting a robust linear relationship between the EASM circulation and summer rainfall anomalies in China. The first two modes and their principal components derived from the wind fields show little change before and after excluding the rainfall anomalies.

The correlation of potential vorticity fields with PCs suggest essential roles of the WPA, SAH and MC in regulating the diversity of the coupling wheels of the EASM (Fig. 2). The temporal correlation coefficients (TCCs) of PC1 with the WPAI, SAHI and MCI are +0.20, +0.68 and −0.74, respectively, suggesting dominant roles of the SAH and MC anomalies in the first coupling wheel of the EASM. The TCC between the SAHI and MCI is −0.45, passing the significance test at the 0.05 level. Thus, the SAH and MC intensity show an opposite change to the meridional fluctuation of the westerly jet at 200 hPa.

To verify the diverse regime of the EASM circulation (Figs. 3b–d), we conduct principal components analysis on the WPAI, SAHI and MCI indicated by the potential vorticity. The three leading modes account for 58%, 24% and 18% of the total variance, respectively, corresponding to the different EASM circulation regimes. The first mode (mode 1) shows an out-of-phase variation between the SAHI and MCI (Fig. 3b). The second mode (mode 2) shows a unified in-phase variation of the three circulation indices. In contrast, the third mode (mode 3) shows an inverse change in the WPAI with the SAHI and MCI. PC1 (PC2) in the MV-EOF is significantly correlated with the time series of mode 1 (mode 2), showing a TCC of +0.84 (+0.55). No significant correlation is observed between the time series of mode 3 and PC3. Therefore, the first coupling wheel of the EASM (Fig. 1a) primarily depends on the MC with its more considerable variance and its interaction with the SAH (Fig. 3b). However, the mutual interactions among the WPA, MC and SAH constitute the second coupling wheel of the EASM (Fig. 1b), in which the WPA is most important due to its significant correlation with either mode 2 or PC2.

Figure 3.  (a) Time series of the standardized PC1 (gray bars) and the anomalous circulation indices of the MCI, SAHI and WPAI during the period 1979–2015. (b–d) Three leading modes of the MCI, SAH and WPAI based on principal components analysis. (e) Time series of the standardized PC2 (gray bars) and three leading modes of the MCI, SAHI and WPAI during the period 1979–2015.

The diversity of the coupling wheels of the EASM indicates the complex interannual variability of the EASM. We use the k-means cluster method (Hartigan and Wong, 1979) to validate the statistical results of the EOF analysis and show the existence of diverse regimes of EASM circulation in the period 1979–2015. Since k-means clustering is a method of vector quantization, which can partition the data space into Voronoi cells (Hartigan and Wong, 1979), this method has been applied in the atmospheric and ocean sciences for decades, and the categories are obtained by grouping the cases for their similarities.

Figure 4 presents the first four clusters of the circulation indices and their associated summer rainfall anomalies in mainland China during 1979–2015. Categories 1 and 3 represent the positive and negative phases of the first mode of the coupling wheel of the EASM, respectively (Figs. 1a and b). In this case, the larger changes in the MCI and SAHI are opposite, with a relatively weaker variation in the WPAI (Figs. 4a and e). The composite anomalous rainfall is also reversed between the two categories, corresponding to the positive and negative phases of this circulation regime of the EASM (Figs. 1a, 4b and 4f). This mode occurs in 24 of 37 samples, suggesting its dominant role in the year-to-year variation of the EASM.

Figure 4.  Left-hand column: k-means cluster analysis of the MCI, SAHI and WPAI during 1979–2015. Right-hand column: composite summer rainfall anomalies (units: mm) in mainland China in each category (anomalies exceeding the 0.05 significance level are stippled).

Cluster 2 and its related rainfall anomalies (Figs. 4c and d) reflect the positive phase of the second coupling wheel of the EASM (Fig. 1b), featured by a tight interaction between the SAH and the WPA and with the anomalous rainfall centered along the Yangtze River (Figs. 4c and d). There are seven samples in cluster 2, suggesting a lower frequency of the second mode of the coupling wheel of the EASM. The residual six samples are classified as cluster 4, with the lowest frequency during 1979–2015, showing the reverse change in the SAHI and WPAI and the less summer rainfall over South China. Therefore, the coupling between the MC and the SAH is the leading target of seasonal forecasts of the EASM and shows the importance of the MC and its related cold air activity in regulating the summer rainfall pattern of the EASM.

