Monsoon Break over the South China Sea during Summer: Statistical Features and Associated Atmospheric Anomalies

The monsoon break usually manifests as a remarkable decrease or even disappearance of rainfall during the summer monsoon season (e.g., Gadgil and Joseph, 2003; Rajeevan et al., 2010). Severe break events adversely impact local agricultural production and the economy since they may cause droughts or heatwaves (Prasanna, 2014; Panda et al., 2017). For instance, in 2009, two prolonged break events caused the worst drought in India since 1972, which led to food insecurity and threatened the lives of millions of people (Xavier et al., 2010; Neena et al., 2011).

In addition to the Indian monsoon break that has been deeply researched (e.g., Ramamurthy, 1969; Gadgil and Joseph, 2003; Rajeevan et al., 2010; Dey et al., 2022), breaks of other monsoons have been gradually identified in recent decades (e.g., Drosdowsky, 1996; Jones and Carvalho, 2002; Chen et al., 2004). For instance, it was found that there are prominent monsoon break phenomena over the Indochina Peninsula (Takahashi and Yasunari, 2006) and the western North Pacific (WNP; Xu and Lu, 2015, 2022), two monsoon domains surrounding the South China Sea (SCS). Takahashi and Yasunari (2006) detected a climatological break in late June during the Southeast Asia monsoon, which features a rainfall reduction of ~2 mm d⁻¹. To the east of the SCS, Xu and Lu (2015) identified a monsoon break significant in climatology over the WNP, which occurs in early August with a reduction in rainfall amount similar to the Indochina break. Subsequent studies identified break events over the WNP and found that the break events can highlight the formation and maintenance of monsoon breaks (Vega et al., 2020; Olaguera et al., 2021; Xu and Lu, 2022). Furthermore, all these previous results, including those on the Indian monsoon break, suggest that the characteristics and mechanisms of the monsoon break are regionally dependent.

The SCS summer monsoon (SCSSM) is a vital component of the Northern Hemisphere summer monsoon. Due to its special geographic location, the SCSSM plays a crucial role in bridging the East and South Asian summer monsoons and the WNP summer monsoon (WNPSM) (Wang and LinHo, 2002). The onset of the SCSSM occurs in mid-May, characterized by a reversal of the low-level easterlies to westerlies and an abrupt convection enhancement (Wang et al., 2009; Kajikawa and Wang, 2012). This SCSSM onset can be regarded as an essential precursor of the establishment of the East Asian summer monsoon and thus has been extensively studied (e.g., Lau and Yang, 1997; Wang and LinHo, 2002; Wang et al., 2009). The SCSSM generally persists for about five months and withdraws around mid-October (Kajikawa and Wang, 2012; Luo and Lin, 2017; Hu et al., 2018, 2019).

Some studies have noticed that the break of the SCSSM also exists during the monsoon period (Chen et al., 1995, 2000; Jin and Tao, 2002; Zheng et al., 2016). These previous studies indicated that the SCSSM break is characterized by lower-tropospheric easterlies, in contrast to westerlies during active periods, and convection suppression. However, the monsoon break has not been specifically investigated in these previous studies on the SCSSM. For instance, Chen et al. (2000) focused on the relationship between the SCSSM and East Asian summer monsoon on intraseasonal time scales, while Zheng et al. (2016) categorized the SCSSM into active, inactive, and break phases, and emphasized the possible role of synoptic disturbances in depicting the SCSSM activities. Therefore, it is necessary to objectively identify the SCSSM break, and investigate the statistical features of the break, which is the major motivation of the present study.

The paper is organized as follows. Section 2 describes the data and methodology used in this study. In section 3, we introduce the definitions of monsoon onset and withdrawal dates. Section 4 identifies break events and explores their statistical characteristics. The change in atmospheric variables during the break compared to adjoining periods is examined in section 5. Section 6 further discusses the origin of circulation and convection anomalies responsible for the monsoon break. Conclusions are presented in section 7.

Daily interpolated outgoing longwave radiation (OLR) data from the National Oceanic and Atmospheric Administration (NOAA) is used as a proxy for convection (Liebmann and Smith, 1996). In addition, daily data of rainfall, sea surface temperature (SST), and multi-level atmospheric variables, including horizontal winds, air temperature, specific humidity, and vertical velocity, are obtained from the fifth generation of the European Centre for Medium-Range Weather Forecasts (ECMWF) atmospheric reanalysis dataset (ERA5; Hersbach et al., 2020).

All these data have a horizontal resolution of 2.5° × 2.5° and span from 1979 to 2020. Before the analysis, daily values on 29 February in leap years are excluded. The daily climatology is defined as the 42-yr mean of each calendar day, and daily anomalies are calculated by removing the daily climatology of that day from the raw data.

