Long-Term Variability in Potential Evapotranspiration, Water Availability and Drought under Climate Change Scenarios in the Awash River Basin, Ethiopia

Abstract

Understanding the hydrological processes of a watershed in response to climate change is vital to the establishment of sustainable environmental management strategies. This study aimed to evaluate the variability of potential evapotranspiration (PET) and water availability in the Awash River Basin (ARB) under different climate change scenarios and to relate these with long-term drought occurrences in the area. The PET and water availability of the ARB was estimated during the period of 1995–2009 and two future scenarios (2050s and 2070s). The representative concentration pathways (RCP4.5 and RCP8.5) simulations showed an increase in the monthly mean PET from March to August in the 2050s, and all the months in the 2070s. The study also identified a shortage of net water availability in the majority of the months investigated and the occurrence of mild to extreme drought in about 40–50% of the analysed years at the three study locations (Holetta, Koka Dam, and Metehara). The decrease in water availability and an increase in PET, combined with population growth, will aggravate the drought occurrence and food insecurity in the ARB. Therefore, integrated watershed management systems and rehabilitation of forests, as well as water bodies, should be addressed in the ARB to mitigate climate change and water shortage in the area.

1. Introduction

Climate change and anthropogenic factors cause long-term variability in the meteorological parameters of a watershed, leading to changes in the hydrological cycle [1]. Precipitation and temperature are the major meteorological parameters that have a direct effect on the hydrological cycle. The change in precipitation and temperature influence water resources, land cover, and ecological sustainability. Consequently, such changes can significantly affect the spatial and temporal availability of water resources and the water balance [2,3,4].

Evapotranspiration (ET) is the critical climate variable that relates evaporation, latent heat flux, and transpiration [5,6,7,8]. The ET represents the amount of water that evaporates from water bodies and transpires from plants. It plays a significant role in surface runoff and groundwater prediction, water balance assessment, crop water requirement, and environmental water demand assessment [9,10,11]. Therefore, it is a vital component in hydrological water balance, global atmospheric system, water resource management, and design of hydrological structures [7,12,13]. The potential evapotranspiration (PET) is an indicator of the environmental demand for evapotranspiration, and is thus the widely accepted method of estimating evapotranspiration for hydrological assessment purposes [14,15,16]. Previous studies used different methods to estimate PET, including empirical equations such as the Hargreaves method and the complicated energy-mass transfer models [17,18,19,20]. The Hargreaves method is a temperature-based method suitable for locations with limited hydro-climatology data [21,22]. This method calculates PET using a simplified equation requiring maximum and minimum temperature, recording year, and latitude [23].

The hydrological cycle illustrates the water movement and its storage within, above and below the Earth surface [24,25,26]. The major components of the hydrological cycle are precipitation, evapotranspiration, runoff and groundwater. The hydrological cycle of a watershed describes the water balance between the inflow of water as precipitation and upstream drainage, and outflow of water as evaporation, transpiration, downstream drainage and any internal storage [8,27]. When there is sufficient precipitation, the water supply may or may not be increased while the net water supply declines due to a decrease in precipitation. Water supply declines when temperature increases, which causes a higher evaporation rate and a rise in water demand. The changes in temperature, evaporation, and water demand have a significant effect on agriculture, energy production, and water usage in a watershed [7,10]. Furthermore, meteorological and hydrological droughts are the major environmental changes that occur due to long-term temperature and precipitation change. Drought indices can be used to investigate the occurrence of meteorological and hydrological drought. The Standardized Precipitation Index is the widely used drought index while Reconnaissance Drought Index and Streamflow Drought Index are the recently developed indices to analyse drought occurrence [28,29,30,31].

