Climate change will exacerbate seasonal flow variability in the Karnali River basin: Implications for water, energy, and agriculture sectors Open Access

The impacts of climate change on water resources in the Himalayan region have been extensively studied (Dimri et al. 2021; Khanal et al. 2021; Nie et al. 2021; Baral et al. 2023). These studies reveal multifaceted and multi-sectoral effects, including floods (Wijngaard et al. 2017), droughts (Sharma et al. 2021), changes in snow and glaciers (Khanal et al. 2021), soil moisture variations (Nepal et al. 2021a, b), impacts on agriculture (Lutz et al. 2022), and ecosystems (Negi et al. 2012). Climate change impacts are cross-sectoral and focusing on any sector individually may not adequately capture the impacts felt across the scales. Therefore, a holistic understanding is necessary to evaluate the impact of climate change. One approach to addressing the cross-sectoral impacts is the water, energy, food, biodiversity, and ecosystem nexus (Herrera-Franco et al. 2023; Holmatov et al. 2023; Wu et al. 2023). The Nexus approach emphasizes the interconnectedness of all sectors and the importance of understanding how impacts in one sector translate to others, which is crucial for designing effective mitigation and adaptation strategies.

Nepal's water, energy, food, and ecosystem (WEFE) sectors are closely interconnected, forming a complex nexus. Managing this nexus is key to ensuring resource security, sustainable development, and environmental protection. Nepal has advanced integrated resource management through policies such as the Water Resources Strategy (MoEWRI 2002), National Water Plan (MoEWRI 2005), Agriculture Development Strategy (MoALD 2015), Climate change policy (MoHP 2019), Water Resources Policy (MoEWRI 2020), and Irrigation Policy (MoEWRI 2024). These policies focus on efficient, coordinated resource use to improve water productivity, food security, and ecosystem preservation.

The Karnali River basin in central Himalaya, shared by China, Nepal, and India, is a vital lifeline for millions. Snow and glacier melt significantly contribute to its dry season flow. According to Khatiwada et al. (2016), 77% of the basin's rainfall occurs during the monsoon, with snowfall accounting for 10% of annual precipitation. Dhami et al. (2016) noted that snowmelt runoff makes up 12% of total runoff. In Nepal, it remains the last free-flowing river, supporting ecosystems, agriculture, and communities, while in India (known as the Ghaghra), it is crucial for irrigation.

The Karnali River basin is experiencing rising temperatures and shifting precipitation patterns due to climate change. From 1981 to 2012, maximum and minimum temperatures increased by 0.05 and 0.01 °C/year, respectively, with higher elevations warming faster. The annual precipitation was decreasing at the rate of 4.91 mm/year for the basin, however, the monsoon season in the mountain region of the basin showed an increasing trend (Khatiwada et al. 2016). While annual precipitation is projected to increase in the future, it will not be uniform across seasons, with the post-monsoon season – already the driest – expected to receive even less precipitation (Dahal et al. 2020).

The hydrology of the Himalayan rivers is shaped by seasonal precipitation, steep topography, forested, and glaciated areas. Climate change studies projected increased streamflow in the future. Lamichhane et al. (2024) projected up to a 55% increase in streamflow at Chisapani under SSP245 and 90% under SSP585 by 2074–2100. Dahal et al. (2020) projected a rise in annual discharge by 8.2 (2021–2045), 12.2 (2046–2070), and 15% (2071–2095), though post-monsoon flows are expected to decrease. Pandey et al. (2020) indicated that higher altitudes in the northern mountains are more vulnerable to water availability changes than the southern flatlands. Rising temperatures and shifting precipitation patterns may lead to both floods during the monsoon and reduced flows in the dry season, worsening water scarcity and affecting agriculture, hydropower, and ecosystems. Streamflow projections vary widely, from a 15% increase (Dahal et al. 2020) to 90% (Lamichhane et al. 2024), making it difficult for policymakers to plan effective mitigation and adaptation strategies.