This study mainly uses composite and correlation analysis. The Student’s t-test with an effective degree of freedom is applied to examine the statistical significance of composite and correlation results. The effective degree of freedom Ne is estimated as follows (Zwiers and von Storch, 1995): Ne=N(1−r1)/(1+r1), where N is the original sample size and r1 is the lag-one autocorrelation. Specifically, the composite days are compared with all 42-yr summer monsoon days to gain statistical significance.

There are many indices based on different atmospheric variables to define the monsoon onset, including low-level winds (Wang et al., 2004; Kajikawa and Wang, 2012), rainfall (Wu, 2010), high cloud amount (Tanaka, 1992), and a combination of thermal and dynamic atmospheric variables (Ueda and Yasunari, 1998; Gao et al., 2001; Shao et al., 2015). These indices highlight different aspects of characteristics associated with the monsoon onset. In this study, we follow Wang et al. (2004) and Kajikawa and Wang (2012) to define the SCSSM onset by using the 850-hPa zonal wind averaged over 5°–15°N, 110°–120°E (here named the SCSSM circulation index, abbreviated Uscs). Specifically, the monsoon onset date is defined as the first day that satisfies the following criteria: (1) on the onset day and during the following five days, the average Uscs must be greater than 0 m s⁻¹; (2) in the subsequent 20 days, the Uscs must be positive for at least 15 days; (3) the cumulative 20-day mean Uscs must be greater than 1 m s⁻¹. This definition signifies a firm establishment of steady westerlies, which is the most prominent feature of the SCSSM onset (Wang et al., 2004).

The monsoon withdrawal can be viewed as an opposite process to the monsoon onset, namely, the low-level winds transform from westerly to easterly wind, which has been confirmed by several previous studies (Kajikawa and Wang, 2012; Luo and Lin, 2017; Hu et al., 2018). Therefore, we also use the Uscs to determine the monsoon withdrawal date. Simply we reverse the time series of Uscs and calculate the “onset day” based on the above three criteria, which actually means the last day when steady westerly wind dominates the SCS. Then the day after this day is defined as the monsoon withdrawal date.

Figure 1a gives an instance for the year 2018. According to the three criteria, the monsoon onset occurs on 1 June. Before the monsoon onset, easterly wind governs the SCS, indicated by negative Uscs. After the monsoon onset, Uscs drastically turns positive and maintains for about four months (albeit with some transient fluctuations in between), suggesting the establishment of stable westerly wind over the SCS. Figure 1b shows the reversed series of Uscs, and the “onset day” is determined to be on 1 October. Therefore, the monsoon withdrawal date is 2 October based on the definition. As shown in Fig. 1a, the monsoon withdrawal (marked by a star) is characterized by the decay of stable westerly wind and the establishment of easterly wind. Of note is that the monsoon withdrawal is more gradual compared to the drastic monsoon onset, as documented in previous studies (Wang and Wu, 1997; Wang et al., 2009).

Figure 1.  (a) Daily series of Uscs (m s⁻¹) in 2018. The green (red) line denotes the averaged Uscs during 5 days (20 days) after the onset day, and the blue dots present the positive Uscs during the 20 days after the onset day. (b) As in (a), but for reversed series of Uscs. The cross and stars mark the monsoon onset and withdrawal dates, respectively.

According to the above definition, monsoon onset and withdrawal dates are calculated over 1979–2020. The average monsoon onset and withdrawal dates are 22 May and 12 October, in good agreement with previous studies (Kajikawa and Wang, 2012; Luo and Lin, 2017). Also, we compared the yearly withdrawal dates with those in Luo and Lin (2017) and found that the correlation coefficient between the two is 0.64. Luo and Lin (2017) also defined the monsoon withdrawal by the zonal wind but used the cumulative Uscs index [refer to Luo and Lin (2017) for details]. Despite the discrepancies in the yearly onset and withdrawal dates, it should be mentioned that the conclusions drawn in this paper are valid. We have repeated the main analyses using the daily SCSSM onset and withdrawal dates defined by Luo and Lin (2017) and Hu et al. (2020a) and obtained similar results (not shown).

