The Teesta basin is shared by Bangladesh and India, holds significant importance in the bilateral relationship, and sustains the livelihoods of over 30 million people in Bangladesh. Employing a cellular-automata model (CA), we accurately estimate LULC for the 2020s and projected for the 2050s and 2080s. A semi-distributed hydrological model, Soil Water Assessment Tool (SWAT), is used to generate flow for the base period (1995–2014), the near future (2035–2064), and the far future (2071–2100). SWAT model is forced by eight general circulation models (GCMs) under two socioeconomic pathways (SSP245 and SSP585). The CA-Markov prediction indicates LULC changes, especially increased agriculture and settlements by 76 and 42%, and decreased forest and water by 13 and 36%, respectively, which are expected by 2050s and will influence discharge patterns. This results in additional discharge increases of 4% (–8 to 5%) for SSP245 and 5% (–8 to 10%) for SSP585 scenarios during wet seasons. In the far future, monsoon flow will increase by 13% (0.4 to 23%) for SSP245 and 52% (–29 to 151%) under SSP245 and SSP585. A marginal change in winter flow was shown by –6% (–16 to 4%) reduction under SSP245 and –13% (–64 to 63%) under SSP585 reduction in the 2080s.

  • Land use and land cover (LULC) and climate change will affect the future flow of the Teesta basin.

  • The cellular automata-Markov model is used for LULC and Soil and Water Assessment Tool (SWAT) for hydrological modeling.

  • An increase in settlements and agricultural land, with reductions in forest and water, is predicted.

  • Monsoon flow will increase by 4 and 5% in the 2050s and by 13 and 52% in the 2080s under SSP245 and SSP585 scenarios.

  • The peak flow of the 100-year return period will increase by 33% in the 2080s under the SSP585 scenario.

Water constitutes a fundamental element crucial for sustaining all forms of livelihood and represents the predominant natural resource distributed across the globe. Regions reliant on agriculture within a transboundary river basin encounter a multitude of challenges. Teesta, a crucial transboundary river shared by Bangladesh and India, holds significant importance in the bilateral relationship and sustains the livelihoods of over 30 million people in Bangladesh through subsistence agriculture (Momen & Rahman 2023). The basin, in particular, grapples with a multifaceted hydrological crisis characterized by the occurrence of recurring floods during the monsoon season and, conversely, droughts or acute water scarcity in the winter months (Goyal & Goswami 2018; Das et al. 2022; Tamang 2023). Additionally, the lower part of the basin has experienced numerous floods, particularly devastating ones occurring in 1988, 1998, 2008, and more recently in 2021 (Shampa et al. 2023). This complex situation is further exacerbated by extensive anthropogenic activities and the impact of climate change, leading to a concerning increase in the frequency and intensity of hydrological events on a daily basis (Knoesen 2012; Ficklin et al. 2022; Tamang 2023). Land use and land cover (LULC) change within river basins results from a complex interplay of factors such as population growth, technological advancements, socioeconomic structures, and ecological conditions (Leta et al. 2021; Sun et al. 2022). Human development activities in these areas have profound effects on hydrological cycles, water balance, radiation budget, and biodiversity, impacting climate, biogeochemical cycles, energy flow, and livelihoods (Mollel et al. 2023).

In the past, many studies exemplified the significant influence of LULC changes in basin flow patterns (Gaur & Singh 2023; Mohseni et al. 2023) while exploring the impact of climate change in the flow across diverse basins (Mohammed et al. 2017, 2018; Jiménez-Navarro et al. 2021; Sun et al. 2022; Mohseni et al. 2023; Mollel et al. 2023; Ullah et al. 2023; Yalcin 2023). Recent research that combines climate change and LULC effects reveals a notable impact on river basin flow, predicting decreased flow during the dry period and increased flow during the wet season flow (Bal et al. 2021; Fields et al. 2021; Mohseni et al. 2023). Although any specific studies for the Teesta basin are lacking, research in the broader Indian sub-continent, such as that conducted by Mohseni et al. (2023), underscores the substantial impact of LULC and climate change on future flow, indicating a significant increase in peak flow.

This study encompasses answering the following three research questions: (i) What are the changes in LULC and runoff for the Teesta basin during the baseline period? (ii) How will the LULC be changed in the near future (2035–2064) and the far future (2071–2100)?, and (iii) What is the combined effect of LULC and climate change scenarios (SSP245 and SSP585) on runoff? These findings will enhance the understanding of potential flow pattern changes, aiding water resource management and adaptation strategies in the Teesta basin. To address these research questions, we have applied the cellular automata-Markov (CA-Markov) model for LULC modeling and the Soil and Water Assessment Tool (SWAT) for basin-scale flow modeling. Climate model output from the Coupled Model Intercomparison Project Phase 6 (CMIP6) has been used here to derive the SWAT model. This dataset is widely used for robust simulations in advancing our understanding of complex interactions within river basins (Aksoy & Kaptan 2022; Girma et al. 2022; Hossein et al. 2023; Nath et al. 2023).

The Teesta River basin, integral to the Brahmaputra basin, spans elevations ranging from 30 to 8,600 m above mean sea level (MSL) (Fields et al. 2021; Figure 1). Characterized by high mountains with perpetual snow cover, including glaciers like Talung and Zemu, two-thirds of the basin consist of such terrain (Fields et al. 2021). Originating in north Sikkim at around 5,280 m, the Teesta basin is fed by glaciers like Pahunri, Khangse, and ChhoLhamo Lake (Alam et al. 2021). Flowing through mountainous terrain and then the alluvial plains of North Bengal, India, it adopts a braided pattern before crossing into Bangladesh. The basin, with its intricate interplay of climate, soil, and vegetation, influences slope equilibrium, contributing to heightened extreme events during the monsoon. The southwest monsoon, predominant from May to October, brings substantial and evenly distributed precipitation, totaling about 2,534 mm annually in Sikkim (Fields et al. 2021). The basin's discharge varies significantly between dry and wet periods and depends on precipitation. At the Dalia Point, upstream of the Teesta barrage, water flow ranges from 10,000 to 280,000 cusecs (Khan & Ali 2019; Fields et al. 2021).
Figure 1

(a) Administrative map of Teesta basin and (b) spatial variables for LULC model training.

Figure 1

(a) Administrative map of Teesta basin and (b) spatial variables for LULC model training.

Close modal

The transboundary river basin spans 12,159 km2, with 10,155 km2 in India and 2,004 km2 in Bangladesh. Of the total, 4,108 km2 lie in the plains, with West Bengal covering 2,104 km2 and Bangladesh 2,004 km2 (Momen & Rahman 2023). Around 9,667 km2, encompassing 35 Upazilas and 5,427 villages, rely on the river for water. As per the 2011 census, approximately 9.15 million people depend on the Teesta River for freshwater.

Satellite and other geospatial data

Landsat satellite images from the 2000s, 2010s, and 2020s have been obtained from the official website of the US Geological Survey (USGS). The study areas are covered by Landsat Paths 138 and 139, and Rows 41 and 42 cover the study area. The collected satellite images are projected in the Universal Transverse Mercator (UTM) system within Zone 46 using the World Geodetic System (WGS84) Datum, and the pixel size is 30 m. The reference data were collected from Google Earth Pro and OpenStreetMap for calibration and validation. The future LULC controlling drivers' data – road network, rail network, and city center data – were collected from OpenStreetMap.

Hydrological data

The discharge data for Dalia Point, located upstream of the Teesta Barrage, was obtained from the Bangladesh Water Development Board (BWDB) for the years 1997 to 2013. The station-based water level data are required to analyze the trend and to calibrate and validate the model. The model calculates the discharge based on precipitation and snowmelt, and its outputs must be checked with a set of observed discharge data. The details of the collected data are shown in Table 1.

