Evaluating the potential impact of climate change on the hydrology of Ribb catchment, Lake Tana Basin, Ethiopia

GCM: General Circulation Model; HRUS: Hydrological Response Units; IPCC: Intergovernmental Panel on Climate change; RCPs: Representative Concentration Pathways SWAT: Soil and Water Analysis Tool; SWAT-CUP: SWAT-Calibration and Uncertainty program; ALPHA_BNK: Baseflow alpha factor for bank storage; CANMX: Maximum canopy storage; CN2: Initial SCS CN II value; BIOMIX: Efficiency of soil biological mix; ESCO: Soil evaporation compensation factor; REVAPMN: Threshold water depth in the shallow aquifer for ‘revap’; RCHRG_DP: Deep aquifer percolation fraction; GW DELAY: Groundwater delay; GWQMN: Threshold water depth in the shallow aquifer for flow; GW_REVAP: Groundwater ‘revap’ coefficient; SOL_AWC: Available water capacity.

Climate change refers to the change of climate variables, temperature, and precipitation. The average global temperature has risen by about 0.74 °C from 1906 to 2005 (IPCC 2007). This climate change happened in two phases, from the 1910 to 1940s and more strongly from the 1970s to 2005. An increase in temperature causes higher evapotranspiration and alters the global precipitation situation (Urrutia & Vuille 2009; Paparrizos et al. 2016). Increasing temperature also affects the occurrence, property, distribution, and movement of water resources and causes a higher frequency of hydrological events such as floods and droughts in the catchment level. A high concentration of greenhouse gases in the atmosphere causes increase in temperature and variability of rainfall distribution and amount (Houghton 2001). Rainfall and temperature (climate) change impacts surface hydrology and water resources which can change regional water balances and hydrological regimes.

The beginning of the 21st century is also the warmest (IPCC 2013). Global climate change impact will continue by increasing temperatures and variability of rainfall (Alexander et al. 2006; Zwiers & Wehner 2013). Climate change impact is the main issue in Africa, as the income of the continent depends on agriculture which is susceptible to climate change. The continent is one of the most impacted regions to climate change (IPCC 2007). Similarly, future climate change is shown to increase in Ethiopia. Conway and Schipper (2011) testified future climate change using medium to high Special Report on Emission Scenarios (SRES) of annual warming across Ethiopia could be 0.7–2.3 °C and 1.4–2.9 °C in the 2020 and 2050s respectively. Future climate change at the Lake Tana basin level showed variability of rainfall and an increase in temperature (Abdo et al. 2009; Gebremariame 2009; Adem et al. 2015).

IPCC reports showed the presence of climate change at the international level based on different scenarios data. The current report of IPCC, which is used for different impact assessment studies, is based on the Representative Concentration Pathways (RCPs) scenarios. It is the latest generation that provides input to climate models, which are time and space-dependent trajectories of concentrations of greenhouse gases and pollutants resulting from human activities. These scenarios provide a quantitative description of concentrations of climate change pollutants in the atmosphere over time, as well as their radioactive forcing in 2100. Four main scenarios have been defined with different targets of radioactive forcing. The RCP 8.5: It is considered increasing greenhouse gas emissions over time; will reach over 8.5 W/m² by 2100 and will continue to rise onwards after some time. RCP 6.0: total radioactive forcing is stabilized after 2100 without adopting any technologies and strategies to reduce greenhouse gas emissions. RCP 4.5: total radioactive forcing is stabilized before 2100 by applying a range of technologies and strategies to reduce greenhouse gas emissions. RCP 2.6: radioactive forcing peaks at approximately 3 W/m² before 2100 and then declines to approximately 2.6 W/m² in 2100.

Several empirical, physically based and conceptual models could be used for hydrological studies. Soil and Water Assessment Tool (SWAT) model, physically-based semi-distributed watershed model, uses modified Soil Conservation Service-Curve Number (SCS CN) method to estimate the runoff. This model is simple, user-friendly, used worldwide, and freely available. SWAT showed a good application performance in Ethiopia. It was calibrated and validated frequently in the Blue Nile basin and showed good performance (Kassa & Foerch 2007; Awulachew et al. 2008; Setegn et al. 2008; Shimelis 2008; Adamu 2013). Modeling of Gumara watershed in Lake Tana basin (Awulachew et al. 2008) indicated that streamflow and sediment yield simulated with SWAT was reasonably accurate.

