1 Introduction

Intraseasonal variation (ISV), mainly including the 30–60- and 10–20-day oscillations [also known as the boreal summer quasi-biweekly oscillation (QBWO)], is a prominent phenomenon in the monsoon region, and the onset, active, and break phases of the monsoon show clear ISV characteristics (Li and Li, 1997; Long and Li, 2002). Moreover, it is well known that ISV has important effects on weather and climate changes in the monsoon region.

Since Madden and Julian (1971, 1972) first observed 41–53-day oscillation in the tropical atmosphere, an increasing number of studies have focused on ISV in the Asian monsoon region. Through analyzing MONEX (Monsoon Experiment) data, Krishnamurti and Subrahmanyam (1982) indicated that 30–50-day oscillation exists in the Indian monsoon region and presents northward propagating characteristics. Murakami et al. (1984) and Murakami and Nakazawa (1985) found that 45-day perturbations over the Asian monsoon region show clear features of eastward and northward propagation. Mao and Chan (2005) clarified that 30–60- and 10–20-day oscillations controlled the behavior of the South China Sea summer monsoon (SCSSM) activities for most years during 1975–2002. Zhou and Murtugudde (2014) further pointed out that 30–60-day oscillation of the Indian summer monsoon promotes early outbreak of the Asian summer monsoon. Krishnamurti and Ardanuy (1980) also found 10–20-day westward-propagating oscillation in the Indian monsoon. In addition, the QBWO can also regulate the active and break phases of the Indian summer monsoon (Chen and Chen, 1993), the SCSSM (Chen and Chen, 1995; Chan et al., 2002), and the East Asian summer monsoon (EASM) (Mao and Wu, 2006; Jia and Yang, 2013; Yang et al., 2014). As a result, Chan et al. (2002) studied the evolution of the SCSSM in 1998 and suggested that 30–60-day oscillation controlled the maintenance and the interruption of the SCSSM, and at the same time, the 10–20-day mode also had a modifying effect. Jia and Yang (2013) indicated that the QBWO influences the EASM through regulating the extratropical circulation and subtropical monsoonal flow.

The Tibetan Plateau (TP) is situated in the Asian monsoon region and its mechanical and thermal effects are key regulators of the Asian monsoon systems (Yeh, 1950; Flohn, 1957; Yeh et al., 1957; Manabe and Terpstra, 1974; Hahn and Manabe, 1975; Wu et al., 2007; Koseki et al., 2008). In addition, the TP monsoon is an important component of the Asian monsoon system (Tang and Reiter, 1984; Duan et al., 2013). Some previous studies have suggested that the TP is an active region of ISV in the extratropical atmosphere (Sun and Chen, 1988; Sun et al., 1994), and the wind, precipitation, atmospheric diabatic heating, and so on, all show significant multiscale ISV modes (Krishnamurti and Subrahmanyam, 1982; Zhang et al., 2014; Wang and Duan, 2015). Nitta (1983) demonstrated that the heat sources over the TP present 10–15- and 30-day periods, and Feng et al. (1985) pointed out that 10–15-day periodicity also exists in the latent heat over western TP. A 10–30-day oscillation has also been found for TP vortices, and their clustering in the summer of 1998 was likely modulated by this periodicity (Zhang et al., 2014). In addition, the ISV modes over the TP can affect large-scale atmospheric circulation and local climate changes (Zhou et al., 2000; Zhang et al., 2014).