Table 1

Details of the data collection

Data typeSourceStation namePeriod timeRemarks
DEM Shuttle Radar Topography Mission (SRTM) –  DEM with the resolution of 90 m 
River network USGS Hydrosheds – –  
Land use Model generated data – 2000; 2010; 2020 30 m resolution 
Soil data FAO/UNESCO soil map of the world –  The soil types based on the FAO classification system were manually incorporated into the SWAT database 
Soil slope data Generate data from DEM using GIS tools    
Discharge data Bangladesh Water Development Board (BWDB) Dalia station 1997–2013 The discharge and water level data from Dalia Point were collected from the BWDB 
Data typeSourceStation namePeriod timeRemarks
DEM Shuttle Radar Topography Mission (SRTM) –  DEM with the resolution of 90 m 
River network USGS Hydrosheds – –  
Land use Model generated data – 2000; 2010; 2020 30 m resolution 
Soil data FAO/UNESCO soil map of the world –  The soil types based on the FAO classification system were manually incorporated into the SWAT database 
Soil slope data Generate data from DEM using GIS tools    
Discharge data Bangladesh Water Development Board (BWDB) Dalia station 1997–2013 The discharge and water level data from Dalia Point were collected from the BWDB 

The digital elevation model (DEM) was collected from the Shuttle Radar Topography Mission (SRTM) at 90 m × 90 m resolution. Besides topography, the DEM is used extensively to delineate and determine a given watershed's drainage pattern. The SWAT model is set up for a large-scale basin, with a size of 12,159 km2 using a DEM of 90-m resolution. This resolution allows for the effective delineation of most rivers within the basin. As the selected DEM has a resolution of 90 m, all other spatial data were also resampled to a 90-m resolution to maintain consistency across datasets. The soil map data were collected from the FAO–UNESCO soil map.

LULC data

LULC is the initial watershed behavior parameter in the SWAT model. Basin soil properties and characteristics are changed by LULC, such as forest, urban, or agricultural areas. There are several sources of land-use data, but this study used freely available satellite image analysis data and model-driven future LULC data.

Weather and climate data

The SWAT model necessitates diverse meteorological data, including maximum and minimum temperature, daily precipitation, relative humidity, solar radiation, and wind speed, for accurate hydrological simulations. Due to the unavailability of historical data for relative humidity, solar radiation, and wind speed in the base area, we use the WXGEN database to run the model (Sharpley & Williams 1990). In this study, rainfall and temperature data (2035–2065 for the 2050s and 2071–2100 for the 2080s) were obtained from freely available bias-corrected datasets by Mishra et al. (2020). They employed the empirical quantile mapping (EQM) technique to downscale daily temperatures and precipitation for South Asia and Indian sub-continental river basins. Bias correction was made using daily precipitation at 0.25°, which was obtained from the India Meteorological Department (IMD) for the Indian region. These gridded daily precipitation data for India, using station observations from more than 6,000 stations located across India, were developed by Pai et al. (2014). The datasets cover 13 CMIP6 general circulation models (GCMs) under five scenarios: historical, SSP126, SSP245, SSP370, and SSP585, with a spatial resolution of 0.25°. Data for different periods were accessed from the Zenodo Sarver platform (https://zenodo.org/record/3874046#.YWvmdRpBxnJ). The CMIP6 scenarios correspond to various shared socioeconomic pathways (SSPs) with target radioactive forcing levels for the 21st century. Eight out of the 13 models were used in this study, which are better performing and widely used (Table 2; Jiménez-Navarro et al. 2021; Sun et al. 2022).

Table 2

Climate model data used in this study

S. No.Model nameInstituteResolution (degree)
ACCESS-CM2 Commonwealth Scientific and Industrial Research Organization (CSIRO) and Bureau of Meteorology (BOM), Australia Australian Community Climate and Earth System Simulator 0.25° × 0.25° 
BCC-CSM2-MR Beijing Climate Center (BCC) 0.25° × 0.25° 
CanESM5 Canadian Centre for Climate Modelling and Analysis 0.25° × 0.25° 
EC-Earth3 European Earth System Model 0.25° × 0.25° 
INM-CM5-0 Institute for Numerical Mathematics 0.25° × 0.25° 
MPI-ESM1-2-HR Max Planck Institute for Meteorology 0.25° × 0.25° 
MRI-ESM2-0 Meteorological Research Institute, Tsukuba, Japan 0.25° × 0.25° 
NorESM2-MM Norwegian Climate Centre 0.25° × 0.25° 
S. No.Model nameInstituteResolution (degree)
ACCESS-CM2 Commonwealth Scientific and Industrial Research Organization (CSIRO) and Bureau of Meteorology (BOM), Australia Australian Community Climate and Earth System Simulator 0.25° × 0.25° 
BCC-CSM2-MR Beijing Climate Center (BCC) 0.25° × 0.25° 
CanESM5 Canadian Centre for Climate Modelling and Analysis 0.25° × 0.25° 
EC-Earth3 European Earth System Model 0.25° × 0.25° 
INM-CM5-0 Institute for Numerical Mathematics 0.25° × 0.25° 
MPI-ESM1-2-HR Max Planck Institute for Meteorology 0.25° × 0.25° 
MRI-ESM2-0 Meteorological Research Institute, Tsukuba, Japan 0.25° × 0.25° 
NorESM2-MM Norwegian Climate Centre 0.25° × 0.25° 

Markov chain model and CA-Markov

The CA model is a discrete, spatially extended dynamic system that operates based on predefined transition rules, linking the new state of the LULC type to its previous state (Bal et al. 2021; Aksoy & Kaptan 2022; Mensah et al. 2022). CA-based models offer the advantage of capturing complex and nonlinear spatially distributed processes, which are capable of assessing LULC change patterns on a regional to local scale. However, cells, transition rules, cell size, time, and cell neighborhood are the significant components that should be considered for the optimum prediction (Liping et al. 2018). The spatial and temporal states of neighboring cells heavily rely on each cell's state (Mathanraj et al. 2021). The CA model can be mathematically expressed (Liping et al. 2018) as follows:
(1)
where S represents the set of states of the cells, t and t + 1 represent different time steps, N represents the number of neighborhood cells, and f represents the transformation rule of local space.
The Markov chain model (MCM) is a probabilistic stochastic automation model that characterizes the transition probabilities of LULC types shifting from one state (St) to another state (St+1) over a defined period (Liping et al. 2018; Mathanraj et al. 2021). The transition probability distribution of the LULC changes between states is typically represented as an MC transition matrix. The transition matrix is an n × n matrix, where n represents the number of possible states, and each entry denotes the probability of transitioning from the state St to the state St+1. Mathematically, the MCM for predicting LULC changes can be expressed using the conditional probability equation as follows (e.g., Subedi et al. 2013):
(2)
(3)
where St represents the LULC status at time t and St+1 represents the LULC status at time t + 1. Pij denotes the transition probability from the LULC type i (at time t) to the LULC type j (at time t + 1), and n is the number of LULC types.

The CA-Markov model, integrating CA and MC approaches, predicts spatiotemporal LULC changes. While it may have limitations in capturing the impact of individual cells on the entire space, CA-Markov excels in accurately forecasting transitions between LULC states based on prior conditions (Hossein et al. 2023; Nath et al. 2023). This integrated model offers advantages over traditional methods, providing more reasonable and accurate LULC transition predictions (Long et al. 2021; Matlhodi et al. 2021).

A two-step validation process was employed to ensure prediction accuracy. Initially, using LULC data from 2000 to 2010 as inputs, the model predicted the LULC map for 2020 and compared it to satellite-derived data. Kappa indices assessed prediction accuracy, with values exceeding 80% indicating excellent performance (Matlhodi et al. 2021). Following validation, the IDRISI software incorporated land-use maps for 2000 and 2010 to create a transition matrix. This matrix was then used to project land-use patterns for 2020, 2050, and 2080.

Soil and Water Assessment Tool

Several hydrological models are available, among them MIKE SHE and SWAT, which are widely used. MIKE SHE requires extensive physical parameters and simulates surface and groundwater movement. In contrast, the freely available, open-source SWAT model, developed by the USDA, is a semi-distributed physical model that assesses rainfall-runoff and flow prediction. Moreover, the SWAT model is widely used in the ungauged basin because of its flexibility of operation in the basin where there are data constraints (Arnold et al. 1998; Qi et al. 2020; Gwal et al. 2023). It is known for its computational efficiency, auto-calibration, and uncertainty analysis techniques, which facilitate high-quality water resource management.

