1. Introduction

Subseasonal variability over a 10–90-day period is a typical and dominant component of atmospheric variations in the tropics. Generally, two major modes, the 10–20-day quasi-biweekly oscillation (QBWO; e.g., Krishnamurti and Ardanuy 1980; Chen and Chen 1993, 1995; Fukutomi and Yasunari 1999, 2002; Wen and Zhang 2007; Jiang and Lau 2008; Kikuchi and Wang 2009; Wen et al. 2011; Jia and Yang 2013; Li et al. 2020) and the 30–60-day intraseasonal oscillation (ISO; Madden and Julian 1971, 1972; Krishnamurti and Subrahmanyam 1982; Zhang 2013), are distinguished according to the power spectra of many convection and circulation parameters. As indicated in previous studies, both the QBWO and ISO influence the active and break cycles of the Indian summer monsoon (Krishnamurti and Ardanuy 1980; Yasunari 1981; Goswami et al. 2003; Roman-Stork et al. 2019) and South China Sea monsoon (Chen et al. 2000; Chan et al. 2002; Mao and Chan 2005; Zhou and Miller 2005; Lin and Li 2008). However, compared with the widely documented planetary-scale ISO [see reviews in Madden and Julian (1994), Zhang (2005), and Li (2014)], systematic explorations of the much higher-frequency QBWO are still limited (Li et al. 2020), probably due to its complex regionality (Kikuchi and Wang 2009).

The earliest records of the QBWO, which were related to the Indian summer monsoon, can be traced back more than a half-century (e.g., Keshavamurty 1972). Subsequently, extensive studies have been conducted on the features, mechanisms, and influence of the QBWO over the Indian monsoon region (Krishnamurti and Ardanuy 1980; Chen and Chen 1993; Li et al. 2022), the western North Pacific (WNP; Mao and Chan 2005; Chen and Sui 2010; Wang et al. 2017; Li et al. 2020, 2021), the South China Sea (Chen and Chen 1995; Goswami et al. 2003), the Indo-China subtropical region (Yang et al. 2010; Jia and Yang 2013), the North American monsoon region (Mullen et al. 1998; Kiladis and Hall-McKim 2004; Jiang and Waliser 2009; Wen et al. 2011), and other subtropical monsoon regions. In fact, Kikuchi and Wang (2009) identified eight QBWO modes from a global perspective, involving three boreal summer modes in the Asia-Pacific, Central America, and subtropical South Pacific regions and five austral summer modes in the Australia–southwest Pacific, southern Africa (SA)–Indian Ocean (IO), South America–Atlantic, subtropical North Pacific, and North Atlantic–North Africa regions.

Due to the most striking quasi-biweekly variability in the WNP, a detailed exploration of the QBWO in terms of its initiation, movement, development, and dissipation has been conducted in this region during boreal summer (Chen and Sui 2010; Chatterjee and Goswami 2004; Chen and Sui 2010; Zhang et al. 2020; Tao et al. 2009; Wang and Duan 2015; Li et al. 2020; Yang and Li 2020; Yang et al. 2022). Chen and Sui (2010) claimed that the QBWO in the WNP originates from the equatorial western Pacific and propagates westward along the off-equatorial belt with a typical ∼6000-km zonal wavelength (Chatterjee and Goswami 2004; Chen and Sui 2010; Zhang et al. 2020). Pronounced poleward propagation toward East Asia has also been observed and explained (Tao et al. 2009; Wang and Duan 2015; Li et al. 2020; Yang and Li 2020; Yang et al. 2022). The boreal summer QBWO is also prominent over the tropical Indian Ocean, although its characteristics and propagation routine are inconsistent in recent studies. One group of researchers believes that the IO QBWO is closely related to its counterpart in the WNP, which propagates northwestward as a southwest–northeast-tilting system (Lee et al. 2013; Wang and Duan 2015; Ortega et al. 2017; Wang et al. 2017; Gui and Yang 2020; Liu et al. 2022). Another group believes that the life cycle of the IO QBWO is independent of the WNP, which could start from the western IO and mature in the eastern IO (Wen and Zhang 2008; Wen et al. 2010; Qian et al. 2019). Most recently, Li et al. (2022) summarized these two dominant boreal summer QBWO modes in the IO and suggested that the IO-independent QBWO is regulated by the mean clockwise circulation in the tropical IO.

Until now, most analyses regarding the QBWO have been conducted either in the WNP or in the northern IO during boreal summer. Studies focusing on the QBWO in the southwestern IO (SWIO) during austral summer are still scarce, even though a few researchers have noticed convective disturbances on a quasi-biweekly time scale in the SWIO in one form or another (Jury and Pathack 1991; Jury et al. 1991; Kikuchi and Wang 2009; Fukutomi and Yasunari 2013, 2014; Ghatak and Sukhatme 2021). In fact, the SWIO also favors the development of the QBWO during austral summer (Fig. 1). The subseasonal variability during austral summer [December–February (DJF)] is much stronger than that in boreal summer [June–August (JJA)] in the SWIO. During the austral summer, the 10–20-day mode, which plays a dominant role in this region, is much stronger than the 30–60-day mode. Therefore, a comprehensive understanding of the SWIO QBWO is urgently needed.