The degree of climate change is spatially and temporally variable (Droogers & Aerts 2005). It is needed to study the impacts at a small catchment scale which has not been given much attention in developing countries like Ethiopia. Similarly, the previous studies in the Lake Tana sub-basin and also in Ethiopia mostly used Global Climate Models (GCMs) data of the third phase Coupled Model Intercomparison Project (CMIP3) downscaled to the local level based on A2 and B2 SRES. However, currently, available Representative Concentration Pathways (RCPs) are used for impact assessment and showed a better agreement to observed climate data (Fentaw et al. 2018). Estimation of future streamflow using projected changes of local-scale climate variables is necessary to properly plan for future climate change impact adoption and mitigation measures. Planning for sustainable utilization of water resources based on climate change impact assessment is critical for Ethiopia since the income of the country depends on agriculture which is highly vulnerable to climate change. Despite the significance of the Lake Tana basin for continuous settlement of the people, little has been done on future climate change impact assessment (Gebremariame 2009). Ribb River is among the main tributaries of Lake Tana, the origin of the Blue Nile. Blue Nile basin is the largest basin in Ethiopia with large population pressure, degraded land, and highly dependent on agricultural (Tsegaye 2006), and the Blue Nile (Abbay) River supplies water for many downstream African countries. The river is highly vulnerable to climate change which affects streamflow but the effects of climate change in water resource analysis and management at the Lake Tana sub-basin and the Blue Nile basin has not been adequately addressed (Setegn et al. 2011).

The main objective of this study was to estimate the potential impact of climate change on the water flow of Ribb River with the specific objectives: (1) Representing catchment hydrology through calibration and validation of Soil and Water Assessment Tool (SWAT); (2) Quantifying the potential impact of climate change on future streamflow relative to the present situation based on RCP4.5 and RCP8.5 scenarios data.

Ribb catchment is located in the Lake Tana basin, Ethiopia. The catchment stretches between 11°40′0″–12°10′0″N latitude and 37°30′0″–38°14′0″E (Figure 1). The total area of the catchment (gauged part) is about 131, 250 ha. Ribb River originates from the higher elevation of Guna Mountains, Southern Gondar. The catchment supports the total number of the population about 888,260 based on Statistical, (2012), and about 88.5% of the population mainly engaged in agriculture.

The elevation of a catchment developed from ArcGIS by using 30 m × 30 m resolution DEM; higher elevation is found at Guna Mountains, the southeastern part of the catchment, and also north and northeastern tips (Figure 2 left). The lower elevation is found at the western part of the catchment around Fogera plain. The catchment has also various slope features (Figure 2 right). The dominate slope is from 15–30% which covers 30% of the total area, about 22.5% of the area has a slope greater than 30% and only 8% has a slope of 0% to 3%.

Major soil types found in the catchment were Alisols, Luvisols, Nitosols, Vertisols, Regosols, Cambisols, Fluvisols, and Leptosols (Figure 3 left). Land use is changed to agricultural land due to rapid population growth. Since the catchment currently supports a large population, the dominant land use is for agriculture which covers about 95.5% (Figure 3 right).

The temperature of the catchment varies spatially and temporally. The mean annual maximum and minimum observed temperatures were about 24 °C and 8.77 °C respectively. The mean monthly maximum highest temperature record was found in March. April, February, October, November, and December also showed higher temperature records. The lowest maximum temperature was recorded in August and July. In monthly minimum temperature, the highest temperature was recorded in May and July and the lower record was found in December, November, and January.

Rainfall of the catchment also varies spatially, temporally, and seasonally. The mean annual rainfall of the catchment was about 1,350 mm per year. The mean monthly rainfall distribution showed maximum rainfall in July and August. The daily average discharge of the river was about 14.75 m³/s, peak discharge of 160 m³/s and minimum flow 0 m³/s. The mean monthly flow is highest in August.