Understanding the impacts of climate change on the hydrological processes of a watershed is vital to the establishment of sustainable environmental management strategies aimed at improving water management [1]. The spatial and temporal variability of hydrological processes has a direct influence on agricultural and economic development. ET-based methods are used to provide valuable information on broad aspects such as ecology, hydrology, atmospheric science, agronomy, and carbon science of a watershed [7]. ET-based research has been conducted in different countries [32,33,34]; however, limited studies have been undertaken in Ethiopia, especially in the Awash River Basin (ARB). Therefore, this research aims to (a) evaluate the variability of potential evapotranspiration in ARB under different climate change scenarios, (b) predict the change in water availability in ARB, and (c) assess the long-term drought occurrence and relate it with the variability in PET and water availability. The investigative outcome of this research will help to design and establish sustainable water resources management in ARB.

2. Methodology

2.1. Study Area

The study was carried out in the Awash River Basin (ARB), Ethiopia (latitude 7°53′ N–12° N; longitude 37°57′ E–43°25′ E) (Figure 1). The ARB is one of the 12 river basins in Ethiopia, with an area of 112,000 km². A previous study by Ethiopia’s Ministry of Environment, Forest and Climate Change (MEFCC) [35] indicated that the annual flow from this river basin was 4.9 billion cubic metres (BM³). Most of the Ethiopian rivers show seasonality of water flows and heavy sediment load, and about 90% of their annual runoff occurs in the main rainy season, from July to September. The Awash River originates from Ginchi Town and ends at Lake Abee. The altitude of the ARB ranges from 210–4195 m above mean sea level, and the main Awash River has a length of 1250 km. The annual rainfall of the ARB ranges from 200–1800 mm [35], and the temperature ranges from 12–37 °C [36].

2.2. Dataset

The daily data of precipitation, temperature, and streamflow were collected from the National Meteorology of Ethiopia and Ethiopia’s Ministry of Water, Irrigation, and Energy. In this study, 3, 10, and 22 stations were selected for streamflow, temperature, and precipitation data, respectively (Figure 1). The selection of the stations and the duration of the historical period were based on the availability of long-term data and minimum missing values. Fifteen years (1995–2009) was used for PET and water availability analysis as there was no decent data available after 2009, while 30 years (1986–2015) was taken for drought analysis. All the selected stations had less than 10% missing data. Thus, data from at least two or more representative stations in the Upper, Middle, and Lower parts of the basin were used in the study. The meteorological stations used in this study are shown in

Table 1. Details of the meteorological stations in the Awash River Basin.

No	Stations	Longitude	Latitude	Elevation (m)	Variables	Duration
1	Addis Ababa Bole	38.75	9.03	2354	Precipitation, temperature	1995–2009
2	Addis Alem	38.38	9.04	2372	Precipitation	1995–2009
3	Akaki	38.79	8.87	2057	Precipitation	1995–2009
4	Abomassa	39.83	8.47	1630	Precipitation	1995–2009
5	Aliyu Amba	39.78	9.55	1805	Precipitation	1995–2009
6	Debre Zeyit	38.95	8.73	1900	Temperature, precipitation	1995–2009
7	Debre Sina	39.75	9.87	2800	Precipitation	1995–2009
8	Dire Dawa	42.53	9.97	1180	Temperature, precipitation	1995–2009
9	Ginchi	38.13	9.02	2132	Precipitation	1995–2009
10	Gewane	40.63	10.15	568	Temperature, precipitation	1995–2009
11	Holetta	38.38	9.03	2400	Streamflow, temperature, precipitation	1986–2009, 1986–2015
12	Koka Dam	39.15	8.47	1618	Streamflow, temperature, precipitation	1986–2009, 1986–2015
13	Kulumsa	39.23	8.07	2211	Precipitation	1995–2009
14	Kimoye	38.34	9.01	2150	Precipitation	1995–2009
15	Kombolcha	39.71	11.08	2341	Temperature, precipitation	1995–2009
16	Metehara	39.92	8.86	944	Streamflow, temperature, precipitation	1986–2009, 1986–2015
17	Majete	39.85	10.5	2000	Precipitation	1995–2009
18	Melkasa	39.32	8.4	1540	Temperature, precipitation	1995–2009
19	Merssa	39.67	11.66	1578	Temperature, precipitation	1995–2009
20	Mojo	39.11	8.61	1763	Precipitation	1995–2009
21	Shola Gebeya	39.55	9.22	2500	Precipitation	1995–2009
22	Tulu Bolo	38.21	8.65	2190	Precipitation	1995–2009