The SWAT, a semi-distributed hydrological model developed in the 1990s, operates at watershed-to-river basin scales, simulating daily surface and groundwater dynamics. Utilizing physiographical data and meteorological information, it predicts environmental impacts from land use, management, and climate change. The model integrates key hydrological processes, such as evapotranspiration, runoff, infiltration, and percolation. Sub-basins, segmented into hydrological response units (HRUs) based on uniformity in land cover, soil, and climate, enable detailed analysis (Arnold & Allen 1996). SWAT's adaptability extends to GIS platforms like MapWindows (MWSWAT) and QGIS (QSWAT), facilitating efficient modeling and assessment.

The hydrological cycle, as simulated by the SWAT, is governed by the water balance equation as follows:
(4)
where represents the final soil water content, represents the initial soil water on day I, t represents the time in days, represents the amount of water entering the vadose zone from the soil profile on day I, represents the amount of precipitation on day I, represents the total surface runoff on day I, represents the amount of evapotranspiration on day I, and Qgw represents the amount of flow on day I (all units, except time, are presented as millimeters of water).

The weather generator, WGNMaker 4.1 (http://swat.tamu.edu/), determined parameters using data from 13 evenly distributed weather stations across the basin. Daily maximum and minimum temperature and precipitation were used as input. The rest of the weather parameters (solar radiation, humidity, and wind speed) have been taken from the model base value from the WXGEN model. The model uses the Hargreaves method to estimate evapotranspiration (ET) (López-Ramírez et al. 2021). The SWAT model integrated land-use scenarios for the 2020s, 2050s, and 2080s. The model was calibrated and validated using observed discharge in Dalia Point in Bangladesh. The model evaluation employed R2 and Nash–Sutcliffe efficiency (NSE) coefficients. Simulations were conducted with outputs from eight GCMs under SSP245 and SSP585 scenarios, considering 2020s LULC and projected LULC data for the 2050s and 2080s. These simulations assessed the impacts of LULC change on runoff dynamics.

Accuracy assessment

The LULC layers for 2000, 2010, and 2020 were classified using the maximum likelihood method and refined through auxiliary data editing. Evaluation metrics, including producer and user accuracies, total accuracy, and kappa values, were employed, with results detailed in Table 3. User accuracy exceeded 80% for most LULC classes in 2000, 2010, and 2020, except for water/river due to the braided Teesta River characteristics. Overall classification accuracy was 87, 85, and 85% for 2000, 2010, and 2020, respectively. The kappa values, which measure the agreement between the classified map and the reference data, were calculated as 0.85, 0.84, and 0.82 for 2000, 2010, and 2020, respectively (Table 4).

Table 3

Accuracy assessment of LULC mapping

LULC classesLULC 2000
LULC 2010
LULC 2020
Producer accuracyUser accuracyProducer accuracyUser accuracyProducer accuracyUser accuracy
Agriculture 82% 90% 79% 93% 80% 92% 
Settlements 94% 84% 95% 87% 94% 84% 
Forest 88% 90% 91% 88% 87% 84% 
Water/river 94% 78% 91% 72% 91% 72% 
Sandbar 88% 85% 82% 90% 82% 90% 
Snow cover 77% 100% 82% 100% 77% 100% 
Barren land/exposed rock 85% 88% 90% 80% 85% 71% 
Overall accuracy 87% 87% 85% 
Kappa 0.85 0.84 0.82 
LULC classesLULC 2000
LULC 2010
LULC 2020
Producer accuracyUser accuracyProducer accuracyUser accuracyProducer accuracyUser accuracy
Agriculture 82% 90% 79% 93% 80% 92% 
Settlements 94% 84% 95% 87% 94% 84% 
Forest 88% 90% 91% 88% 87% 84% 
Water/river 94% 78% 91% 72% 91% 72% 
Sandbar 88% 85% 82% 90% 82% 90% 
Snow cover 77% 100% 82% 100% 77% 100% 
Barren land/exposed rock 85% 88% 90% 80% 85% 71% 
Overall accuracy 87% 87% 85% 
Kappa 0.85 0.84 0.82 
Table 4

Predicted LULC in the 2020s, 2050s, and 2080s

LULC2000s2010s2020s2050s
2080s
Area (km2)Area (km2)Area (km2)Area (km2)Area change (km2)% of changeArea (km2)Area change (km2)%of change
Agriculture 658.97 662.2 680.55 1,199.13 518.58 76.2% 1,510.40 829.85 121.9% 
Settlements 288.92 293.75 318.97 455.91 136.94 42.9% 1,254.29 935.32 293.2% 
Forest (mixed) 5,272.21 5,195.99 5,125.47 4,417.33 −708.14 −13.8% 3,421.71 −1,703.76 −33.2% 
Water 64.87 67.62 72.83 46.02 −26.81 −36.8% 46.02 −27.82 −38.2% 
Sandbar 224.83 212.48 210.83 213.92 3.09 1.5% 141.98 −67.84 −32.2% 
Exposed rock 1,844.47 1,928.87 1,925.42 2,048.62 123.20 6.4% 1,966.68 41.25 2.1% 
Snow cover 883.66 877.02 903.86 857.01 −46.85 −5.2% 896.86 −7.00 −0.8% 
LULC2000s2010s2020s2050s
2080s
Area (km2)Area (km2)Area (km2)Area (km2)Area change (km2)% of changeArea (km2)Area change (km2)%of change
Agriculture 658.97 662.2 680.55 1,199.13 518.58 76.2% 1,510.40 829.85 121.9% 
Settlements 288.92 293.75 318.97 455.91 136.94 42.9% 1,254.29 935.32 293.2% 
Forest (mixed) 5,272.21 5,195.99 5,125.47 4,417.33 −708.14 −13.8% 3,421.71 −1,703.76 −33.2% 
Water 64.87 67.62 72.83 46.02 −26.81 −36.8% 46.02 −27.82 −38.2% 
Sandbar 224.83 212.48 210.83 213.92 3.09 1.5% 141.98 −67.84 −32.2% 
Exposed rock 1,844.47 1,928.87 1,925.42 2,048.62 123.20 6.4% 1,966.68 41.25 2.1% 
Snow cover 883.66 877.02 903.86 857.01 −46.85 −5.2% 896.86 −7.00 −0.8% 

Comparison of LULC changes

The prediction of future land use is influenced by two fundamental factors: transition probability and transition area. The transition probability matrix shows the likelihood of transitioning to other classes for each LULC class within the given time frame. The matrix was derived by multiplying the likelihood matrix with the corresponding areas of the LULC classes during the designated period. The transition area and the transitional matrix were developed for 2020 using LULC layers from 2000 and 2010, and the predicted LULC of 2020 was then validated with an analyzed 2020 LULC image to ensure accuracy >80%.

Comparison of LULC scenarios in 2020, 2050, and 2080

The predicted LULC of 2020 was obtained from the LULC of 2000 and 2010 using the CA-Markov model. To assess the accuracy and reliability of the predictions, the transition areas, transition potential, and compatible and non-compatible areas were determined by comparing the predicted LULC with the reference layers of the actual LULC in 2020. The CA-Markov model exhibited an accuracy of 86.21% when considering the composition of the transition potential. The accuracy level indicates that the model can provide reliable projections for future land use dynamics (Aksoy & Kaptan 2022). Based on the satisfactory accuracy (Kappa > 80%), the CA-Markov model has been run for the years 2050s and 2080s.

Substantial changes in LULC are found over the study period. Agricultural expansion is prominent, increasing from 680.55 km2 in 2020 to 1,199.13 km2 in 2050, and further to 1,510.40 km2 in 2080, representing growth rates of 76.2 and 121.9%, respectively. Settlement areas are expected to surge by 42.9% (136.94 km2) in 2050 and 293.2% (935.32 km2) in 2080. Concurrently, forest areas will decrease to 13.8% (708.14 km2) and 33.2% (1703.76 km2) in 2050 and 2080. The substantial sandbar is expected to decrease in the future, potentially due to factors such as increased river flow, erosion, river management and infrastructure, changes in sediment load, encroachment, and land-use changes (Haque et al. 2023; Hasanuzzaman et al. 2023). Water bodies are projected to decrease, indicating a reduced wet perimeter due to siltation. The LULC map of 2010 serves as a baseline for estimating scenarios in the 2020s, 2050s, and 2080s (Figure 2), with Table 4 detailing area statistics for each sub-watershed, highlighting notable changes in agriculture, settlements, and forests. Minor shifts are expected in water, bare land, snow cover, and exposed rock categories, resembling patterns observed in 2000 and 2010.
Figure 2

Predicted LULC maps of 2020s, 2050s, and 2080s.