Study results obtained at Lake Tana sub-basin and the Blue Nile basins done by Conway (2000), Kebede et al. (2006), Sutcliffe & Parks (1999), and Tarekegn & Tadege (2005) indicated the hydrological year of the study areas are characterized by one main rainy season (summer) between June to September, in which 70% to 90% of the annual total rainfall occurs.

The observed meteorological data were rainfall, maximum and minimum temperatures, wind speed, relative humidity, sunshine hours, and solar radiation. These meteorological data were collected from the Ethiopian Meteorological Agency, Bahir Dar branches. However, only Debre Tabor meteorological station is found inside the catchment with the required quantity and quality of data. Taking only one meteorological station data may not rightly represent the entire area due to the diverse topographic feature of the study area. To alleviate this problem, additional meteorological data were collected from the surrounding stations: Addis Zemen, Amed Ber, Yifag, and Agere Genet meteorological stations (Table 1). Debre Tabor is the only principal station that has all the required data. Solar radiation was prepared from daytime sunshine hour data using the Angstrom method. The other model input data, the weather generator which is used to provide data for stations having missing data, was prepared manually by using PCPSTAT and dewPOINT2.0 version programs on 21-year rainfall, temperature, and wind data obtained from Debre Tabor meteorological station.

Digital Elevation Model (DEM), land use, and soil map data were spatial data used for this study. These data were obtained from Amhara Design and Supervision Works Enterprise (ADSWE) office. The observed meteorological and spatial data were used as input to SWAT to see its performance to represent the catchment hydrology. The observed discharge (flow) data of the river, collected from Amhara Water, Irrigation and Energy office, Ethiopia, was used for sensitivity analysis, calibration, and validation of the model.

The projected rainfall and temperature data at RCP4.5 and RCP8.5 scenarios were used to simulate the future discharge of the river. Nile Basin Initiative partnership prepared climate change projections data set for the Nile basin studies. It was established for the aims of the provision of well-vetted climate change projection data at an appropriate spatial and temporal resolution for key target end-users (Asefa 2019). Historical (1976–2005) and projected (2011–2098) temperature and rainfall data from SMHI-RCA4 Regional Climate Model (RCM) by 0.25-degree of resolution which derived from the MPI-M-MPI-ESM-LR Global Climate Model (GCM) were used for this study (Table 2).

Data quality assessment, such as filling the missed data, data consistency checking, and data homogeneity and trend analysis, is needed to perform hydrological studies. The presence of missing data is a common and serious problem and producing bias in study results. No matter how much the data measuring system is well designed, the data recorder is professional and instruments are accurate; there are missing meteorological and hydrological data. Similarly, using inaccurate missing data filling methods also affects the accuracy of the study results. To perform a good analysis and simulation by using long time series continuous data, handling this problem by using the accepted missing data filling method is a pre-request. Many methods are employed to fill in missing data. Arithmetic mean, Normal ratio, Regression, and distance power methods are commonly used. Choosing an appropriate method depends on method simplicity to apply, length of missing data, and accuracy. The normal ration method was used based on the upper criteria.

The data homogeneity test was checked by Addinsoft‘s XLSTAT2016. It was evaluated based on P values estimated by the Monte Carlo method at a 95% confidence interval. As computed p-value is greater than alpha value, 0.05, it indicates data homogeneity. If a computed P-value is less than the Alpha value, it indicates the data is not homogeneous. Trend analyses of the observed data were evaluated by using the Mann-Kendall trend test in XLSTAT2016. The algorithm of this method is when computed P-value is greater than Alpha value; data has no significant trend (Arun & Sananda Kundu 2012) and vice versa. The data consistency was checked by the double mass curve method.

The meteorological and hydrological data were prepared based on the SWAT input data format by Mintab18. The meteorological data from 1988 to 2015 were used for SWAT modeling and from 1976 to 2005 for baseline period data to generate the streamflow which is used as a reference. The observed discharge data from 1990 to 2013 (1990–2005 for model calibration and 2006–2013 to model validation) was used with 2 years model warm-up period.