. The monthly average precipitation of the ARB was calculated by the Thiessen polygon method in ArcMap 10.4 using the twenty-two rainfall stations (Figure 1). The future precipitation and temperature were taken from two different General Climate models (HadGEM2_AO and GFDL_CM3) and three different Regional Climate models (CCLM, CNRM_RCA4, and MPI_RCA4) (

Table 2. List of climate models (General Climate Models (GCMs) and Regional Climate Models (RCMs)) used in this study [39].

No	Climate Model	Abbreviations (for This Study)	Organization	Resolution (°)
1	CNRM-CERFACS-CNRM-CM5_CLMcom-CCLM4-8-17	CCLM	Climate Limited Area Modeling (CLM) Community	0.44
2	MPI-M-MPI-ESM-LR_SMHI-RCA4	MPI_RCA4	MPI (Max Planck Institute), Germany	0.44
3	CNRM-CERFACS-CNRM-CM5_SMHI-RCA4	CNRM_RCA4	SMHI (Sveriges Meteorologiska och Hydrologiska Institute), Sweden	0.44
4	GFDL_CM3	GFDL_CM3	Geophysical Fluid Dynamics Laboratory	2.0 × 2.5
5	HadGEM2_AO	HadGEM2_AO	Met Office, Hadley Centre, United Kingdom	1.3 × 1.9

). The Regional Climate Models (RCMs) and General Climate Models (GCMs) data were obtained from CORDEX Africa and WorldClim Version-1, respectively. The future period includes two scenarios: mid of 2041–2060 and 2060–2080 under two representative concentration pathways (RCP4.5 and RCP8.5). The GCM data was statistically downscaled and bias-corrected [38], while the RCM data was bias corrected using Climate Model data for hydrologic modelling (CMhyd) tool [39].

2.3. Data Analysis

2.3.1. Evapotranspiration

The daily historical temperature and precipitation data were arranged and checked for consistency and missing values in Microsoft Excel^®. The summarized monthly and annual data were used in the historical and future period analysis. In this study, the DrinC (Drought Indices Calculator) software [31] was used to calculate the PET using the Hargreaves method. The DrinC is a user-friendly software widely used to calculate different drought indices and PET. The Hargreaves method uses temperature data to estimate PET, as shown in Equation (1) [40].

$$PET=0.0023\times RA\times \left(T^{°}\mathrm{C}+17.8\right)\times T{D}^{0.05}$$

(1)

where $PET$ represents a daily PET and RA is extra-terrestrial radiation in mm/day; $TD$ is the mean of maximum temperature minus mean of minimum temperature (T_max − T_min) in degree Celsius, while $T^{°}\mathrm{C}$ denotes the mean of maximum and minimum temperature (T_max + T_min)/2.

The historical PET (1995–2009) was calculated by the Hargreaves method in DrinC software for each of the 10 stations. The mean PET of the 10 stations was taken to analyse the variability of PET in the ARB. The maximum and minimum temperature data from five different GCMs and RCMs models were used to analyse the future PET of ARB. Statistically downscaled and bias-corrected GCMs data downloaded from WorldClim and the bias-correction for the RCMs data was performed by distribution mapping method using CMhyd tool [39]. The temporal variability of monthly and seasonal historical and future PET was evaluated using a graphical method in Microsoft Excel^®.

2.3.2. Water Availability

The general water balance equation can be used to identify and model the water system using inflow and outflow of water in an area [8,24,27]. In this study, the balance between precipitation and potential evapotranspiration (Equation (2)) was used to estimate the amount of water that could be returned to the atmosphere, that is, net water available (NWA).

$$NWA=P-PET$$

(2)

where $NWA, P, and PET$ denote the net water available, precipitation, and potential evapotranspiration, respectively.