To reduce uncertainties, ensembles of climate models using multiple emissions scenarios should be applied, accounting for natural variability, especially in regional projections for the Karnali River basin. In Nepal, most of the climate change and its impact on water resources are based on CMIP5 climate projections (Rajbhandari et al. 2016; Dahal et al. 2020; Pandey et al. 2020; Pradhan et al. 2021). While CMIP5 models have an advanced understanding of future climate, CMIP6 is introduced for three key reasons: First, CMIP6 offers a broader range of future temperature and precipitation changes, providing a more detailed view of potential climate outcomes (Fan et al. 2020; Hamed et al. 2022; DMI 2023). Second, it improves the simulation of precipitation variability and amplitude, addressing a key limitation of CMIP5 (Gulakhmadov et al. 2025). Lastly, CMIP6 includes higher resolutions, updated physical processes, reduced uncertainties, and additional Earth system dynamics, making models more accurate and reflective of real-world complexities (Guo et al. 2022; Martel et al. 2022; Wang et al. 2022; IPCC 2023). These improvements highlight the importance of adopting CMIP6 for more reliable climate projections and better-informed policy decisions.

Foresight analyses are essential for resource management, particularly in addressing the impacts of environmental and climate change. By illustrating probable future scenarios, foresight analyses play a crucial role in exploring solutions to socioeconomic and environmental challenges while guiding effective policy interventions (Gallouj et al. 2015; Wiebe et al. 2018). Understanding how climate change affects seasonal variability in water availability and its impact on various sectors is a major gap in knowledge regarding the Karnali River basin. In this paper, we assess the impacts of climate change on the water resources of the Karnali River basin using downscaled global climate projections from CMIP6 to a 9 km resolution, a significant improvement over previous studies (e.g. Lamichhane et al. (2024) at 25 km and Dahal et al. (2020) at 50 km). Unlike earlier research that utilized the SWAT model, we employed the more advanced SWAT+ model to project impacts under various scenarios aligned with potential future pathways of IPCC's AR6 report. We analyze changes in water balance components – such as snowmelt, evaporation, surface runoff, and groundwater flow – across historical and future periods. Future hydrological simulations based on climate projection data were used to examine the implications of water availability, energy production, and irrigation water demand in the basin. These are vital for informed decision-making in the Karnali River basin.

The Karnali River basin (44,000 km²) spans western Nepal, stretching from the Tibetan plateau in the north to India in the south. It hosts Nepal's longest river, the Karnali, which originates in the Trans-Himalayas at altitudes of 5,500 to 7,726 m and flows down to the southern Terai. Fed by snowmelt, the river passes through key watersheds, including West Seti, Kawari, Humla Karnali, Mugu Karnali, and Bheri. The basin contains 1,459 glaciers (1,023 km²) and 1,128 glacial lakes (48.93 km²). The basin's glacier area has decreased by 26% in the last three decades (Bajracharya et al. 2014). The basin is 12% snow-covered, with 2% permanent glaciers, 16% agricultural land, and 33% forest.

The basin's precipitation is mainly influenced by the summer monsoon from the Indian Ocean, with 71% occurring between June and September. Winter precipitation is affected by the westerlies. The basin receives an average of 1,479 mm of precipitation annually. The temperature decreases from the southern lowlands to the northern highlands, with an average minimum of 13 °C and a maximum of 25 °C. The northern mountains have a polar tundra climate, while the southern plain experiences a temperate climate with hot summers and dry winters.

The Karnali basin's diverse ecosystems, which support freshwater fish, amphibians, reptiles, birds, and mammals, are largely due to altitude variation. It includes three national parks and a hunting reserve and hosts several globally threatened species (Khatiwada et al. 2021). Two of Nepal's 10 listed wetlands are located here. Hydropower projects under construction and planned raise concerns about ecosystem impacts and biodiversity (Sharma et al. 2020).

Most agricultural land in the southern plains of the basin is rain-fed, with key crops including paddy, wheat, maize, and barley. However, the Rani Jamara Kuleriya Irrigation Project irrigates 38,300 ha, downstream of the Chisapani by drawing 55 m³/s from the Karnali Rivers (Nepal et al. 2024). The minimum monthly flow in the river is about 300 m³/s. The basin houses most of the national pride energy and irrigation projects.