Thus far, the ISV modes over the South and East Asian monsoon regions have been well documented. However, far less attention has been paid to ISV over the TP monsoon region, mainly because of insufficient observations. Furthermore, previous studies have largely focused on the 30–60-day oscillation over the tropical monsoon region to the south of 20°N. In contrast, little work has been conducted on ISV with respect to the 10–20-day oscillation in the entire Asian monsoon region, including the TP area. Wang and Duan (2015) investigated the characteristics of the QBWO over the TP and its link with the Asian summer monsoon. They pointed out that the QBWO is the dominant mode over the TP, and the equatorial western Pacific may be the main origin. In their paper, they presented the entire propagation pathway and proposed two possible mechanisms underpinning the northward propagation of QBWO signals over the Asian summer monsoon region: (1) barotropic vorticity and (2) the effect of moisture advection. Based on Wang and Duan (2015), we further examine the vertical structure and physical mechanisms in the process of the propagation of the QBWO signals. We also pay additional attention to the role of surface sensible heating (SSH) during the propagation process. Then, using an atmospheric general circulation model (AGCM) and designing numerical experiments, we attempt to verify our diagnostic results.

The paper is organized as follows. In Section 2, the datasets, the model, and the experimental designs are described. The propagation characteristics and vertical structures of the QBWO over the Asian monsoon region are given in Section 3. Section 4 compares the contributions of northward-propagating mechanisms of the QBWO signals. Section 5 presents the model results to illustrate the role of SSH in the northward propagation of the QBWO. Finally, a summary of the results is given in Section 6.

2 Data, model, and experiments 2.1 Datasets

The datasets employed in this study are as follows.

(1) The ECMWF Interim reanalysis (ERA-Interim) dataset (Simmons et al., 2007; Dee et al., 2011), with a resolution of 1.5° latitude by 1.5° longitude (). The analysis here covers the 33 years from 1979 to 2011. We derive the daily mean field by simply averaging the original 6-hourly data.

(2) The GPCP Pentad Precipitation Analyses (Huffman et al., 1997), constructed on 2.5° × 2.5° grids, over the whole globe, for the 33 years from 1979 to 2011 (). The data were produced by merging precipitation estimates calculated from some precipitation-related satellites, and precipitation gauge analyses.

Based on the ERA-Interim dataset, the vertically integrated atmospheric apparent heat source (< Q₁>) is derived by residual budget analyses via the thermodynamic and moisture equations (Yanai et al., 1973):

${c_p}\left[ {\frac{{\partial T}}{{\partial t}} + {V} \cdot \nabla T + {{\left({\frac{p}{{{p_0}}}} \right)}^k}\omega \frac{{\partial \theta }}{{\partial p}}} \right] = {Q_1}.$

(1)

The above equation can be integrated vertically:

$\left\langle {{Q_1}} \right\rangle =\frac{1}{g}\mathop {\int\limits_{100}^{{p_0}} {{Q_1}{\rm{d}}p \approx L \times P + {Q_{\rm{s}}} + \left\langle {{Q_{\rm{R}}}} \right\rangle } }, $

(2)

where T, V , and p denote the temperature, horizontal wind vector, and pressure, respectively; p₀ = 1000 hPa, and k =R/c_p ; c_p and R represent the specific heat of dry air at constant pressure and the gas constant, respectively; θ is potential temperature; ω is the vertical p-velocity; L is the latent heat of condensation; and P is precipitation rate. In addition, Q_s is the surface sensible heat, and < Q_R> denotes the vertical integration of radiation heating (cooling).

Using the Lanczos filtering method (Duchon, 1979) with cut off periods at 10 and 20 days and 21 weights, we extract the 10–20-day QBWO signal from “raw” observational datasets and model results. Moreover, the filtered data are highly auto-correlated, so the degrees of freedom are much lower than the sample size of the “raw” data. The effective degrees of freedom can therefore be calculated according to the approach of Bretherton et al. (1999) for the statistical significance test (t test) to obtain the confidence level of the correlation coefficient.

