Spatial and temporal variations in water balance components of a Class-II river system in Nepal: a SWAT-based analysis Open Access

Physical and biological deterioration causing soil erosion and numerous socio-economic impacts is very evident in Nepal (Ojha et al. 2021). Problems related to water stress and resource availability are soaring and expected to increase more due to the ongoing and future impacts of climate change (Gurung et al. 2019; Dahal et al. 2020). The growing imbalance between water supply and demand amid climate change has intensified the water crisis in the country (Dahal et al. 2019; Prakash & Molden 2020; Singh et al. 2020). As a result, most of the areas, especially the urban, experience intermittent or persistent water scarcity (Pandey 2021). Climate change has increased the challenges associated with water resource management in this region (Aryal et al. 2019). The topic of water scarcity and stress due to natural and anthropogenic drivers arises frequently in national development discussions (Dixit et al. 2009; NCVST 2009; GoN-WECS 2011).

Nepal's precipitation patterns present significant challenges due to their seasonal and geographic variability. The country receives an average of approximately 1,500 mm of rainfall annually. However, around 80% of this total precipitation falls during the monsoon season, which spans from June to September. This concentrated rainfall often leads to temporary water excess and flooding. Conversely, the remaining 8 months of the year experience much lower levels of rainfall, resulting in periods of drought and water scarcity (Merz et al. 2003). The future demand for freshwater is expected to rise due to population growth and changing lifestyles (Liu et al. 2008).

The river system of Nepal is primarily divided into three classes. The first-class rivers (Class-I), namely Koshi, Gandaki, Karnali, and Mahakali that originate from the Himalayas, are perennial rivers, and receive water from snow melt. Similarly, second-class rivers (Class-II), namely Mechi, Kankai, Bagmati, West Rapti, and Babai originate from the Mahabharat range, also perennial, but do not receive water from snowmelt. Furthermore, numerous rivers originating from the Chure range which are ephemeral are categorized as class III rivers (WECS 2011). Several studies focus on the Class-I river basins of Nepal, with most research examining the impact of climate and land use changes on water availability (Dahal et al. 2020; Pandey et al. 2020a, b; Bajracharya et al. 2023; Baral et al. 2023; Devkota et al. 2024). However, in light of the anticipated water crisis, the assessment of Class-II rivers should not be overlooked, as these rivers are crucial for the overall water availability of the nation and serve numerous lives within their basins.

The West Rapti Basin, a Class-II river system, is a subject of study in this research. This basin has been facing issues related to watershed deterioration in the wet seasons and consequently water availability problems during the dry season. The situation can be worsened by increasing water demands due to population growth and agricultural development. The basin has been the subject of very few studies earlier. Most of those are concerned with rainfall–runoff (Talchabhadel et al. 2015; Thapa et al. 2020), erosion (Talchabhadel et al. 2020a, b), evaluating extreme precipitation (Talchabhadel et al. 2021), projection of stream flow (Shrestha & Zeng 2017), gross run-of-river hydropower potential (Bista et al. 2021), sensitivity analysis of groundwater parameters (Talchabhadel et al. 2020a, b), etc. focusing on the whole or part of the basin. However, a comprehensive study to understand the spatial and temporal variation of water balance components using the latest available hydro-meteorological data has not yet been conducted.

Detailed coverage of hydrological observations at high spatial and temporal resolutions is a challenging task. These observations are resource-intensive which is costly, labor and capital intensive. Hydrological simulation models like the Soil and Water Assessment Tool (SWAT) can offer relatively reliable estimates of water yield and availability across various watershed conditions and climatic scenarios, even when observational data are limited. The development of hydrological models for the targeted basins is essential to depict the spatio-temporal distribution of hydrology and water resources (Pandey et al. 2020a, b). Measurement and simulation of water availability with accuracy at detailed temporal and spatial levels are crucial to ensure water is available for various purposes, including hydropower, irrigation, water supply, domestic use, municipal supply, recreation, fisheries and industrial activities. This study, therefore, aims to fill this gap by calibrating and validating the SWAT model within the West Rapti River Basin. The fully calibrated and validated hydrological model can be utilized to study water availability under historic and future climates.