The Karnali and Mahakali basins have a theoretical hydropower potential of 36,180 MW, with a technical potential of 26,570 MW and an economic potential of 25,125 MW (KC et al. 2011). However, only 92 MW of projects are completed, 1,337 MW are under construction, and 3,432 MW have survey licenses (Department of Electricity Development 2024). In the mid hills, many farmer-managed irrigation systems use water from many tributaries of the Karnali River systems. At the same time, the southern plains have the Rani Jamara Kulariya irrigation system, covering 38,300 ha. Since 2023, a single intake near Chisapani has fed three irrigation systems, including 2.5 MW of hydropower (Nepal et al. 2024).

SWAT+ is an updated open-source version of the Soil and Water Assessment Tool (SWAT), designed to handle current and future challenges in water resources modeling and management. It aims to make code development and maintenance easier, improve data handling and visualization, and enhance the model's ability to represent different elements and processes within watersheds. The biggest change is the addition of landscape units, which allows for better tracking and detailed spatial representation of the movement of water across the landscape. SWAT+ is also more flexible than SWAT when it comes to defining management plans and linking human-made water systems to natural streams (Bieger et al. 2017). It is a semi-distributed model where watersheds are discretized into hydrological response units (HRUs) based on areas of unique properties of DEM, land use, and soil type. Water from the HRUs/landscape units is routed to the sub-basin's main channel. This includes surface runoff, lateral flow (shallow subsurface flow), and groundwater flow. SWAT+ tracks how water moves within each landscape unit before it enters the channel, enhancing spatial representation compared with the original SWAT model (Supplementary Figure S1). Once water reaches the main channel of a sub-basin, it is routed downstream following the river network. Water from multiple sub-basins accumulates as it moves downstream, reaching the outlet of the watershed. The following steps were involved in setting up the model.

The spatial data – digital elevation model (DEM) (HydroSHEDs), land use/land cover (Copernicus), and soil (HWSD) – were resampled at 90-m resolution. The SWAT+ extension in QGIS (QSWAT+) has the function of delineating the watershed and HRUs. Follow parameters were used during sub-basin delineation (5,000 cells), minimum HRU (500 cells).

The hydrometeorological input for SWAT+ includes files for precipitation, temperature, and relative humidity. These variables were extracted from ERA5-Land with a spatial resolution of 9 km and hourly temporal resolution (Copernicus Climate Change Service 2024). ERA5-Land data is a high-resolution, global atmospheric reanalysis dataset produced by the European Centre for Medium-Range Weather Forecasts (ECMWF) (Muñoz Sabater 2024). The daily precipitation and mean temperature were derived from hourly temperature data for the period 1998 to 2019. ERA5-Land is a reliable and detailed source of data for hydrological modeling, offering the flexibility to simulate various water processes over different time scales. The observed hydrological data for the outlet of the basin – Chisapani – was obtained from the Department of Hydrology and Meteorology (DHM), Nepal, for the period 2000 to 2010.

The model was set from 2000 to 2004 for the calibration period and 2006–2010 for the validation period with two years as a warm-up period (1998–1999). Hydrological models for catchments need to be calibrated, and this is often done by comparing the model's output with observed discharge data at the catchment's outlet. To determine if the model has been successfully calibrated, objective measures such as Nash–Sutcliffe Efficiency (NSE), the coefficient of determination (R²), and percent bias (PBIAS) are used. These measures help evaluate how well the model's results match the observed data (Moriasi et al. 2007). The automatic calibration tool, SWAT+ Toolbox was employed for automatic calibration. The range for the parameters was determined based on the study area and literature review. One thousand sets of parameters within the range were generated by the tool based on Latin-Hypercube sampling (McKay et al. 2000). The tool would then select the best-performing parameter set based on the highest NSE value.

