Decreasing number of propagating mesoscale convective systems in Bangladesh and surrounding area during 1998–2015

Mesoscale convective systems (MCSs) are important precipitation‐bearing phenomena in the Tropics. Hence, it is important to elucidate the temporal variation of MCSs. This study investigated the interannual variation of MCSs in Bangladesh and surrounding area during June–September using 3‐hr TRMM3B42 precipitation data. After objective detection of propagating MCSs (PMCSs), their statistical features were analyzed. It was found that the annual number of PMCSs decreased significantly during 1998–2015. The most remarkable decrease was in southward PMCSs generated over land, consistent with the observed decrease in precipitation. Recent weakening of the low‐level southwesterly and stabilization of the atmosphere probably contributed to the decrease of PMCSs. Because southward PMCSs generated over land were present on about 40% of days with heavy rain, PMCSs clearly have a crucial role in precipitation variability. The findings of this study suggest that investigation of MCSs is essential for understanding precipitation response to climate change in South Asia.


| INTRODUCTION
Monsoon precipitation in the vast tropical regions is vital as a huge fresh water resource, although it can cause devastating flood disasters. Therefore, elucidating the reasons for its variation on various timescales has long been a principal objective in tropical meteorology. Because of global climate change, both the duration of the rainy season and the amount of associated precipitation are projected to increase in regions dominated by the monsoon (Christensen et al., 2013). However, the predicted response of precipitation to elevated concentrations of greenhouse gases contains large uncertainty, especially in the Tropics (Kent et al., 2015). This is partly related to the difficulty in simulating mesoscale atmospheric processes in global climate models and partly attributable to the limited understanding of the response of mesoscale precipitation systems to a changing climate. Therefore, it is crucial to elucidate the multiyear variations of mesoscale precipitation systems based on archived observations. The focus of this study was Bangladesh and surrounding area, comprising the Himalayan ranges to the north, flat lowlying land of Bangladesh in the middle, narrow Arakan Mountains to the east, and Bay of Bengal to the south ( Figure 1a). The Meghalaya Plateau lies in the northern part of the domain. These complex topographic features enhance heavy precipitation on the windward side of the mountain slopes (Xie et al., 2006;Romatschke and Houze, 2011;Mahanta et al., 2013;Sato, 2013;Fujinami et al., 2017), which can cause devastating floods during the monsoon season . Severe weather and heavy precipitation events in this region are often associated with mesoscale convective systems (MCSs) (Ohsawa et al., 2000). Therefore, MCSs have been the subject of intensive study through observational data analysis (Ohsawa et al., 2000;Webster et al., 2002;Zuidema, 2003;Rafiuddin et al., 2013) and numerical modeling (Kataoka and Satomura, 2005;Jain et al., 2018). Hoyos and Webster (2007) noticed that MCSs are often embedded within large-scale envelopes associated with intraseasonal oscillation (Madden and Julian, 1972;Yasunari, 1979;Annamalai and Slingo, 2001). This denotes a crucial role of scale interaction between the synoptic circulation and MCSs, which might regulate the interannual variation of precipitation in South Asia (Fujinami et al., 2011). However, most previous studies on MCSs (e.g., Miyakawa and Satomura, 2006;Liu et al., 2008;Rafiuddin et al., 2010) have considered periods of only several years, which are insufficient to determine the role of MCSs on the interannual variation of precipitation under the effects of climate change.
The objective of this study was to reveal the interannual variation of MCSs during the monsoon season (June-September) using satellite-derived surface precipitation data. We further examined the contribution of MCSs to heavy precipitation events, which proved that understanding of mesoscale features is essential for the reliable interpretation of long-term precipitation change over South Asia.

| Data
The Tropical Rainfall Measuring Mission (TRMM) TRMM3B42v7 satellite-derived precipitation dataset (Huffman et al., 2007) was analyzed to detect MCSs in the region (15 -30 N, 87 -93 E) denoted by the black box in Figure 1a. The high spatial (0.25 × 0.25 grid) and temporal (3-hr) resolutions of the TRMM3B42v7 dataset are desirable for capturing the characteristics of MCSs. To investigate the synoptic conditions during the occurrence of the MCSs and to determine the interannual variation in MCS numbers, the 6-hr Japanese 55-year reanalysis (Kobayashi et al., 2015) dataset on a 1.25 × 1.25 grid was used. The period of analysis was June-September during 1998-2015, which comprises the period of full coverage of the TRMM observations.