2.2 Model

Several experiments are carried out to examine the SSH effect. The AGCM used (SAMIL2.4.7) is the current version of the Spectral Atmospheric Model developed by the State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics, Institute of Atmospheric Physics (LASG/IAP) (Bao et al., 2013). This model uses a horizontal resolution of R42 (approximately 1.66° latitude × 2.81° longitude), with 26 sigma-pressure hybrid vertical layers that extend from the surface up to 2.19 hPa. The model radiative framework employs the Sun–Edwards–Slingo scheme (Edwards and Slingo, 1996; Sun and Rikus, 1999; Sun, 2011), and the direct effects of aerosols are included (Li et al., 2012). Convective precipitation is calculated according to the mass flux cumulus parameterization developed by Tiedtke (1989), with a modified closure assumption and organized entrainment and detrainment (Nordeng, 1994; Song, 2005). The planetary boundary layer scheme is then on-local first-order closure scheme proposed by Holtslag and Boville (1993).

The land surface model is the Community Land Model version 3.0 (CLM3.0), from NCAR (National Center for Atmospheric Research). A detailed technical description can be found in its manual (Oleson et al., 2004).

2.3 Experimental design

To verify the propagation characteristics and SSH mechanism of the QBWO signals over the Asian monsoon region, three sets of AGCM ensemble runs are carried out (Table 1). A more detailed description of the experimental design is as follows.

(1) Control experiment (CTRL). According to the experiment protocol of the Atmospheric Model Intercomparison Project (AMIP; ), a set of standard AMIP experiments during 1979–2004 are conducted with monthly outputs. We then take the fields on every 1 January from 1979 to 2004 as the restart fields to run the standard experiments again with the daily outputs.

(2) Heating experiment (Heating). On the basis of CTRL, a heating profile is added from 1 to 5 June in the equatorial western Pacific (Fig. 1a), which is diagnosed by observations as the QBWO source area. This set of experiments has the following purposes: 1) to verify whether the propagation pathway of the QBWO forced by the heating in the model is consistent with that revealed by observations; 2) to confirm whether the equatorial western Pacific is the source area for the QBWO over the Asian monsoon region; and 3) to simplify the analysis of the QBWO under this unified setting. Additionally, we also conduct the same experiments in July and August, and find that their results are highly consistent. In fact, the number of QBWO cases propagating from the low latitudes to the TP is comparable during June, July, and August, according to the statistics (table omitted). Thus, the results of the numerical experiments in this context are consistently provided for June.

(3) Heating and sensible heat experiment (Heating-0.5SH). On the basis of the Heating experiment, the SSH in northern Indian and southern rim of the TP (20°–30N, 70°–100°E; Fig. 1b) is reduced to a half of its original value (see Fig. S1 in supplemental material of the online version) to enable the investigation of the specific role of the SSH in the northern shift of the QBWO signals.

3 Propagation characteristics of the QBWO over the Asian monsoon region

The Asian monsoon system includes the EASM, South Asian monsoon, and TP monsoon subsystems. However, previous studies have always separated the TP monsoon region from the other systems because of the lack of observational data. Therefore, understanding ISV does not provide comprehensive information for the whole Asian monsoon region. In this study, to acknowledge the propagation features of the QBWO over the whole Asian monsoon region (including the TP monsoon region), we analyze the QBWO life cycle to examine its evolution using the extended empirical orthogonal function (EEOF) technique (Weare and Nasstrom, 1982). We use EEOF analysis on the filtered < Q₁> data with 2- and 4-day lags for the whole Asian monsoon area, and the first two leading EEOFs indicate a half cycle of the QBWO mode (Fig. 2). On day0-EEOF2, the positive convection anomalies with the QBWO mode exist first in the equatorial western Pacific, indicating this area is likely to be the source of QBWO signals over the Asian monsoon region. It then shows a prominent northwestward propagation and becomes a convection band covering a wide area from western Pacific to India from day2-EEOF2 to day4-EEOF1. This northwestward propagation is highly consistent with previous studies (Krishnamurti and Bhalme, 1976; Annamalai and Slingo, 2001; Chatterjee and Goswami, 2004).