The general circulation models (GCMs) from the sixth phase of Coupled Model Intercomparison (CMIP6) were used for climate projection of the Karnali River basin. This has so far included 37 GCMs from the 53 modeling centers (Durack et al. 2020). The CMIP6 included Shared Socioeconomic Pathway (SSP) experiments based on various levels of socioeconomic challenges for adaptation and mitigation. The models were selected based on their ability to represent historic seasonal and annual variations compared with the ERA5-Land reference dataset (Nepal et al. 2021a). The model selection included the following steps:

• The first step involved assessing the model's capability to capture the seasonal cycle in the historical period for mean temperature and precipitation. The four seasons are defined as winter (December, January, February), pre-monsoon (March, April, May), monsoon (June, July, August, September), and post-monsoon (October, November). For precipitation, seasonal contribution to the annual total precipitation biases were calculated for each season, whereas for the mean temperature, a difference of each season to the annual average is calculated for the reference and GCM datasets for the historical period. The absolute bias between the reference dataset and the GCM dataset is calculated for each season for both precipitation and temperature. These biases are then ranked, average ranks are calculated, and the top 24 models with the least bias are selected for step 2.

• In the second step, relative bias for annual precipitation and absolute bias for annual mean temperature are calculated for the reference and GCM datasets for the historical period. The biases are ranked, and the average rank is calculated for the selected 24 GCMs. For the next step, 16 different models with the lowest bias are selected.

• The third step involved selecting representative models from the four corners of four future potential climatic conditions (warm-wet, cold-dry, warm-dry, and cold-wet). This envelope approach ensures that the whole spectrum of the models is considered for future scenarios (Lutz et al. 2016; Nepal et al. 2021a). We calculated the changes in annual precipitation and annual mean temperature between the historical period (1985–2014) and the end-century (2071–2100) for the 12 selected GCMs. The delta change method selected the representative models from each corner. The precipitation is calculated in percentage, whereas the mean temperature is calculated in the difference between the future and historical periods. The delta changes are then calculated, and the nearest model to each of the four corners representing four future potential climatic conditions is selected for SSP245 and SSP585. Supplementary Figure S2 provides the monthly climatological plot for all three steps and four selected models compared with the reference dataset.

The eight selected models were interpolated to match the resolution of reference datasets using bilinear interpolation. The empirical–statistical downscaling method, quantile mapping, was used for downscaling and bias correction. Monthly correction factors were derived from empirical cumulative distribution functions (ECDFs) between raw GCM and reference datasets (1985–2014) and applied to the entire 1985–2100 period. A similar approach was used by Lutz et al. (2016), Kaini et al. (2020), and Nepal et al. (2021a, b). The bias-corrected temperature and precipitation data were then extracted for all GCMs, and SWAT+ model input files were prepared for future simulations.

The future projection data (2015–2100) alongside the reference data (1985–2014) was fed into the validated SWAT+ model to derive the continuous future hydrology from 1985 to 2100. A comparison of future hydrology was made between the historical period (1985–2014) with the mid-century (2036–2065) and the end-century (2071–2100), depicting changes in monthly, seasonal, and annual discharge, floods, and water balance components.

The projected monthly energy generation for the project is calculated using the simulated monthly discharge from the SWAT+ model at the intake site for the mid-century (2036–2065) and the end-century (2071–2100) and compared with the historical period (1985–2014).

Future irrigation demand for the Rani Jamara Kuleriya irrigation scheme was assessed using CROPWAT 8.0. In order to model the crop using the CROPWAT, ET₀, rain, crop, and soil parameters are required. The ET₀ is calculated based on the minimum and maximum temperatures. Bias-corrected downscaled climate datasets for both historical and future periods (mid-century and end-century) are used to evaluate these changes. The sand, silt, and clay % for the study area are determined from the NARC soil database.¹ The soil is identified as red loamy soil for the area. The parameter file for red loamy soil is thus used for the calculation. The inbuilt crop parameter file developed by FAO is used for rice and wheat while altering the sowing and harvest dates for the plants. The information for cropping length and planting date is obtained from the FAO Crop Calendar for rice and wheat and for the country Nepal. The model is then set up to estimate the required crop water requirement and irrigation water requirement for paddy (May–September) and wheat (October–December) throughout their growth cycles. The length of the crop development stages for paddy and wheat is provided in Supplementary Table S1. Similar estimation throughout the world has been conducted using CROPWAT for crops like paddy, wheat, sugarcane, and others (Solangi et al. 2022; Soomro et al. 2023).