The net water available in the watershed can be expressed as the total available water that can be in the form of runoff, soil moisture, and groundwater recharge. In this study, the monthly precipitation of ARB calculated using Thiessen polygon method was taken as precipitation of the area, while the monthly PET calculated using DrinC was considered as measure of evaporation. Then, these variables were used to estimate the net water available in the area, which includes the runoff and change in storage. The net monthly water availability was computed by deducting the monthly PET of the area from its corresponding monthly mean precipitation. The net water availability was calculated for the historical (1995–2009) and future periods (2050s and 2070s under RCP4.5 and RCP8.5) to analyse the monthly and seasonal variation of water availability in the ARB.

2.3.3. Drought Indices

Drought indices can be assessed using different indicators. In this study, a standardized precipitation index (SPI), reconnaissance drought index (RDI), and streamflow drought index (SDI) were used to analyse the occurrence and severity of hydrological drought at three locations within the ARB (Holetta, Koka Dam, and Metehara). The three locations were selected based on long-term data availability, and they represent the upper and middle part of ARB. The historical drought indices at Holetta, Koka Dam, and Metehara were calculated using DrinC software. The drought analysis was carried out in 3, 6, and 12 months’ time scale at the selected locations. The RDI and SPI were calculated for 30 years (1986–2015), while the SDI was analysed for 24 years (1986–2009) due to data availability.

Reconnaissance Drought Index (RDI)

The RDI is one of the accurate methods for agricultural drought analysis that incorporates two meteorological parameters, measured precipitation and calculated PET [41,42]. The RDI was used to detect the water deficit in a water system considering the inflow and outflow of water [43]. The RDI was calculated using Equations (4) [31].

$${\alpha }_k^{\left(i\right)}= \frac{{\sum }_{j=1}^k{P}_{ij}}{{\sum }_{j=1}^{k}PE{T}_{ij}}, i=1 to N and j=1 to k$$

(3)

where $PE{T}_{ij} and P_{ij}$ denote the potential evapotranspiration and precipitation in the $j$th of month and $i$th of year, respectively, and N is the total number of years of available data.

Previous studies showed that ${\alpha }_k$ (Equation (3)) follow satisfactory lognormal and gamma distributions in wide range of locations and timescales [44,45] and by assuming lognormal distribution $RD{I}_{st}$ can be calculated as:

$$RD{I}_{st}^i=\frac{y^{\left(i\right)}-\overline{y}}{{\widehat{\sigma }}_y}$$

(4)

where $y^{\left(i\right)}$ is the ln (${\alpha }_k^{\left(i\right)})$, and $\overline{y} \& {\widehat{\sigma }}_y$ denote its mean and standard deviation, respectively.

Standardized Precipitation Index (SPI)

The SPI has been used in a number of studies [46,47,48,49]. It uses the gamma distribution and cumulative probability to detect a drought event and is calculated at several time scales [31]. In this study, SPI was calculated using DrinC software.

Streamflow Drought Index (SDI)

The SDI is a recently developed drought index used to detect hydrological drought. In this study, the SDI was calculated using Equations (5) and (6) [31].

$$V_{i,k}={{\displaystyle \sum }}_{j=1}^{3k}{Q}_{i,j} i=1,2,3,\dots ;j=1,2,3,\dots ,12 and k=1,2,3,4$$

(5)

$$SD{I}_{i,k}=\frac{V_{i,k}-{\overline{V}}_k}{s_k} i=1,2,\dots ,k=1,2,3, and 4$$

(6)

where $Q_{i,j,}and V_{i,k}$ denote the monthly streamflow and cumulative streamflow in the $i$th year and $k$th reference period, respectively; $i and j$ denote the hydrological year and month of the data, respectively. The $k$ period refers to $k$ = 1 (October–December), $k$ = 2 (October–March), $k$ = 3 (October–June), and $k$ = 4 (October–September); ${\overline{V}}_k$ is the mean, and $s_k$ is the standard deviation of cumulative streamflow.