SWAT consists of sets of processes including project setup, watershed delineation, hydrologic response unit (HRU) definition, writing up the input table, editing SWAT input (if necessary), and SWAT simulation to get the final output. Watershed delineation was done by watershed delineator tool in ArcSWAT interface on ArcGIS by using Digital Elevation Model (DEM) as input. A watershed is delineated into many sub-basins, which are further divided into Hydrological Response Units (HRUs) that consist of homogeneous land use, slope, and soil characteristics. A detailed description of telling how SWAT is working is found in the user manual. SWAT was run by monthly data.

Many parameters affect the hydrologic system of a specific catchment but all parameters do not have equal effects. Therefore, sensitivity analysis was carried out to determine the number of optimized parameters to obtain a good fit between SWAT simulated and observed streamflow data.

Sensitivity analysis helps to know the relative ranking of which parameters mostly affect the output (Griensven et al. 2006), it reduces the uncertainty of model outputs and provides parameter estimation guidance for model calibration and validation. It was undertaken in SWAT-CUP by using the global sensitivity design method. Sensitivity analyses were conducted by using 23 flow parameters (Table 3) which were selected based on the reviewing different articles which were done at the Blue Nile and specifically Lake Tana sub-basin.

SWAT-CUP uses t-test and p-value to rank sensitive parameters that correspond to a greater change in output response. A t-stat provides a measure of sensitivity (larger in the absolute value of t-stat are more sensitive) and p-value determined significance of sensitivity, a value close to zero has more significance. Eleven most sensitive parameters were selected and model calibration and validation were conducted. SWAT_CUP using the SUFI_2 algorithm is commonly used in the Lake Tana sub-basin and Blue Nile system and showed the best-fitted result for Gilgel Abay, Gumara, Ribb, and Megech catchments (Setegn et al. 2009) for model calibration and validation. This algorithm was used for this study.

The level of goodness to fit of models has been evaluated by an objective function that measures the level of agreement between observed and model-simulated output. Model Performance was evaluated by using Nash-Sutcliffe efficiency (⁠⁠) and Coefficient of determination (R²) (Table 4).

Future streamflow of the river was simulated by using the SWAT model after verifying the model performance by the observed data. The Nile Basin Initiative partnership projected rainfall and temperature data set for Nile basin studies were used as input to the model. This data set was developed based on the recently reported Representative Concentration Pathway scenarios. These projected data have a finer resolution but it was further downscaled to the nearest meteorological stations observed data. The downscaling was done by the linear scaling bias correction method. The future streamflow was simulated on three time horizons: 2020s (2011–2040), 2050s (2041–2070), 2080s (2071–2098), and baseline period (1976–2005) as a reference. The model simulated discharge was used to analyze the effect of future climate change on the streamflow of the river on annual, monthly, seasonal, total volume, individual months, and on mean maximum and minimum discharge (Figure 4).

The data quality checking was done and showed good results in all used parameters. SWAT sensitivity analysis on six iterations; 50, 300, 500, 800, 1,000, and 1,200 number of monthly simulations using 15 years (1990–2004) streamflow data were done and accepted which is 0.74 and R², and = 0.79, were obtained. Then eleven sensitive parameters, namely: GW_DELAY, CN2, GWQMN, GW_REVAP, RCHRG_DP, CANMX, ALPHA_BNK, REVAPMN, REVAPMN, SOL_AWC, ESCO, BIOMIX were selected based on t-test and P-value as shown in Figure 5.

Calibration helps to improve the result of model simulation and validation for result verification. Model calibration on four iterations; 100, 250, 400, 500, and 600 numbers of simulations provide acceptable results (R² = 0.83 and = 0.72). Model validation was also provided a very good result 0.71 and R² = 0.72) after four iterations with 100, 200, 300 and 400 simulations (Figure 6).

However, many research findings at Blue Nile Bain level obtained better and R² results during model calibration and validation. It is due to the Ribb River overflowing in the rainy season (SMEC 2007) and a traditional flow measuring instrument was used until seven years ago which may have inaccurate recorded flow data. However, the calibration and validation results showed SWAT can estimate the discharge of the river to see the climate change impact on the streamflow.

Monthly streamflow of the river was shown to decrease/fall from March to August and started to slightly rise/increase from September to February, from the baseline to 2080s, based on both RCP4.5 and RCP8. 5 scenarios data as shown in Figures 7 and 8.