It is clear that, after reaching the Indian subcontinent, the negative convective anomaly weakens gradually from day2-EEOF2 to day0-EEOF1. After this, it begins to move northward and finally penetrates into the TP area on day2-EEOF1 and day4-EEOF1. It is clear that day0-EEOF2 and day4-EEOF1 present an opposite distribution (Fig. 2). From day0- to day4-EEOF2, the QBWO over the TP seems to propagate to the East Asian areas. Furthermore, another northeastward propagation can be seen over northern SCS. It then joins the southward QBWO signals from midlatitudes, and eventually reaches western North Pacific.

After examining the source and propagation pathway of the QBWO over the Asian monsoon region, we next make further efforts to reveal the vertical structure in its propagating stages. For such a clockwise propagation pathway, we need to analyze the vertical structures in three stages: westward-propagating (equatorial western Pacific to the Indian subcontinent), northward-propagating [Indian subcontinent and Bay of Bengal (BOB) to the TP], and eastward-propagating (TP to the East Asian areas). For the first two stages, Fig. 3 shows the Hovmöller diagrams of the 10–20-day filtered daily vorticity and vertical velocity along 10°–12°N, 70°–100°E, respectively, for day0-EEOF2, from Fig. 2. During the westward-propagating stage of the QBWO (Fig. 3a), the vorticities in the high and low troposphere have opposite signs with strong ascending/descending motion, indicating that the QBWO has a baroclinic vertical structure. In addition, the speed of westward propagation calculated from Fig. 2 is about 6 m s^–1, which is likely to be the equatorial Rossby waves in the background of mean flow (Chen and Chen, 1993; Kikuchi and Wang, 2009).

Once reaching the BOB and Indian subcontinent, the QBWO turns to propagate to the north. In this northward-propagating stage, it is clear that the nature of the QBWO has not changed and keeps its baroclinic structure (Fig. 3b). However, it is noteworthy that some previous studies (Chen et al., 1991; Xu and Zhu, 2000) have shown that maintenance and change of baroclinic structure over the south rim of the TP is closely related to the steep terrain in this area. Moreover, Fig. 3b also indicates that the QBWO over northern TP presents an equivalent barotropic structure. This is because the QBWO mainly originates from the wave train in mid—high latitudes and actually represents the extratropical Rossby wave (Kikuchi and Wang, 2009). It is therefore clear that the vertical structures of the QBWO that arrives at the southern and northern boundaries of the TP are opposite. Furthermore, the QBWO signals over the TP can then continue propagate to the downstream region, with their original vertical structures. That is, the QBWO over the southern area of the TP, mainly originating from low latitude areas, can keep its baroclinic vertical structure as it propagates to East Asia, while the QBWO over the northern area of the TP keeps its equivalent barotropic structure (see Fig. S2 in supplemental material of the online version).

4 Relative contributions of northward propagation mechanisms

A large volume of previous research (Wang and Xie, 1997; Jiang et al., 2004; Bellon and Sobel, 2008; Abhik et al., 2013; DeMott et al., 2013) has examined the essential atmospheric dynamical process for the northward propagation of the 10–90-day boreal summer intraseasonal oscillation. Demott et al. (2013) analyzed the relative contribution of several mechanisms to the northward propagation of the 20–60-day boreal summer oscillation. They found that the northward propagation is dominated by boundary-layer moisture advection and the barotropic vorticity effect, and the other mechanisms are relatively minor. In the current study, we conduct similar analyses for the 10–20-day oscillation to examine the relative importance of several northward-propagating mechanisms of the QBWO signals. This has been conducted by very few previous studies, especially for the specific role of SSH, which was not clearly clarified by Wang and Duan (2015).