| MCS detection method
Following the methodology of Carbone et al. (2002), we objectively detected propagating MCSs (PMCSs) that appeared in time-latitude plots of TRMM3B42v7-derived precipitation averaged over 87 -93 E (see Figure 1 for an example). The two-dimensional autocorrelation function (2DAF; Carbone et al., 2002) was superimposed on each pixel of the time-latitude plot and rotated at 10 intervals. This was performed to determine the orientation having the maximum correlation coefficient (r) between the 2DAF and the precipitation rate that satisfied the condition r > 0.5. Here, the threshold precipitation amount required for the detection of a PMCS was set to 3 mm/hr based on the mean precipitation rate along the axis of each PMCS episode (black lines in Figure 1b). We removed episodes with an angle of 90 , 270 , 180 , or 360 and those for which a certain direction of propagation was not identified in the animation. Small or short-lived PMCSs were subjectively identified by inspecting the animation of precipitation distribution. After the above screening processes, 5.7% of the objectively detected PMCS cases were removed, which left 700 PMCS cases for further analysis. The example of detected PMCSs shown in Figure 1b demonstrates the clear 1-day cycle in the timing of their genesis, similar to the diurnal cycle of precipitation induced by the thermally driven local circulation (Sato, 2013;Fujinami et al., 2017).

| PMCS CLIMATOLOGY
Prior to analysis of the temporal variation, we confirmed the basic features of the detected PMCSs because this study constituted the first attempt to compile the climatology of PMCSs in the studied region. The genesis location and propagation direction of the PMCSs are summarized in Table 1.
The total genesis number of southward PMCSs is more than twice that of northward PMCSs. Over land, the number of southward PMCSs is three times larger than northward PMCSs. Over ocean, the ratio of southward to northward PMCSs is close to double. The average propagation speed of  Figure 2 illustrates the genesis frequency with respect to latitude. The number of southward PMCS shows two peaks located around the northernmost Bay of Bengal and the Meghalaya Plateau. The former is probably related to the land-sea contrast, whereas the latter is more likely an outcome of orographic effects. In contrast, northward PMCSs appear insensitive to such forcings because they do not show clear peaks. The genesis of southward PMCSs is dominant over northward PMCSs at all latitudes except near the southern boundary. In Figure 2 (left), the composite of the lowlevel wind anomaly with respect to the climatology (June-September of 1998-2015) was calculated for the day of initiation of southward PMCSs generated over land (183 samples). The southwesterly wind was dominant over the south of the Meghalaya Plateau when land genesis of southward PMCSs occurred. This supports the findings of earlier case studies that found the counter flow toward PMCSs intensifies convergence near the front edge of the cold pool, which encourages MCS propagation (Kataoka and Satomura, 2005). These results suggest the hypothesis that temporal variation of the low-level southwesterly wind is fundamental to understanding the temporal variation of PMCS genesis (and hence, precipitation variability) on daily-interannual timescales.

| Interannual variation and trend of PMCSs
Changes in PMCSs might contribute to the interannual variation of precipitation over the study region. Figure 3a shows the interannual variation of the number of PMCSs during the monsoon season (June-September). The number of PMCSs decreased remarkably during 1998-2015, which was found statistically significant (p < 0.01) when examined using the Note. Occurrence frequency of PMCSs (number/year) with respect to genesis location (land or ocean) and propagation directions (northward or southward) are shown. The number in the parenthesis indicates percentage fraction of each type relative to the total genesis number.
Number ( nonparametric Mann-Kendall tau test. Among the types of PMCS, southward PMCSs generated over land (p < 0.01) and northward PMCS generated over ocean (p < 0.02) showed significant trends of reduction (Figure 3b). Figure 4a depicts the monthly mean 925-hPa wind regressed onto the interannual variation of the number of southward PMCSs generated over land during June-September. The number of annual occurrences during the monsoon season increased with stronger westerlysouthwesterly winds around Bangladesh (r = 0.6, p < 0.05), consistent with the composite analysis on the daily timescale (Figure 2, left). This implies the recent reduction in PMCS numbers (Figure 3a) is regulated by change in the low-level southwesterly wind. Figure 4b illustrates the distribution of the linear trend of low-level winds during 1998-2015. Significant weakening of the speed of the southwesterly winds exceeding 1.0 m/s/decade was detected during the monsoon season (p < 0.05), as indicated by the southwestward vectors around the Meghalaya Plateau and the southern Himalayas. Therefore, it is probable that weakening of the local monsoonal low-level flow could be attributed to the reduction in PMCS genesis.
We further analyzed dynamic and thermodynamic environmental conditions that might contribute to the PMCSs reduction. In the monsoon season (June-September), the convective available potential energy (CAPE) decreased significantly (p < 0.01) over the period of 1998-2015 while the convective inhibition (CIN) showed significant increase (p < 0.05) (Figure 5a). These results are consistent with observed decreasing trend of severe weather events across China in association with the CAPE reduction (Zhang et al., 2017). The CAPE does not change significantly for days with land-genesis southward PMCSs. It is speculated that the days with low CAPE tend to dominate in recent years.
The precipitable water vapor increased weakly for monsoon period (p < 0.10). The trend was more prominent for days with occurrence of land-genesis southward PMCSs (p < 0.05) (Figure 5b). There was no significant change in vertical wind shear between 850-hPa and 925-hPa ( Figure 5c) because weakening of the southerly to southwesterly flow around Bangladesh (Figure 4) occurred simultaneously at both two levels. Therefore, recent reduction in PMCS genesis is likely attributed to two drivers: stabilization of the atmosphere which weakens thermodynamic preconditioning, and weak low-level winds which reduced dynamic forcing for convective initiation.