The streamflow change is a function of temperature and rainfall changes. The above result showed decreasing streamflow from March to August may be due to temperature and rainfall change. The monthly temperature trend analysis of the catchment showed a high increase of maximum temperature from March to July but comparatively lower changes in the remaining months as is shown in Figure 9.

However, the trend analysis of future monthly rainfall change showed a decreasing trend from March to July (except a slight rise in April based on RCP4.5 scenario at the three time horizons and RCP 8.5 scenario at 20250s) and showed more or less a rise from August onwards, based on RCP 4.5 and RCP 8.5 (Figure 10).

The study conducted by DeBoor (2007) at upper Nile basins showed a decrease of rainfall in May, June, and July whereas it was found to increase in September, October, and November which is more or less similar to this study result. The same result is obtained from the study conducted at Lake Tana (Abdo et al. 2009). This decrease of rainfall and increase of temperature leads to a decrease in the streamflow of the river because increasing temperature causes high evapotranspiration which leads to a decrease of the streamflow.

Concerning individual months streamflow change (Figure 11), June could show the highest magnitude of decrease (63.3% and 55.45% at RCP8.5 and RCP4.5 scenarios respectively) from baseline to the 2080s. A higher magnitude of decrease of streamflow would also observe in May (24.38% and 58.64% at RCP4.5 and RCP8.5 scenarios respectively) and July (22.31% and 51.33% at RCP4.5 and RCP8.5 scenarios respectively). The minimum magnitude of decrease would observe in August based on the RCP8.5 scenario which is found to be 6.64% and April (1.21%) based on the RCP4.5 scenario. Streamflow showed a comparatively maximum increase in October (15.87% and 14.66% at RCP 4.5 and RCP8.5 scenarios respectively). The minimum increase would observe in September (6.06% and 4.01% at RCP4.5 and RCP8.5 scenarios respectively).

Percentage change of mean monthly streamflow was obtained by multiplying the differences between the 2080s and baseline period by 100.

Based on the changes of maximum temperature and rainfall Figures 9 and 10, the highest magnitude of the decrease is expected to happen in July but it is in June. It may be due to the streamflow is affected by other factors such as the infiltration rate of the soil, land preparation, and the crop cover of the area. The soil infiltration rate increases at beginning of rain and gradually decreases when the rain continued. In the study area, rainfall mostly started in May and continued to rain up to the end of September/October. So, the result of the study may be due to the higher infiltration rate on the first rainy month cause for less runoff formation. Similarly, in the study area, land preparation for cultivation is mostly carried out in May and June. So, it may cause low runoff formation due to the high infiltration rate.

In total volume reduction (Figure 12), obtained by deducting baseline streamflow from streamflow obtained in the 2080s, July is the most impacted month in both scenarios. The total volume of water reduced in July is found to be 21.08 cubic meters per second and 51.22 cubic meters per second at RCP4.5 and RCP8.5 scenarios respectively. It is due to the maximum temperature increment and highest rainfall decrement in July. It may also be due to the total volume of the river in the baseline period is highest compared with other months. The maximum increase happen in October which is found to be 10.31 and 11.26 cubic meters per second at RCP4.5 and RCP8.5 scenarios respectively.

Concerning seasonal streamflow change of the river in the 2080s from the baseline period, it would decrease in summer (June, July, and August) and spring (March, April, and May) seasons. But, it would increase in the autumn (September, October, and November) and winter (December, January, and February) seasons of the country in both scenarios (Figure 13). The highest magnitude of the decrease is shown in spring based on the RCP8.5 scenario and summer based on the RCP4.5 scenario. The highest magnitude of increase should be seen in autumn rather than winter in both scenarios.

Streamflow would decrease by 69.87% and 30.13% in summer and spring seasons, respectively, based on the RCP4.5 scenario whereas 43.82% in summer and 56.18% in spring based on the RCP8.5 scenario in the 2080s from the current situation. Streamflow would increase by 61.29% and 38.71% in autumn and winter season respectively based on the RCP4.5 scenario whereas 55.19% in autumn and 44.81% in winter based on the RCP8.5 scenario in the 2080s from the baseline period.