View Full Size

Standard deviation of the (a),(b) original, (c),(d) 10–20-day filtered, and (e),(f) 30–60-day filtered outgoing longwave radiation (OLR; W m⁻²) in (left) December–February (DJF) and (right) June–August (JJA) in 1979–2018. The blue box marks the OLR averaged region (15°–25°S, 40°–60°E) for identifying individual QBWO events.

Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-22-0766.1

• Download Figure
• Download figure as PowerPoint slide

Among the many underlying uncertainties related to the SWIO QBWO, understanding the general characteristics is still a top priority. Kikuchi and Wang (2009) analyzed the leading mode of the QBWO based on extended empirical orthogonal function (EEOF) analysis, in which two major convective centers showing different propagation paths were identified. One convective center over Botswana moves northeastward with a zonal phase speed of approximately 5 m s⁻¹, whereas another center in the central IO moves southeastward. After merging in the SWIO, the convective anomaly then moves eastward and disappears over the central southern IO. On the other hand, a recent study by Ghatak and Sukhatme (2021) finds only a regular poleward-propagating QBWO over the SWIO during austral summer. In fact, a preliminary analysis of 10–20-day filtered outgoing longwave radiation (OLR) anomalies has already detected eastward propagation (eastward in Fig. 2a with slightly northward in Fig. 2d) during austral summer in 2000/01 and poleward propagation (southward in Fig. 2c with weak westward in Fig. 2b) during austral summer in 1999–2000.

View Full Size

Example of diversified propagation patterns of the QBWO, with 10–20-day OLR anomalies (W m⁻²) averaged between (a),(b) 15°–25°S and (c),(d) 40°–60°E for the 1999/2000 and 2000/01 austral summers. The blue contours indicate −10 W m⁻² contours. The thick black dashed lines indicate the propagating routines.

Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-22-0766.1

• Download Figure
• Download figure as PowerPoint slide

To identify the diversity in the propagation routine and obtain an in-depth understanding of the underlying mechanism of the austral summer QBWOs in the SWIO, K-means cluster analysis and the moisture budget are conducted in the present study. The rest of the paper is organized as follows. The data and methods used in the study are introduced in section 2. Section 3 reveals the diversity in QBWO propagation using an objective and systematic method. In section 4, a moisture budget analysis is conducted to understand the underlying mechanisms. A summary and discussion are presented in the final section.

2. Data and methods

3. Diversity of QBWO propagation and the associated circulation features

As introduced above, the K-means analysis classifies the 103 QBWO events in austral summer (DJF) into two optimal clusters. The first cluster includes 69 QBWO events, and the second cluster has 34 events. Then, the composite horizontal evolutions of OLR anomalies, 850-hPa wind anomalies, and specific humidity anomalies averaged between 300 and 1000 hPa of the two clusters at 2-day intervals are provided in Figs. 3 and 4.

View Full Size

Time evolution of OLR, specific humidity, and low-level circulation for the EM QBWO mode. Composited 10–20-day OLR anomalies (contours; W m⁻²), 850-hPa wind anomalies (vectors; m s⁻¹), and 10–20-day specific humidity anomalies averaged between 300 and 1000 hPa (shading; g kg⁻¹) from day −8 to day 8 are shown. Only the specific humidity anomalies and wind vectors above the 95% confidence level are shown. The OLR anomalies above the 95% confidence level are represented by magenta points. Arrows highlight the propagation of a convective anomaly.

Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-22-0766.1

• Download Figure
• Download figure as PowerPoint slide

View Full Size

As in Fig. 3, but for the PM mode.

Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-22-0766.1

• Download Figure
• Download figure as PowerPoint slide

For the first cluster (Fig. 3), a convective center coupled with a low-level cyclonic circulation first emerges in the SA region on day −8. At that time, the SWIO was controlled by a well-developed anticyclonic circulation with positive OLR anomalies and high pressure. The convective center strengthens quickly and extends to the east at days −6 and −4. Moreover, the anticyclonic circulation in the SWIO moves slowly to the southeast and diminishes gradually during this period. On day −2, the convective center passes eastward across the SA region and reaches the SWIO. Then, the convection becomes strongest at day 0, with a low pressure center located east of Madagascar and a new suppressed convection initiates in the SA region. Thereafter, the second half of the QBWO life cycle begins in a similar manner as that described in the first half. Overall, the convective anomalies associated with the first cluster of the QBWO generally depict an eastward propagation in the subtropical region. Hereafter, the first cluster is referred to as the eastward mode (EM) for simplicity.