The framework begins with the collection of two key data sources: spatial data, which includes the digital elevation model (DEM), soil data, and land use/land cover (LULC), and observed meteorological data, consisting of precipitation (PPT), maximum temperature (T_max), and minimum temperature (T_min). These datasets are integrated into the SWAT model to simulate hydrological processes.

The model undergoes a parameter estimation process, ensuring it reflects the observed hydrological behavior, specifically using daily observed discharge data (m³/s). This is followed by an iterative process of calibration and validation. During calibration, the model's parameters are adjusted to closely match simulated outputs with observed data. Once calibration is complete, the model is validated to ensure its reliability and accuracy in simulating the hydrological processes of the basin. If either calibration or validation fails, the parameter estimation process is repeated until the model is well calibrated and validated. After successful calibration and validation, the framework moves on to analyze the spatio-temporal distribution of water balance components, providing valuable insights into water balance dynamics in the study area.

The DEM used was a 30-m resolution Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) Global Digital Elevation Model Version (GDEM) version 3, developed by the Ministry of Economy, Trade, and Industry (METI) of Japan and the National Aeronautics and Space Administration (NASA) of the USA. Daily observed precipitation and temperature data for the years 2010–2019 were obtained from the Department of Hydrology and Meteorology (DHM), Government of Nepal. Daily average river discharge values for two-gauge stations were obtained from DHM. Eleven precipitation and two temperature stations were selected for model input data. A land use map with a spatial resolution of 30 m for the year 2010 was obtained from the International Centre for Integrated Mountain Development (ICIMOD). The soil map for the year 2009 was downloaded from the Soil & Terrain Database (SOTER) for Nepal. Daily solar radiation, wind speed, and relative humidity data were generated using SWAT's inbuilt weather generator functions. Table 1 provides the data and data sources for the SWAT model.

The SWAT model is a process-based continuous hydrological model that predicts the impact of land management practices on water, sediment and agricultural chemical yields in complex basins with varying soils, land use and management conditions (Arnold et al. 1998). The Arc-SWAT is a semi-distributed hydrological model incorporated into an ArcGIS interface. The main components of the model include climate, hydrology, erosion, soil temperature, plant growth, nutrients, pesticides, land management, and channel and reservoir routing. The SWAT model divides the basin into sub-basins, which are connected through a stream channel and further divided into hydrologic response units (HRUs). An HRU is a unique combination of soil and vegetation in a sub-watershed, and the model simulates the hydrology, vegetation growth, and management practices at the HRU level. Maintaining a continuous water balance allows the model to reflect differences in evapotranspiration for various crops and soils, resulting in more accurate runoff predictions for each sub-basin and a better physical description of the water balance.

A DEM was imported into the SWAT model. Flow direction and accumulation were calculated, and the threshold area for stream generation was set at 18,000 ha. The stream network and outlet were then created, with the downstream outlet at Kusum selected for the delineation of the river basin. The WRR basin was divided into 20 sub-basins and four slope classes (0–5, 5–15, 15–25 and >15%). The LULC map and soil maps were imported with the slope into the model for the overlay to obtain a unique combination of land use, soil and slope. Multiple HRUs with 20% LULC, 20% soil and 20% slope thresholds were set to eliminate minor land uses and slope classes in each watershed. Basin was further subdivided into a total of 155 HRUs. Additionally, to account for orographic effects, 500 m range elevation bands were generated in each sub-basin. Weather station meteorological data such as daily rainfall, minimum and maximum temperature, relative humidity, wind speed and solar radiation were also imported into the model as recommended in the SWAT user manual (Neitsch et al. 2002). SWAT's inbuilt weather generator functions were utilized to generate daily solar radiation, wind speed, and relative humidity data. Precipitation data from 11 rain gauge stations and temperature data from two climatic stations were fed into the model. Observed flow data of Nayagaon (mid-stream) and Kusum (downstream) stations from 2010 to 2019 were used in the study. A warm-up period of a year (2010) was excluded from the analysis to stabilize the model initially. The model was calibrated using 6 years (two-thirds) of the study period (2011–2017) and validated using the remaining one-third, i.e., 3 years (2017–2019).