The classification of drought indices (

Table 3. Reconnaissance drought index (RDI)- and standardized precipitation index (SPI)-based drought classification.

RDI and SPI Range	Description
≤−2.0	Extremely dry
−2.0 to −1.5	Severely dry
−1.5 to −1.0	Moderately dry
−1.0 to 1.0	Near Normal

;

Table 4. Streamflow drought index (SDI)-based hydrological drought classification.

SDI Range	Description
<−2.0	Extreme drought
−2.0 to −1.5	Severe drought
−1.5 to −1.0	Moderate drought
−1.0 to 0	Mild drought
≥0	Non-drought

) was undertaken according to previous studies [31,46,50,51,52]. In this study, the near-normal category for RDI and SPI were separated into two groups as normal wet from 0 to 1.0, and mild dry from 0 to −1.0.

3. Results

3.1. Potential Evapotranspiration (PET)

Figure 2a shows the percentage of change in potential evapotranspiration (PET) in 2050s RCP4.5 using five different climate models. Most of the climate model simulations have a similar pattern with historical PET. The change in PET showed a similar temporal trend for more than four model simulations in half of the twelve months. The positive value indicated an increase in PET, while the negative value showed a decrease in PET. The 2050s RCP4.5 simulation indicated an increase in mean monthly PET from March to August (Figure 2a). However, a decrease in the mean monthly PET was indicated from September to December. During this period, the increase in mean PET ranged from 3–43%, while the decrease rate ranged from 2–10%. The higher increase occurred during June, whereas a higher decrease was detected in December. The percentage of change in PET in 2050s RCP8.5 using different climate model simulations are shown in Figure 2b. In this period, the change in PET showed a similar temporal trend for more than four model simulations in 8 of the 12 months, while 4 of the months showed a different temporal trend. The 2050s RCP8.5 monthly PET showed an increase from March to September and a decrease for the other months (Figure 2b). The increase in mean PET ranged from 3–46%, and the highest increase was detected in June. Conversely, the highest decrease was identified in December (−12%).

Figure 2c shows the change in monthly PET (%) in 2070s RCP4.5 using GCMs and RCMs simulations. The change in PET showed a similar temporal trend for more than four model simulations in eight of the twelve months. The 2070s RCP4.5 simulation showed a decrease in mean monthly PET only in December (−9%) and February (−3%). The increase in mean monthly PET was detected in the rest of the months with a range of 2–35%. The highest decrease and increase were indicated in December and June, respectively. The 2070s RCP8.5 monthly change in PET of different GCMs and RCMs simulations are shown in Figure 2d. The change in PET showed a similar temporal trend for more than four model simulations in nine of the twelve months. For only three of the months, the future simulations indicated a different temporal trend. The 2070s RCP8.5 simulation revealed an increase in mean monthly PET in all the months except a slight decrease in December (Figure 2d). The increase in mean monthly PET ranged from 3–55%, and the highest increase was detected in June.

3.2. Net Water Availability

Figure 3 and Figure 4 show the net monthly water availability of ARB using different GCMs and RCMs simulations. The net water availability using all the five climate models indicated a similar pattern with the historical period. Almost all five model simulations indicated a positive value in net water availability during July and August (Figure 3). However, the negative value was shown in the rest of the months in most of the simulations and historical period. In September, the net water availability showed a negative value but higher than the historical period. Similarly, March and April exhibited a negative value but lower than the historical period in most of the simulations. The highest decrease in water availability was 143.94 mm in 2050s RCP4.5, 128.82 mm in 2050s RCP8.5, and 145.87 mm in 2070s RCP8.5 simulation (Figure 4). However, the higher increase in water availability was 67.73, 54.79, and 71.04 mm in the 2050s RCP4.5, 2050s RCP8.5, and 2070s RCP8.5 simulations, respectively. On the other hand, the 2070s RCP8.5 showed the highest deficit of 135.46 mm in November and the highest increase of 72.42 mm in July.