Figure 2

Predicted LULC maps of 2020s, 2050s, and 2080s.

Close modal

SWAT model calibration and validation

Conducting sensitivity analysis for hydrological model calibration is vital for accurately representing field, subsurface, and channel conditions. This study employs the Latin hypercube one-factor-at-a-time (LH-OAT) method proposed by Van Griensven & Meixner (2003) to rank prediction parameters for each sub-basin based on sensitivity analysis using the SWAT-CUP tool. Twenty parameters have been assessed for the calibrations of the model, among which nine parameters are crucial for flow conditions, including soil properties, surface runoff, groundwater, and evaporation (Table 5), which were assessed, with P-values below 0.05 considered significant for basin flow calibration (Thavhana et al. 2018). The Sequential Uncertainty Fitting II (SUFI-2) algorithm in SWAT-CUP was used for model calibration, employing nine sensitive parameters. Considering the availability of the observer discharge, model calibration and validation were performed on distinct time segments (1997–2005 and 2006–2013) to ensure optimal model performance. The NSE served as the objective function, with a minimum threshold set at 0.5. Monthly averaged data from 1997 to 2013 were utilized for calibration and validation, demonstrating a significant agreement between simulated and observed flow, affirming the model's accuracy in capturing hydrological system dynamics (Figure 3).
Table 5

List of most sensitive parameters according to their P-value

S.L.Parameter nameP-value
SLSUBBSN.hru 0.514774 
OV_N.hru 0.355168 
SOL_AWC(..).sol 0.347581 
ESCO.hru 0.347171 
HRU_SLP.hru 0.345003 
CN2.mgt 0.343957 
ALPHA_BF.gw 0.343875 
GWQMN.gw 0.336482 
GW_DELAY.gw 0.253192 
10 GW_REVAP.gw 0.004999 
11 REVAMPM.gw 0.004875 
12 SHALLST_N.gw 0.003982 
13 GW_SPYLD 0.002954 
14 GDRAIN 0.001872 
15 SOL_ROCK 0.000980 
S.L.Parameter nameP-value
SLSUBBSN.hru 0.514774 
OV_N.hru 0.355168 
SOL_AWC(..).sol 0.347581 
ESCO.hru 0.347171 
HRU_SLP.hru 0.345003 
CN2.mgt 0.343957 
ALPHA_BF.gw 0.343875 
GWQMN.gw 0.336482 
GW_DELAY.gw 0.253192 
10 GW_REVAP.gw 0.004999 
11 REVAMPM.gw 0.004875 
12 SHALLST_N.gw 0.003982 
13 GW_SPYLD 0.002954 
14 GDRAIN 0.001872 
15 SOL_ROCK 0.000980 
Table 6

Evaluation of the performance of the SWAT model

ParametersCalibrationValidationPerformance (Moriasi et al. 2015)
R2 0.85 0.84 Very good 
NSE 0.81 0.81 Very good 
PBIAS 18.0 16.9 Satisfactory 
ParametersCalibrationValidationPerformance (Moriasi et al. 2015)
R2 0.85 0.84 Very good 
NSE 0.81 0.81 Very good 
PBIAS 18.0 16.9 Satisfactory 
Figure 3

Observed and simulated monthly flow at the Dalia Point of Teesta basin: (a) calibration (1997–2005) and (b) validation (2006–2013).

Figure 3

Observed and simulated monthly flow at the Dalia Point of Teesta basin: (a) calibration (1997–2005) and (b) validation (2006–2013).

Close modal

Performance of the SWAT model

Upon the completion of the calibration and validation process, the model output is depicted in Figure 3. This process involved conducting 400 simulations across four iterations. Notably, the NSE value of 0.81 obtained during the calibration and validation periods indicates a strong agreement between the observed and simulated discharges. Furthermore, the coefficient of determination (R2) values of 0.85 for calibration and 0.84 for validation signify a high level of goodness of fit between the observed and simulated data as shown in Table 6. The high R2 values provide evidence of the model's reliability and robust performance. The PBIAS values of 18.0% for calibration and 16.9% for validation indicate a slight overestimation of the simulated discharges compared with the observed values. However, these values fall within an acceptable range, suggesting that the model adequately captures the overall discharge behavior. Based on these statistical measures, it can be concluded that the SWAT model performed exceptionally well in both the calibration and validation stages, utilizing historical data from the Dalia Point Teesta River basin. The model demonstrated a strong ability to reproduce the observed discharge patterns and capture the hydrological dynamics of the basin.

Climate change impact on flow

Climate change impact on flow in the near future (2050s)

The Teesta River's future flow at the Dalia station in Nilphamari for the near future was simulated to capture the uncertainties stemming from varying climate scenarios. Figure 4 depicts monthly mean flow simulations for SSP245 and SSP585 scenarios, considering eight GCMs. Most models accurately track dry period flows (December to May), with ACCESS-CM2, EC-Earth3, and MRI-ESM2-0 showing similar peak flows to the base period. Conversely, BCC-CSM2-MR, MPI-ESM1-2-HR, and NorESM-MM predict lower peak flows for both scenarios (Figure 4).
Figure 4

Mean monthly discharge (m³/s) at the Dalia station for the Teesta River basin in the near future (2050s) and the far future (2080s).

Figure 4

Mean monthly discharge (m³/s) at the Dalia station for the Teesta River basin in the near future (2050s) and the far future (2080s).

Close modal

Diverse projections for future flow characteristics emerge from the considered GCMs. ACCESS-CM2 and MPI-ESM1-2-HR suggest fewer flood peaks, with flow dynamics more pronounced in the worst scenario (SSP585) in the near future period. While the simulation anticipates higher peak flow at Dalia Point for SSP585, the dry period between SSP245 and SSP585 scenarios exhibits no significant changes. Among the GCMs, EC-Earth3 predicts a substantial increase in flow exceeding 100% during the pre-monsoon period (MAM) for both scenarios, whereas ACCESS-CM2, BCC-CSM2-MR, and CanESM5 models show relatively unchanged discharges.

The seasonal flow pattern shown in Figure 5 for SSP245 and SSP585 scenarios across eight GCMs indicates a projected decrease in winter flow and an increase in monsoon flow compared with the base period (1995–2014). Most models, excluding CanESM5, predict reduced winter flow for both scenarios. However, in the monsoon, all models anticipate increased flow for SSP585, while, except for BCC-CSM2-MR, other models project increased flow for SSP245 in the future. The study foresees potential water scarcity for farmers in winter and heightened flood risks in the near future due to climate change.
Figure 5

Changes in seasonal flow (a) SSP245 scenarios for the near future, (b) SSP585 scenarios for the near future, (c) SSP245 scenarios for the far future, and (d) SSP585 scenarios for the far future.

Figure 5

Changes in seasonal flow (a) SSP245 scenarios for the near future, (b) SSP585 scenarios for the near future, (c) SSP245 scenarios for the far future, and (d) SSP585 scenarios for the far future.

Close modal

The Teesta River's future flow patterns, as projected by various models, exhibit disparities. While some models anticipate a decrease post-monsoon, the multimodel average flow shows an overall reduction in flow for the near future. CanESM5 predicts a significant increase, contrasting with INM-CM5-0 and MPI-ESM1-2-HR, which suggested a notable reduction. Pre-monsoon predictions vary, but the multimodel average indicates an overall increase. Comparing the baseline (1995–2014) to the near future, most models project drier winters and wetter wet seasons, with BCC-CSM2-MR and CanESM5 showing similar changes. SSP245 indicates a higher winter flow increase than SSP585, while in the pre-monsoon, SSP585 exhibits a higher rate of increase than SSP245.

Table 7 presents the multimodel mean of seasonal flow changes from the baseline to the near future, indicating wetter pre-monsoon and monsoon seasons and a drier winter. In the near future, about −46% (−72 to 21%) and −49% (−84 to 27%) flow reductions are projected under SSP245 and SSP585 scenarios in winter. In contrast, pre-monsoon flow is expected to increase by up to 15% (−68 to 194%) for SSP245 and 13% (−73 to 199%) for SSP585 scenarios. Monsoon flow is projected to rise by 26% (−40 to 120%) for SSP245 and 38% (−36 to 151%) for SSP585 scenarios. On the contrary, the post-monsoon flow will decrease by −9% (−50 to 69%) for SSP245 while increasing by 5% (−38 to 65%) for SSP585 scenarios.