The SWAT model was simulated for the first year (2010) as a warming period. Then, the streamflow was calibrated for the period of (2011–2016) and validated for the period of (2017–2019). Two stream gauging stations at mid-stream and downstream were used for multiple site calibration and validation processes. SUFI-2 (Abbaspour et al. 2004), a semi-automated approach, was employed using the SWAT–Calibration Uncertainty Procedure (SWAT–CUP), the most widely used approach to carry out parameterization, sensitivity and uncertainty analysis, calibration, and validation of input parameters.

The sensitivity analysis of model parameters, calibration, and validation were performed with SWAT–CUP tools (http://swat.tamu.edu/software/swat-cup/) using the Sequential Uncertainty Fitting Algorithm (SUFI-2). In SWAT–CUP, global sensitivity analysis for the perturbation of the parameters was carried out (Abbaspour 2015). In total, 31 input parameters were selected based on the SWAT and SWAT–CUP user manual as suggested by Abbaspour et al. 2007.

Considering the entire streamflow 1,000 simulations were performed for two iterations, meaning a total of 2,000, which helped to optimize the ranges of selected parameters and understand sensitive parameters of the streamflow. The parameter is more sensitive when the absolute t-stat is higher and the p-value is smaller (Abbaspour 2015). Further, for model calibration and validation, as flow is the main controlling variable, parameters sensitive to flow only were selected (Abbaspour et al. 2007) and the calibration was done. After the model was calibrated successfully, the same calibrated parameters were used for model validation.

Seven parameters were identified as the most sensitive among the selected 31 parameters (p-value <0.05) for the study area. A list of parameters sensitive to flow based on t-stat and p-value is presented in Table 2. The results show that the PLAPS is the most sensitive parameter (Rank 1) followed by LAT_TTIME, ALFA_BNK, CH_K2, ALPHA_BF, CH_N2, GW_DEALY. The remaining parameters have no significant effect on the streamflow.

After performing sensitivity analysis, the seven most sensitive parameters: PLAPS, LAT_TTIME, ALFA_BNK, CH_K2, ALPHA_BF, CH_N2, and GW_DEALY were optimized. PLAPS is related to the elevation band (.sub), LAT_TTIME corresponds to lateral flow (.hru), while ALFA_BNK, CH_K2, and CH_N2 represent channel water routing (.rte). Additionally, ALPHA_BF and GW_DEALY are associated with groundwater (.gw). All seven parameters are of the ‘v’ type, indicating that their existing values are to be replaced with specified values. Within the selected 31 parameters, some belong to the ‘r’ type, suggesting that existing parameter values are to be multiplied by (1+ a given value). Table 3 presents the maximum, minimum, and fitted values for all selected parameters, highlighting sensitive parameters with their final calibrated range and fitted value.

Table 3

Selected 31 parameters with their final range and fitted values

View Large

The analysis of hydrological data for the study area reveals critical insights into the water balance dynamics. In terms of precipitation, the area received a total of 1,367.1 mm of rainfall. Evapotranspiration (ET) accounts for a significant portion of this water, with a value of 78.8 mm. The potential evapotranspiration (PET) is slightly higher at 85.7 mm. Surface runoff, a critical component of water balance, was measured at 901.98 mm, indicating substantial overland flow in the study area. This high runoff rate indicates that a significant portion of the precipitation contributes directly to surface flow. High rainfall intensity or low infiltration capacity in certain sub-basins could be the reason for high surface runoff. Lateral flow is 260.56 mm representing subsurface movement of water. This flow is essential for sustaining base flow in rivers and streams, particularly during dry periods. Variation in the distribution of lateral flow across sub-basins is observed which might be influenced by soil properties and topography. The study area exhibited notable groundwater recharge dynamics. The return flow is 27.05 mm, indicating the small amount of water that returns to the system. Furthermore, percolation to the shallow aquifer is measured at 65.4 mm, while revap from the shallow aquifer is 9.57 mm. Recharge to the deep aquifer is observed at 14.91 mm, reflecting the replenishment of the deeper groundwater reservoir.

