Abstract
Due to global warming, extreme hydroclimatic events (e.g., floods) are expected to happen more frequently and last longer. This study investigated such an extreme flood in the transboundary Teesta River that occurred in October 2021. We attempted to quantify the event's impact using data from time series flood levels, precipitation-related satellite images, and two-dimensional hydromorphological modeling. We found it challenging for people to cope with such a hazardous event since the depth of the flooding increased 6.98-fold in just 24 h. Our simulation results indicate that a sand-filled sediment measuring 0.27 m thick covered more than 33% cropland, and the velocity increased by almost 2.5 times. 136,000 individuals were marooned in the water. Compared to previous flooding events in its basin, which occurred in India and Bangladesh, the river appears to have some natural shock absorption features, i.e., a wide braided plain. We propose impact-based forecasting with a proactive early response as a valuable tool for managing such extreme events.
HIGHLIGHTS
Evidence of an extreme climatic event and its impact.
Identifies natural shock absorbing phenomena of a river.
Looks into the flood from hydromorphological consequences and human impact.
Suggests how to manage such an extreme event in a better way.
INTRODUCTION
Flash flooding, due to the complex, multifaceted processes of the catchment, can cause severe damage to societies. Globally, it is evident that extreme rainfall events have increased in recent years due to climate change, and flash flooding is quickly becoming one of the most common types of natural disaster worldwide (IPCC 2022; Yang et al. 2022). Managing flash flooding is very challenging because of its suddenness, complex pattern, and difficulty in forecasting (Yang et al. 2022). It is widely believed that flash floods are highly localized, short-duration events with a very high peak, with less than 6 h between the occurrence of the rainfall and the occurrence of the peak flood (Foody et al. 2004; Thayyen et al. 2013; Archer & Fowler 2018). However, such floods had a substantial impact on the most downstream part of the river where the valley gradient is low, which has received little attention.
Several studies have been conducted over the last two decades to improve flash flood forecasting (O'Donnell 2002; Asghari & Nasseri 2015; Alipour et al. 2020; Zanchetta & Coulibaly 2020; Nanditha & Mishra 2021; Yordanova et al. 2022). However, damages have not been significantly reduced (Montz & Gruntfest 2002). For proactive flash flood risk management strategies, it is crucial to understand the flash flood characteristics and possible impacts (Hapuarachchi et al. 2011). Numerous studies have been conducted around the world to characterize flash floods (e.g., Bonacci et al. 2006; Rusjan et al. 2009; Rizzetto 2020) and risk management (e.g., Xia et al. 2011; Abuzied et al. 2016; Ma et al. 2021). There have been few studies on flash flood extreme events (e.g., Barrera et al. 2006; Marchi et al. 2010; Kumar et al. 2018). Because many of the study areas are located within a single country, management and information dissemination were simple. Since extreme events pose new challenges to risk management, process understanding becomes crucial. In this research, an extreme flash flood event has been investigated that occurred in October 2021 in the transboundary braided river, Teesta. The Teesta is an important source of fresh water for both India (especially West Bengal) and Bangladesh, as agricultural farming is the main income source for most of the people in its basin (Sarker et al. 2011a, 2011b).
Photographs of the damage due to the flash flood in the Teesta River basin. (a) River bank damage by flood, (b) flood bypass road collapsed after recent flood, (c) after flood farmer preparing their land for cultivation, (d) a woman indicating how far the flood water level rose, (e) a road damages by recent flood, and (f) damaged houses due to flooding (at Hatibanda, Lalmonirhat, Bangladesh, November 2021. Photo source: author).
Photographs of the damage due to the flash flood in the Teesta River basin. (a) River bank damage by flood, (b) flood bypass road collapsed after recent flood, (c) after flood farmer preparing their land for cultivation, (d) a woman indicating how far the flood water level rose, (e) a road damages by recent flood, and (f) damaged houses due to flooding (at Hatibanda, Lalmonirhat, Bangladesh, November 2021. Photo source: author).
The Teesta's catchment is well known for having a complex hydrological regime. The river is fed by groundwater as well as precipitation, melting glaciers, and snow (Wiejaczka et al. 2014; Mandal & Chakrabarty 2016). Large post-flood boulders with a diameter of up to a few meters are present in the Teesta River's channel, similar to other Himalayan rivers (Wiejaczka et al. 2014). However, in Bangladesh, only sand is discernible. One of the causes of the Teesta flash flood is sudden cloud bursts or orographic rainfall in the Darjeeling and Sikkim Himalayas during the pre- and post-monsoon periods. For instance, a cloud burst in 2015 caused the flood, forcing the flood embankments to be breached by flow through the channel (Pal et al. 2016). Landslides in Darjeeling significantly impact Teesta plain floods, particularly in the Indian portion (Jalpaiguri district). Landslides can occasionally obstruct river flow, increasing the likelihood of flash floods. However, there is not much solid proof linking landslides to the occurrence of flood events. The flood of 1968 was one exception (Wiejaczka et al. 2014; Pal et al. 2016).