Table 7

Seasonal changes of flow at Dalia Point of Teesta River for the near future and far future

SeasonNear future (2050s)
Far future (2080s)
SSP245SSP585SSP245SSP585
Winter (DJF) −46% (−72 to 21%) −49% (−84 to 27%) −40% (−84 to 51%) −38% (−86 to 48%) 
Pre-monsoon (MAM) 15% (−68 to 194%) 13% (−73 to 199%) 16% (−68 to 200%) 40% (−65 to 235%) 
Monsoon (JJAS) 26% (−40 to 120%) 38% (−36 to 151%) 30% (−42 to 144%) 45% (−30 to 167%) 
Post-monsoon (ON) −9% (−50 to 69%) 5% (−38 to 65%) 1% (−45 to 60%) 22% (−41 to 100%) 
SeasonNear future (2050s)
Far future (2080s)
SSP245SSP585SSP245SSP585
Winter (DJF) −46% (−72 to 21%) −49% (−84 to 27%) −40% (−84 to 51%) −38% (−86 to 48%) 
Pre-monsoon (MAM) 15% (−68 to 194%) 13% (−73 to 199%) 16% (−68 to 200%) 40% (−65 to 235%) 
Monsoon (JJAS) 26% (−40 to 120%) 38% (−36 to 151%) 30% (−42 to 144%) 45% (−30 to 167%) 
Post-monsoon (ON) −9% (−50 to 69%) 5% (−38 to 65%) 1% (−45 to 60%) 22% (−41 to 100%) 
Table 8

Seasonal flow changes considering both climate and LULC changes

SeasonNear future
Far future
SSP245SSP585SSSP245SSP585
Winter (DJF) 0.3% (−1 to 3%) 0.8% (−1 to 5%) −6% (−16 to 4%) −13% (−64 to 63%) 
Pre-monsoon (MAM) 3% (0.4 to 8%) 4% (0.8 to 7%) −1% (−4 to 4%) 24% (−36 to 116%) 
Monsoon (JJAS) 4% (−8 to 5%) 5% (−8 to 10%) 13% (0.4 to 23%) 52% (−29 to 151%) 
Post-monsoon (ON) −0.04% (−5 to 5%) −0.3% (−4 to 7%) −0.5% (−18 to 28%) 1% (−72 to 99%) 
SeasonNear future
Far future
SSP245SSP585SSSP245SSP585
Winter (DJF) 0.3% (−1 to 3%) 0.8% (−1 to 5%) −6% (−16 to 4%) −13% (−64 to 63%) 
Pre-monsoon (MAM) 3% (0.4 to 8%) 4% (0.8 to 7%) −1% (−4 to 4%) 24% (−36 to 116%) 
Monsoon (JJAS) 4% (−8 to 5%) 5% (−8 to 10%) 13% (0.4 to 23%) 52% (−29 to 151%) 
Post-monsoon (ON) −0.04% (−5 to 5%) −0.3% (−4 to 7%) −0.5% (−18 to 28%) 1% (−72 to 99%) 
Table 9

Return period of 2050s flow and 2080s flow of average

Return period% of changes 2050s
% of changes 2080s
SSP245SSP585SSP245SSP585
14% (−9 to 30%) 24% (−3 to 56%) 24% (−9 to 62%) 43% (2 to 100%) 
10 5% (−16 to 19%) 14% (−11 to 43%) 14% (−17 to 49%) 32% (−6 to 85%) 
20 5% (−17 to 20%) 13% (−12 to 42%) 13% (−18 to 47%) 32% (−7 to 84%) 
50 6% (−16 to 23%) 14% (−12 to 43%) 13% (−18 to 48%) 33% (−6 to 86%) 
100 6% (−17 to 24%) 13% (−12 to 42%) 12% (−19 to 47%) 33% (−7 to 86%) 
Return period% of changes 2050s
% of changes 2080s
SSP245SSP585SSP245SSP585
14% (−9 to 30%) 24% (−3 to 56%) 24% (−9 to 62%) 43% (2 to 100%) 
10 5% (−16 to 19%) 14% (−11 to 43%) 14% (−17 to 49%) 32% (−6 to 85%) 
20 5% (−17 to 20%) 13% (−12 to 42%) 13% (−18 to 47%) 32% (−7 to 84%) 
50 6% (−16 to 23%) 14% (−12 to 43%) 13% (−18 to 48%) 33% (−6 to 86%) 
100 6% (−17 to 24%) 13% (−12 to 42%) 12% (−19 to 47%) 33% (−7 to 86%) 

Climate change impact on flow in the far future (2080s)

In the far future, the average monthly flow projections for the Teesta basin under SSP245 and SSP585 scenarios are shown in Figure 4. The model outputs exhibit diverse trends, encompassing decreasing, increasing, and flat patterns with random fluctuations for both scenarios. BCC-CSM2-MR, MPI-ESM1-2-HR, MRI-ESM2-0, and NorESM-MM models indicate an increase of peak flow in the far future, surpassing the base period, while ACCESS-CM2, CanESM5, and EC-Earth3 models indicate higher peak flow for both base periods and in the far future.

The SSP585 scenario, associated with fossil-fuel-based development and high radioactive energy emissions, shows a more pronounced increase in peak flow compared with the SSP245 scenario. Except for BCC-CSM2-MR, most of the models project higher peak flow than the SSP245 scenario. Notable variations in peak flow, including rightward shifts (higher peaks) in NorESM-MM and leftward shifts (lower peaks) in INM-CM5-0, are observed.

Changes in the seasonal flow pattern are evident, with most models depicting a consistent flow pattern during the dry period, particularly from December to March. The model outputs indicate future flow increases in pre-monsoon and monsoon periods and a decrease in winter. Figure 5(a) illustrates projected changes in mean monthly flow for the pre-monsoon (MAM) and monsoon (JJAS) periods in the near future, while Figure 5(b) shows corresponding changes in the post-monsoon (ONDJ) period for the far future. Among the eight GCMs, INM-CM5-0 predicts a flow increase exceeding 100% during the pre-monsoon period (MAM) in the near future for both scenarios, while ACCESS-CM2, BCC-CSM2-MR, and CanESM5 models indicate relatively unchanged discharges for both scenarios.

The monsoon period (JJAS) shows that the flow will decrease from the base period, while BCC-CSM2-MR CanESM5, INM-CM5-0, MRI-ESM2-0, MPI-ESM1-2-HR, and NorESM-MM models indicate that flow will be significantly increasing for both scenarios, The EC-Earth3, MPI-ESM1-2-HR, and NorESM-MM models show significantly decreasing flow for the SSP585 scenario in the far future. In the post-monsoon season, mixed results have been found; some models predict that flow will decrease, and others indicate an increasing pattern. While the CanESM5 model predicts that flow will increase significantly in the far future for both scenarios, the flow will substantially decrease based on the INM-CM5-0 and MPI-ESM1-2-HR models. However, average data from multiple models show that the average flow will decline in the near future in the post-monsoon period.

Most models project a drier dry season and a wetter wet season. BCC-CSM2-MR and CanESM5 exhibit similar change patterns. The increased winter flow is more pronounced in the SSP245 scenario than SSP585, while, in the pre-monsoon season, SSP585 displays a higher rate of increase than SSP245.

Climate change's impact on the water balance in the Teesta basin is severe, as indicated in Table 7, showcasing multimodel seasonal variability for both scenarios in the far future compared with the baseline period (1997–2013). Winter flow is projected to reduce by −40% (−84 to 51%) for SSP245 and −38% (−86 to 48%) for SSP585. Contrastingly, pre-monsoon flows are expected to increase by 16% (−68 to 200%) and 40% (−65 to 235%), respectively, for SSP245 and SSP585. Similarly, monsoon flow will be increased by 30% (−42 to 144%) and 45% (−30 to 167%) for SSP245 and SSP585. However, post-monsoon flow is anticipated to increase by 1% (−45 to 60%) in the near future but will increase by 22% (−41 to 100%) in 2080.

