Response of climate change impact on streamflow: the case of the Upper Awash sub-basin, Ethiopia

Climate change refers to a change in the state of the climate that can be identified by changes in the mean and variability of its properties and that persists for an extended period, typically for a decade or more (IPCC 2007). Climate may be changed by internal processes such as changes in solar radiation and/or by external forcing within the climate system (IPCC 2007, 2014). Currently, the effect of climate change is the major global concern as is evidenced by the increase of the greenhouse gases concentration in the atmosphere (Seneviratne et al. 2012; IPCC 2013). Based on different climate models and emission scenarios, the temperature is expected to increase continuously while precipitation shows mixed change (IPCC 2014; Negash et al. 2020). The IPCC (2013) also suggested that the global mean air temperature will likely increase, ranging from 1.4 to 4.8 °C, under all emission scenarios by the end of the 21st century. The high latitudes regions, the equatorial Pacific and the mid-latitudes of wet regions will probably experience an increase in precipitation under the RCP8.5 scenario while in the mid-latitudes and sub-tropical dry regions it is expected to decrease (Elena et al. 2019). Climate change will have a profound effect on streamflow availability and variability due to frequent changes in precipitation and temperature patterns which are now becoming the research areas that need attention (IPCC 2007; Babur et al. 2016; Tadese et al. 2019). Precipitation and temperature are highly affected by climate change among the components of the hydrological cycle that alter the temporal and spatial availability of streamflow magnitude, variability and timing of occurrence (Babur et al. 2016). Regional hydrology is highly sensitive to changing climate conditions that make climate change projections essential to assess the likely availability of water resources to meet future demands (Brekke et al. 2009; Claudia 2012; Negash et al. 2020). Understanding the climate change effect enhances the flexibility and has net benefits in water resources management (IPCC 2014). The uncertainty of the availability of streamflow in the future could affect agricultural production, socioeconomic systems and threaten environmental sustainability (Daba et al. 2017; Elena et al. 2019). The impact of climate change has a significunt effect on developing countries, especially those whose economy is dependent on agriculture (IPCC 2014). Ethiopia's economy is directly subjected to climate change effects as a large portion of the land is arid and semi-arid and is dependent on rainfall. Hence it is practical to understand the effect of climate change on future streamflow changes to adapt different mitigation strategies. Numerous studies have been conducted on the impact of climate change on potential availability of water resources using different climate secenarios and climate models, for instance Babel et al. (2014), Mulligan (2015), Wanders & van Lanen (2015), Babur et al. (2016), Daba et al. (2017) and Gizaw et al. 2017). Many of them used Special Report Emmission Scenarios (SRES) linked with GCMs outout in different regions, including the present study area. The outcomes of these studies suggested that the projected temperature shows continuously increasing trends under all emissions while the precipitation and respective streamflow show mixed patterns (increasing and decreasing). Currently, SRES are known to be old emission scenarios and may not possess the present climate conditions (Höök et al. 2010). To overcome the shortcomings of the SRES scenarios, new emission scenarios (representative concentration pathways, RCPs) have been developed by integrating technological enhancement, economic development, demography, policy and future institutional challenges and adaptation and mitigation linked to exclusive socioeconomic assumptions which shows the present and future climate conditions (IPCC 2007; Getahun 2015). The RCPs allow better resolution over SRES which helps in performing local and regional comprehensive studies (Höök et al. 2010). In spite of all these advantages, no study has so far been conducted in the Upper Awash by combining the new emission scenarios (RCPs) with a high resolution regional climate model (RCM). Hence, investigation of the response of climate change impact on future streamflow availability using the new emission scenarios (RCPs) incorporated in a regional climate model (RCM) is essential. The main scientific question to be addressed in this work is whether the streamflow in Upper Awash sub-basin could be affected by climate change and how much the change could be. A common approach to achieve this goal is to use hydrological models with scenarios provided by climate model outputs downscaled at catchment level (Daba et al. 2017; Negash et al. 2020). To minimize the uncertainties that may occur in climate projections, the application of bias correction is very important for better hydrological simulation (Claudia 2012; Fanget et al. 2015; Min et al. 2018). Thus, the intent of this study was to investigate the potential response of climate change on streamflow in the future. This finding will be useful to plan, design and apply suitable water management and mitigation strategies to minimize the impact of the change.