To further illustrate the temporal–spatial evolution of the zonal wind and convection around the SCSSM monsoon onset and withdrawal, we perform a composite analysis with respect to the onset and withdrawal dates (designated as day 0) during the 42 years. Figures 2a and 2b show the meridional evolution of the 850-hPa zonal wind and OLR over 110°–120°E. Before the monsoon onset (Fig. 2a), the SCS domain is dominated by easterlies, when the band of deep convection (represented by OLR < 230 W m⁻²) is limited to the south of 5°N. After the monsoon onset, the zonal wind transitions to westerlies over 5°–15°N, increasing drastically by more than 10 m s⁻¹ from days −5 to +5. Concurrently, the deep convection band rapidly expands northward to 20°N. On the contrary, as the monsoon withdraws (Fig. 2b), the steady westerlies with a magnitude of >6 m s⁻¹ weaken sharply and transition to easterlies over 5°–15°N, and the strong convection retreats from 20°N southward to ~10°N. The above meridional evolution clearly shows that the convection change occurs ~5° north of the zonal wind transition. Accordingly, the zonal wind and OLR averaged, respectively, along 5°–15°N and 10°–20°N are taken to investigate the zonal evolution in wind and convection (Figs. 2c and 2d). Specifically, the westerlies and deep convection are limited to the west of 105°E prior to the monsoon onset, and then extend rapidly to ~130°E as the monsoon becomes established (Fig. 2c). Dissimilar to the monsoon onset, around the withdrawal day, the westerly wind and deep convection concurrently disappear over a large area of 90°–150°E without showing a noticeable zonal shift (Fig. 2d). This synchronized monsoon retreat in the zonal direction was also documented in Hu et al. (2019), despite the fact that they derived their conclusion based on the climatological evolution of the SCSSM.

Figure 2.  (a) Latitude–time section of 850-hPa zonal wind (contours; the interval is 2 m s⁻¹) and OLR (colors; W m⁻²) averaged over 110°–120°E concerning the (a) onset day and (b) withdrawal day of the SCSSM. The negative and positive days represent the days before and after the onset and withdrawal day, respectively. (c, d) As in (a, b), except for the longitude–time section of the 850-hPa zonal wind and OLR. The zonal wind is averaged over 5°–15°N and OLR is averaged over 10°–20°N. The black contours denote 0 m s⁻¹, and the solid and dashed grey contours are for positive and negative zonal wind, respectively.

Many studies have revealed that there are prominent intraseasonal fluctuations of precipitation during the monsoon period, manifested as active spells with abundant rainfall and inactive spells with deficit rainfall (Chen and Chen, 1995; Krishnamurthy and Shukla, 2000; Yuan et al., 2010). Those inactive spells with extreme rainfall reduction are termed the “monsoon break” (e.g., Ramamurthy, 1969; Takahashi and Yasunari, 2006). Therefore, it is common to use rainfall (Gadgil and Joseph, 2003; Rajeevan et al., 2010; Olaguera et al., 2021) or OLR (Krishnan et al., 2000; Xu and Lu, 2015) to identify monsoon break events.

Figure 3a shows the correlation map of the OLR with regard to the U_scs index during the SCSSM. Significant negative correlations are found along 10°–20°N, covering the entire SCS and part of the western Pacific. The largest correlation is over 10°–20°N, 110°–120°E, which is thereby defined as the key region of convection, located north of the domain of Uscs by 5° latitudes. Figure 3b shows the correlation coefficients of the 850-hPa zonal wind with respect to the OLR averaged over the key region, and the largest negative correlation appears along 5°–15°N over the SCS. This spatial inconsistency between convection and zonal wind (as also noted in Figs. 2a and 2b) can be explained well by the Gill model (Gill, 1980), i.e., a Rossby wave induced by increased (reduced) diabatic heating due to enhanced (suppressed) tropical convection generates anomalous westerlies (easterlies) to the south of the convection center. The close relationship between the zonal wind and convection has been documented in several previous studies on the SCSSM (Wang et al., 2004, 2009). Consequently, it is reasonable to use the OLR over the key region to represent the evolution of the SCSSM and identify monsoon break events. This key region is exactly where the convection fluctuation is the most prominent during the SCSSM (not shown). An alternative approach to identifying break events is through using 850-hPa zonal winds. However, we found that the break events identified through this approach are manifested by much weaker convection suppression in the SCS (not shown).

Figure 3.  The correlations of (a) OLR at each grid with respect to the Uscs and (b) 850-hPa zonal wind at each grid with regard to the OLR averaged over the key region during the SCSSM. Shadings indicate a 99% confidence level based on a Student’s t-test. Rectangles denote the key region of convection and zonal wind, respectively.

A monsoon break event is defined as a spell during which the area-mean OLR exceeds 230 W m⁻² for three consecutive days or more, following Xu and Lu (2022), who studied monsoon break events during the WNPSM. This threshold of 230 W m⁻² corresponds to +0.52 standard deviations of the daily OLR time series during the 42 monsoon seasons. A difference from Xu and Lu (2022) is that they identified the monsoon break at every single grid due to the wide time–space range of the WNPSM, whereas we directly use the area-mean OLR since the SCSSM’s domain is much smaller and its evolution is concentrated in the key region. Besides, if two break events are at only a one-day interval, they are considered one event. Based on the method, we identify 214 break events during 1979–2020, including 1442 break days that account for 23.9% of the total summer monsoon days. The maximum value of the area-mean OLR is defined as the break event’s intensity, and the corresponding day is designated as the peak day of the event.

