The variability and distribution of rainfall are uniquely significant for climatic risk prediction. This study aims to assess spatiotemporal rainfall variability and flash flood intensity events in the Sylhet haor region of Bangladesh by analyzing rainfall data from April for the period 1995–2022. For this, we used both Climate Hazards Group InfraRed Precipitation with Station (CHIRPS) data and Bangladesh Meteorological Department (BMD) daily rainfall data. Rainfall patterns were studied using zonal statistics in ArcGIS and graphical illustration. The results revealed that the rainfall pattern was erratic and showed a range of spatiotemporal variability. If the average rainfall exceeds 250 mm in Meghalaya and Assam and 400 mm in Sylhet, severe flash floods may occur in the Sylhet haor region. An increase in pre-monsoon rainfall and its shift from May to April may increase the intensity of flash floods and consequently damage the rice crop. This finding might help flood management agencies to develop flood management strategies, prepare flood contingency plans, provide real-time and advanced warnings to strengthen flood warning and forecasting systems, and schedule seasonal agricultural activities.
Comparative analysis of April rainfall trends in the Assam, Meghalaya, and the Sylhet haor region between 1995 and 2022.
Correlation between rainfall from Assam and Meghalaya and flash flood probability in the Sylhet haor region.
Flash flood severity forecast in the Sylhet haor region.
The variability in rainfall over time and space is one of the most significant features of Bangladesh's climate. The predominant climate characteristic in the northeastern Sylhet haor region of Bangladesh is rainfall (Hasan et al. 2012). The Sylhet haor region has a subtropical monsoon climate, and precipitation increases during the pre-monsoon season (April–May), with rainfall ranging from about 490 mm in the southwest to roughly 1,290 mm in the northeast (Nowreen et al. 2014; DBHWD 2017). The haor area, which is located directly below the hills of the Indian states of Assam, Meghalaya, and Tripura, has some of the worst hydrological conditions, such as extremely high rainfall and subsequent flooding (Kartiki 2011; Hossain & Naser 2014). The flooding in the haor area is associated with the amount of rainfall received in the upstream catchment in India (Meghalaya, Barak, and Tripura Basins in India) and often occurs on short notice (Adnan et al. 2019). This rapid flooding with a high discharge and velocity is called a flash flood. The haor region receives water from the catchment slopes of the Shillong Plateau over the borders in India to the north and the Tripura Hills in India to the southeast (Uddin et al. 2013; Bari et al. 2015).
The sudden flash flood in April is the biggest threat to the agriculture of the haor region (Kamruzzaman & Shaw 2018). Flash flood damages were maximum in 1997, 2000, 2002, 2004, 2007, 2010, 2012, 2015, 2016, 2017, 2019, and 2022, and in these years, the flash flood occurred in the haor area in April and also because the duration of the flash flood was the maximum (Salauddin 2010; Roy et al. 2017; Abedin et al. 2022; Rahman et al. 2022). The rainfall pattern in the upstream catchment has a significant impact since floods in the haor area are greatly influenced by rainfall on the nearby Indian side (Basher et al. 2018). A lot of rain falls in Meghalaya and the adjoining haor region of Sylhet when moisture-rich monsoon air sweeps up the hills and mountains of Meghalaya. Additionally, the local rainfall runoff from the Barak valley, located in southern Assam, with that from nearby mountainous areas, runs down the river Barak and its many tributaries before being discharged into Sylhet, Bangladesh (Dey et al. 2021; Haque et al. 2021).
The Bangladesh Water Development Board (BWDB) has revealed that more floods are occurring in the haor area than ever before. Numerous tracts of land in Sylhet and neighboring haors have been overwhelmed by flash floods brought by torrents of water flowing upstream from Meghalaya and heavy rainfall, trapping hundreds of people there (Jahan 2019). Additionally, there is a high chance of getting more floods because most of the sub-districts close to the Indian border may get flooded by the onslaught of water from upstream (Rashidin et al. 2019). Masood et al. (2015) projected that pre-monsoon rainfall in haor place may be extended by 11.4% in the near future (2015–2039), and the accelerated quantity is 33.6% from 2075 to 2099. The haor region has a greater dominance of rainfall variability than other regions of Bangladesh (Chowdhury et al. 2016). Due to inconsistent rainfall, several studies predicted that the flash flood depth would increase in the haor region in the future (Choi et al. 2021). Additionally, there is a possibility of shifting the flash flood time in the future (Masood & Takeuchi 2016). It inevitably requires a thorough understanding of the local rainfall pattern in this situation.
It is crucial to comprehend the characteristics of rainfall variability to reduce damages from the disaster. Now, researchers are concentrating on predicting the rainfall during the pre-monsoon and monsoon seasons (Rana et al. 2007; Terao et al. 2013). The periodicity and trends in the pattern of precipitation are apparent and have been demonstrated in many studies in broader aspects, both spatially and temporally. However, adequate studies are lacking in the regional context. The impacts of dynamic rainfall patterns on northeast India or, comparatively, a smaller state like Assam are less explored and less known till now, making the future scenario more uncertain for flash flood risk management in Bangladesh. However, this study aims to establish a relationship between flash flood probabilities in the Sylhet haor region with rainfall from Assam and Meghalaya and the prediction of flash flood intensity in haor that would provide suggestive information for flood risk managers and climate forecasters.