Rivers in Ethiopia are mostly used for agricultural purposes in spring, winter, and autumn but rainfed agriculture is commonly adopted in summer. Future water resources planning and implementation for irrigation projects at Ribb River are advisable in autumn and winter since a slight increase of streamflow in the future allows production more than the current situations. In spring and summer, streamflow could decrease that shall construct reservoirs and other structures in the cause of different projects are planning to implement at the river based on current streamflow information.

Planning and implementing of projects at rivers shall be based on future maximum (peak) and minimum streamflow to increasing the life span, sustainability, and profitability of projects. Maximum streamflow may cause distraction of constructed structures if the constructions do not consider the peak discharge. It may also cause the construction of large structures which leads to unnecessary cost of construction. Knowing future minimum streamflow is also helpful to know the nature and type of projects to implement in the future.

Monthly mean maximum streamflow showed a decreasing trend from January to July and October whereas increasing in August, November, and December in both scenarios from baseline to 2080s. However, there is not the same result at RCP4.5 and RCP8.5 scenarios in September (Figure 14). The highest peak monthly mean streamflow is observed in August therefore structures constructed in the future should give consideration to the fact that peak flow could increase above that shown today. The same is true in November.

Besides, monthly mean minimum stream flow could change through time (Figure 15). The mean monthly minimum streamflow of the river could not show a significant change from January to May based on both scenarios. It showed a slightly increasing trend in September which could increase by 6.44 m³/s and 2.65 m³/s at RCP4.5 and RCP8.5 scenarios respectively. However, it would show a decreasing trend in July, August, October, November, and December at both scenarios. The maximum magnitude of the decrease is found in July and August which is 24.70 m³/s at RCP4.5 and 35.19 m³/s at RCP8.5 scenarios in July and by 22.45 m³/s at RCP4.5 and 53.30 m³/s at RC8.5 scenarios in August.

On the annual streamflow basis, the mean annual streamflow of the river was 42.78 m³/s on the baseline period. It could be 40.75 m³/s, 40.83 m³/s, and 40.24 m³/s at the RCP4.5 scenario and 41.73 m³/s, 43.42 m³/s, and 37.58 m³/s at RCP8.5 scenario in the 2020s, 2050s, and 2080s respectively. This showed a decreasing trend of streamflow from baseline to the 2080s by 2.54 m³/s and 5.2 m³/s at RCP4.5 and 8.5 scenarios. The simulated streamflow in the 2080s is lower based on RCP8.5 scenario data than RCP4.5. It is due to RCP8.5 is increasing and RCP4.5 is stabilization scenarios.

The reduction of the streamflow from the baseline to the 2080s is due to increasing temperatures and decreasing rainfall (Figures 16a and b).

When looking at the changes between the consecutive time horizons (from the baseline to 2020s, the 2020–2050s, and the 2050–2080s), streamflow would not show a consistent trend. Streamflow would decrease from the baseline to the 2020s but increased from the 2020s to the 2050s then decreased again from the 2050s to the 2080s on both scenarios (Figure 17(a)).

The changes from baseline to each time horizon (the 2020s, 2050s, and 2080s) the streamflow becomes decreased in all scenarios but increased from baseline to 2050 based on the RCP8.5 scenario (Figure 17(b)). It seems out of logic to see the different results on RCP 4.5 and RCP 8.5. It may be due to climate projected data used which comes from a single model. So, it is recommended to see other models for future studies.

The annual average maximum streamflow of the river showed a decreasing trend from the baseline to the 2080s based on both scenarios simulation results (Figure 18). It was 294.60 m³/s on the current situation and would be 234.8 m³/s, 269.5 m³/s, and 236.8 m³/s at the RCP4.5 scenario and 288.1 m³/s, 272.1 m³/s, and 240.5 m³/s at RCP8.5 scenario in the 2020s, 2050s, and 2080s. Therefore, the magnitude of annual peak streamflow showed a decreasing trend between the consecutive time horizons at both scenarios except showed a slightly increasing trend from the 2020 to 2050s at the RCP4.5 scenario.