The meridional propagation characteristics of the filtered daily < Q₁> averaged over 70°–100°E for 6 years are shown in Fig. 4, in which the northward propagation of the QBWO can be clearly detected. In addition, the QBWO signals are able to propagate from the TP area to the high and/or low latitudes and from the midlatitudes to the TP area. We carry out a composite analysis by choosing 48 northward-propagating cases from 1979 to 2011 to illustrate the evolution and underlying mechanism of the QBWO. The composite is constructed such that on day 0 the QBWO convection takes place at exactly 27°N, where it is already located over the TP. The temporal evolution of the composite < Q₁> data and the leading northward propagation mechanisms from days –18 to 0 (with a 2-day interval) is shown in Figs. 5 and 6. The anomalous positive QBWO convection (contours) first appears near the SCS on day –12 and then it propagates northwestward to the Indian subcontinent and the BOB on day –4, with a somewhat weakened signal. Later, on day –2, the convection turns northward and reaches the southern slope of the TP. We next compare the relative importance of physical mechanisms on the northward propagation of anomalous QBWO convection.

Evaluating the relative contribution of the different mechanisms with different units in the process of the northward propagation of QBWO signals is complicated. Therefore, we address this complexity by constructing composites on normalized time series of all leading northward propagation mechanisms, which are shown in Figs. 5, 6. Of course, the contributions of two linear components ( $ - \bar v\displaystyle\frac{{\partial q'}}{{\partial y}}$ and $ - v'\displaystyle\frac{{\partial \bar q}}{{\partial y}}$ ) of boundary-layer moisture advection are different (Fig. 5), and the latter has almost no association with the northward propagation of the anomalous convection. From days –8 to –4, the anomalous moisture advection by the mean meridional wind ( $ - \bar v\displaystyle\frac{{\partial q'}}{{\partial y}}$ ) gathers on the north side of the QBWO convection, which may affect the northward shift of the convection and may even arrive at the TP. Figure 6 shows the evolution of mean column-integrated vorticity between 1000 and 100 hPa and SSH with the propagation of anomalous convection. From days –10 to –6, the maximum convection center (MCC) is located in the area south of 20°N, and there is always a significant positive vorticity anomaly situated to the north of the MCC, with an equivalent barotropic structure, which is consistent with Jiang et al. (2004) and Wang and Duan (2015), in that, when the QBWO signals are situated to the south of 20°N, the generation of barotropic vorticity induced by the easterly vertical shear is mainly responsible for the northward movement of the convection. As illustrated in Wang and Duan (2015), the boundary moisture advection is the main mechanism of the northward propagation of the QBWO located to the north of 20°N. In this study, we further indicate that its functional component is the anomalous moisture advection by the mean meridional wind ( $ - \bar v\displaystyle\frac{{\partial q'}}{{\partial y}}$ ).

It is worth noting that there is also significant SSH located in the north of the MCC on day –6, and this may also contribute to thenorthern shift of the QBWO signals. On day –4, however, the MCC is situated at roughly 20°N, and there is a positive SSH anomaly rather than vorticity situated north of the MCC, which can lead the northern propagation of the QBWO to the TP area by making the lower troposphere less stable (Webster, 1983; Hsu et al., 2004). For the propagation of the QBWO to the north of 20°N, the distribution of the SSH seems wider than boundary moisture advection, which may indicate that the former plays a larger role in the northern propagation of the QBWO signals than the latter, especially for propagation to the TP area.

In brief, when the QBWO signals originating from the equatorial western Pacific move northwestward to the BOB and northern Indian, it shows a significant northward propagating feature. For the northward propagation of the QBWO to the south of 20°N, the key process is the generation of barotropic vorticity because of the presence of the vertical shear of the mean flow. In addition, the SSH located north of the MCC may also play a certain role. Then, in the area from 20°N to the southern rim of the TP, the easterly shear of the mean flow is not obvious (figure omitted). Therefore, from day –4, the main mechanisms that influence the northward propagation of the QBWO are the anomalous moisture advection by the mean meridional wind ( $ - \bar v\displaystyle\frac{{\partial q'}}{{\partial y}}$ ) and the SSH over northern India and the southern rim of the TP, and the latter possibly plays a greater role in the further northward propagation of the QBWO. In the next section, we present the specific scope and role of the SSH in detail, based on numerical simulation.