Out of 37 global climate models, a comparison was made with the historic climate cycle from the ERA5-Land dataset (1985–2014). In the first step, 24 models with the lowest biases were selected, while 13 models with precipitation bias over 7% and temperature bias above 1.2 °C were excluded (Supplementary Table S3). In the second step, 12 models with annual biases exceeding 29% for precipitation and for temperature 2.3 °C were discarded (Supplementary Table S4). The remaining 12 models were used to assess projected changes in annual precipitation and annual mean temperature by the end of the century compared with the reference period. This analysis provided insight into four scenarios: warm-wet, warm-dry, cold-wet, and cold-dry. In the end, we chose one model from each probable future scenario, and altogether, four models were selected for future hydrological assessment for each scenario (Supplementary Table S5).

We further downscaled and bias-corrected the selected global climate models using quantile mapping to a higher resolution of 9 km, which matches the resolution of the ERA5-Land reference datasets. This process resulted in 427 grids of 9 km resolution, which were then used in the SWAT+ hydrological model.

For wheat, a winter crop (October–December), cold-wet conditions favor growth, while warm-dry conditions are unfavorable. IWRs will increase under all climate conditions compared with the reference period. The highest rise in IWR occurs during the mid-season growth stage for both scenarios and periods. In the mid-season growth stage, IWR is projected to increase by 44% mid-century and 64% end-century under warm-dry conditions. In the late-season growth period, IWR will rise by 130% mid-century under warm-dry conditions. End-century warm-wet conditions show IWR increases of 208–241% for SSP245 and SSP585 scenarios, respectively.

Climate change in the Karnali River basin is expected to increase flow during the monsoon and post-monsoon seasons while reducing flow in the winter and pre-monsoon seasons (Figure 8 and Supplementary Table S6). The combined effect of a decrease in precipitation combined with an increase in actual evapotranspiration from December to April is the probable cause of the decrease in the discharge of the basin. However, despite the increase in the actual evapotranspiration, the discharge is going to increase from August to October due to a higher increase in precipitation during these months. These findings are consistent with other studies showing similar trends in the Karnali basin (Pandey et al. 2020) and across Nepal and the Himalayas (Khanal et al. 2021; Lutz et al. 2022).

However, the climate projections based on CMIP6 from this study indicate that precipitation would decrease in the winter and spring seasons, compared with a decrease in only the spring season as indicated in CMIP5 (Rajbhandari et al. 2016; MOFE 2019). This may be due to improved physical processes (references) included in the recent version of the CMIP6 projections. This additional insight will be important for water resource planning and development.

There are eight operational hydropower projects, 17 under construction and 27 with survey licenses. As these projects are run-of-the-river in nature, they may not greatly affect basin-wide water availability. However, the water diversion will create dewatered zones between the intake and the powerhouse site. In case of the Upper Karnali hydropower project, a dewatered zone of more than 50 km will be created between the proposed dam and the powerhouse site (Figure 1). The reduced discharge, especially during the dry period, will hamper the ecological health of that stretch of the river. The environmental impact assessment (EIA) should include an environmental management plan to mitigate these effects, particularly on aquatic life for all hydropower projects. A fish ladder could be one of the mandatory solutions necessary for fish movement from downstream to upstream and vice versa. To sustain biodiversity and ecosystems, the Hydropower Development Policy 2001 prescribed releasing at least 10% of the minimum monthly discharge, or a higher amount specified by the EIA, to maintain the environmental flow. This has been the common practice in all the hydropower projects that are developed in Nepal. However, critics argue that the flat 10% minimum flow is not a true environmental flow and is most likely not enough during the dry period. The suggestion is to make the release more dynamic, reflecting the long-term natural variability of the river's flow.