3.3. Drought Index

According to the United Nations Environment Programme [53] aridity index, Koka Dam (aridity index = 0.45) and Metehara (aridity index = 0.24) are classified as semi-arid, while Holetta (aridity index = 0.64) is identified as dry sub-humid. The 3 month, 6 month, and 12 month timescale SPI and RDI showed a similar pattern, with correlation ranging from 0.93–0.99 in all the three locations.

The 12 months Drought Indices

Figure 5 shows the 12 months SPI, RDI, and SDI at the three locations (Holetta, Koka Dam, and Metehara). The 12 months SPI showed a dry period of mild to extreme dry at 43% of the analysed years at Holetta and Metehara while it was 47% at Koka Dam. Moreover, the 12 months RDI showed a dry period at 40%, 50%, and 43% of the analysed years at Holetta, Koka Dam, and Metehara, respectively. At Holetta, the severe drought occurred in 2009 (RDI = −1.9) and the extreme drought occurred in 2015 (RDI = −3.3) (Figure 5a). The SPI and RDI at Holetta showed a continuous dry period from 2002 to 2005. From 1997 to 2001, a relatively continuous wet period was shown at Koka Dam while a severe and extreme drought occurred in 1987 (RDI = −1.6) and 1986 (RDI = −2.4), respectively (Figure 5b). Considering the RDI value, the extreme drought at Metehara occurred in 2002 (RDI = −2.5) and 2015 (RDI = −2.8) (Figure 5c). Furthermore, the 12 month SDI indicated a mild to moderate drought at 58%, 50%, and 46% of the analysed years at Holetta, Koka Dam, and Metehara, respectively. The extreme drought occurred in 1988 (SDI = −2.15) and 2009 (SDI = −2.44) at Koka Dam, and in 1987 (SDI = −2.03) and 1988 (SDI = −2.3) at Metehara. A continuous mild drought occurred at Metehara from 1987–1989 and 2002–2005.

The 6 months Drought Indices

The 6 month SPI, RDI, and SDI at Holetta, Koka Dam, and Metehara are shown in Appendix A. The 6 month analysis considered two seasons: (i) October–March and (ii) April–September. The 6 month SPI, RDI, and SDI indicated a mild to extreme drought at 40–50% of the analysed years at Holetta, Koka Dam, and Metehara, except for the SDI at Holetta, which showed a dry period at 54–58% of the analysed years. At Holetta, a continuous dry period was shown from 2003–2007 during the first season, and an extreme drought occurred in 2012 (RDI = −3.3) and 2015 (RDI = −3.0) during the first and second seasons, respectively (Appendix A). At Koka Dam, the extreme drought occurred in 2012 (RDI = −2.2) and 1986 (RDI = −2.4) during the first and second seasons, respectively. Moreover, a severe drought was shown in 2015 (RDI = −1.9) during the first season. At Metehara, a continuous mild drought was shown from 1986–1988 during the first season, and then, an improvement has been observed. However, in 2000, a mild drought started and became severe in 2009 (RDI = −1.79), 2011 (RDI = −1.67), 2012 (RDI = −1.73), and 2015 (RDI = −1.74) (Appendix A). In addition, an extreme drought was shown in 1994 (RDI = −2.1), 2002 (RDI = −2.7), and 2015 (RDI = −2.6). A continuous mild to moderate drought occurred at Holetta from 1999–2005, at Koka Dam from 2003–2007, and Metehara from 2002–2005. An extreme streamflow drought occurred in 2002 (SDI = −2.4) at Holetta, in 1988 (SDI = −2.1) and 2009 (SDI = −2.2) at Koka Dam, and in 1988 (SDI = −2.8) at Metehara. The 6 month hydrological drought analysis showed the majority of the dry periods at Holetta, Koka Dam, and Metehara were mild droughts.