The basin has anthropogenic interventions, some of which are under-designed. In 1993, the barrage (in India) could not adequately control the floodwater due to the intense water pressure upstream caused by heavy rainfall. Consequently, the sudden release of water resulted in a flash flood. However, the flooding was not so severe in Bangladesh. The flood characteristics of the braided plain of Teesta (in Bangladesh) are one of the poorly investigated regions (Mondal et al. 2020). In 1968, during the catastrophic flood in India's Teesta basin, the water level was flowing 0.69 m below the ‘DL’ (52.40 m) at Dalia in Bangladesh (FFWC 2015). In 2015, the water level was flowing 0.8 m above the DL. But how these flooding shocks were reduced in Bangladesh was not documented in previous literature. Glacial lake outburst floods (GLOFs) are another potential threat, but they have not been quantified for Bangladesh (Aggarwal et al. 2016).
In comparison to the Indian section of the basin, there are few studies on the Teesta flood in Bangladesh's portion. Mondal et al. (2020) examined the flood risk of riverine households from the perspective of socioeconomics using a qualitative survey. Sarker et al. (2011a, 2011b) investigated the impact of climate change on barrage irrigation. Norms, practices, and gendered vulnerabilities were the focus of Ferdous & Mallick (2019). Rahman et al. (2011) discussed design flow characteristics while examining historical maximum and minimum flow trends. Similarly, Islam et al. (2018) used hydrologic modeling to investigate the impact of climate change on flooding in the Teesta River basin. The literature review suggests that there is a gap in understanding flood hydraulics in previous literature.
Therefore, this research's primary objective is to investigate the October 2021 flood event. Our research questions were as follows: (1) Was it an extreme climate-driven event? (2) If yes, then how severe is this flood in terms of depth, duration, velocity, and sedimentation? (3) Lastly, how much damage did the flood cause in terms of population exposure and crop losses?
STUDY AREA
The focus area of this study is the portion of the river following through Bangladesh, as shown in Figure 2. It crosses into Bangladesh at the Kharibari border crossing in the Rangpur division. The Teesta River flows year-round as a rain- and snow-fed river (Mondal & Islam 2017). The amount of water that flows down the Teesta is highly variable between dry (December–February) and rainy periods (June–October). Near Chilmari, Bangladesh, it meets the Jamuna River after a 121-km journey. The Teesta sub-catchment in Bangladesh is about 2,000 sq. km (Rahman et al. 2011). The riverbed is made up of fine to medium sand that is typical of alluvial floodplains. The river's average gradient is between 0.47 and 0.55 m/km (Rahman et al. 2011). Every year, the Teesta sub-catchment is at risk of flooding. The Teesta River is a major contributor to flash flooding, and the heaviest flooding happens when the maxima of the Jamuna and Teesta Rivers coincide.
MATERIALS AND METHODS
Data type and source
Details of the data used in this study
Type . | Data . | Source . | Period . |
---|---|---|---|
Time series data | Water level | Bangladesh Water Development Board (BWDB) at Dalia, Kaonia, and Haripur stations | 1961–2021 |
Discharge | Bangladesh Water Development Board (BWDB) at Dalia and Kaonia stations | 1961–2021 | |
Satellite-based data | Precipitation | Global Precipitation Measurement (GPM) | 17 October to 21 October 2021 |
Vegetation Index | Visible Infrared Imaging Radiometer Suite (eVIIRS) | 17 October to 21 October 2021 | |
Land use land cover (LULC) | Landsat 4–9 | 1991–2021 | |
Population | Population exposure | WorldPop (2022) (Database of population counts with a 100-m resolution) | 2022 |
Primary data | Depth, velocity, duration, and exposure | Key Informant Interview (KII) with affected people | November 2021 |
Type . | Data . | Source . | Period . |
---|---|---|---|
Time series data | Water level | Bangladesh Water Development Board (BWDB) at Dalia, Kaonia, and Haripur stations | 1961–2021 |
Discharge | Bangladesh Water Development Board (BWDB) at Dalia and Kaonia stations | 1961–2021 | |
Satellite-based data | Precipitation | Global Precipitation Measurement (GPM) | 17 October to 21 October 2021 |
Vegetation Index | Visible Infrared Imaging Radiometer Suite (eVIIRS) | 17 October to 21 October 2021 | |
Land use land cover (LULC) | Landsat 4–9 | 1991–2021 | |
Population | Population exposure | WorldPop (2022) (Database of population counts with a 100-m resolution) | 2022 |
Primary data | Depth, velocity, duration, and exposure | Key Informant Interview (KII) with affected people | November 2021 |
Methodology
Time series data analysis
The BWDB measured data on 60 years of water levels (Wls) and discharges (from 1961 to 2021) were procured and analyzed to determine whether there is a pattern of change in peak flooding. The analysis of daily precipitation percentiles to observe the extreme precipitation increase under global warming is quite common among researchers (i.e., Manton et al. 2001; Roy & Balling 2004; Chen & Sun 2017; Myhre et al. 2019). Following the same example, the variations in daily Wl, intensity, and frequency were analyzed. The relationship between the maximum annual flood level and its corresponding rate of rise was examined. In addition, the WorldPop (2022) database of population counts with a 100-m resolution was retrieved to assess the exposure.