Figure 4a shows the 42-yr accumulated frequency of break days. The monsoon break barely appears before May, and the occurrence frequency grows rapidly in late May as the SCSSM becomes established. From June to September, the frequency of break days keeps generally constant, albeit reducing slightly after August. During these four months, the number of break days accounts for 86.1% of the total. The frequency declines remarkably around early October and almost disappears after mid-October due to the monsoon withdrawal. Figure 4b records the 214 break events during 1979–2020. As it shows, all 42 monsoon seasons saw break events, but the occurrence time, lengths, and frequency of break events differ considerably from year to year, which confirms that it is necessary to investigate the monsoon break as individual events to understand its occurrence regularity and physical process. Years with six break events are the most frequent (28.6% of the years), followed by five break events (26.2%) and four break events (21.4%). Figure 4c further depicts the yearly series of accumulated break days, which vary greatly with an average of 34.3 days and a standard deviation of 12.6 days. The maximum was 63 break days in 1979, and the minimum was only 10 days in 2007 and 2018. This yearly variation in break days is highly related to the length of the monsoon season. The correlation coefficient between them is 0.46, significant at a 99% significance level, indicating that the longer the monsoon period, the more break days. This relationship was also detected in the break of the WNPSM (Vega et al., 2020). In contrast to the strong interannual variation, no statistically significant trends in break days are found during 1979–2020.

Figure 4.  (a) Seasonal distribution of the 42-yr accumulated frequency of break days. The dashed lines mark the climatological SCSSM onset and withdrawal dates. (b) The record of the monsoon break events (brown lines) during 1979–2020. The red and blue curves denote, respectively, the SCSSM onset and withdrawal dates for each year. (c) Yearly frequency of break days. The dashed line indicates the 42-yr mean.

Figure 5a shows the duration distribution of break events. The lengths of break events vary in a wide span of 3–24 days, and the number of events generally decreases as the duration increases. In particular, the break events lasting for three and four days are far more frequent than others, with a total of 86 events. The average duration is 6.71 days, slightly longer than the 6.1 days of the WNPSM break (Xu and Lu, 2022). Nevertheless, break events over the SCS tend to be more durable than those over the WNP. For instance, 26% of WNPSM break events persist more than one week (Xu and Lu, 2022), while this percentage increases to 32.7% for SCSSM break events. The difference is more evident for those extremely long events. Only one break event exceeds two weeks over the WNP, in sharp contrast to 14 events over the SCS, among which two events are even no less than three weeks.

Figure 5.  Frequency distribution of break events with different (a) durations and (b) intensities. Vertical dashed lines represent the mean duration and intensity of all break events.

The intensity distribution of break events is shown in Fig. 5b. Interestingly, only a few events are found near 230 W m⁻² (the threshold used to define the monsoon break), and instead, the majority (67.3%) are concentrated between 250 W m⁻² and 270 W m⁻², i.e., 20–40 W m⁻² above the threshold. This distribution indicates that the OLR should increase rapidly as the monsoon break begins, given that most break events are short-lived, implying that synoptic disturbances or intraseasonal oscillations may play essential roles, which will be further explored in a separate work. Still, this speculation can be partially confirmed by the WNPSM break events, which display a similar intensity distribution under the great influence of tropical intraseasonal oscillations (Xu and Lu, 2022). In addition, the mean intensity of SCSSM break events is 258.6 W m⁻², which is almost identical to that of WNPSM break events, corresponding to +1.5 standard deviations of the daily OLR time series during the 42 monsoon seasons.

To reveal the changes of convection around the monsoon break, the OLR and OLR anomalies during the break period are compared with those in adjoining periods (seven days before and after the break). Before the monsoon break (Figs. 6a and 6d), deep convection, with a low center of <200 W m⁻², controls the SCS domain, along with significant negative OLR anomalies, indicating that the convection at this stage is anomalously enhanced compared to the climatology. Meanwhile, anomalously suppressed convection is found east of 130°E over the western Pacific. In contrast, during the break period (Figs. 6b and 6e), the contour of 230 W m⁻² shrinks sharply southward from north of 20°N to ~10°N in the SCS longitudes. Specifically, the value of OLR increases from below 200 W m⁻² to as high as >260 W m⁻², signifying a dramatic convection inhibition. Such a suppression in convection is also manifested in significant positive OLR anomalies, which are distributed in a banded region along 10°–20°N, covering the entire SCS and part of the western Pacific. It is noted that the amplitude of the positive OLR anomalies (40 W m⁻²) is even stronger than that of the WNPSM break events (30 W m⁻²; Xu and Lu, 2022). As the monsoon break ends (Figs. 6c and 6f), the deep convection and negative OLR anomalies reoccur over the SCS, similar to the period before the break, except that positive OLR anomalies are not discerned over the western Pacific.