In contrast, the evolution of the second cluster seems more complicated (Fig. 4). The strong convection in the SWIO is incorporated by two branches from the north and the west. The west convective branch first emerges on the western coast of Africa at approximately 10°–20°S on day −6 (the magenta points shown around western Africa) and propagates poleward along the coast from day −6 to day −2. The east convective branch seems to originate from the east tropical Indian Ocean with a decaying trend (from days −8 to −6), but it quickly strengthens when it reaches the northeast of Madagascar and turns to southward propagation (from days −4 to −2), accompanied by an enhanced, low-level cyclonic circulation. Then, the west convective branch passes across the SA region and merges with the stronger east branch at day 0. The two branches ultimately merge into one major convective center on day 2 and continuously move southeastward until day 6. In the life cycle of the second cluster, a generally poleward propagation of the SWIO is exhibited, although a slight contamination of eastward movement from the SA region is involved. Therefore, the second cluster is referred to as the poleward mode (PM) for simplicity.

Kikuchi and Wang (2009) recognize two convective centers in the leading EEOF, with one located over the SA region and another in the central Indian Ocean. Here, using K-means analysis, these two convective anomalies are distinguished into two modes. Generally, both modes show northwest–southeast-tilted convective patterns, except that the convection band in the EM is narrow, whereas the convection band in the PM is broad. Our EM mode denotes the eastward-moving convective anomalies in their studies, whereas the PM mode represents the southward-moving convection. Therefore, the two modes specified in this study represent a further refinement of their research.

Moisture dynamics have been reported to significantly influence the evolution of the QBWO. As shown in Figs. 3 and 4, the development of convective anomalies is closely related to specific humidity anomalies for both modes. For the EM, the positive (negative) specific humidity anomaly center is coincident with the active (suppressed) convection center on day −6. The positive specific humidity anomalies propagate eastward along 20°S in the following days (from day −4 to day 2). After the moisture anomalies move east of Madagascar, they weaken sharply and move in a southeastward direction to the central IO. On day 6, with the precondition of positive moisture anomalies, new convection is triggered in the SA region. In comparison with the EM, the moisture condition for the PM is somewhat complicated. Although the OLR anomalies in the western center are weak from day −6 to day −2, the positive specific humidity anomalies strongly develop along with the southward movement of convection without any lead in the location. On day 0, these positive specific humidity anomalies remain in the SA and largely weaken, whereas another striking positive specific humidity anomaly grows sharply and occupies almost the whole SWIO. After day 2, the positive specific humidity anomalies gradually decay and move in a southeastward direction, accompanying the OLR anomalies.

To estimate the period and propagation speed of the QBWO, we construct time–longitude and time–latitude Hovmöller diagrams. Specifically, 10–20-day OLR anomalies and specific humidity anomalies averaged between 300 and 1000 hPa over 10°–30°S for the EM shown in Fig. 5a detect an eastward propagation with a mean speed of 3.9° day⁻¹. The convective center moves slowly eastward with a phase speed of approximately 2.5° day⁻¹ when it travels on the eastern Africa continent. The moisture anomalies slightly lead the OLR anomalies during this period, which indicates the eastward movement of convection. The phase speed accelerates up to approximately 7.5° day⁻¹ when the convection leaves eastern Africa and enters the Mozambique Channel. As the sea surface temperature (SST) tends to enhance the surface turbulent heat fluxes ahead of the convective center and suppress them behind the convective center (Li et al. 2020), the warmer SST in the Mozambique Channel (Han et al. 2019; Mawren et al. 2022) may play an essential role in accelerating the phase speed, which is worth detailed examination in future work. On the other hand, the 10–20-day OLR anomalies and 300- to 1000-hPa averaged specific humidity anomalies averaged over 40°–60°E for the PM shown in Fig. 5b exhibit a significant southward propagation speed of approximately 2.8° day⁻¹. The zonally averaged (40°–60°E) results also show a westward phase speed of 2.4° day⁻¹ (figure not shown), which is weaker than that estimated from previous studies in this area (Ghatak and Sukhatme 2021; Bessafi and Wheeler 2006). The time period of both modes is nearly the same (i.e., approximately 14–18 days).

View Full Size

(a) Time–longitude diagram (averaged over 10°–30°S) and (b) time–latitude diagram (averaged over 40°–60°E) of 10–20-day OLR anomalies (contours; W m⁻²) and specific humidity anomalies averaged between 300 and 1000 hPa (shading; g kg⁻¹) from the composite in Fig. 3. The black line in (a) corresponds to an eastward movement of 3.9° day⁻¹. The black line in (b) corresponds to a southward movement of 2.8° day⁻¹. The red asterisks represent the minimum OLR anomalies on each day.

Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-22-0766.1

• Download Figure
• Download figure as PowerPoint slide

4. Moisture budget analysis for the eastward and poleward modes

The two distinct QBWO modes in the SWIO during austral summer have been thoroughly described in the previous section, but we also need to examine how the structural differences in the circulation can lead to the diverse propagation of the QBWO. The well-collocated moisture and convection anomalies in Figs. 3 and 4 suggest that we investigate the moisture budget to see what processes generate eastward or poleward propagation tendencies and how these processes are associated with the circulation differences between the two modes.

Moisture processes have been considered to play vital roles in the maintenance and propagation of the QBWO (Tao et al. 2009; Wang et al. 2017; Wang and Li 2020; Li et al. 2020, 2021, 2022; Hu et al. 2021). Here, the spatiotemporal pattern of the moisture time tendency (∂q′/∂t) and the OLR anomalies are shown in Fig. 6. Unsurprisingly, the collocation of positive (negative) moisture tendency anomalies in front of the active (suppressed) convection center, which favors QBWO propagation, is found in both modes. In particular, the positive moisture tendency anomalies are always east of the active convection center for the EM, and the positive moisture tendency anomalies are south of the active convection center for the PM. Notably, the positive moisture tendency anomalies for the PM on the west coast of Africa decay sharply with the southeastward movement from day −5 to day −3 and diminish on day −1. Moreover, the positive moisture tendency anomalies in the SWIO develop strongly with southward movement during the same period.

View Full Size

Time evolution of moisture time tendency and OLR for the two QBWO modes. Composited 10–20-day specific humidity time tendency averaged between 300 and 1000 hPa (contours; g kg⁻¹ day⁻¹), 10–20-day OLR anomalies (shading; W m⁻²), and 850-hPa wind anomalies (vectors; m s⁻¹) from day −5 to day 4 for the (a) EM and (b) PM modes are shown. Only the wind vectors and OLR anomalies above the 95% confidence level are shown. The specific humidity time tendencies above the 95% confidence level are represented by black points. The red pentagrams denote the convective centers in each subplot.

Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-22-0766.1

• Download Figure
• Download figure as PowerPoint slide

The vertical sections of the moisture and circulation anomalies associated with the two QBWO modes are further examined in Fig. 7. Here, the zonal–vertical (meridional–vertical) sections are averaged between 10° and 30°S (40°–60°E) on day 0. For the EM, the deep-layer ascending flow in the troposphere coincides with the positive moisture anomalies (Fig. 7a), and a striking zonal contrast appears in the moisture tendency with positive anomalies to the east and negative anomalies to the west of the active convection center, which indicates an eastward propagation. The positive moisture anomalies are slightly eastward phase-leading of the convective center, which preconditions an eastward propagation. Similarly, the above statement is true for the PM but regarding the meridional–vertical section (Fig. 7d).

View Full Size

Vertical structures of the two QBWO modes based on the composites in Figs. 3 and 4 on day 0. (a),(c) The zonal–vertical sections (10°–30°S averaged), where the shading represents the 10–20-day specific humidity (g kg⁻¹), vectors represent the 10–20-day zonal velocity (m s⁻¹) and vertical velocity (0.01 Pa s⁻¹), and contours represent the 10–20-day specific humidity time tendency (g kg⁻¹ day⁻¹). (b),(d) The meridional–vertical sections (40°–60°E averaged). The blue triangles represent the convective centers, and the red squares represent the maximum upward motion positions.

Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-22-0766.1

• Download Figure
• Download figure as PowerPoint slide

To further explore the relative importance of each moisture budget term on the right-hand side of Eq. (1), the horizontal patterns of each moisture budget term on day −4 and day 0 are constructed and shown in Fig. 8. These two days are selected according to Figs. 3 and 4, which denote the growing and mature phases of convection, respectively, and might capture the mechanisms of different locations. For the PM mode, as the convection mainly comes from the northeast center, we chiefly focus on this center. Clearly, notable positive moisture tendency anomalies are located eastward (southward) and ahead of the convective center of the EM (PM) (Figs. 8a,f,k,p). These positive moisture tendency anomalies are mainly contributed by horizontal moisture advection (Figs. 8b,g,l,q), which matches the moisture tendency anomalies in both spatial pattern and magnitude. Although the amplitude of the vertical moisture advection and moisture sink terms are nearly 3 times as large as that of the horizontal moisture advection, they are almost offset with each other, and neither term shows noticeable phase leading. Furthermore, their combined net positive effect can play more important roles on day −4 than that on day 0. This kind of contribution mainly comes from the moisture source or sink. However, this combined net positive effect is much weaker than the effect from horizontal moisture advection resulting in the propagation of convection. Therefore, this study concludes that horizontal moisture advection fundamentally promotes the successive propagation of moisture anomalies for the two types of QBWOs in the SWIO. Interestingly, in the PM, the strongest convection is located in the northeast sector of the cyclonic circulation on day 0 (Figs. 4 and 8p). This kind of shift is possibly due to both the convergence between the gyres and the horizontal rotational poleward advection of moist air out of the equatorial region (Ghatak and Sukhatme 2021), which is reminiscent of observations of equatorial Rossby (ER) waves (Wheeler et al. 2000; Molinari et al. 2007) and has been noted in recent idealized moist shallow water simulations (Suhas and Sukhatme 2020).