Satellite image analysis
To understand why the locals were misled about the early warning signs of such devastating flooding, the next step was to analyze the basin-wide daily accumulated rainfall from the Global Precipitation Measurement (GPM) from 17 October to 21 October 2021. The Normalized Difference Vegetation Index (NDVI), which is derived from data from the Visible Infrared Imaging Radiometer Suite (eVIIRS), was used to evaluate crop damage. Ha et al. (2022) used the difference in NDVI before and after hail damage on several types of crops and compared it with the ground data. They found that the NDVI value is well correlated with ground estimates of crop damage. Therefore, we followed a similar technique in crop damage estimation. The flooding event was classified as ‘low’, ‘moderate’, and ‘high’ damage in accordance with the local people's perception. Besides this, the river planform changes from the years 1973 to 2021 were also performed.
The changes in land use and land cover (LULC) over a 40-year period (from 1991 to 2021) were also quantified. In this case, dry season images (December–February) of Landsat 4 through 9 were used. Here, an atmospherically corrected and georeferenced level 2 product was used. To begin, unsupervised image classification techniques based on statistical differences in pixel spectral characteristics were used. Based on previous maps and field samples, the classes are later assigned to rocks, snow cover, sand, water bodies, agriculture, and dense forests. The accuracy of classified results was determined using Google Earth Pro images. According to Monserud & Leemans (1992), the accuracy assessment was performed using the Kappa statistic, which yielded a value of 0.774, indicating very good agreement.
Two-dimensional hydromorphic simulation
Governing equations
A physics-based 2D morphodynamic model of the Teesta River was developed to evaluate the effects of flooding. The open-source Delft3D platform was used to run the numerical model (flow version 4.00.01.000000) (Lesser et al. 2004). Here, the fundamentals of the hydromorphology model are described briefly. The model uses Boussinesq approximations in the hydrodynamic section to solve two-dimensional depth-averaged shallow water equations (derived from Navier–Stokes' equations) for incompressible free surface flow.

































The schematization of the hydromorphic model.
The grid and bathymetry
Model calibration and boundary conditions
Model parameters
Parameter . | Unit . | Base condition . |
---|---|---|
Mean grain size, ![]() | μm | 200 |
Density of sediment, ![]() | kg/m3 | 2,650 |
Density of water, ![]() | kg/m3 | 1,000 |
Van Rijn's reference height factor | – | 2 |
Horizontal eddy viscosity | m2/s | 1 |
Hydrodynamic time step | s | 6 |
Roughness (Manning's) | sm−1/3 | 0.027 |
Morphological acceleration factor, m | – | 2 |
Threshold sediment thickness | m | 0.05 |
Drag coefficient of pier, ![]() | – | 1 |
Parameter . | Unit . | Base condition . |
---|---|---|
Mean grain size, ![]() | μm | 200 |
Density of sediment, ![]() | kg/m3 | 2,650 |
Density of water, ![]() | kg/m3 | 1,000 |
Van Rijn's reference height factor | – | 2 |
Horizontal eddy viscosity | m2/s | 1 |
Hydrodynamic time step | s | 6 |
Roughness (Manning's) | sm−1/3 | 0.027 |
Morphological acceleration factor, m | – | 2 |
Threshold sediment thickness | m | 0.05 |
Drag coefficient of pier, ![]() | – | 1 |
Spatial calibration of Wl for the flash flood happened on 9 September 2021.