The interannual variability of the three-dimensional circulation of the EASM shows diverse interactions among the MC, SAH and WPA in the anomalous potential vorticity fields. The first two dominant modes act as the coupling wheels and induce the year-to-year variation in the EASM. The first mode shows the meridional coupling between the MC and the SAH and the westerly jet at 200 hPa, corresponding to a dipole of more and less rainfall over China south and north of the Yangtze River. The second mode shows the vertical interaction between the WPA and the SAH south of 30°N, associated with a tripole rainfall pattern with a positive center over the Yangtze River. The first mode occurs in 24 years, in contrast with 7 years contributed by the second mode, during 1979–2015. Because of the dominant role of the first coupling wheel of the EASM and its frequent occurrence, we should pay more attention to this mode in seasonal forecasts of the rainfall anomalies in China.

ENSO is regarded as a crucial source for seasonal forecasts of the EASM, and the related SST anomalies in the northern tropical Indian Ocean (NTIO) can affect the EASM through either the WPA in the mid-to-lower troposphere (Zhang et al., 1996, 2016; Wang et al., 2000; Xie et al., 2009, 2016) or the SAH in the upper troposphere (Yang et al., 2007; Huang et al., 2011; Liu et al., 2017). The tripole and dipole modes (NATM and NADM) of the North Atlantic SST anomalies are also responsible for changes in the EASM (e.g., Lau et al., 2004; Wu et al., 2009; Zuo et al., 2013; Cui et al., 2015). However, the first mode of the coupling wheel of the EASM shows a weak correlation with the SST anomaly and index of the Niño3.4, NTIO, NATM and NADM in the regions of tropical Pacific, North Atlantic and Indian Oceans in the preceding months (Fig. 5a). By contrast, the second mode is closely associated with the Niño3.4, NTIO and NADM indices, with a maximum positive correlation in April, May and July, respectively (Fig. 5b). Such correlation suggests that the El Niño in winter, the NTIO SST anomaly in spring, and the Atlantic SST anomalies in summer may affect this mode via a relay process, suggesting its higher potential predictability on interannual time scales. However, the first mode shows very low predictability due to its weak correlation with the SST anomalies, particularly for the MC (figure not shown). The prediction of the first mode of the coupling wheel of the EASM is, therefore, a great challenge in forecasting the seasonal climate over China based only on the SST anomalies.

Figure 5.  Lead–lag correlation coefficients of PC1 and PC2 of the coupling wheel of the EASM with the SST anomaly index of the Niño3.4, NTIO, and tripole and dipole modes (NATM and NADM) of the North Atlantic SST anomalies during the period 1979–2015. The SST anomaly index of the NTIO is defined as the SST anomaly averaged over (5°–25°N, 40°–100°E) following Xie et al. (2016). The NATM and NADM indexes are defined as the first and second principal components of the SST anomalies over (0°–60°N, 80°W–0°) during the period 1979–2015 without a linear trend.

Acknowledgements. The authors acknowledge the anonymous reviewers’ helpful suggestions. This study was jointly supported by the National Natural Science Foundation of China (Grant Nos. 41830969, 41775052, 42005011, 41776023 and 42076020), the National Key R&D Program (Grant No. 2018YFC1505904), the Scientific Development Foundation of the Chinese Academy of Meteorological Sciences (CAMS) (Grant No. 2020KJ012 and 2020KJ009), the Basic Scientific Research and Operation Foundation of CAMS (Grant Nos. 2018Z006), and Youth Innovation Promotion Association CAS (Grant No. 2020340). This study was also supported by the Jiangsu Collaborative Innovation Center for Climate Change. The authors declare that they have no conflicts of interest. The NCEP–DOE AMIP-II reanalysis dataset and ERSST.v4 data were provided by NOAA/OAR/ESRL PSD, Boulder, Colorado, USA (https://www.esrl.noaa.gov/psd/), and the 160 in-situ station rainfall records were downloaded from http://cmdp.ncc-cma.net/.