The 3 months Drought Indices

The 3 month drought analysis was performed for four seasons: (i) October–December, (ii) January–March, (iii) April–June, and (iv) July–September. The 3 month SPI and RDI result indicated a negative value at 40–53%, 43–53%, and 33–60% of the analysed years at Holetta, Koka Dam, and Metehara, respectively (Appendix B). At Holetta, the extreme drought occurred in 1989, 1991, 2000, 2002, 2004, 2008, 2009, 2012, and 2015 during one of the seasons. The severe and extreme drought existed at least in one of the seasons in 1986, 1987, 1988, 1991, 2007, and 2009 at Koka Dam. At Metehara, the severe and extreme drought occurred in 1987, 1989, 1995, 2000, 2002, 2008, 2012, and 2015 during one of the seasons. Considering the hydrological drought, about 54–67%, 50–63%, and 33–58% of the analysed years indicated a mild to extreme drought at Holetta, Koka Dam, and Metehara, respectively (Appendix B). A majority of the dry periods in all the three locations were mild drought and extreme drought occurred in 2002 (SDI = −2.6) at Holetta, in 1988 (SDI = −2.1) and 2009 (SDI = −2.1) at Koka Dam, and in 1987 (SDI = −2.6), 1988 (SDI = −3.2) and 1989 (SDI = −2.4) at Metehara. The variability of 3 month hydrological drought indicated a continuous mild drought in the majority of the analysed years, at least in one of the seasons in all the three locations.

4. Discussion

The change in monthly PET showed a similar pattern and temporal trend in most of the simulations undertaken using GCMs and RCMs. However, there was still some difference in the temporal trend in a few months under different model simulations. This indicated the existence of uncertainty in model simulations and the importance of using more than one model in order to increase the accuracy of the predictions. The 2050s RCP4.5 analysis indicated an increase in mean monthly PET from March to August. Similarly, in the 2050s RCP8.5, the monthly PET increased from March to September. The periods of March–May (MAM) and June–September (JJAS) are the two main rainfall seasons as well as the main rain-fed agricultural seasons in Ethiopia [54,55]. The 2050s RCP4.5 and RCP8.5 simulations showed an increase in the monthly PET during these seasons in ARB. This is likely to have a negative effect on agricultural productivity and irrigation management of the basin. Moreover, the crop water requirement and ecological water demand of the ARB will be influenced due to changes in PET. With inadequate water availability to satisfy the crop water demand, the growth of crops would be affected especially in the early and development stages which might result in crop failure due to agricultural drought.

Furthermore, the 2070s RCP4.5 simulations indicated an increase in mean monthly PET in all months except December and February. The mean monthly PET continued to increase in 2070s under RCP8.5 in all the months except December. The increase rate ranged from 3% to 55% in the 2050s and 2070s simulations, while the decrease rate ranged from 3% to 12%. This showed that an increased rate was about four times the decrease rate. Therefore, the increase in mean monthly PET was more significant than the decrease. The increase of monthly PET during most of the months might have a negative impact on the water management system, including irrigated and rain-fed agriculture in the ARB. The Awash basin is one of the most densely populated and intensively used basins in Ethiopia [35]. In the basin, large irrigation schemes and hydropower generation plants provide hydropower energy and water supply for different towns and cities. Previous studies showed the vulnerability of ARB to climate variability and water stress during agricultural seasons [56,57,58]. Moreover, the study by Tadese et al. [37] noted a 25% loss of forest and an expansion of cropland by 15% in 2018, which will have a significant influence on the hydrological process in ARB. Thus, the climate variability and land-use change in ARB is highly likely to intensify the increase in PET in the basin. An increase in PET could be a challenge in crop production especially in dry areas as it leads to an increase in crop water requirement and water demand. Similarly, previous studies have indicated the critical role of evapotranspiration on surface runoff, water budget, irrigation management, and environmental water system [7,10,13,14]. Integrated water management can help to boost crop productivity and improve water availability of the watershed by integrating land and water resources management, social and economic development. Therefore, integrated water management systems and sustainable climate change mitigation strategies are necessary for ARB to cope with these changes in PET.