Awash River basin is one of the 12 Ethiopian River Basins which drains though the central and eastern highlands part of the country, covering a total catchment area of about 110,000 km². The river starts from Ginchi town, west of Addis Ababa, travels though Rift Valley, and ends in Lake Abe on the border between Ethiopia and Djibouti. The basin altitude ranges from 250 to 3,000 m above mean sea level. The total average annual rainfall ranges from 160 mm in the lowlands to 1,600 mm in the highlands. According to the Awash Basin Authority (2017) report, the basin's total annual water demand for irrigation, domestic water supply, livestock and industry is estimated to be around 3.4 BMC (billion metric cube). Based on the agricultural activities, socio-economic system, climatological, physical and water resources characteristics, Awash River basin is divided into upper valley, middle valley and lower valley (Edossa et al. 2010). The Upper Awash sub-basin (study area) is one of the most highly populated sub-basins in the western highland part of the Awash River basin, including the capital city of Ethiopia. Large mechanized and private irrigated agricultural farms and rapidly expanding industries are located in this sub-basin. The increasing rate of population together with other factors will intensify the water consumption rates in the future (Edossa et al. 2010; Tadese et al. 2019). Hence, investigation of streamflow changes in the future in response to climate change is economically and environmentally important for the development of the sub-basin as well as the whole country. The geographical location lies between latitude 8°10′57″–9°13′54″N and longitude 37°57′–39°11′E. The total area of the sub-basin is about 11,232 km² (Figure 1).

The daily meteorological data, precipitation, and maximum and minimum temperatures from 1996 to 2015 were collected from the Ethiopian National Meteorological Service Agency (NMSA). The streamflow from 1996 to 2015 for the selected stations were collected from the Ethiopian Ministry of Water, Irrigation and Energy office (MoWIE). The meteorological and hydrological stations were selected based on the standards of the length of record years, completeness (or fewer missing records) and the extent of the catchment network. Hence six rainfall and temperature and three streamflow gauging stations were used for this work. Sometimes, the collected data from both meteorological hydrological gauging stations may have missing or inaccurate records (Daba et al. 2017). In those cases, to improve the quality of the collected data, we applied different data quality test mechanisms such as filling missing data, homogeneity tests and consistency. Predominantly, the rainfall pattern is unimodal with the main rain occurring from June to September and the other months are mainly dry except for small rains from March to May. The total average annual rainfall is about 1,120 mm. The annual average temperature varies from 4.8 to 31.4 °C (Figure 2). Koka dam streamflow gauging station, which is located at the outlet of the sub-basin, was used for calibration and validation of the HBV hydrological model (Figure 3). A digital elevation map (DEM) 30×30 m resolution was obtained from the Ethiopian Ministry of Water Irrigation and Energy Bureau Department of GIS and processed in Arc-GIS to extract the drainage area and drainage network. Table 1 shows the meteorological and streamflow gauging stations used for this study.

The statistical value Z determines the significance of trend at a selected significance level. For this study the 5% significance level was selected. The slope of n pairs of data points was estimated using the Sen's slope estimator (Equation (20)):

T_i is a Sen's estimator of slope which is the median of these N values. T indicates the increasing or decreasing values in trends. The whole methodology and procedures are summarized in the flowchart as shown in Figure 5.

One of the objectives of this study is to evaluate the quality of selected climate model to reproduce precipitation and temperature in the baseline period (1996–2015). Accordingly, Figure 6 shows the average monthly precipitation cycle and percentage change of RCM from that recorded in the baseline period. The result indicates that the precipitation cycle from RCM is overlapped with the recorded precipitation in all months except from the middle of June to September (Figure 6(a)). However, after bias correction the differences were significantly minimized. The difference before bias correction varied from –64.97 to +78.65%. However, this difference reduced to –6.45 to +8.72% after bias correction (Figure 6(b)). Similarly, Figure 7 depicts the average monthly RCM temperature cycle and its change compared to recorded temperature. The model totally underestimated both minimum and maximum temperatures. Nevertheless, after bias-correction, the shape of the simulated and observed graph overlapped well. The changes range from –46.6 to –14.1% before bias correction and from –3.2 to 3.5% after bias correction for maximum temperature. Bias correction significantly improved the quality of climate model prediction which may in turn minimize the uncertainties that may occur in hydrological simulation. Similar results were depicted by Fanget et al. (2015) and Ahmed et al. (2017). The correlation between observed rainfall and RCM rainfall, shown in Figure 8, indicates that strong correlation exists after an application of bias correction.

Table 3 provides a summary of statistical parameters to evaluate the performance of the climate model (RCM).