View Full Size

Horizontal composite of the moisture budget terms for the EM mode and the PM mode on day −4 and day 0. The shading shows (first row) the 10–20-day specific humidity tendency averaged between 300 and 1000 hPa (g kg⁻¹ day⁻¹), (second row) the horizontal advection term averaged between 300 and 1000 hPa, which is divided by 3.0 (g kg⁻¹ day⁻¹), (third row) the vertical advection term averaged between 300 and 1000 hPa, which is divided by 3.0 (g kg⁻¹ day⁻¹), (fourth row) the moisture source or sink term (g kg⁻¹ day⁻¹), and (fifth row) the column processes (vadv + res; g kg⁻¹ day⁻¹). The contours indicate the 10–20-day OLR anomalies (W m⁻²). Vectors represent 10–20-day 850-hPa wind anomalies (m s⁻¹). Only the shaded values and the vectors that are statistically significant at the 95% confidence level are shown.

Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-22-0766.1

• Download Figure
• Download figure as PowerPoint slide

As mentioned above, horizontal moisture advection dominantly contributes to variation in the moisture time tendency and hence to the movement of QBWO convection. Next, the detailed physics involved in horizontal moisture advection are analyzed. The specific subitems indicated in Eq. (2), except for the residual terms, are averaged in the areas 10° east (south), 5° north (west), and 5° south (east) of the convective center for the EM (PM) and are shown in Fig. 9. Note that the sum of all the subitems (red bars) shows nearly the same value as the horizontal moisture advection (black bars). For the EM, the top contributor is clearly the $-{(V^{\prime }\cdot \nabla \overline{q})}^{\prime }$ term, which represents the interaction of quasi-biweekly wind anomalies and the low-frequency humidity on both day −4 and day 0. Other subitems have a much smaller influence and are negligible. However, for the PM, the $-{(\overline{V}\cdot \nabla q^{\prime })}^{\prime }$ term, which represents the interaction of the low-frequency wind anomalies and quasi-biweekly humidity, plays a main role in guiding the southward movement of convention on day −4, and the $-{(V^{\prime }\cdot \nabla \overline{q})}^{\prime }$ term becomes the top contributor again on day 0. Generally, although the EM and PM exhibit different propagation routes, they are similar in their underlying moisture dynamics in the mature phase; that is, the eastward (southward) phase-leading positive moisture tendency with regard to the active convection center is primarily due to the horizontal advection of low-frequency moisture by the QBWO wind anomalies [i.e., $-{(V^{\prime }\cdot \nabla \overline{q})}^{\prime }$]. However, the controlling mechanisms of EM and PM in the growing phase differ; that is, the horizontal advection of low-frequency moisture by the QBWO wind anomalies [i.e., $-{(V^{\prime }\cdot \nabla \overline{q})}^{\prime }$] controls the eastward phase-leading positive moisture tendency, whereas the horizontal advection of quasi-biweekly moisture by the low-frequency wind anomalies [i.e., $-{(\overline{V}\cdot \nabla q^{\prime })}^{\prime }$] is responsible for the southward phase-leading positive moisture tendency although its value is relatively small. The results on days ranging from day −3 to day 2 are similar to the results on day 0 for the EM, whereas the results on days ranging from day −3 to day −2 are similar to the results on day −4 and the results on days ranging from day −1 to day 2 are similar to the results on day 0 for the PM (figures not shown).

View Full Size

Moisture budget terms in the areas 10° east (south), 5° north (west), and 5° south (east) of the convective center for the EM (PM). The green bars denote the 16 subitems, the red bars denote the sums of all 16 subitems, and the black bars denote the horizontal advection terms of the moisture budget.

Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-22-0766.1

• Download Figure
• Download figure as PowerPoint slide

Now, how can the top contributors induce different propagations of the austral summer QBWOs? To address this question, the horizontal moisture advection and the major contribution terms are presented in Fig. 10. Consistent with the result from Fig. 9, the sum of all subitems fit well with the horizontal moisture advection (Figs. 10a,b,f,g,k,i,p,q) for both QBWO modes. For the EM, the $-{(V^{\prime }\cdot \nabla \overline{q})}^{\prime }$ term agrees well with the horizontal moisture advection, which confirms its dominant role in guiding eastward propagation on day 0 (Fig. 10h). The explanation on day −4 is a little complicated. The $-{(V^{\prime }\cdot \nabla \overline{q})}^{\prime }$ term on day −4 can explain the western part of the horizontal moisture advection (Fig. 10c), whereas the eastern part comes from the $-{(\overline{V}\cdot \nabla q^{\prime })}^{\prime }$ term even though it is weak. The sum of these two terms could roughly cover the horizontal moisture advection (Fig. 10e). Besides, there are still some weak high-frequency scale interactions such as the −(V″ ⋅ ∇q″)′ term that can contribute to the eastern part of the horizontal moisture advection (figure not shown). In general, the strongest $-{(V^{\prime }\cdot \nabla \overline{q})}^{\prime }$ term that locates the eastward phase leading of the convective center still can be recognized as the main factor in inducing the eastward movement of convection. For the PM, the $-{(\overline{V}\cdot \nabla q^{\prime })}^{\prime }$ term coincides well with the horizontal moisture advection on day −4 (Fig. 10n), which may be explained as the westward advection of the quasi-biweekly moisture gradient by the low-frequency easterlies. Again, the weak high-frequency scale interactions such as the −(V″ ⋅ ∇q″)′ term, −(V″ ⋅ ∇q′)′ term, and −(V′ ⋅ ∇q″)′ term also play positive roles to some extent. The upscale feedback from higher- to lower-frequency variability has been investigated widely, especially for the interaction between ISOs and synoptic-scale disturbances (Maloney 2009; Hsu and Li 2011; Wang et al. 2018). A phase-leading of the positive synoptic-scale disturbances always favors the propagation of the ISO. The $-{(V^{\prime }\cdot \nabla \overline{q})}^{\prime }$ term is significantly negative south of the convective center and the horizontal moisture advection center (Fig. 10m), which might largely obstruct the development of horizontal moisture advection. This finding might be a reason for the shortage of moisture in the growing phase for PM. The weak positive values locate southeast and southwest of the convective center. When the PM turns to the mature phase on day 0, the $-{(V^{\prime }\cdot \nabla \overline{q})}^{\prime }$ term performs as a northeast–southwest-tilted band, whose pattern is in general accordance with horizontal moisture advection (Fig. 10r). Generally, as the convections are still in their initial developing phase on day −4, the combined effect of the $-{(V^{\prime }\cdot \nabla \overline{q})}^{\prime }$ term, $-{(\overline{V}\cdot \nabla q^{\prime })}^{\prime }$ term, and −(V″ ⋅ ∇q″)′ term contributes to the initial movement for both modes, whereas the $-{(V^{\prime }\cdot \nabla \overline{q})}^{\prime }$ term plays the dominant role for both modes when the convections turn to the mature phase on day 0. As we are exploring the main mechanism for the propagation of these two kinds of QBWOs, we will focus on mechanism explanation on day 0 in the following analysis. For the EM, the positive $-{(V^{\prime }\cdot \nabla \overline{q})}^{\prime }$ is east of the cyclonic circulation, which seems to indicate that low-frequency moisture is transported from the northwest to the east of the QBWO convection. For the PM, the southeast shift of the positive $-{(V^{\prime }\cdot \nabla \overline{q})}^{\prime }$ relative to the cyclonic circulation denotes southwestward transportation of the low-frequency moisture by the quasi-biweekly wind anomalies on day 0. Therefore, the physical meaning of the top contributor terms in the zonal and meridional directions is further revealed in the following paragraph.

View Full Size

Horizontal composite of the moisture budget terms for the EM and PM on day −4 and day 0. The shading shows (first row) the 10–20-day horizontal advection averaged between 1000 and 300 hPa (g kg⁻¹ day⁻¹), (second row) the sum of 16 horizontal advection terms averaged between 300 and 1000 hPa (g kg⁻¹ day⁻¹), (third row) the quasi-biweekly wind times low-frequency specific humidity anomalies term (g kg⁻¹ day⁻¹), (fourth row) the low-frequency wind times quasi-biweekly specific humidity term (g kg⁻¹ day⁻¹), and (fifth row) the sum of row 3 and row 4. The vectors represent 10–20-day 850-hPa wind anomalies (m s⁻¹) in rows 1–3 and low-frequency wind anomalies (m s⁻¹) in row 4; only values that are statistically significant at the 95% confidence level are shown. The blue squares represent the convective centers.

Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-22-0766.1

• Download Figure
• Download figure as PowerPoint slide

As shown in Fig. 11, the calculation demonstrates that meridional moisture transport occurs; that is, the low-frequency moisture advected by the meridional quasi-biweekly wind is basically responsible for moisture phase leading in both QBWO modes. Specifically, for the EM, suppressed convection is located west of the active convection, which arranges the cyclonic and anticyclonic circulation anomalies zonally. Hence, hardly any northerly quasi-biweekly wind is found south of the active convection center, especially when their structures are northwest–southeast tilted. The maximized northerly anomaly appears east of the active convection center. For the PM, the major active convection on day 0 has a “tail” in its southwestern area, which originates from the western branch. Such a structure causes a strong northerly quasi-biweekly wind south of the active convection center. The importance of zonal and meridional moisture transport by the $-{(\overline{V}\cdot \nabla q^{\prime })}^{\prime }$ term on day −4 for PM is nearly equal despite the zonal component being located to the east and the meridional component being located to the west.