5 Results of SAMIL2.4.7

The above results are based on observations, which illustrate the source area of the QBWO in the Asian monsoon region, its propagation pathway, and the relative importance of its northward mechanisms. In this section, we employ an AGCM and sensitivity experiments to deepen our understanding of these processes.

5.1 Evaluation of the simulation of SAMIL

Before analyzing the results of SAMIL2.4.7, we first evaluate its performance in simulating the precipitation, circulation, and ISV fields in the Asian monsoon region, which can provide a reference for subsequent results.

Figure 7 shows the boreal summer (June, July, and August) mean 850-hPa wind and precipitation fields averaged over 33 years (1979–2011) in the Northern Hemisphere. In comparison with observations, the simulation of precipitation is systematically overestimated over 10°– 30°N. For the wind field, SAMIL2.4.7 can simulate the cross-equatorial flow and two monsoon troughs well. In general, SAMIL2.4.7 is able to reproduce the basic precipitation pattern and circulation systems.

The numerical experiments in this study focus on the 10–20-day timescale, so we need to investigate the performance of simulated ISV in SAMIL2.4.7. The distribution of variance of the filtered precipitation in observation and CTRL are shown in Fig. 8. As we can see, the spatial distributions are very similar, with the strongest signals mainly in the north of the BOB and secondarily in the eastern Arabian Sea and SCS.

5.2 Experimental results

According to the experimental design mentioned above, the difference between the Heating and CTRL experiments is referred to as Heating–CTRL, denoting the role of thermal forcing after adding the heating in the source area (equatorial western Pacific) of the QBWO. We conduct the 26-ensemble averaged precipitation and wind field at sigma 0.95 in Heating–CTRL (figure omitted), and the propagation pathway of the QBWO signals is basically reproduced as shown in observations. Considering the interannual variability, and to better reflect this propagation process, we derive the precipitation time series on day 16 over the domain (25°–30°N, 88°–100°E) for 26 years and define the strong year as the precipitation above 0.5 standard deviations. Ten strong cases are selected to be averaged in Fig. 9. On day 0, there is a significant positive precipitation anomaly over the area with added atmospheric heating. Then, in the following four days, the precipitation anomaly becomes gradually strengthened, in agreement with an enhanced cyclonic circulation. From day 6, the precipitation anomaly in the equatorial western Pacific starts to propagate northwest, going through the SCS and arriving at the BOB and Indian subcontinent areas (days 8 and 10). The QBWO signals of precipitation then turn to move northward and reach the south slope of the TP on day 14, with the signals gathering constantly during days 16 and 18. The whole clockwise propagation pathway is essentially consistent with the diagnostic results. This demonstrates that this AGCM can accurately produce the propagation of the QBWO over the Asian summer monsoon region. In addition, this further supports the suggestion that the equatorial western Pacific is indeed the source area of the QBWO.