MATERIALS AND METHODS
Description of the study region
Fieldwork and questionnaire survey
Rainfall data used and analysis
We used the Climate Hazards Group Infrared Precipitation with Station (CHIRPS) data (https://data.chc.ucsb.edu/products/CHIRPS2.0/global_monthly/tifs/) for long-term rainfall analysis. As per requirement, we analyzed CHIRPS global monthly data for April from 1995 to 2022. We produced a year-wise rainfall pattern map with contours. We used Zonal Statistics as a Table, an ArcGIS Spatial Analyst geoprocessing tool, to calculate statistics like the average in defined zones. In ArcGIS, the Zonal Statistics as Table tool generates a summary table based on the specified input value raster, statistical function, and zonal extents.
Selection of flash flood years
Following a search of numerous research papers, a thesis, newsletters, conference proceedings, and reports, flash flood years were selected for this study. The comments of haor farmers on flash flood occurrence years were also taken into consideration. From November 2021 to April 2022, a field survey was conducted for this purpose. However, historical databases indicate that the Sylhet haor region experienced devastating flash floods in 1997, 2000, 2002, 2004, 2007, 2010, 2012, 2015, 2016, 2017, 2019, and 2022. As a result of the prolonged flooding period, considerable crop losses (mostly boro rice) occurred (Salauddin 2010; Rashid & Yasmeen 2017; Roy et al. 2017; Ferdushi et al. 2019; Abedin et al. 2022; Awal 2022). They stated that Sylhet haor was the most often affected area by flash flooding in April.
BMD rainfall data analysis
Comparison of CHIRPS satellite data and BMD rainfall data
Long-term variability of rainfall trends
The contours plotted at 30 mm intervals in the given Figure 5 represent the rainfall gradient. Closely spaced counts are a sign of orographic rainfall, which is the primary cause of flash floods in the Sylhet haor regions. The variation in monthly maximum rainfall over the past three decades was significantly higher than monthly minimum rainfall. Total monthly minimum rainfall varied from 11 mm in 1999 to 228 mm in 2004, while maximum rainfall varied from 161 mm in 1999 to 1,309 mm in 2004.
Spatiotemporal functional relationship of rainfall
Prediction of flash flood occurrence in the Sylhet haor region
Seasonal rainfall trend
Flash flood intensity prediction in the Sylhet haor region
Significant upstream rainfall results in flash floods, which rapidly raise river water levels and overbank discharges. Rainfall is an important meteorological component, and its intensity is taken into account when assessing flash floods. Understanding the characteristics of rainfall variability is essential to reducing damage from flash floods (Bagchi et al. 2020). In this study, we have tried to establish a correlation between rainfall in Assam and Meghalaya and the probability of flash floods in the Sylhet haor region by analyzing rainfall trends in Assam, Meghalaya, and the Sylhet haor region from 1995 to 2022 for April. We observed high-intensity flash floods in the Sylhet haor region during heavy rainfall years. The southern and western foothills of Assam and Meghalaya and the south and southwest of Assam receive heavy rainfall, with an average annual rainfall of more than 12,000 mm (BWDB 2016). Bangladesh receives extraordinarily high flows in northeastern rivers such as Surma, Kushiyara, Manu, Khowai, and Someswari due to the high rainfall in the Indian regions during the pre-monsoon (Adnan et al. 2019).
Based on time-series CHIRPS data from 1995 to 2022, the Sylhet haor region, Assam, and Meghalaya experienced unusually high rainfall levels, even more than 1,000 mm in April. There is also a spatial variability in rainfall in the three regions. Such variability may be due to complex interactions between dynamic systems operating at different spatial and temporal scales (Dey et al. 2021). Assam mainly consists of the Brahmaputra and Barak plain valleys, surrounded by numerous hills, and the Meghalaya mountainous province receives the highest average monsoon rainfall (Goyal & Gupta 2014). With the alignment of these two mountains, warm and moist air from the Bay of Bengal during the monsoon season is the main reason for Meghalaya's high rainfall (Prokop & Walanus 2015). In general, Sylhet and Sunamganj districts received more rainfall than Assam and Meghalaya. This can be attributed to the additional effect of the Meghalaya layer (Bhattacharya & Suman 2012; Rahman et al. 2018). Intense cloud activity over the Meghalaya Plateau creates a cool air mass covering the lower Sylhet layer (Kataoka & Satomura 2005). Murata et al. (2011) reported the existence of cool air masses, which are responsible for pre-monsoon rains.