Figure 6.  Composite (left) OLR and (right) OLR anomalies (colors; W m⁻²) averaged during (a, d) the seven days before the onset day of break events, (b, e) the break period, and (c, f) the seven days after the end day of break events. The thick contours in (a)–(c) denote 230 W m⁻². Stippled regions in (d)–(f) indicate a 95% confidence level based on a Student’s t-test. Rectangles mark the key region of the SCSSM as in Fig. 3a.

Since rainfall reduction is the most pivotal characteristic of the monsoon break, it is necessary to quantify the change in rainfall around the break of the SCSSM. The evolution of rainfall (Fig. 7) resembles that of convection, confirming that OLR can represent tropical precipitation well. Before the monsoon break (Fig. 7a), abundant rainfall covers the SCS domain with a maximum center of >15 mm d⁻¹. The corresponding positive rainfall anomalies (Fig. 7d) indicate anomalous rainfall enhancement. Negative rainfall anomalies occur to the east of 130°E, collocated with the positive OLR anomalies (Fig. 6d). In comparison, during the monsoon break (Fig. 7b), the contour of 6 mm d⁻¹, which is a key threshold representing monsoon rainfall over tropical regions (Lau and Yang, 1997; Wu and Wang, 2001), retreats remarkably southwestward. The entire key region is exposed to low rainfall, and the minimum center is merely 0–3 mm d⁻¹. That is, the rainfall reduction is more than 10 mm d⁻¹ during the break compared to before. Correspondingly, significant negative precipitation anomalies are detected spanning a large zonal scale from the Indochina Peninsula to the Philippine Sea along 10°–20°N, and the minimum value of <−8 mm d⁻¹ appears over the key region (Fig. 7e). Considering that the climatological summer mean rainfall is 8.9 mm d⁻¹, the amplitudes of the negative rainfall anomalies and the rainfall variation around the break are pretty substantial. Concurrently, significant positive rainfall anomalies are detected to the south and north of the key region, i.e., over the equatorial region and southern China, and are related to abnormal water vapor transport that will be discussed later. After the monsoon break (Figs. 7c and 7f), the monsoon rainfall resumes over the SCS, and the rainfall amount and anomalies are similar to those before the break.

Figure 7.  As in Fig.6, but for composite (left) rainfall and (right) rainfall anomalies (colors; mm d⁻¹). The black contours in (a)–(c) denote 6 mm d⁻¹.

Previous studies have suggested that large-scale circulations play essential roles in the monsoon break (Chen and Chen, 1995; Gadgil and Joseph, 2003; Xu and Lu, 2022). As shown in Fig. 8, before the monsoon break, the SCS is governed by the monsoon trough (Fig. 8a), and the trough line extends southeastward from the Indochina Peninsula to the Philippine Sea. This monsoon trough is abnormally enhanced, indicated by a cyclonic anomaly over the SCS (Fig. 8d), coinciding with the intensified convection and rainfall. During the monsoon break, the monsoon trough weakens notably with the trough line retreating to near the Indochina Peninsula (Fig. 8b). This contraction of the monsoon trough corresponds to a significant anticyclonic anomaly centered over the SCS (Fig. 8e), i.e., to the northwest of the convection suppression (Fig. 6e), suggesting a Rossby wave response to reduced diabatic heating (Gill, 1980). In turn, the pronounced northeasterly (southwesterly) anomalies south (north) of the anomalous anticyclone decrease (increase) the water vapor transported into (out of) the SCS by the mean southwesterly flow, and consequently lead to suppressed (enhanced) convection and rainfall over the SCS (near the equator and southern China). Moreover, this anticyclonic anomaly tends to have a baroclinic structure, and the SCS is dominated by a significant cyclonic anomaly at 200 hPa (not shown). It is interesting to note that these changes in large-scale circulations around the monsoon break are similar to those related to the SCSSM withdrawal (Luo and Lin, 2017; Hu et al., 2019). A crucial difference is that after the monsoon withdrawal, the monsoon westerlies in the lower troposphere transform into easterlies over the SCS (Fig. 3 in Hu et al., 2019), while during the monsoon break, the monsoon trough still prevails over the SCS, despite having weak intensity. After the monsoon break (Figs. 8c and 8f), the monsoon trough recovers over the SCS, thereby the convection and precipitation become enhanced again.

Figure 8.  As in Fig.6, but for composite 850-hPa (left) wind and (right) wind anomalies (vectors; m s⁻¹). The orange lines in (a)–(c) represent the axis of the monsoon trough, which is defined the same way as in Feng et al. (2020). The shadings indicate that the wind anomalies are significant at a 95% confidence level based on a Student’s t-test.