Climate and land-use change impact on the flow

Runoff considering both climate and LULC change

Under developmental pressures, the Teesta basin is undergoing rapid LULC changes, dominating by agriculture, settlements, and forests. Simulations predict significant future transformations, projecting decreased agriculture and forest cover and increased settlements. Modeled Teesta River flow at the Dalia station, calibrated under predicted LULC and climate change scenarios, indicates substantial impacts in the near and far future. Anticipated changes lead to increased additional flow during the wet season and either reduced or unchanged flow in the dry period. The simulation suggests a 10% mean annual flow change under SSP245 and SSP585 scenarios across eight models, with BCC-CSM2-MR, CanESM5, and EC-Earth3 showing the most significant influence on flow patterns (Figure 5). LULC changes, especially an increase in agriculture and settlements by 76 and 42%, respectively, and a decrease in forest and water by 13 and 36%, are expected by the 2050s, influencing discharge patterns. This results in additional discharge increases of 4% (−8 to 5%) for SSP245 and 5% (−8 to 10%) for SSP585 scenarios during wet seasons, with marginal changes in winter and post-monsoon periods. Emphasizing afforestation, reforestation, and sustainable development, these measures could mitigate future flood risks in the lower Teesta basin.

Results using multimodal ensemble climate projection showed a decrease of −6% (−16 to 4%) in winter season flow for SSP245 and a −13% (−64 to 63%) decrease for SSP585 in the 2080s. Conversely, the monsoon will witness an increase in flow, and post-monsoon seasons will witness increased flow under SSP245. Far-future predictions indicate more significant discharge changes than in 2050, predominantly impacting dry to pre-monsoon periods.

Anticipated LULC shifts by the 2080s, including a 121.9% increase in agriculture and a 293.2% surge in settlements, in contrast with declines of −33.2 and −38.2% in forest and water areas, respectively. As a result, flow is projected to rise by 13% (0.4 to 23%) for SSP245 and 52% (−29 to 151%) for SSP585 in the 2080s. Monsoon seasons are expected to witness higher flows, while winters might experience a −6% (−16 to 4%) reduction under SSP245 and a −13% (−64 to 63%) reduction under SSP585 in the 2080s as shown in Table 8.

Impact of LULC change on hydrological processes

Changes in LULC can significantly impact hydrological processes, particularly ET, infiltration, and canopy interception. Using the LULC map in the 2020s as a baseline and the simulated LULC maps of 2050s and 2080s, we evaluated absolute changes in mean annual ET, water percolation (PERC), surface runoff (SURQ), and lateral flow contribution (LATQ) during the near future. Precipitation (PREC) patterns showed a varying impact, particularly in the monsoon period, reflecting afforestation or deforestation trends. Notably, the prominence of deforestation was observed, as indicated by rising PREC rates. Additionally, runoff components showed a 2 mm decrease in SURQ during the monsoon, countered by a 7 mm increase in LATQ due to LULC changes. Temporal variation in potential evapotranspiration (PET) revealed June's highest change (0.4 mm) and December's lowest (−0.1 mm). Significantly higher changes were observed in SSP585 scenarios compared with SSP245 scenarios.

This study encompasses eight GCMs and highlights considerable variability in hydrological parameters due to complex interactions between location-specific factors. Catchment slope, soil permeability, vegetation characteristics, and settlements contribute to unique watershed responses. Soil depth emerges as a critical factor shaping how watershed discharge responds to changes in LULC. The simulations at Dalia Point in the Teesta River reveal a significant decrease in discharge under evolving LULC scenarios, emphasizing the far-reaching impact of such changes.

Projected LULC changes in the Teesta basin are poised to increase winter runoff but decrease monsoon runoff, resulting in a slightly elevated mean annual flow. LULC alterations significantly influence surface runoff (SURQ), lateral flow (LATQ), PET, and ET processes, as depicted in Figure 6. The study underscores the direct impact of LULC changes on ET and infiltration, affecting the water balance of basin aquifers. Utilizing the 2000 LULC as a baseline and the 2080 LULC as the future scenario, the assessment reveals noteworthy alterations in mean annual ET, water percolation, surface runoff, and lateral flow for both SSP245 and SSP585 scenarios. The intricate variations in hydrological components at the sub-basin level underscore the profound influence of LULC changes on the far-future flow regime, particularly under the SSP585 scenario.
Figure 6

LULC change impact on hydrological parameters of Teesta basin in (a) the near future (2050s) and (b) the far future (2080s).

Figure 6

LULC change impact on hydrological parameters of Teesta basin in (a) the near future (2050s) and (b) the far future (2080s).

Close modal

Changes in flood frequency

Extreme floods are a common scenario in the Teesta floodplain area of Bangladesh. To understand the future peak flow scenario of the basin, we use the generalized extreme value (GEV) distribution to analyze the simulated discharge between the 2050s and 2080s. Figure 8 shows the multimodal average results for the eight GCM models while Table 9 displays the return periods of peak flow under two scenarios in the 2050s and 2080s. The result indicates that the peak flow will increase in the future. In the 2050s, the average peak flow will increase by 14, 5, and 6% for a 5-, 20-, and 100-year return period flood under the SSP245 scenario. However, the average peak flow will increase to 24, 13, and 13% of 5-, 20-, and 100-year return period flows under the SSP585 scenario. In the far future, peak flow will increase by 24, 13, and 12% under the SSP245 scenario and by 43, 32, and 33% under the SSP585 scenario for a 5-, 20-, and 100-year return period flood. The result indicates that the peak flood magnitude will be higher for high-emission scenarios.

The annual peak flow of 100-year return periods includes two climate change scenarios, and eight GCMs are plotted in Figure 7. The return period plot shows that the flood will be more intense in the future, and people living in the basin will face more devastating floods. The higher intensity flood might cause more significant damage to agriculture, land, infrastructure, households, and other properties. Therefore, proper precautions and countermeasures need to be taken to cope with probable extreme future floods. Participatory flood management may be an option to cope with these extremes in the future.
Figure 7

The annual peak flow of 100-year return periods includes two climate change scenarios and eight GCMs.

Figure 7

The annual peak flow of 100-year return periods includes two climate change scenarios and eight GCMs.

Close modal
Figure 8

Probability density estimates of monthly mean streamflow (m3/s) of eight GCMs of two scenarios in the 2050s and 2080s at the Dalia station of Teesta River.

Figure 8

Probability density estimates of monthly mean streamflow (m3/s) of eight GCMs of two scenarios in the 2050s and 2080s at the Dalia station of Teesta River.

Close modal

The probability distribution function (PDF) of the peak flow in the Teesta River basin at Dali Point under two climate change scenarios and land-use change scenarios in the 2050s and 2080s is shown in Figure 8. The PDFs are obtained by applying a GEV to simulate peak flow values from eight GCMs and two emission scenarios. PDFs are shown for different combinations of the following scenarios: SSP245 in the 2050s, SSP245 in the 2080s, SSP585 in the 2050s, and SSP585 in the 2080s. Figure 8 shows model probability density plots where the x-axis denotes the discharge in cumec and the y-axis indicates the probability of occurrence. Most of the models showed a rightward shift of the extreme flood flows. Extreme flows will be increased for the SSP585 scenarios than SSP245 scenarios in both near and far future periods.

There are several limitations of this study. Teesta is a transboundary river, and most of the basin lies in India. Ground-based observational data in the Indian part of the catchment was limited. The satellite images are mostly high (>30%) cover around the year; for the lower cloud coverage area, this study had to compromise with the time of image collection. This study did not consider upstream interventions like dams and barrages and their impact on the flow of the Teesta.

This study attempted to quantitatively assess the impacts of LULC changes and climate change on the water balance of the Teesta River basin for the two future periods 2050s (2035–2064) and 2080s (2071–2100). The analysis involves an examination of LULC classification in the Teesta River basin during the baseline period of 2020, as well as predictions for two future periods: the near future 2050 and the far future 2080. For the classification, we use a supervised classification technique and the CA-Markov model used for the LULC prediction. The LULC changes and climate change impacts on the water balance are evaluated using the widely used semi-distributed hydrological model, SWAT. Additionally, SWAT-CUP has been used for calibration and validation purposes.