Previous studies on the northward mechanisms of ISV generally focus on the timescale of 30–60-day oscillation, based on theoretical deduction and data analyses. As mentioned above, this study analyzes the northward mechanisms of the QBWO in the Asian monsoon region and their relative importance through observational data. We next confirm the specific role of SSH through the numerical simulation according to the experimental scheme in Table 1. The difference map between the Heating-0.5SH and CTRL experiments (denoted as Heating-0.5SH–CTRL) embodies the evolution of the QBWO when adding the heating in the equatorial western Pacific and reducing the SSH in northern Indian and over the southern slope of the TP. To clarify more intuitively the impact of the SSH on the propagation of the QBWO signals, we compare the characteristics of the QBWO signals in the Heating–CTRL and Heating-0.5SH–CTRL experiments through the time–latitude section of filtered precipitation (Fig. 10). Figure 10a indicates that the QBWO signals from the low latitudes are able to propagate north to the TP area. In the process of the propagation across 20°N, the QBWO signals become weakened. However, the signals strengthen again as the propagation approaches the southern slope of the TP. When the SSH in northern India and the southern slope of the TP (20°–30°N, 70°–100°E) is reduced to half of the original value, the propagation process of the QBWO signals in the area south of 20°N does not change obviously, as illustrated in Fig. 10b. This is closely related to the easterly vertical shear background field that has no obvious change with SSH weakening (see Fig. S3 in supplemental material of the online version). To the north of 20°N, the northward propagation of the QBWO signals gets weaker than that in Fig. 10a, and is even almost interrupted. Although the QBWO signals get strengthened when arriving at the southern slope of the TP, compared to that in Fig. 10a, the significant signals are reduced distinctly and its strength is also weakened. In summary, the SSH in northern India and the southern slope of the TP can greatly benefit the northward propagation of the QBWO to the north of 20°N, especially for the significant signals moving as far as the TP area.

6 Discussion and summary

Some studies involving the northward-propagating mechanisms of the 30–60-day oscillation (Vecchi and Harrison, 2002; Jiang et al., 2004) have proposed the possible role of air–sea interaction in its northward propagation. Vecchi and Harrison (2002) analyzed the air–sea coupling of precipitation anomalies and SST in the BOB, noting that ocean coupling in the coupled general circulation model experiments improved the northward propagation of the intraseasonal oscillation, compared with the atmosphere-only simulations with the same model. This result was also confirmed by a few later studies (Fu and Wang, 2004; Roxy and Tanimoto, 2007). We also make an attempt to recognize whether the air–sea interaction can affect the northward-propagating QBWO, using a coupled model named SAMIL-OMLM [an ocean mixed layer model (Noh and Kim, 1999) that is coupled with SAMIL2.4.7 over the region (15°S– 23°N, 60°–180°E)]. The results show that there is no clear change in the entire propagation pathway and northward-propagating intensity of the QBWO relative to those in the AMIP (see Fig. S4 in supplemental material of the online version). This may indicate that the air–sea interaction is not the leading mechanism for the northward propagation of the QBWO.

The functions of several mechanisms for the northward propagation of the QBWO signals over the Asian monsoon area are examined, and their relative importance is addressed by constructing normalized time series for their evolution. Importantly, based on the data diagnosis and numerical simulation, the specific effect of surface heating on the northward propagation of the QBWO has become much clearer. The main conclusions of this paper are as follows.

(1) Diagnosis on the origin of the QBWO signals over the Asian summer monsoon region indicates that the QBWO mainly comes from the equatorial western Pacific and moves northwestward to the BOB and northern India, and then turns northward to the southeastern TP. The QBWO signals in the TP can also propagate eastward to the East Asian region. The results of the numerical simulations confirm this source area and the clockwise propagation pathway.

(2) During the process of the westward propagation of the QBWO over the source area, the QBWO shows a baroclinic vertical structure, with a 6-m s^–1 speed, which is likely to represent equatorial Rossby waves. When the QBWO reaches the BOB and Indian subcontinent, it turns to move northward to the southern rim of the TP and still keeps its baroclinic vertical structure. However, the QBWO in the northern TP presents an equivalent barotropic structure and this is because it mainly originates from the extratropical Rossby wave train in the middle–high latitudes.

(3) In the areas south of 20°N, the barotropic vorticity induced by vertical easterly shear may be the main mechanism accounting for the northward shift of the QBWO signals. To the north of 20°N, the main mechanisms that influence the northward propagation of the QBWO are the anomalous moisture advection by the mean meridional wind $ - \bar v\displaystyle\frac{{\partial q'}}{{\partial y}}$ and the SSH over northern India and the southern rim of the TP. The numerical simulations further verify the specific scope and function of the SSH that strongly contributes to the northward propagation of the QBWO, even as far as the TP area.