The vertical profiles of atmospheric anomalies over the key region during the break period are shown in Fig. 9. Consistent with the suppressed convection during the monsoon break, the vertical velocity (Fig. 9a) shows anomalous subsidence throughout the troposphere and is most significant at 300 hPa. In addition, significant negative anomalies of pseudo-equivalent potential temperature ($ {\theta }_{\mathrm{s}\mathrm{e}} $) are observed at all levels (Fig. 9b). $ {\theta }_{\mathrm{s}\mathrm{e}} $ in the lower troposphere has been commonly used to depict the warm-humid airmasses that signify the summer monsoon (e.g., Ding and Chan, 2005; Hu et al., 2019). Therefore, negative $ {\theta }_{\mathrm{s}\mathrm{e}} $ anomalies during the monsoon break indicate abnormally cold-dry airmasses over the SCS, which also denotes a weakening of the summer monsoon. Decreased $ {\theta }_{\mathrm{s}\mathrm{e}} $ in the middle and lower troposphere is closely related to moisture anomalies (Fig. 9c), which have a vertical profile similar to that of $ {\theta }_{\mathrm{s}\mathrm{e}} $, while air temperature anomalies are insignificant below 600 hPa (Fig. 9d). To explain the negative moisture anomalies, we further investigate horizontal and vertical moisture advection ($ -{{\bf{V}}}_{h}\cdot{{\boldsymbol{\nabla}} }_{h}q $ and $ -\omega {\partial }_{p}q $, respectively) and find that the latter plays an overwhelming role compared to the former. Namely, the downward transportation of dry air by strong subsidence dominantly contributes to the decrease in specific humidity. In contrast, the air temperature anomalies that are significant only in the middle and upper troposphere show the strongest cooling around 300 hPa (Fig. 9d), which can be explained by the decrease in diabatic heating due to significant rainfall reduction.

Figure 9.  Vertical profiles of anomalous (a) vertical velocity (Pa s⁻¹), (b) pseudo-equivalent potential temperature (K), (c) specific humidity (10⁻³ kg kg⁻¹), vertical and horizontal moisture advection (2×10⁻⁸ s⁻¹), and (d) air temperature (K) averaged over the key region during the break period. Circles denote the anomalies are significant at a 95% confidence level based on a Student’s t-test.

On the other hand, it is noted that $ {\theta }_{\mathrm{s}\mathrm{e}} $ anomalies decrease with height below 500 hPa (Fig. 9b), indicating that atmospheric instability in the lower troposphere is strengthened during the break period. However, the anomalously dry environment makes it difficult to release potential unstable energy, and the accumulation of unstable energy may eventually trigger the recovery of convection after the break. This atmospheric stratification during the monsoon break is analogous to that before the SCSSM onset, characterized by a pronounced dry layer in the middle troposphere (Ding and Liu, 2001). Conversely, during the convective period before the break, atmospheric instability is anomalously decreased (not shown), probably due to the release of convection instability related to the successive development of convective clouds (Ninomiya, 1999), and the gradual stabilization of the atmospheric stratification paves the way for the subsequent occurrence of the monsoon break.

Figure 10 further illustrates the spatial distribution of the above atmospheric anomalies. The vertical velocity at 300 hPa shows significant subsidence anomalies over the SCS and Philippine Sea with a center at the key region (Fig. 10a), a pattern highly resembling that of the anomalously suppressed convection and precipitation (Figs. 6e and 7e). This horizontal distribution remains consistent in the entire troposphere, although the amplitudes are smaller at other levels (not shown). The thermodynamic anomalies are examined in the lower troposphere, and here we show the result at 850 hPa, the same as in previous studies on the SCSSM (Hu et al., 2019; Fan et al., 2022). Significant negative anomalies of $ {\theta }_{\mathrm{s}\mathrm{e}} $ are observed in a band-like region (Fig. 10b), roughly corresponding to abnormal subsidence, which is because the strengthened descending flow results in a substantial decrease in specific humidity (Fig. 10c), consistent with our discussion in Fig. 9. As also shown by the original value of $ {\theta }_{\mathrm{s}\mathrm{e}} $, the extent of warm-humid airmasses, represented by the contour of 344 K, covers the key region under normal conditions but shrinks considerably to be near the Indochina Peninsula during the break. In contrast to specific humidity, air temperature anomalies are insignificant over the key region (Fig. 10d), indicating a weak contribution to $ {\theta }_{\mathrm{s}\mathrm{e}} $ anomalies. Anomalous cooling (warming) is found to the south (north) of the key region.

Figure 10.  Anomalies of (a) 300-hPa vertical velocity (colors; Pa s⁻¹), and 850-hPa (b) pseudo-equivalent potential temperature (colors; K), (c) specific humidity (colors; 10⁻⁴ kg kg⁻¹), (d) air temperature (colors; K) during the break period. Stippled regions indicate a 95% confidence level based on a Student’s t-test. The isolines in (b) represent 344-K values of pseudo-equivalent potential temperature, which depicts the extent of warm-humid airmasses. Dashed and solid isolines are for the composite mean during the break periods and during the 42-yr summer monsoon seasons, respectively. Rectangles mark the key region of the SCSSM as in Fig. 3a.