Hydromorphic model validation
Evaluation statistics for discharge and water level
Parameter . | RRMSE . | KGE . | PBIAS . | NSE . |
---|---|---|---|---|
Discharge | 0.021 | 0.728 | −8.524 | 0.695 |
Water level | 0.041 | 0.939 | −5.061 | 0.617 |
Parameter . | RRMSE . | KGE . | PBIAS . | NSE . |
---|---|---|---|---|
Discharge | 0.021 | 0.728 | −8.524 | 0.695 |
Water level | 0.041 | 0.939 | −5.061 | 0.617 |
The validation of numerical simulation. (a) Comparison between the simulated and observed discharge. (b) Comparison between the simulated and observed water level.
The validation of numerical simulation. (a) Comparison between the simulated and observed discharge. (b) Comparison between the simulated and observed water level.
RESULTS
Floods of Teesta in Bangladesh
Changes in river water level extremes. (a) Daily water level at Dalia station for 60 years. (b) Daily water level percentiles frequency and intensity pattern.
Changes in river water level extremes. (a) Daily water level at Dalia station for 60 years. (b) Daily water level percentiles frequency and intensity pattern.
Number of extreme events per year and the magnitude of the corresponding highest flood.
Number of extreme events per year and the magnitude of the corresponding highest flood.
Flooding
Precipitation
The daily accumulated rainfall derived from the Global Precipitation Measurement (GPM) mission along the catchment (dates from 17 to 21 October 2021).
The daily accumulated rainfall derived from the Global Precipitation Measurement (GPM) mission along the catchment (dates from 17 to 21 October 2021).
Flooding depth and duration
Flooding velocity and sedimentation
The depth-average velocity distribution before and during the event. The top two figures show the depth-average velocity magnitude on 19 and 20 October 2021, respectively. Subfigures (a) and (b) show the velocity distributions near the Teesta barrage before and during the event.
The depth-average velocity distribution before and during the event. The top two figures show the depth-average velocity magnitude on 19 and 20 October 2021, respectively. Subfigures (a) and (b) show the velocity distributions near the Teesta barrage before and during the event.
Exposed population and crop damage
Spatial distribution of exposed populations and crop damage during October 2021 flood.
Spatial distribution of exposed populations and crop damage during October 2021 flood.
DISCUSSIONS
Flooding is already one of humanity's most serious threats, and it's becoming more common and affecting more people (UNDRR 2020). Flooding in flood-prone areas around the world is increasing in magnitude and changing in pattern as a result of global warming (Zhang et al. 2021). This study attempts to understand the potential effects of a severe flood on a transboundary river. The study of floods, especially flash floods, should contain the analysis from source to sink, but due to data unavailability, we concentrated on the flooding in Bangladesh. But despite such limitations, here we look into the extreme flood of Teesta with its morphological consequences and damages, which might be the first initiative so far to our knowledge.
The flood level time series analysis shown in Figures 8–10 reinforces that there is strong evidence that climatic variations are altering the Teesta flooding pattern in Bangladesh. As the driver of such change, the study done by Das et al. (2022) can be mentioned. They investigated the trend of climatic change drivers in the same basin, such as temperature, precipitation, potential evapotranspiration, and so on. Their analysis revealed that during the observed period (1971–2010), the maximum temperature in the central and lower parts of the basin increased significantly during the monsoon season. The minimum temperature, on the other hand, increased throughout the basin during the pre-monsoon, monsoon, and annual series, despite the fact that they found no significant trend in precipitation. However, using the Indian Meteorological Department (IMD) measured maximum rainfall in a single day data, Chaubey et al. (2023) showed that heavy to extreme rainfall in the basin is increasing. It is reflected in Figures 9 and 10 as well.
We posed the simple question, ‘Was the Teesta flooding in October 2021 evidence of an extreme climate-driven event or not?’ The flood level time series analysis supports this, and there is strong evidence that climatic variations are changing Teesta's flooding pattern in Bangladesh. There was some evidence of significant extreme flooding in the Indian portion of the catchment for the years 2003, 2000, 1996, 1993, 1978, 1976, 1975, 1973, 1968, and 1950. Moreover, using Coupled Model Intercomparison Project Phase 6 (CMIP6) projections, Das et al. (2022) showed that under 1.5 and 2 °C warming, an overall significant increasing trend is found in monsoon precipitation, which is similar to our observation. However, the observed data showed that the water level in Bangladesh during that time crossed the DL twice, in 1996 and 2015, by 0.2 and 0.108 m, respectively (Figure 10). Therefore, it would seem that there should be some sort of natural mechanism for absorbing shock. The river's width varies between 200 m in mountainous areas and 1,000 m in braided plains in the Indian portion (Pal et al. 2016). In contrast, in Bangladesh, the natural width ranges from 6,800 to 3,900 m (with some constricted width, i.e., 700 m near the bridge section). This large, braided plain could effectively reduce the effects of catastrophic flooding. Any development activities that interfere with this mechanism could magnify flooding. Non-intervention and maintenance of this large, braided plain can be used to reduce future risk, and this can be applied to the other most downstream areas of the flash flood-affected region in other countries.