Change in storage is the sum of groundwater recharge, change in soil moisture storage in the vadose zone and model inaccuracies. The negative value of change in storage may reflect that most of the precipitation in a year in the WRR basin is contributing to the groundwater recharge. Throughout all sub-basins, there is variation in the value of change in storage. Change in storage is positive in sub-basins 12, 16, 17, and 20 while other sub-basins have negative value of change in storage. In sub-basin 1, change in storage is around 20% (negative) of the precipitation, reflecting that NWY (535 mm) is lower than the difference between PPT and AET. Similarly, sub-basins 15 and 20 have the lowest change in storage about 1% of the precipitation. Eventually, average change in storage of the entire basin is 6% (negative). Overall, Δ storage is negative across most sub-basins, indicating substantial groundwater recharge.

The NWY across the WRR sub-basins varies from 546 to 2,358 mm. Sub-basin 8 with the highest PPT (2,595 mm) also has the highest NWY of 2,358 mm, whereas sub-basin 6 with the lowest PPT (655 mm) also has the lowest NWY of 546 mm. In 17 (or 85%) sub-basins, NWY is more than 80% of PPT and in 20 (or 100%) sub-basins the NWY is more than half of PPT. The surface runoff has the major share in the NWY across most of the sub-basins while the contribution of groundwater and lateral flow fluctuates.

The sum of NWY and AET does not equal to PPT across all sub-basins. This might be because NWY is the combination of surface runoff, lateral flow, and groundwater flow, with a reduction in transmission losses and pond abstractions rather than simply the difference between PPT and AET. The NWY is affected by factors such as rainfall intensity, soil properties, and land use/cover characteristics and is inconsistent with the precipitation trend (Bharati et al. 2019). It is observed that NWY is lower than the difference between PPT and AET in most sub-basins due to various reasons.

The study reveals that most sub-basins experience high-intensity rainfall, where lateral flow significantly influences water movement across the landscape. Steep topography leads to high surface runoff, forming streams and rivers. During short and intense rainfall periods lateral water movement is increased due to low infiltration capacity in certain sub-basins. Sub-basins with steep slopes are prone to natural calamities such as flooding, landslides, and erosion. The spatial variation in water balance components within the study area is driven by the interrelationship between precipitation, land cover, and topography. Understanding this correlation is crucial for developing water resource management and policy interventions aimed at sustainable water use and mitigating hydrological extremes.

The negative storage value in April, May, June, July, and August suggests significant groundwater recharge within the basin during these months. Similarly, Δ storage is positive in September, October, November and December, indicating the contribution of groundwater to the water yield. Large values of Δ storage during the monsoon can be attributed to high groundwater recharge, which is responsible for groundwater flow and ultimately baseflow during the dry period of the year. Similarly, a decreasing trend of Δ storage is notable from the monsoon to the dry period. In August, the delta storage is lower than in July as the soil is fully saturated and in September, a positive storage value is observed. This reflects that there is a reduction in groundwater recharge in August while the groundwater is contributing to the water yield after September. The Δ storage is positive during the post-monsoon season until December and from January it is negative. In the post-monsoon (ON) season, the highest Δ storage value is 26 mm (positive). Recharge during the monsoon season (JJAS) and discharge of that recharge water in the post-monsoon could be the reason for the highest positive value. As a result of winter, from January there are minimal negative values.

During the monsoon period (JJAS), sub-basin 8 receives the highest rainfall compared to other sub-basins, at 2,164 mm, while sub-basin 6 experiences the lowest rainfall at 538 mm. Additionally, AET is higher in sub-basins 7 and 8, at 70 mm, and very low in sub-basin 16, at 34 mm. Moreover, the water yield in sub-basins 7 and 8 is significantly higher than in other sub-basins, measuring 1,864 and 1,949 mm respectively, whereas sub-basin 1 has a very low water yield of 389 mm. These variations underscore the monsoon's dominant influence on water availability and distribution across the basin. Following the monsoon, there is a marked decline in precipitation, AET, and water yield. The post-monsoon (ON) and winter (DJF) periods are characterized by reduced hydrological activity, as the basin transitions into a drier phase. Δ storage values during the dry months are positive, indicating that groundwater contributes to water yield during this time. This contribution is crucial for maintaining baseflow in rivers and streams, sustaining the basin's hydrology during periods of low rainfall.