A composite analysis is constructed concerning the peak day (day 0) of each break event (see definition in section 4) to reveal the evolution of anomalous convection and circulation associated with the monsoon break. As shown in Fig. 11a, on day −8, positive OLR anomalies begin to emerge near the equatorial western Pacific, at which time negative OLR anomalies are observed over the Philippine Sea. These areas of anomalously suppressed and enhanced convection both move northwestward in the following days, with the positive OLR anomalies gradually enhancing but the negative ones decaying. On day −6 (Fig. 11b), an anticyclonic anomaly forms to the northwest of the suppressed convection, which can be explained as an atmospheric response to the reduced diabatic heating, and this anticyclone in turn inhibits convection, as we have discussed in Fig. 8. This convection suppression and the coupled anticyclone get stronger over the WNP on day −4 (Fig. 11c), and their western parts penetrate into the SCS on day −2 (Fig. 11d). At this time, the suppressed convection and the anomalous anticyclone possess the largest zonal scale, dominating the SCS and WNP domain. They further migrate westward and center at the SCS on day 0 (Fig. 11e), reaching their peak intensity. The maximum amplitude of OLR anomalies exceeds 45 W m⁻². Meanwhile, the monsoon westerlies are remarkably reduced by the easterly anomaly in the south of the anomalous anticyclone. After day 0, the positive OLR anomalies and the anomalous anticyclone turn to weaken and extend westward (Figs. 11f and g). Concurrently, anomalously enhanced convection occurs south of the key region and gradually moves northward. On day +6 (Fig. 11h), the positive OLR anomalies over the SCS almost vanish.

Figure 11.  Composite evolution of OLR anomalies (colors; W m⁻²) and 850-hPa wind anomalies (vectors; m s⁻¹) from days −8 to +6 with an interval of 2 days. Stippled regions denote a 95% confidence level based on a Student’s t-test. Only the vectors significant at a 95% confidence level are shown. The dashed line in (a) represents the general propagation path of anomalous convection, which is used in Fig. 12. Rectangles mark the key region of the SCSSM as in Fig. 3a.

Figure 12a more clearly illustrates the northwestward shift of the anomalous convection by compositing OLR anomalies along the dashed line in Fig. 11a. It shows that the suppressed convection propagates along this track after it forms near the equatorial western Pacific on day −8. Of note is that when the suppressed convection arrives at the SCS, its amplitude reaches the peak and its horizontal scale enlarges remarkably, but the propagation slows down. As can be seen, the center of the positive OLR anomalies stays at the key region from days −1 to +4. Coupled with the suppressed convection, negative vorticity anomalies in the lower troposphere present a quite similar propagation and development process (Fig. 12b). These negative vorticity anomalies, corresponding to the anomalous anticyclone, are located ahead of (i.e., to the northwest of) positive OLR anomalies, consistent with the results shown in Fig. 11. The anomalous anticyclone intensifies local divergence of moisture flux (not shown) and thus promotes the suppressed convection to shift toward the northwest. This indicates that the internal atmospheric dynamics favor the northwestward migration of the suppressed convection. This convection–circulation-coupled mechanism was also used to explain the northwestward propagation of quasi-biweekly and 30–60-day oscillations over the WNP (Hsu and Weng, 2001; Yang and Li, 2020). It should be noted that the phase difference between convection and circulation is more evident before day 0, which may contribute to the more rapid shift of the suppressed convection in early stages.

Figure 12.  Composite evolution of (a) OLR anomalies (W m⁻²), (b) 850-hPa vorticity anomalies (10⁻⁶ s⁻¹), and (c) SST anomalies (K) along the dashed line shown in Fig. 11a. Stippled regions indicate a 90% confidence level based on a Student’s t-test.

On the other hand, tropical SST also shows significant changes around the monsoon break (Fig. 12c). Negative SST anomalies appear over the SCS from days −8 to −3, a period when enhanced convection dominates the SCS, indicating that the atmosphere influences the ocean probably through cloud-radiation or wind-evaporation effect (Wang et al., 2005; Yang and Li, 2020). In turn, this SST cooling can stabilize the atmosphere above, which is favorable for propagating the suppressed convection into the SCS. As the monsoon break takes place around day 0, the enhanced solar radiation and reduced wind speed lead to SST warming, which increases the convective instability in the atmosphere (Wang et al., 2005) and finally hinders the maintenance of the monsoon break. These results suggest that the atmosphere–ocean interaction also contributes to the northwestward propagation of the suppressed convection (e.g., Liu and Wang, 2014) and modulates the evolution of the monsoon break over the SCS.