Generally, in Bangladesh, chars are naturally resourceful and almost unintervened (Mondal et al. 2015). Therefore, increasing their ecosystem services should be a priority in future planning. In situations where the results of this study can be instructive, some non-structural risk management strategies are suggested as an alternative to structural ones. A combination of structural and non-structural measures should be used in delicate situations where the river reaches are highly prone to erosion, particularly along the banks.
Our next question was why the locals and authorities missed the early signs of flooding. Rainfall is classified differently in India and Bangladesh (as shown in Supplementary Table S1) (Mannan & Karmakar 2008; IMD 2021; BMD 2022). India sets a limit of 115.6–204.5 mm/day for ‘Very Heavy Rain’, whereas Bangladesh requires that it be greater than or equal to 89 mm/day. On 19 October 2021, when this torrential rainstorm hit the Indian portion of the basin, locals in Bangladesh received a moderately heavy to heavy rain warning (see Figure 9 and DDM 2021b). It could cause the local population and authorities to be misled. A basin-wide warning system is therefore required. Nevertheless, it is essential to remember that simply increasing weather forecast accuracy would not help vulnerable people fare any better. We suggest specializing the communication and response capabilities in the most vulnerable communities as part of climate change adaptation efforts.
According to flood depth, duration, velocity, and sedimentation pattern, the upstream portion of the river was more affected than the downstream portion, which is controversial for other rivers in the country, such as the Brahmaputra-Jamuna or the Arial Khan, as observed by Shampa et al. (2022) and Roy et al. (2021). This could be caused by the geophysical circumstances and the timing of the flooding. The downstream river was already in a recession phase because this flood happened at the end of the monsoon season. But in the future, it might not always be the case. As a result, preparation is required.
The forecasts or early warning information must be completely thorough to manage such an unpredictable event. Especially for flood events, we have never seen before, ‘impact-based’ warnings that translate warnings into impacts and scenarios may be the answer (Coughlan et al. 2022). In this instance, the disaster bulletin from DDM on 19 October 2021 included information about heavy precipitation in the Indian portion but failed to project and pursue the likely harm to the community. Second, we recommend creating appropriate plans and policies to respond to extreme events. Since there is currently no ‘warning number’ for flooding disasters in Bangladesh, it was time-consuming and laborious for locals and government officials to take prompt action. These plans ought to specify who is responsible for acting in response to a forecast. In addition to capacity, an effective response also needs a good backup plan, pre-arranged funding, and consensus on how to start the response.
CONCLUSIONS
Living in a time of climate change, we can observe changes in the intensity, frequency, and pattern of extreme flooding. Bangladesh's Teesta braided plain serves as an illustration of that. We can see that during such a severe event, the magnitude of the flood was 1.8 times greater than what people had ever experienced. Within 24 h, there was a 6.98-fold increase in flooding, making it extremely challenging for people to cope. The 223 sq. km of cropland experienced a nearly 2.5 times increase in velocity compared to a typical day, and a layer of sandy sediment 0.27 m thick covered the area. For several days, almost 136,000 people were stranded in the water. We found gaps in cross-country information exchange and inaccurate local perceptions of flooding, both of which increased the damage. We observed that the river has some built-in shock absorption features, such as a wide braided plain, which should not be changed. With better risk management techniques that vigorously focus on early actions, we feel that we may be able to manage the catastrophe more wisely than before. The forecasts or early warning information must be exhaustive. It is vitally important to develop appropriate plans, policies, and community education to respond to extreme events. In addition to capacity, effective response and pre-arranged funding may be good alternatives to reducing suffering.
ACKNOWLEDGEMENTS
The authors pay their gratitude to the Ministry of Science and Technology, Govt. of Bangladesh for funding this research.
DATA AVAILABILITY STATEMENT
Most of the relevant data are included in the paper or its Supplementary Information. For the other secondary data readers should contact the corresponding author for details.
CONFLICT OF INTEREST
The authors declare there is no conflict.