It has been found that the basin has experienced significant LULC changes from 2000 to 2010. During this decade, agriculture, settlement, and water/river areas increased, while forest and sand bar areas decreased. A similar trend was observed from 2010 to 2020. Using the CA-Markov model with 2020 LULC as the base year, the projection indicates that these trends will continue in 2050 and 2080. The prediction results show that agriculture and settlements will increase by 76.2 and 42.9% from 2020 to 2050 and by 121 and 293% by 2080. Conversely, forest, water, and snow cover will decrease by 13, 36, and 5.2% by 2050 and by 33, 38, and 0.8% by 2080.

To fit the model for the Teesta River basin, calibration and validation were performed for the periods 1997–2005 and 2006–2013, respectively. The sensitivity analysis, using the NSE, R², and PBIAS parameters, yielded values of 0.85, 0.81, and 18 for the calibration period, and 0.84, 0.81, and 16.9 for the validation period. These values indicate a good fit between the simulated and observed discharge. The prediction result illustrates that annual flow will be increasing for the near future period 2050s for all scenarios. The monsoon flow may increase by 26% for SSP245 and 38% for SSP585, while the dry season might see a 46% decrease for SSP245 and 49% for SSP585 in the near future. Similar patterns are expected in the far future, with monsoon flow increasing by 30% for SSP245 and by 45% for SSP585, while winter flow is projected to decrease by 40% for SSP245 and by 38% for SSP585. With climate change, LULC changes bring additional stress to the water balance of this region. As settlements increase, pre-monsoon flow may rise by 3% and monsoon flow by 4% under SSP245 in the near future. Under SSP585, pre-monsoon flow may increase by 4.5% and monsoon flow by 5%. However, post-monsoon flow will decrease by 0.04% for SSP245 and 0.3% for SSP585. In the far future, LULC changes could lead to a 13% increase in monsoon flow under SSP245 and a 52% increase under SSP585, while dry period flow might decrease by 6 and 13%, respectively. The peak flow frequency analysis indicates an increase in future peak flows. By the 2050s, peak flow is expected to rise by 14, 5, and 6% for 5-, 20-, and 100-year return periods under SSP245 and by 24, 13, and 13% under SSP585. In the far future, peak flow is projected to increase by 24, 13, and 12% under SSP245 and by 43, 32, and 33% under SSP585 for the same return periods.

The overall results indicate that the Teesta River basin will face a severe water crisis during dry periods and increased flooding during the monsoon season. Higher emission scenarios are expected to exacerbate these impacts. The return period analysis shows that future floods will be more intense, leading to more devastating effects on agriculture, land, infrastructure, households, and other properties. To mitigate these impacts, it is recommended to adopt a comprehensive and participatory approach, including implementing integrated watershed management practices to balance water resources and reduce flood risks and strengthen cooperation with countries sharing the Teesta River by negotiating equitable water-sharing agreements and joint management initiatives to ensure sustainable water use and disaster risk reduction. Additionally, integrating climate change adaptation into regional planning processes and enforcing sustainable land-use practices to prevent excessive deforestation and unplanned urbanization may ease the crisis in the future.

The work has been supported by the research project on ‘PROVATi3 (Promoting Resilience of Vulnerable through Access to Infrastructure, Improved Skills, and Information’ carried out by the Institute of Water and Flood Management (IWFM) of Bangladesh University of Engineering and Technology (BUET). It is funded by the Local Government Engineering Department (LGED) and sponsored by the International Fund for Agricultural Development (IFAD). The authors would like to acknowledge the access to observed water level, discharge, and bathymetric data from the Bangladesh Water Development Board (BWDB).

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.