Another specific objective of this study was to analyze the projected changes in precipitation and temperature based on the RCP4.5 and 8.5 scenarios for the period of 2021–2040 and 2041–2060 compared to the baseline period 1996–2015. Figure 9 depicts the general pattern of average monthly projected rainfall. The projected rainfall for both scenarios is expected to increase in the months of February, June, July, August and November, while in March, April, May, October, September, December and January it is expected to decline. Relatively, in summer (June–August) rainfall is projected to increase more than the other months (Figure 9(a) and 9(b)). Figure 10 shows the percentage change in precipitation between the reference period (1996–2015) and projected periods of 2021–2040 and 2041–2060. Overall, average annual rainfall is projected to increase by 11.97 and 20.67% under RCP4.5 and 8.5, respectively. For the period of 2021–2040, the projected rainfall is expected to increase to 3.22% under RCP4.5 and 16.21% under RCP8.5. The maximum increment will be expected in November (69.51%) under RCP4.5 and in February (66.6%) under RCP8.5. The maximum reduction will be expected in March under RCP4.5 and in April under RCP8.5 (Figure 10(b)). Likewise, for the period of 2041–2060, mean annual rainfall is projected to increase by 20.72 and 25.13% under RCP4.5 and 8.5, respectively. Relatively, in this time period, the rainfall increment will be high compared to the period 2021–2040. The maximum increment in this time period will be expected every November and the maximum reduction will be in March under both scenarios. Generally, the result clearly indicates that for both future periods and under both scenarios, the projected mean annual rainfall is expected to increase over the study area. However, the increment will vary among the scenarios and periods. The IPCC (2013) also suggested that, in the case of increasing temperature, the mean annual precipitation is expected to increase up to 20%.

Figure 11 shows monthly projected average temperature cycle and percentage change under both scenarios. As can be seen under both scenarios, both maximum and minimum temperature is expected to increase. On average, the projected minimum temperature is expected to increase by 9.6 and 11.7% under RCP4.5 and 8.5, respectively. Maximum temperature change could be +6.3% under RCP4.5 and +7.6% under RCP8.5. Moreover, the increment of minimum temperature is relatively high compared to the maximum. The results clearly show that the projected air temperature under RCP8.5 is higher than that of under RCP4.5 which confirms that RCP8.5 is a higher greenhouse carbon concentration emission scenario with a higher degree of global warming. The forecasted temperature in both the near and far future period is within the range forecast by IPCC (2007), i.e. average temperature increases from +1.4 to +20% at the end of the 21st century.

Table 4 shows the monthly average change in percentage of precipitation, maximum and minimum temperature for the periods 2021–2040 and 2041–2060 compared to the baseline period (1996–2015). The result indicated that the temperature and precipitation increment is higher for the far future period (2041–2060) than the near future period (2020–2040) under both scenarios. The percentage increment of the maximum temperature would be less than that of minimum temperature under both scenarios. In general, both the projected temperature and precipitation would increase for the study area.

The HBV model calibration and validation was carried out manually by optimizing the model parameters in each subroutine that has a significant effect on the model performance. Based on this, several runs were performed to select the most optimum parameter sets in order to compare the observed to simulated hydrograph. The results of each run were evaluated based on: visual inspection by comparing the simulated and observed hydrograph, NSE, P_BIAS and R². The calibration and validation was conducted in monthly time series. Figure 12 shows monthly recorded and simulated runoff cycles for the calibration and validation period over the reference period. The shape of the simulated hydrograph is well captured by the shape of recorded flow using the optimized model parameters during the calibration period (2002–2010) and validation period (2011–2015). However, some of the peak and low flows were underestimated and overestimate recorded flow. All performance evaluation variables are within acceptable ranges for both calibration and validation periods, as displayed in Figure 12. The most sensitive parameters and their optimum values sets during the calibration and validation periods are shown in Table 5.