For paddy's IWR, for the development, mid-season and late-season growth stages, irrigation water requirement is low as the period falls during the monsoon season. As the monsoon and post-monsoon precipitation is projected to increase, the IWR will either remain the same or decrease in both the mid-century and end-century for these stages. The highest increase in IWR is during the nursing and land preparation stage (May) as well as the initial growth stage, which is also the stage where the crop requires the most irrigation water as these stages fall during the pre-monsoon stage. For wheat's IWR, the projected decrease in winter precipitation and increase in mean temperature is the probable cause of increasing IWR. The higher percentage increase in IWR during the mid-season and late-season growth periods is primarily due to the relatively low irrigation demand during the reference period.

Similar studies in other countries – Bangladesh, China, Malaysia, and South Korea – report a similar increase in irrigation water requirement for rice in comparison with the reference period. Kaini et al. (2022) projected the irrigated water requirement for wheat in the Terai Region of the Koshi basin to increase by 3% by the end of the century for RCP4.5 scenarios and decrease by 8% for RCP8.5 scenarios. The differences in the projection are most likely to be due to spatial variability of the precipitation in the Koshi and the Karnali basins.

Our findings indicate a general increase of 6–10% in crop water demand under climate change scenarios. This additional demand is unlikely to pose a significant challenge to the irrigation system, given the intake capacity of 55 m³/s, which exceeds the required crop water demand. Moreover, the minimum flow in the Karnali River basin, at 300 m³/s, is well above the intake capacity of 55 m³/s. However, as indicated by Nepal et al. (2024), maintaining the intake slope during the dry season to facilitate river water diversion to the irrigation fields may present challenges, especially to the tail section of the irrigation scheme. Also, increased water stress on crops, particularly during the dry season, may lead to lower yields for staple crops like paddy and wheat. Higher evapotranspiration rates and shifting rainfall patterns could reduce soil moisture availability, forcing farmers to adjust sowing and harvesting periods, which may disrupt crop growth cycles. Additionally, water shortages may limit crop diversity, reducing dietary diversity and contributing to nutritional deficiencies. Addressing these challenges will require enhancing irrigation efficiency, adopting climate-resilient farming practices, and expanding intake capacity through improved irrigation infrastructure.

These changes in water availability will have significant implications for social, economic, and ecological systems within the basin. Reduced water availability in dry seasons could impact drinking water supply, increase drought and irrigation stress, and affect riverine ecosystems, while increased monsoon flow, driven by increased precipitation, will likely lead to more floods, soil erosion, and landslides in both mid- and end-century periods disrupting local economies and livelihoods. Shifting temperatures and precipitation may force rural-to-urban migration of local communities and wildlife species to move to higher altitudes, affecting biodiversity. The Karnali River exhibits shifting flow patterns within its braided channels, influenced by pronounced climate variability across its diverse topography and geographical features. Seasonal fluctuations further amplify these shifts, contributing to the complexity of flow regimes in this river system. This could affect water delivery to the irrigation system if there is a significant shift toward the eastern bank in the future. Natural and artificial water storage can be effective strategies for enhancing water security in regions like the Karnali River basin, where climate variability impacts river flow through seasonal fluctuations and altered precipitation and snowmelt patterns. These storage solutions help stabilize water supplies, provide drought resilience, reduce flood risks, and support hydroelectric power generation and groundwater recharge. However, large-scale storage projects can disrupt ecosystems and affect communities, necessitating careful environmental and social impact assessments. Integrated water resource management, aligned with agricultural, energy, and community needs, and regular capacity building in storage maintenance ensure sustainable and balanced use of resources while protecting the environment.