| Contribution of PMCSs to heavy precipitation in South Asia
To investigate the contribution of PMCSs to heavy precipitation events, analysis was conducted using daily precipitation computed from the 3-hr TRMM3b42v7 dataset. Here, areaaveraged (21 -27 N, 87 -93 E) daily precipitation covering Bangladesh and Northeast India was used to identify the target days. A heavy rain day was assigned when the daily precipitation exceeded a threshold value. We adopted the threshold value of 23.0 mm/day which is derived from one standard deviation of daily precipitation added onto the climatological mean precipitation over the study period. It was found that PMCSs were present for 72% of the heavy rain days, indicating the fundamental role of PMCSs to heavy precipitation days. The percentages of heavy rain days associated with southward PMCSs generated over land, northward PMCSs generated over ocean, northward PMCSs generated over land, and southward PMCSs generated over ocean were 40, 15, 6, and 5%, respectively. The remaining events (6%) were associated with multiple PMCSs. The largest share of southward PMCSs generated over land suggests that reduction of the genesis of this type (Figure 3b) is relevant to precipitation variability. Figure 3c displays the interannual variation of heavy rain days. There is positive correlation (r = 0.6, p < 0.01) between the annual number of southward PMCSs generated over land and heavy rain days, although the number of heavy rain days does not exhibit a recent decreasing trend. Ground-based precipitation measurements suggest that monsoon precipitation around the study area has decreased in recent years, especially around the area to the north of 20 N (Naidu et al., 2009;Kumar et al., 2010). The occurrence of heavy rainfall events over northeast India (Mahanta et al., 2013) and Bangladesh  has also decreased during the previous few decades. Further studies are needed to enhance our understanding of the response of PMCSs, and the resultant precipitation, to various synoptic backgrounds characterized by intraseasonal oscillations and atmosphere-ocean coupled modes that are dominant on subseasonal-multidecadal timescales.

| SUMMARY AND DISCUSSIONS
In this study, we detected PMCSs using a satellite-based precipitation dataset for the monsoon season of 1998-2015. The genesis number of southward PMCSs generated over land is high around the Meghalaya Plateau and the southern Himalayas, indicating the importance of topographic effects to initiate the MCSs. The present study revealed the number of PMCSs, in particular southward PMCSs generated over land and northward PMCSs generated over ocean, has decreased since 1998. Among the various PMCS types, southward PMCSs generated over land contribute most to the heavy precipitation events that occur over northeast South Asia. The annual number of southward PMCSs generated over land during the monsoon season is modulated by the intensity of the low-level southwesterly wind. The decreasing trend in the number of southward PMCSs generated over land is likely due to weakening of the low-level southwesterly wind over the Meghalaya Plateau and the southern Himalayas as well as the recent stabilization of the atmosphere in Bangladesh and surrounding area.
This study found that PMCSs were present for more than 70% of heavy rain days around Bangladesh. Therefore, the recent decrease in the number of PMCS events, revealed in this study, is consistent with the decreasing tendencies in precipitation intensity and the number of heavy precipitation events found in previous studies. Anomalous warming of the Indian Ocean modulates the monsoon circulation, which subsequently weakens the strength of the southwesterly monsoon flow toward the Indian subcontinent (Naidu et al., 2009;Roxy et al., 2014). Weakening of the low-level southwesterly around the Meghalaya Plateau and the southern Himalayas appears associated with change in the monsoon circulation.
Despite the weakening of the monsoon circulation, precipitation is projected to increase in the future over South Asia during the monsoon season because of increased moisture flux from the ocean toward the land due to increasing atmospheric moisture content (Christensen et al., 2013). Our findings indicate that a large fraction of the monsoonal precipitation in the study area is the result of MCSs whose horizontal scale is smaller than that currently resolvable by existing atmosphere-ocean general circulation models. The low-level southwesterly wind has weakened in recent years and it is projected to decrease further in the years ahead; therefore, the number of PMCSs might decrease in the future. Consequently, predictions of the future change in precipitation around the study area could include considerable uncertainty associated with the behavior of MCSs under the effects of climate change. Further study is needed to elucidate the characteristics of MCSs, in particular the long-term changes in their intensity and life span, which will help further the understanding of the mechanism of precipitation variability.
MSc, from Edanz Group (www.edanzediting.com./ac) for editing a draft of this manuscript.