View Full Size

(a)–(h) Horizontal composite of the moisture budget terms for the EM and PM on day −4 and day 0. The shading in (a), (c), and (g) shows the zonal component of the quasi-biweekly wind times the low-frequency specific humidity term (g kg⁻¹ day⁻¹), whereas the shading in (b), (d), and (h) is for the meridional component. The shading in (e) shows the zonal component of the low-frequency wind times the quasi-biweekly specific humidity anomaly term (g kg⁻¹ day⁻¹), whereas the shading in (f) is for the meridional component. The contours in (a)–(d), (g), and (h) indicate the quasi-biweekly wind times the low-frequency specific humidity anomaly term (g kg⁻¹ day⁻¹). The contours in (e) and (f) indicate the low-frequency wind times the quasi-biweekly specific humidity anomaly term (g kg⁻¹ day⁻¹). The vectors in (a)–(d), (g), and (h) represent 10–20-day 850-hPa wind anomalies (m s⁻¹). The vectors in (e) and (f) represent low-frequency 850-hPa wind anomalies (m s⁻¹). Only values that are statistically significant at the 95% confidence level are shown. The blue squares represent the convective centers.

Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-22-0766.1

• Download Figure
• Download figure as PowerPoint slide

Thus, anomalous moisture advection can be easily understood when we consider the austral summer mean moisture distribution (Fig. 12). During austral summer, the southwest–northeast tilted mean moisture associated with the intertropical convergence zone (ITCZ) exhibits two positive centers, with one located in central Africa (western center) at approximately 10°S, 25°E and the other in the equatorial eastern IO expanding to the central south tropical IO (eastern center). For the EM, as deep convection is generated in the subtropics and propagates south of the western moisture center (blue line in Fig. 12), the northerlies of the cyclonic wind response bring higher mean moisture levels east of the convective center, which consequently leads to eastward movement of the QBWO. After the convection moves across Madagascar, the strong mean easterlies may hinder its eastward propagation, and the mean subsidence related to the Mascarene high appearing ahead may gradually weaken the convective system. Similarly, the deep convection of the PM initiates at approximately 10°S, which appears as a mean moistening center in the meridional direction. As such, the northeasterlies south of the active convection center bring higher mean moisture levels to the south and consequently lead to the southward movement of the QBWO. This convection also seems to be hindered by the Mascarene high and diminished gradually when it moved to the area south of Madagascar.

View Full Size

The austral summer mean low-frequency specific humidity averaged between 300 and 1000 hPa (shading with green solid contours; g kg⁻¹) and 850-hPa wind (vectors; m s⁻¹). The blue (dark red) line indicates the propagation routine of the EM (PM), with squares (diamonds) denoting the convective centers.

Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-22-0766.1

• Download Figure
• Download figure as PowerPoint slide

In summary, although the two QBWO modes originate in different regions and exhibit distinct evolutionary features, they have similar moisture dynamics. The phase-leading relationship between the positive moisture tendency and the active convection center is observed in both modes, which preconditions the movement of the QBWO. Horizontal moisture advection, particularly for the low-frequency moisture advected by quasi-biweekly wind anomalies [i.e., $-{(V^{\prime }\cdot \nabla \overline{q})}^{\prime }$], is the primary physical process behind the scenes. Further calculation demonstrates that meridional moisture transport occurs; that is, the low-frequency moisture advected by the meridional quasi-biweekly wind is basically responsible for the moisture phase leading in both QBWO modes.

5. Summary and discussion

The 10–20-day QBWO is active in the SWIO and significantly influences the local weather and climate system. Compared with comprehensive analyses of the QBWO in the Asian monsoon regions during boreal summer, studies that focus on the QBWO in the SWIO during austral summer are still scarce. Based on K-means cluster analysis, the diversity of the austral summer QBWO in the SWIO is examined in the current study. The QBWO is objectively classified into two archetypes, namely, the EM and PM, which correspond to the southwest and northeast convective centers, respectively, as mentioned in Kikuchi and Wang (2009). Therefore, the two modes specified in this study represent a further refinement of their research. Both QBWO modes show distinct circulation structures and movement routes. For the EM, the active convection center mainly comes from the subtropical ocean and exhibits an eastward propagation path. For the PM, the active convection center mainly comes from the tropical ocean and exhibits a poleward propagation path. Moisture dynamics play an important role in governing the QBWO. The moisture time tendency anomalies show a phase-leading relationship relative to both QBWO convection centers, which potentially guides convection movement. Moisture budget analysis indicates that this phase-leading in moisture tendency anomalies is mainly due to horizontal moisture advection, whereas the vertical moisture advection term and the moisture source/sink term are nearly out of phase with QBWO convection.