Mean annual minimum streamflow also showed a decreasing trend from the baseline to the 2080s in both scenarios. It would also show a decreasing trend between consecutive time horizons except would show a slight increasing trend from the 2050s to the 2080s at the RCP4.5 scenario (Figure 19).

This study evaluates the impacts of climate change on the Ribb River streamflow using the bias-corrected output of climate change projection data set for Nile Basin studies and the hydrological model simulation approach of the SWAT model. So the study reached the following conclusions.

SWAT calibration and validation were verified that the model can effectively estimate future streamflow of the river. The monthly streamflow of the river was used to calibrate and validate the model using performance measuring parameters of the coefficient of determination (R²) and Nash Sutcliff (NSE). R² and NSE values were 0.72 and 0.83 for calibration and 0.71 and 0.72 for validation, respectively, which are deemed acceptable.

Climate change projections of RCM data set for Nile Basin studies were found to be biased and the dataset was bias corrected before it was used. A linear scaling method of bias correction was used for this study and was found to be effective in matching model data with observed data on a monthly time scale.

RCP4.5 and RCP 8.5 scenarios resulted in a somewhat similar projection of future streamflow conditions. The streamflow changes were looked at monthly, seasonal, and annual time scales. Annual streamflow showed a decreasing trend from the baseline to the 2080s. The simulated mean annual streamflow was 42.78 m³/s for the baseline period whereas it became 40.75 m³/s, 40.83 m³/s, and 40.24 m³/s at the RCP4.5 scenario and 41.73 m³/s, 43.42 m³/s, and 37.58 at the RCP8.5 scenario for the corresponding 2020s, 2050s, and 2080s time horizons, respectively. It showed a consistent decrease from the baseline to the 2080s but a slight rise and fall between successive time horizons.

Based on the monthly simulated flow increasing trend is shown from September until February and a decreasing trend from January until August. The maximum increase in magnitude is in October that showed an increase of 10.31 and 11.26 cubic meters per second in RCP4.5 and RCP8.5 scenarios, respectively. Whereas the maximum decrease was for July which is shown to be reduced by 21.08 and 51.22 cubic meters per second at RCP4.5 and RCP8.5 scenarios, respectively.

Based on the seasonal change in river flow, summer flows are shown to have a potential decrease by 69.87% and spring by 30.13% according to the RCP4.5 scenario whereas a decrease by 43.82% in summer and by 56.18% in spring is shown based on the RCP8.5 scenario from baseline to the 2080s. However, streamflow could increase by 61.29% and 38.71% in autumn and winter season respectively based on the RCP4.5 scenario and by 55.19% in autumn and 44.81% in winter based on the RCP8.5 scenario.

Monthly peak streamflow from the baseline to 2080s shows a decreasing trend from January to July and October and an increasing trend in August, November, and December in both scenarios. The minimum streamflow did not show a significant change from January to May in both scenarios. The slightly increasing trend of minimum streamflow in September was found (6.44 and 2.65 m³/s on RCP4.5 and RCP8.5 scenarios, respectively). However, it showed a decreasing trend from July to December in both scenarios. The maximum decrease in magnitude was in July and August. The annual maximum streamflow showed a decrease from baseline to 2080s in both scenarios. The mean annual minimum streamflow change also shows a decreasing trend in the future. Besides, both monthly maximum and minimum streamflow change have a decreasing trend in the future.

A result reported in this study is based on a single RCM and two emission scenarios. The generalization of these findings may have to be confirmed by looking at additional RCM scenarios. This will also allow an understanding of uncertainties associated with General Circulation Models for the study area.

SWAT is found to be a very important tool for simulating historical and future impacted hydrological processes at a watershed scale. However, in addition to climate change, future land-use change will also play a critical role in streamflow generation. The impact of potential future land-use change on streamflow for the study area was not within the scope of this study. Therefore, future studies should include both climate change impacts and land-use changes to have a complete picture of future stream flows.

Besides, there are also other factors affecting the streamflow generation such as soil type, evapotranspiration, solar radiation, etc., that may also be impacted by climate change and were not explicitly included. Hence, the results of this study should be taken with care and be considered as an indication for further studies than accurate predictions.

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