This paper assesses the impact of climate change on water balance components, hydropower production, and IWRs. The SWAT+ hydrological model was calibrated and validated using discharge data. Select global climate models, chosen for their accuracy in representing historical climate patterns, were downscaled to high resolution. Future climate data was applied to the validated model to project future hydrology and assess future energy production and IWRs. Key findings from this research include:

• The SWAT+ model provided valuable insights into current water balance components and future hydrological changes, showing that 34% of rainfall is lost to evaporation, with 46% of water yield available for various uses.

• Climate change will cause too much and too little water in the basin. Decreased rainfall during the winter and pre-monsoon seasons and increased rainfall during the monsoon and post-monsoon seasons will disrupt current water management practices. Floods and landslides will likely rise in the monsoon, while droughts may intensify in the dry season.

• Discharge variability may impact hydropower production, especially during dry periods. However, wet season energy is expected to increase, leading to a slight overall rise in annual energy production in the future.

• Future irrigation water demand is expected to rise, while this may not significantly impact the Rani Jamara Kulariya system due to its design discharge (50 m³/s) being lower than the minimum monthly flow (300 m³/s) in the Karnali basin. However, the Karnali River system is changing its river channel due to a high sediment load, which poses a risk that diverting toward the east side might cause water disturbances entering the irrigation systems through an integrated intake structure.

• The insights from CMIP6 climate projections – particularly the projected decrease in rainfall during the winter and pre-monsoon seasons and an increase during the monsoon and post-monsoon seasons – offer critical information for water resource planning and management in the basin. These patterns differ notably from earlier CMIP5-based projections, highlighting the importance of using updated climate data to inform more accurate and adaptive water management strategies.

This study offers a novel, integrated assessment of how climate change could impact water balance, hydropower production, and irrigation demand. It combines high-resolution climate projections with the SWAT+ hydrological model to provide a deeper understanding of the interconnected water–energy–food–ecosystem (WEFE) systems – something earlier studies often overlooked. The findings emphasize the need for cross-sector collaboration and offer practical insights for managing water resources more effectively in a changing climate. The research contributes valuable guidance for developing sustainable and resilient water management strategies by shedding light on emerging risks and complex interlinkages.

This foresight analysis of climate change's impact on water resources involves multiple uncertainties. Future temperature and precipitation projections are uncertain, mainly due to the realization of different socioeconomic development pathways which are also reflected in the future hydrological projections. Uncertainties also arise from input data quality, model validation, and simplifications in representing hydrological processes, compounded by the model's calibration only at the basin outlet, though upstream calibration could improve accuracy. While projections based on four selected climate models indicate increased flood and drought risks, the models may not fully capture extreme events. There is a possibility of policy changes in water and energy sectors or basin development in near future in this developing region that could significantly alter impacts, which are not fully accounted for. To address these uncertainties and improve confidence in results, further detailed studies, enhanced modeling approaches, and tradeoff analyses of competing water uses are recommended. Decisions based on these results should account for these potential uncertainties and limitations.

The study highlights climate change implications for water availability, energy production, and irrigation demand, all crucial to Nepal's water, energy, and food security, as outlined in the National Agricultural Policy (MoALD 2004), Water Resources Policy (MoEWRI 2020), and National Irrigation Policy (MoEWRI 2024). However, coordination and cooperation among these sectors are hindered by the dominance of sectoral silos in development discourse. In addition, the population is also likely to reach 38 million by 2065 (current population 28 million in 2023) (UN 2024), and there will be more competition for water, agriculture, and energy in the future. Therefore, an integrated, holistic approach considering the water, energy, food, and ecosystem (WEFE) nexus is essential. Climate change impacts are cross-sectoral and cooperation among the sectors is critical for effective mitigation and adaptation.

This work was carried out under the CGIAR NEXUS Gains Initiative, which is grateful for the support of CGIAR Trust Fund contributors: www.cgiar.org/funders. We thank Mr Arshad Ansari who supported us on the global model selection for the Karnali River basin. Thanks to Jonathan Lautze of IWMI for their critical input on the draft manuscript.

S.P. and S.N. conceptualized the study, wrote, reviewed, and edited the article. S.K. wrote, reviewed, and edited the article, and M.H. reviewed and edited the article. 

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.