A detailed examination is performed to explore the underlying physical processes, revealing that horizontal moisture advection primarily results from the interaction between quasi-biweekly wind anomalies and low-frequency moisture [i.e., $-{(V^{\prime }\cdot \nabla \overline{q})}^{\prime }$] in both the growing phase and the mature phase for the EM but only the mature phase for the PM. The horizontal moisture advection in the growing phase of PM primarily results from the interaction between the low-frequency wind and the quasi-biweekly moisture [i.e., $-{(\overline{V}\cdot \nabla q^{\prime })}^{\prime }$], but the effects are very weak. As the convections are still in their initial growing phase on day −4, the combined effect of the $-{(V^{\prime }\cdot \nabla \overline{q})}^{\prime }$ term, $-{(\overline{V}\cdot \nabla q^{\prime })}^{\prime }$ term, and $-{(V^{″}\cdot \nabla q^{″})}^{\prime }$ term contribute to the initial movement for both modes, whereas the $-{(V^{\prime }\cdot \nabla \overline{q})}^{\prime }$ term plays the dominant role for both modes when the convections turn to the mature phase on day 0. As we are exploring the main mechanism for the propagation of these two kinds of QBWOs, we only focus on the mechanism’s explanation on day 0. Further analysis demonstrates that meridional moisture transport (i.e., the low-frequency moisture advected by the meridional quasi-biweekly wind) is basically responsible for moisture phase leading in both QBWO modes. For the EM, as deep convection is generated in the subtropics and propagates south of the moisture center, the northerlies of the cyclonic wind response bring higher mean moisture levels east of the convective center, which consequently leads to the eastward movement of the EM. Similarly, the deep convection of the PM initiates at approximately 10°S, which also appears as a mean moistening center in the meridional direction. As such, the northeasterlies southeast of the active convection center bring higher mean moisture levels south of the convective center and ultimately result in the southward movement of the PM. In other words, although the two QBWO modes originate in different regions and exhibit distinct evolutionary features, they are regulated by similar moisture dynamics.

These results support the conclusions of previous studies (Kikuchi and Wang 2009; Ghatak and Sukhatme 2021) for the two distinct kinds of QBWOs in the SWIO during austral summer. As discussed in the above studies, the eastern convective center (equatorial anomaly) is associated with an equatorial Rossby wave, whereas the western convective center (subtropical anomaly) is an extratropical Rossby wave (Kikuchi and Wang 2009). The equatorial Rossby wave from the tropics that dominates poleward propagation in the poleward mode has been discussed in detail by Ghatak and Sukhatme (2021) and is consistent with the propagation of the boreal summer QBWO (Chatterjee and Goswami 2004; Kikuchi and Wang 2009; Chen and Sui 2010; Wang and Chen 2017). However, the extratropical dynamics (e.g., extratropical Rossby waves) that control propagation in the eastward mode are still relatively unknown.

The phase speed of the EM accelerates up to approximately 7.5° day⁻¹ when the convection moves from eastern Africa and enters the Mozambique Channel. As the sea surface temperature (SST) tends to enhance the surface turbulent heat fluxes ahead of the convective center and suppress them behind the convective center (Li et al. 2020), the warmer SST in the Mozambique Channel (Han et al. 2019; Mawren et al. 2022) may play an essential role in accelerating the phase speed, which is worth detailed examination in future work. The air–sea interactions between the QBWOs and SST in the Seychelles–Chagos thermocline ridge also deserve to be investigated.

The upscale feedback from higher- to lower-frequency variability has been investigated widely, especially for the interaction between ISOs and synoptic-scale disturbances (Maloney 2009; Hsu and Li 2011; Wang et al. 2018). A phase-leading of the positive synoptic-scale disturbances always favors the propagation of the ISO. Although the interaction between QBWOs and synoptic-scale disturbances is weak, it also deserves to be examined carefully in the future.

Another interesting phenomenon is that the seasonal variation in the QBWOs might be associated with the background circulations. Thus, the frequency distribution of the two QBWO modes is provided in Fig. 13. The results suggest that the eastward mode tends to occur throughout the whole austral summer but mostly develops in the early summer (December; Fig. 13a) and the poleward mode mostly occurs in the middle and late summer (January and February; Fig. 13b). The variation in the background low-level moisture content or the vertical wind shear might be factors contributing to this phenomenon. However, more detailed studies are needed in the future.

View Full Size

Seasonal dependence of the occurrence of the (a) EM and (b) PM. The bars in each subplot represent the frequency distribution of the EM or PM in each month.

Citation: Journal of Climate 36, 23; 10.1175/JCLI-D-22-0766.1

• Download Figure
• Download figure as PowerPoint slide

Data availability statement.

The data that support the findings of this study are derived from the following sources: ERA5 reanalysis (https://www.ecmwf.int/en/forecasts/datasets/reanalysis-datasets/era5), OLR (https://psl.noaa.gov/data/gridded/data.olrcdr.interp.html), and GPCC precipitation data (https://climatedataguide.ucar.edu/climate-data/gpcc-global-precipitation-climatology-centre).