Aksoy
H.
&
Kaptan
S.
(
2022
)
Simulation of future forest and land use/cover changes (2019–2039) using the cellular automata-Markov model
,
Geocarto International
,
37
(
4
),
1183
1202
.
https://doi.org/10.1080/10106049.2020.1778102
.
Alam
S.
,
Ali
M. M.
,
Rahaman
A. Z.
&
Islam
Z.
(
2021
)
Multi-model ensemble projection of mean and extreme streamflow of Brahmaputra river basin under the impact of climate change
,
Journal of Water and Climate Change
,
12
(
5
),
2026
2044
.
https://doi.org/10.2166/wcc.2021.286
.
Arnold
J. G.
&
Allen
P. M.
(
1996
)
Estimating hydrologic budgets for three Illinois watersheds
,
Journal of Hydrology
,
176
(
1–4
),
57
77
.
https://doi.org/10.1016/0022-1694(95)02782-3
.
Arnold
J. G.
,
Srinivasan
R.
,
Muttiah
R. S.
&
Williams
J. R.
(
1998
)
Large area hydrologic modeling and assessment Part I: Model development
,
Journal of the American Water Resources Association
,
34
(
1
),
73
89
.
https://doi.org/10.1111/j.1752-1688.1998.tb05961.x
.
Bal
M.
,
Dandpat
A. K.
&
Naik
B.
(
2021
)
Hydrological modeling with respect to impact of land-use and land-cover change on the runoff dynamics in Budhabalanga river basin using ArcGIS and SWAT model
,
Remote Sensing Applications: Society and Environment
,
23
,
100527
.
https://doi.org/10.1016/j.rsase.2021.100527
.
Ficklin
D. L.
,
Null
S. E.
,
Abatzoglou
J. T.
,
Novick
K. A.
&
Myers
D. T.
(
2022
)
Hydrological intensification will increase the complexity of water resource management
,
Earth's Future
,
10
(
3
),
e2021EF002487
.
https://doi.org/10.1029/2021EF002487
.
Fields
A. J.
,
Bhattacharjee
J.
&
Chatterjee
N.
(
2021
)
Reservoir bathymetry and riparian corridor assessment in two dammed sections of the Teesta River in Eastern Himalaya
,
Environmental Monitoring and Assessment
,
193
(
10
),
640
.
https://doi.org/10.1007/s10661-021-09433-z
.
Goyal
M. K.
&
Goswami
U. P.
, (
2018
)
Teesta River and its ecosystem
. In:
Sen Singh
D.
(ed.)
The Indian Rivers: Scientific and Socio-Economic Aspects
.
Singapore
:
Springer
, pp.
537
551
.
López-Ramírez
S. M.
,
Mayer
A.
,
Sáenz
L.
,
Muñoz-Villers
L. E.
,
Holwerda
F.
,
Looker
N.
,
Schürz
C.
,
Berry
Z. C.
,
Manson
R.
,
Asbjornsen
H.
,
Kolka
R.
,
Geissert
D.
&
Lezama
C.
(
2021
)
A comprehensive calibration and validation of SWAT-T using local datasets, evapotranspiration and streamflow in a tropical montane cloud forest area with permeable substrate in central Veracruz, Mexico
,
Journal of Hydrology
,
603
,
126781
.
Haque
S.
,
Ikeuchi
K.
,
Shrestha
B. B.
,
Kawasaki
A.
&
Minamide
M.
(
2023
)
Relationship between residential house damage and flood characteristics: A case study in the Teesta River basin, Bangladesh
,
International Journal of Disaster Risk Reduction
,
96
,
103901
.
Hasanuzzaman
M.
,
Shit
P. K.
,
Bera
B.
&
Islam
A.
(
2023
)
Characterizing recurrent flood hazards in the Himalayan foothill region through data-driven modelling
,
Advances in Space Research
,
71
(
12
),
5311
5326
.
Hossein
M. K.
,
Ahmed
A.
,
Uddin
M. J.
,
Saifullah
A. S. M.
,
Babu
M. A. H.
&
Sarker
T.
(
2023
)
Prediction of land use and land cover changes (LULC) and urban growth analysis in Dhaka Metropolitan Area (DMA) using CA-Markov model and geospatial techniques
,
Arabian Journal of Geosciences
,
16
(
8
),
468
.
https://doi.org/10.1007/s12517-023-11570-3
.
Jiménez-Navarro
I. C.
,
Jimeno-Sáez
P.
,
López-Ballesteros
A.
,
Pérez-Sánchez
J.
&
Senent-Aparicio
J.
(
2021
)
Impact of climate change on the hydrology of the forested watershed that drains to Lake Erken in Sweden: An analysis using SWAT+ and CMIP6 scenarios
,
Forests
,
12
(
12
),
1803
.
https://doi.org/10.3390/f12121803
.
Khan
I.
&
Ali
M.
(
2019
)
Potential changes to the water balance of the Teesta River basin due to climate change
,
American Journal of Water Resources
,
7
(
3
),
95
105
.
https://doi.org/10.12691/ajwr-7-3-2
.
Knoesen
D. M.
(
2012
)
Integrating Hydro-Climatic Hazards and Climate Changes as a Tool for Adaptive Water Resources Management in the Orange River Catchment
.
Doctoral dissertation. University of KwaZulu-Natal, Pietermaritzburg. School of Bioresources Engineering and Environmental Hydrology. http://hdl.handle.net/10413/8628
.
Leta
M. K.
,
Demissie
T. A.
&
Tränckner
J.
(
2021
)
Hydrological responses of watershed to historical and future land use land cover change dynamics of Nashe watershed, Ethiopia
,
Water (Switzerland)
,
13
(
17
),
2372
.
https://doi.org/10.3390/w13172372
.
Long
H.
,
Zhang
Y.
,
Ma
L.
&
Tu
S.
(
2021
)
Land use transitions: Progress, challenges, and prospects
,
Land
,
10
(
9
),
903
.
https://doi.org/10.3390/land10090903
.
Mathanraj
S.
,
Rusli
N.
&
Ling
G. H. T.
(
2021
)
Applicability of the CA-Markov model in land-use/land cover change prediction for urban sprawling in Batticaloa Municipal Council, Sri Lanka
,
IOP Conference Series: Earth and Environmental Science
,
620
(
1
),
012015
.
https://doi.org/10.1088/1755-1315/620/1/012015
.
Matlhodi
B.
,
Kenabatho
P. K.
,
Parida
B. P.
&
Maphanyane
J. G.
(
2021
)
Analysis of the future land use land cover changes in the Gaborone dam catchment using CA-Markov model: Implications on water resources
,
Remote Sensing
,
13
(
13
),
2427
.
https://doi.org/10.3390/rs13132427
.
Mensah
J. K.
,
Ofosu
E. A.
,
Yidana
S. M.
,
Akpoti
K.
&
Kabo-bah
A. T.
(
2022
)
Integrated modeling of hydrological processes and groundwater recharge based on land use land cover, and climate changes: A systematic review
,
Environmental Advances
,
8
,
100224
.
https://doi.org/10.1016/j.envadv.2022.100224
.
Mishra
V.
,
Bhatia
U.
&
Tiwari
A. D.
(
2020
)
Bias-corrected climate projections for south Asia from Coupled Model Intercomparison Project-6
,
Scientific Data
,
7
(
1
),
1
13
.
https://doi.org/10.1038/s41597-020-00681-1
.
Mohammed
K.
,
Islam
A. S.
,
Islam
G. T.
,
Alfieri
L.
,
Bala
S. K.
&
Khan
M. J. U.
(
2017
)
Extreme flows and water availability of the Brahmaputra river under 1.5 and 2 °C global warming scenarios
,
Climatic Change
,
145
,
159
175
.
https://doi.org/10.1007/s10584-017-2073-2
.
Mohammed
K.
,
Islam
A. K. M. S.
,
Islam
G. M. T.
,
Alfieri
L.
,
Khan
M. J. U.
,
Bala
S. K.
&
Das
M. K.
(
2018
)
Future floods in Bangladesh under 1.5°C, 2°C, and 4°C global warming scenarios
,
Journal of Hydrologic Engineering
,
23
(
12
),
04018050
.
https://doi.org/10.1061/(ASCE)HE.1943-5584.0001705
.
Mohseni
U.
,
Agnihotri
P. G.
,
Pande
C. B.
&
Durin
B.
(
2023
)
Understanding the climate change and land use impact on streamflow in the present and future under CMIP6 climate scenarios for the Parvara Mula Basin, India
,
Water (Switzerland)
,
15
,
9
.
https://doi.org/10.3390/w15091753
.
Mollel
G. R.
,
Mulungu
D. M. M.
,
Nobert
J.
&
Alexander
A. C.
(
2023
)
Assessment of climate change impacts on hydrological processes in the Usangu catchment of Tanzania under CMIP6 scenarios
,
Journal of Water and Climate Change
,
14
(
11
),
4162
4182
.
https://doi.org/10.2166/WCC.2023.542/1330345/JWC0144162.PDF
.
Momen
M. N.
&
Rahman
M. M.
(
2023
)
The impoverishment of political discourse and legal rights in transboundary Teesta water management
,
World Water Policy
,
9
(
1
),
11
28
.
https://doi.org/10.1002/wwp2.12093
.
Moriasi
D. N.
,
Gitau
M. W.
,
Pai
N.
&
Daggupati
P.
(
2015
)
Hydrologic and water quality models: Performance measures and evaluation criteria
,
Transactions of the ASABE
,
58
(
6
),
1763
1785
.
Nath
N.
,
Sahariah
D.
,
Meraj
G.
,
Debnath
J.
,
Kumar
P.
,
Lahon
D.
,
Chand
K.
,
Farooq
M.
,
Chandan
P.
,
Singh
S. K.
&
Kanga
S.
(
2023
)
Land use and land cover change monitoring and prediction of a UNESCO World Heritage Site: Kaziranga Eco-Sensitive zone using cellular automata-Markov model
,
Land
,
12
(
1
),
151
.
https://doi.org/10.3390/land12010151
.
Qi
J.
,
Zhang
X.
,
Yang
Q.
,
Srinivasan
R.
,
Arnold
J. G.
,
Li
J.
,
Waldholf
S. T.
&
Cole
J.
(
2020
)
SWAT ungauged: Water quality modeling in the Upper Mississippi River Basin
,
Journal of Hydrology
,
584
,
124601
.
Shampa
,
Nejhum
I. J.
,
Hussain
M. M.
,
Islam
M. M.
&
Zoha
R. H.
(
2023
)
Evaluating 2021 extreme flash flood of Teesta River
,
Hydrology Research
,
54
(
10
),
1095
1114
.
https://doi.org/10.2166/nh.2023.122
.
Sharpley
A. N.
&
Williams
J. R.
(
1990
)
EPIC-Erosion/Productivity Impact Calculator. I: Model Documentation. II: User Manual
, vol.
1768
.
Technical Bulletin
.
Washington, DC: United States Department of Agriculture
.
Subedi
P.
,
Subedi
K.
&
Thapa
B.
(
2013
)
Application of a hybrid cellular automaton–Markov (CA-Markov) model in land-use change prediction: A case study of Saddle Creek Drainage Basin, Florida
,
Applied Ecology and Environmental Sciences
,
1
(
6
),
126
132
.
https://doi.org/10.12691/aees-1-6-5
.
Sun
J.
,
Yan
H.
,
Bao
Z.
&
Wang
G.
(
2022
)
Investigating impacts of climate change on runoff from the Qinhuai River by using the SWAT model and CMIP6 scenarios
,
Water (Switzerland)
,
14
(
11
),
1778
.
https://doi.org/10.3390/w14111778
.
Tamang
P.
(
2023
)
Economic valuation and benefit transfer of restoring the Teesta riverine ecosystem
,
Nature Environment and Pollution Technology
,
22
(
1
),
189
197
.
https://doi.org/10.46488/NEPT.2023.V22I01.017
.
Thavhana
M. P.
,
Savage
M. J.
&
Moeletsi
M. E.
(
2018
)
SWAT model uncertainty analysis, calibration, and validation for runoff simulation in the Luvuvhu River catchment, South Africa
,
Physics and Chemistry of the Earth, Parts A/B/C
,
105
,
115
124
.
Ullah
B.
,
Fawad
M.
,
Khan
A. U.
,
Mohamand
S. K.
,
Khan
M.
,
Iqbal
M. J.
&
Khan
J.
(
2023
)
Futuristic streamflow prediction based on CMIP6 scenarios using machine learning models
,
Water Resources Management
,
37
(
15
),
6089
6106
.
https://doi.org/10.1007/S11269-023-03645-3/METRICS
.
Van Griensven
A.
&
Meixner
T.
(
2003
) ‘
Sensitivity, optimisation and uncertainty analysis for the model parameters of SWAT
’,
Proceedings of 2nd international SWAT conference
, pp.
162
167
.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/).