One of the key objectives of this study was to evaluate simulated streamflow under RCP4.5 and 8.5 scenarios for two consecutive periods (2021–2040 and 2041–2060) compared to the baseline period (1996–2015). Thus, the time series of bias corrected precipitation, temperature and estimated evapotranspiration of the periods 2021–2040 and 2041–2060 under both RCPs were used as input for streamflow simulation. The optimized model parameters during calibration and validation were transferred with the intention to simulate future streamflow to investigate the changes from the baseline period. Projected streamflow in June, July and August is expected to increase more rapidly than the rest of the months under both scenarios (Figure 14(a) and 14(b)). Overall, streamflow is expected to increase under both scenarios and time periods (Figure 13(c) and 13(d)). On average, annual streamflow is expected to increase 28.5% under RCP4.5 and 23.9% under RCP8.5. In the near future period (i.e. 2021–2040), it is expected to increase up to 40.1% under RCP4.5 and 29.4% under RCP8.5. Likewise, for the far future period (2041–2060) it will increase to 18.5 and 16.9% under RCP4.5 and 8.5, respectively. However, for this period there will be a slight decrease in streamflow compared to the near future yet it will relatively increase compared to the reference period. Most of the change in streamflow is in line with the changes in rainfall and temperature (see also Figure 10). The maximum increment will be expected in October under both scenarios while the maximum reduction will be expected in March under RCP4.5 and in January under RCP8.5. The increment suggested by the RCP4.5 scenario is more marked than under RCP8.5. This is due to the fact that increasing temperature under RCP8.5 increases evaporation which would lead to a reduction in streamflow. The projected seasonal change of streamflow from the baseline period was computed for winter, spring, summer and autumn seasons. In winter (DJF) and spring (MAM) the streamflow is expected to decrease except for the spring season under RCP4.5 for the period 2041–2060. Similarly, in summer (JJA) and autumn (SON) the projected seasonal streamflow is expected to increase under both climate scenarios and time periods (see Figure 14(a) and 14(b)). Generally, the average expected change in streamflow shows variations between months, seasons and annually. The average seasonal and annual change streamflow is relatively less compared to monthly change. In some months like October and July a higher magnitude of streamflow change may be expected which could lead to floods. On the contrary, in some months like January, February and March a significant reduction in streamflow may lead to water scarcity in the area. Table 6 shows the average change in streamflow for the periods 2021–2040 and 2041–2060 compared to the baseline period (1996–2015). The projected streamflow in all cases (monthly, seasonal and annual) is expected to increase. The increment of streamflow under RCP4.5 is higher than that under RCP8.5.

This study was also aimed at examining trends. Accordingly, trends were tested in terms of annual time series for observed rainfall and streamflow and projected rainfall and corresponding streamflow using the non-parametric Mann–Kendall trend test at a significance level of 5%. The time series (trends) of observed annual streamflow and rainfall are presented in Figure 15 and projected rainfall and corresponding simulated streamflow for the period of 2021–2060 under RCP4.5 and 8.5 are presented in Figures 16 and 17. The observed mean annual flow and rainfall shows decreasing trends. The reduction was –7.73 mm/year for rainfall and –5.6 m³/s/year for streamflow (Figure 15). However, the forecasted rainfall and streamflow under both RCP scenarios shows an increasing trend. The projected rainfall is expected to increase about 0.35 mm/year under RCP4.5 and 1.86 mm/year under RCP8.5. The forecasted streamflow under RCP4.5 is expected to increase by 6.34 and 7.63 m³/s/year under RCP8.5. However, as presented in Table 7, the results of the calculated |Z| values are less than the tabulated |Z| at the significance level of 5%, which is 1.96. This shows that the trends for both recorded and simulated precipitation and streamflow are statistically insignificant under both RCPs.

The impact of climate change studies on streamflow by coupling climate models (RCM/GCM) with hydrological models involves a range of uncertainties (Graham et al. 2007; Fanget et al. 2015; Ahmed et al. 2017). Selection of a GCM/RCM model and plausible scenarios are the greatest sources of uncertainty on climate change analysis in addition to hydrological model simulation (Yacouba et al. 2017; Min et al. 2018). Despite these limitations, in this work every effort was made to minimize the uncertainties on climate prediction and hydrological simulation to understand the potential impact of climate change on future streamflow changes, such as best climate model (RCM) selection, most plausible climate scenarios (RCP4.5 and 8.5), selection and bias correction climate model output. This study coupled a single climate model (RCM) under RCP4.5 and 8.5 emission scenarios with a semi-distributed hydrological model (HBV) to investigate future streamflow changes in the Upper Awash sub-basin. However, it is often recommended to apply different RCMs and ensemble models with all possible emission scenarios, including RCP2.6 and RCP6.0, to explore a wide range of climate changes. However, the results of this work are consistent with other similar studies (Getahun 2015; Mulligan 2015; Ahmed et al. 2017; Gizaw et al. 2017). Overall, projected temperature and precipitation is expected to increase in the study area. The increment in minimum temperature would be higher compared to maximum temperature and such changes are common globally. Similar findings were reported by IPCC (2013), Tekleab et al. (2013) and Negash et al. (2021). The temperature change is more severe under RCP8.5 than RCP4.5, confirming that RCP8.5 is the highest carbon emission scenario and this creates an active hydrological cycle which leads to heavy rainfall (IPCC 2013). The application of bias correction improved the projected temperature and precipitation accuracy. This finding is also in harmony with the results of Fanget et al. (2015), Ahmed et al. (2017) and Daba et al. (2017). On average, annual projected streamflow is expected to increase by 28.5% under RCP4.5 and 23.9% under RCP8.5. This result seems to be in line with similar studies that have been conducted by Mulligan (2015), Babur et al. (2016), Daba et al. (2017), Gizaw et al. (2017), Tadese et al. (2019) and Negash et al. (2021). However, it is in contradiction to the study by Kefeni et al. (2019). This may be due to differences in hydrological model, climate model resolution and bias-correction methods. Apart from this, further investigation is required to verify the results using a multi-climate model with different resolution and possible climate scenarios. The projected streamflow is expected to be high in July and October under both RCPs while a significant reduction is expected in January and April for the period 2021–2040 and in March for the period 2041–2060 under both RCPs emmision scenarios. A similar finding was depicted by Tadese et al. (2019). This study used an HBV hydrological model, which requires relatively less input data compared to the others to simulate streamflow. Thus, this method can be adapted and may be helpful especially to data limited regions to estimate likely streamflow changes in future (see also Endalkachew & Asfaw (2019)). Generally, on average, the expected change in streamflow varies among months, seasons and years. The expected streamflow variations in months and seasons are relatively less compared to annual changes.