The simulated net water availability in ARB indicated a negative value in most of the months using different climate models. This might be due to the topography and vegetation coverage of ARB. Moreover, about half of the basin area is covered with shrubland, barren areas, and drylands having higher evaporation losses, especially in the Lower and Middle Awash. Results from this study showed decreased net water availability in March and April and continues to decline under future simulations. Similarly, May and June showed a negative value during historical and future periods. A higher decrease was detected in most of the model simulations during May. The lower water availability from March to May is likely to have a significant impact on crop productivity as this is one of the main seasons for both irrigated and rain-fed agriculture in the ARB. Similarly, the study by [58] identified a decrease of water availability in ARB during most of the months.

On the other hand, the net water availability in July and August was positive and showed an improvement under future simulations. This suggests that during these months, the amount of water available for runoff, soil moisture, and groundwater recharge would be more than what would be expected in the other months. Furthermore, most of the runoff occurred during July and August, which coincides with Ethiopia’s main rainy season. The excess water may be diverted, harvested, or stored and made available for future use and groundwater recharge. Thus, small and large diversions, dams, and reservoirs are essential in areas like the ARB, which have only a few months with higher rainfall and no rainfall or minimal rainfall during most of the other months. Future water management strategies in the ARB should focus on these important aspects.

The 6 month and 12 month RDI and SPI analyses indicated a dry period from mild to extreme drought at 40–50% of the analysed years. The SDI analysis also showed a hydrological drought at about half of the analysed years. The majority of these droughts were mild, and occurred after mostly 2001. Even though it was a mild drought, the continued occurrence might lead to severe drought and water stress in the long-run unless an appropriate water management system is established at the early stages. The ARB is one of the Ethiopian basins that is highly susceptible to hydrological variability [59] with frequent occurrence of drought [35,60,61]. The result of this is that there could be significant negative impact on crop production and food security for the increasing population in the area. The decrease in water availability and an increase in PET, combined with population growth, will aggravate the drought occurrence and food insecurity in the ARB, especially in the Lower and Middle Awash, as the wider area was classified as arid and semi-arid with minimum forest coverage and high evaporation loss. Therefore, integrated watershed management systems and rehabilitation of forests, as well as water bodies, should be addressed in order to mitigate climate change and water shortage in the area.

5. Conclusions

This study aimed to evaluate the variability of potential evapotranspiration (PET) and water availability in the Awash River Basin (ARB) under different climate change scenarios and to relate these with long-term drought occurrences in the area. The PET and water availability of ARB were estimated using DrinC software and water balance approach during the periods 1995–2009 and two future years (2050s and 2070s). In addition, the variability of drought in ARB was evaluated using three drought indices (RDI, SPI, and SDI). Based on the findings of this study, the representative concentration pathways (RCP4.5 and RCP8.5) simulations showed an increase in the monthly mean PET from March to August in the 2050s, and almost all the months in the 2070s. The study also found that the increase rate was considerably higher than the decrease rate. The change in PET has a significant role in the water balance system, including surface runoff, water resources management, crop, and ecological water demand. Therefore, the increase in monthly PET will have a critical effect on the overall crop productivity and water management of the ARB.

Furthermore, the monthly net water availability of the ARB showed a shortage of water in the majority of the months investigated, and the higher decrease was detected from March to May. This has a direct influence on irrigated and rain-fed agriculture in the basin. Moreover, the long-term drought analysis showed the existence of mild to extreme meteorological and hydrological drought at about 40–50% of the analysed years at the three study locations (Holetta, Koka Dam, and Metehara). Generally, the change in PET and net water availability showed the deficit of water during most of the months and frequent occurrence of the mild drought periods in ARB. The findings of this study are an indicator of the existence of water deficit and drought in the ARB, and crucial attention should be given to sustainable watershed management and climate change mitigation strategies in the ARB, especially in the Lower and Middle parts. Food security and sustainability of agricultural production in the ARB could be considerably exacerbated with a growing population in the ARB and the demand to produce more from reduced water availability unless necessary action is taken as early as possible.