In this study, we identify monsoon break events during the SCSSM over 42 years (1979–2020) and investigate their statistical characteristics. A sharp contrast is found in atmospheric convection, rainfall, specific humidity, and large-scale circulations, between the break period and before/after the break. Furthermore, the origins and temporal evolutions of the convection and circulation anomalies responsible for the SCSSM break are explored.

During 1979–2020, 214 break events are identified based on the criterion that inactive convection (OLR > 230 W m⁻²) lasts three or more days during the SCSSM. These events include 1442 break days (accounting for 23.9% of the SCSSM days), distributed almost evenly between June and September. Statistical analysis indicates that the break durations span extensively from 3 to 24 days, with the number of short-lived events (≤ 7 days) being double that of long-lived events (> 7 days). Even so, there are still 6.5% of events that can persist longer than two weeks. This proportion of extremely persistent events is much higher than the WNPSM break (Xu and Lu, 2022). On the other hand, the intensity of break events varies from 230 W m⁻² to 280 W m⁻², with 67.3% concentrated between 250 W m⁻² and 270 W m⁻², denoting that the convection undergoes a substantial and drastic suppression during the break period.

A comparison is conducted for atmospheric variables between the break and adjoining periods. Compared with the periods before and after the monsoon break, during the break, both convection and precipitation are significantly suppressed, with the amplitude of OLR increase and rainfall reduction exceeding 60 W m⁻² and 10 mm d⁻¹, respectively. This amplitude of rainfall reduction is pretty substantial, given that the local summer mean rainfall is 8.9 mm d⁻¹. The above convection and rainfall variations around the break correspond well to the change in large-scale circulations. During the monsoon break, a significant anticyclonic anomaly occupies the SCS in the lower troposphere, corresponding to a remarkable westward shift of the monsoon trough from the Philippine Sea to the Indochina Peninsula. This anomalous anticyclone, as an atmospheric response to reduced diabatic heating, further suppresses convection and precipitation by decreasing moisture transportation into the SCS and increasing low-level divergence. Moreover, during the monsoon break, another important atmospheric variable that represents the SCSSM, $ {\theta }_{\mathrm{s}\mathrm{e}} $, exhibits significant negative anomalies, which also signifies the monsoon weakening. This decrease in $ {\theta }_{\mathrm{s}\mathrm{e}} $ is attributable to the reduction in specific humidity, which mainly results from dry air advection by anomalously enhanced subsidence throughout the troposphere.

To understand the formation of the monsoon break, the daily evolutions of convection and circulation are further investigated by performing a time-lag composite analysis concerning the peak phase of break events. The results suggest that the anomalous convection suppression associated with the monsoon break forms near the equatorial western Pacific. This inactive convection gradually propagates northwestward and develops over the WNP, before reaching its maximum intensity when entering the SCS. In the journey, a low-level anomalous anticyclone is located northwest (ahead) of the suppressed convection. Therefore, this anomalous anticyclone-induced moisture divergence promotes the suppressed convection to migrate toward the northwest. Besides, SST cooling is detected in the SCS before the monsoon break, probably in response to the enhanced convection. This SST cooling stabilizes the atmospheric layer and thus favors the suppressed convection that moves into the SCS.

The SCSSM break exhibits significant differences from the WNPSM break, though the former relates to the westward propagation of convection anomalies originating from the WNP. For instance, break events over the SCS appear almost evenly between June and September, while those over the WNP are highly concentrated in late August (Xu and Lu, 2022). Besides, break events over the SCS are much longer-lasting events than those over the WNP. These discrepancies suggest that the SCSSM break is, to a certain extent, independent of the WNPSM break, and has its own maintenance mechanisms, probably attributable to regional differences in seasonal advance, background circulation, air–sea interaction, etc. The relationship between the break events over these two adjacent domains is worthy of further exploration.

Previous studies have demonstrated that tropical intraseasonal oscillations significantly modulate the activity of the SCSSM (Chen and Chen, 1995; Mao and Chan, 2005). Around the monsoon break, the changes in convection and circulation are pretty drastic, implying that intraseasonal oscillations, even synoptic-scale waves, may play a critical role. Whether and how these different time-scale factors affect the SCSSM break remains an interesting subject for future research. On the other hand, the SCS is known as a vital water vapor source for the East Asian monsoon (Chen et al., 2000; Ding and Chan, 2005), thus it can be inferred that the prominent circulation and moisture anomalies during the SCSSM break may influence the precipitation in East Asia. Finally, extratropical oscillations and ENSO, which significantly affect the East Asian summer monsoon, including the SCSSM (e.g., Zhu and Li, 2017; Martin et al., 2019; Chen et al., 2020, 2023; Hu et al., 2020b, c, 2021; Liu et al., 2022), may also play a role in modulating the SCSSM break. These issues will be investigated in the future.

Acknowledgements. This work was supported by the National Natural Science Foundation of China (Grant No. 42275025).