Ethiopia is a country which is highly dependent on agricultural products for development. The availability of water resources in the future could be highly influenced by climate change. In this study, the potential response of climate change impact on streamflow availability in the Upper Awash sub-basin was assessed and quantified. A regional climate model (RCM) under RCP4.5 and 8.5 climate scenarios was used to obtain projected precipitation and maximum and minimum temperature in the future. Bias correction was employed to RCM temperature and precipitation to minimize uncertainties in climate projection before being fed into the HBV hydrological model to simulate streamflow. The application of bias correction improved the performance of climate model output (precipitation and temperature) in reproducing station-based recorded temperature and precipitation. Hence, it is practical to apply bias correction to the climate model output for better streamflow simulation for further study. The HBV hydrological model was calibrated and validated to simulate future streamflow. The streamflow response of two typical future periods (2021–2040 and 2041–2060) was predicted based on the bias corrected projected temperature and precipitation and estimated evapotranspiration was compared to the baseline period to assess the change. The predictive ability of the HBV model was evaluated by R² and NSE for both the calibration and validation periods. The results of R² and NSE were 0.89 and 0.88 for calibration and 0.74 and 0.85 for validation, respectively. Under both climate scenarios (RCP4.5 and 8.5) precipitation and temperature shows increasing trends and the temperature change is most severe under RCP8.5 confirming that RCP8.5 is the highest carbon emission scenario. The change of the maximum temperature is larger than the minimum temperature under both scenarios. Likewise, the change of precipitation is larger under RCP8.5 than the RCP4.5 change and the result agreed with the results of IPCC (2013). The average annual streamflow following an increase in precipitation is also expected to increase to 28.5 and 23.95% under RCP4.5 and 8.5, respectively. In the near future period (2021–2040), it could be increased to 40.1 and 29.4%, and 16.9% and 18.5% for the far future (2041–2060) under RCP4.5 and 8.5, respectively. A large increment of streamflow could be expected in October while a significant decline could be in March under both RCPs. There could be a slight decrease in streamflow in the far future (2041–2060) compared to the near future (2021–2040), yet will relatively increase compared to the reference period. Unlike the precipitation and temperature change, the projected streamflow suggested under the RCP4.5 scenario is increasingly more marked than that for RCP8.5. This may be due to the fact that increases in temperature under RCP8.5 resulted in high evaporation losses which could lead to a reduction in streamflow. The study conducted on the whole Awash River basin to evaluate climate change impact on future streamflow suggested that there will be slight decreases in streamflow. This may be due to the middle and lower part of the basin being historically subjected to water deficiency in all seasons and this could lead to an overall reduction in streamflow of the basin, but in the Upper basin (study area) projected streamflow is expected to increase. This study used the HBV hydrological model, which requires relatively less input data to predict future streamflow. Thus, adapting this kind of approach will be helpful for other basins, especially in data limited regions, to estimate likely future streamflow change. The results of this work can also help to inform the water resources planners and designers to frame appropriate planning and management strategies for the effective use of water in the future.

We express our sincere gratitude to the Ethiopian National Meteorological Services Agency (ENMSA) and Ministry of Water, Irrigation and Energy (MoWIE) for providing meteorological and hydrological data respectively. We would like to thank all colleagues for their encouragement throughout this study.

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