Flash floods in the Sylhet haor region are usually worsened by incessant heavy rains in a large marshy area during April. This is in agreement with the findings of several researchers (Rozalis et al. 2010; Haque et al. 2021; Shuvo et al. 2021). Flash floods inundate low-lying floodplains, destroying crops and infrastructure and often taking lives and property. Significant losses of boro rice due to devastating flash floods are not uncommon (Islam et al. 2012; Sikder 2013; Kamal et al. 2018). We have determined the amount of rainfall in April that usually results in flash floods and especially damage to boro rice. For this, we estimated the occurrence and severity of flash floods in the Sylhet haor region from time-series rainfall data, which are often used to predict flash floods (Islam et al. 2019; Deka et al. 2022). We carefully observed and found strong correlations between the rainfall in Assam and Meghalaya and the haor region of Sylhet, with the aim of improving flash flood forecasting. We found a strong correlation between torrential rains in Meghalaya and severe flash floods in the Sylhet haor area. We also found a similar correlation between accumulated rainfall in Assam and the rise in water levels in haor. Mahtab et al. (2018) found that Meghalaya rainfall intensity is a reliable indicator of flash flood frequency and intensity in Tahirpur, Sunamganj. When it rains heavily in Cherrapunji, it takes 6–8 h to reach Tahirpur, and the rainwater does not drain quickly, spreading and creating flooding (Rashid & Yasmeen 2017; Basher et al. 2018).
We observed heavy rainfall for a short time in the Sylhet region during most of the historical flash flood years. Locally intense to extremely continuous and intense rainfalls increase the water level of the Surma River and exacerbate the flash flood situation (Jakariya & Islam 2017). The flow of upstream water almost inundates most of the sub-district near the Indian border and worsens the flash flood situation in Sylhet (Suman et al. 2014). We observed a decreasing trend in rainfall during the monsoon and an increasing trend during the pre-monsoon. The increasing trend of rainfall in the pre-monsoon season shows that Sylhet's flash flood is getting worse day by day. PRECIS (Providing REgional Climates for Impact assessments), a high-resolution regional climate model, is commonly used to produce climate change projections for impact assessments and adaptations. The PRECIS ensemble model predicts a shift to peak rainfall in May with a decrease in rainfall in August (Mondal et al. 2020). As a result, there will be frequent, brief, and intense rainfall episodes in the haor area, particularly during the pre-monsoon season (i.e., the months of March, April, and May). In their investigation, they compared the recorded rainfall over Bangladesh for a base period of 1986–2005 to the simulations of yearly, monthly, and severe rainfall produced by the PRECIS regional climate model. The measured rainfall was gathered from the 28 BMD gauge stations, and the model rainfall was simulated using the standard, unaltered parameter settings. Early flash floods in the haor region are now a new normal (Rumana et al. 2018). Mondal et al. (2020) reported that between 1901 and 1957, May received noticeably more rain than April, and between 1958 and 2017, the difference in total rainfall between May and April decreased while the total rainfall of April increased. Also, seasonal rainfall trend analysis of the Sylhet region indicated that rainfall is increasing during the pre-monsoon season showing a positive trend throughout the study period (1995–2022). Because of this, it is expected that the April rainfall will increase in the future, resulting in sudden, short, and intense flooding in the haor area. This circumstance is concerning since more precipitation in early April increases the risk of flash floods and destroys boro rice (Rumana et al. 2018). Using the PRECIS model, a significant increase in two extreme weather indices, such as ‘maximum rainfall in one day’ and ‘maximum rainfall in five consecutive days’ is projected during the 2080's pre-monsoon season near Sunamganj by Nowreen et al. (2014). Their estimate indicates the potential for frequent occurrence of massive flash floods. This will have a significant impact on the possibility of an earlier than normal flash flood and the loss of standing crops in haor.
A satellite-based approach to estimating precipitation, CHIRPS global monthly data combines in-situ station data with 0.05° resolution satellite images over a 50°S–50°N (and all longitudes) region. In contrast, the BMD gauge station data of Sylhet only offers point data. Both types of data are highly positively correlated, according to regression and correlation analyses. Bangladesh needs more site-specific meteorological stations to analyze the monthly rainfall pattern more carefully and to accurately predict flash floods in the Sylhet haor region.
The Sylhet haor region often experiences high-intensity flash floods when rainfall is high, especially in the foothills of Meghalaya, the southern and southwest parts of Assam, and the Jainta Hills of Meghalaya. Based on the historical rainfall data, there is a possibility of high-intensity flash floods in the Sylhet haor region if the average rainfall exceeds >250 mm in Meghalaya and Assam, and about 400 mm in Sylhet during the month of April. Flash floods are becoming more frequent and intense in the Sylhet haor region. Heavy rains in early April increase the probability of flash floods and the scenario would get worse as pre-monsoonal rains increase and shift from May to April. This requires early warning systems, farmer advisories, and public awareness of flash floods.
This research was supported by the Ministry of National Science and Technology (NST), Government of the People's Republic of Bangladesh.
DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.
CONFLICT OF INTEREST
The authors declare there is no conflict.