Sub-basins with higher precipitation, sub-basins 7 and 8, also exhibited higher AET rates across all seasons except in winter. Similarly, sub-basins with lower rainfall recorded minimum AET rates except in monsoon. This interrelation between precipitation and AET portrays that AET rates are influenced by rainfall and season.

The study reveals significant groundwater recharge during the monsoon, with negative Δ storage values in May, June, July, and August suggesting active recharge processes. In contrast, during the dry months, Δ storage becomes positive, reflecting groundwater's role in supporting water yield. This seasonal fluctuation in storage highlights the basin's reliance on groundwater during dry periods and its restoration during wetter months. The large Δ storage values observed during the monsoon can be attributed to high groundwater recharge, which is essential for sustaining baseflow throughout the year. The prominent decreasing trend of Δ storage from the monsoon to the dry period further draws attention to the dynamic interplay between precipitation, recharge, and groundwater flow.

Optimal water resource management is possible only with a detailed understanding of the temporal variation of water balance components in the basin. During the monsoon, efforts should focus on capturing and storing excess water to safeguard against dry periods. Alternatively, during the dry season, efficient utilization of groundwater resources is crucial for sustainable groundwater management. The temporal variation in water balance components within the study area is significantly influenced by seasonal patterns, particularly the monsoon. These findings emphasize the need for a comprehensive water management strategy that accounts for the dynamic and seasonal nature of hydrological processes to ensure sustainable water availability throughout the year.

This study aimed to understand the spatial and temporal variations in the water balance within the West Rapti River Basin. The SWAT model was effectively calibrated and validated across multiple gauging sites using observed discharge, achieving acceptable ranges for objective functions. Key parameters such as PLAPS, LAT_TTIME, ALFA_BNK, CH_K2, ALPHA_BF, CH_N2, and GW_DELAY significantly contributed to optimizing the model, reflecting the rain-fed characteristics of the basin. Additionally, both statistical and graphical assessments confirmed the model's satisfactory performance. Hence, this study also contributes by establishing a set of calibrated SWAT model parameters for the West Rapti Basin, facilitating future hydrological studies.

The annual average precipitation in the basin is estimated at 1,367.1 mm which contributes to annual water budget of surface runoff 902 mm (rainfall to runoff ratio: 0.66) with an evapotranspiration total of 78.8 mm (approximately 6% of precipitation), and the NWY is 1,204.5 mm (approximately 88% of annual precipitation). It's important to note that these components of the water balance vary both spatially and temporally. In particular, 22% of the NWY originates from lateral flow, 75% from surface flow, and 3% from groundwater flow. This illustrates significantly high surface flow within the basin.

Most sub-basins experience high-intensity rainfall, leading to significant lateral water movement across the area. Short, intense rainfall over steep topography results in high surface runoff, further increasing lateral water movement. As a result, these regions are prone to natural disasters like flooding, landslides, and erosion. Effective water resource management in such areas requires understanding the dynamics of lateral flow and its relationship with surface and groundwater flows. Strategies such as erosion control, bioengineering, and constructing recharge ponds are essential for mitigating the impacts of extreme weather and optimizing water storage.

The hydrological model is well-calibrated and provides reliable flow estimation at the sub-basin level, including in regions without observed data. This research is useful for studies on water availability for irrigation, hydropower, and water supply projects. Additional model outputs, such as soil moisture storage and potential evapotranspiration at the regional level, can help in determining crop water requirements.

Moreover, its application spans across various sectors, supporting the development of water resources management strategies. The model also facilitates climate and land use change assessments offering an essential understanding of potential shifts in water availability. Thus, the model's outputs would facilitate the planners and policymakers to select an approach for integrated and efficient basin development and management.

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

The authors declare there is no conflict.