This study investigates drought propagation from meteorological to hydrological and streamflow required to recover from drought in four sub-basins: Genale, Tekeze, Awash, and Baro basins of Ethiopia. Due to limited observed streamflow data, the soil moisture accounting and routing (SMAR) model was used to extend the streamflow data for each sub-basin from 1985 to 2017. Drought characteristics in terms of duration, severity, and onset/offset of drought and propagation time at different time scales are investigated using run theory and Pearson correlation, respectively. Two Archimedean copulas (Clayton and Gumbel) are used to identify the joint return period between drought duration and severity and the amount of streamflow required to recover from hydrological drought for each sub-basin. Our results revealed that drought frequency has increased over most sub-basins over the last two decades. The propagation time from meteorological drought to hydrological drought is shorter over the Tekeze sub-basin (1–3 months); however, Genale and Awash sub-basin show 3- to 6-month propagation time. The more extended propagation time is seen over the Baro sub-basin (6–9 months). The required amount of water for drought recovery estimation shows a linear relationship between the duration of the drought and the amount required.
The drought onset and offset months for hydrological and meteorological drought depend on local weather conditions.
The drought propagation analysis helps understand the impact of drought on the stream flows.
The onset of hydrological drought lags for a certain period from meteorological drought.
The amount of flow required for the stream to recover from prolonged drought exhibits a linear relationship with drought duration.
Climate extreme events are inevitable natural events resulting in significant devastation. Adaptation and mitigation measures are the only options countries have to minimise the impacts of such extreme events. Drought affects a wide area and persists for an extended period. It creeps in slowly and remains unnoticed until it causes a wide range of impacts on different sectors, such as agriculture, water supply, and hydropower. In recent decades, drought events have shown an increase in their frequency and intensity in Australia (Kirby et al. 2014), the United States of America (Leeper et al. 2022), East Africa (AghaKouchak 2015; Gebremeskel Haile et al. 2020), and Southeast Asia (ESCAP 2019). Drought is categorised into four types: meteorological drought, hydrological drought, agricultural drought, and socio-economical drought (Mishra & Singh 2010). Meteorological drought is related to the reduction in rainfall compared to its long-term average value, which may trigger other forms of droughts. Rain reduction for a long time causes soil moisture depletion, leading to agricultural drought. Reduction in rainfall may result in the depletion of soil moisture and subsequent decrease in streamflow and groundwater, leading to hydrological drought. The onset of hydrological drought generally lags for a certain period from meteorological and agricultural drought (Mishra & Singh 2010; Van Loon et al. 2012).
Indeed, numerous drought evaluation approaches are already developed. Niemeyer (2008) reviewed more than 80 drought indices and categorised them based on their application. For analysis of meteorological drought, the most commonly used indices are the standardised precipitation index (SPI), rainfall anomaly index (RAI), and drought severity index (DSI). Standardised precipitation evapotranspiration index (SPEI), soil moisture deficit index (SMDI), and Palmer drought severity index (PDSI) are commonly used for the investigation of agricultural drought. Standardised streamflow index (SSI), Palmer hydrological drought index (PHDI), and surface water supply index are some indices used to analyse hydrological drought. In recent years, gravity recovery and climate experiment (GRACE) data provided a new way of estimating the deficit in terrestrial water storage (Sun et al. 2018; Rawat et al. 2022). Indeed, all these approaches contributed significantly and advanced the existing knowledge; however, most of the methods mentioned above are one-dimensional, and drought events are multi-dimensional, requiring an investigation of multi characteristics of drought. The multi characteristics of drought, such as duration, intensity, and severity, are investigated using the Run theory approach based on the selected threshold value. It is recommended to use a threshold value of −0.5 and 0 to define the initiation and termination of meteorological and hydrological droughts. Copula theory is also employed since it allows various correlations, marginal distribution, and estimation of joint return periods to multiple droughts features (Wong et al. 2010; Chang et al. 2016).
Similarly, many studies have been conducted to analyse drought events and assess their impact. Most past studies have mainly focused on evaluating meteorological and agricultural droughts as they directly impact food security. Even if countries built large dams to prevent drought impacts, they failed many times. For example, the significant decrease in the water level in reservoirs affected water delivery to residents of cities in South Africa (Baudoin et al. 2017; Ndlovu & Demlie 2020), the low water level in reservoirs threatened electricity production in the United States of America (Turner et al. 2021). It caused a disruption of the water supply for irrigation in Australia (Kirby et al. 2014). Droughts are common in the Horn of Africa region, significantly impacting the population. According to estimates from Tierney et al. (2013) and Venton (2016), the number of people affected varied across different years, with 13 million, 16 million, 12 million, and 10 million individuals affected during the drought years of 2002/2003, 2008/2009, 2010/2011, and 2015/2016, respectively. Particularly in Ethiopia, the year 1984/1985 was the most severe drought in history recorded in the northern part of the country, affecting more than 30 million people causing a famine that persisted until 1986; and due to this, millions lost their lives (Gebrehiwot Veen & Maathuis 2011). Similarly, in 2002/2003, 2011/2012, and 2015/2016, more than 10 million people contributed to the death of many people and livestock in the southeast part of Ethiopia (Murendo et al. 2011; Venton 2016; Bayissa et al. 2018). In recent years, the impact of drought has diversified across various sectors, including hydropower production, which has been particularly affected. These cases, previously unnoticed in Ethiopia, have become more evident. There has been a notable increase in frequent power cutoffs due to declining water levels at Gilgel Gibe-3 and Tekeze dams (Demissie & Solomon 2016), resulting in significant consequences for the country's economy (Mondal et al. 2017; Abdisa 2018; Girma 2020). Since Ethiopia is located in a region prone to drought, where 85% of electric energy generation relies on hydropower (MoWE 2018), such problems are expected to escalate. Hence, gaining a comprehensive understanding of the time it takes for drought to propagate from meteorological to hydrological conditions and the characteristics of drought in terms of duration and intensity will be crucial for effective water management during drought events. This study investigates single and multi-drought characteristics over four sub-basin of Ethiopia. The drought characteristics include its duration, severity, intensity, and joint return period of duration and severity. The propagation time from meteorological to hydrological drought is also studied. Furthermore, the amount of water required for the streams to recover from drought is also investigated in this study. This information is crucial in devising strategies for adapting and mitigating drought impact on water resource management.
STUDY AREA AND DATA
The Awash basin is the only non-transboundary basin in the country, with flows starting in the central highlands to the eastern lowlands with a mean annual flow of 4.6 BMC (Awulachew et al. 2007). The basin is characterised by a bi-modal rainfall pattern with a mean annual rainfall range of 100–1,700 mm (Figure 2(c)) and a mean air temperature of 17 and 26 °C. Due to the regulated flow downstream created by the Koka hydropower dam, our study was limited to a 7,621 km2 drainage area at the Hombole gaging station (Figure 1).
Baro-Akobo is located in the country's southwest, comprising three rivers: Baro, Gilo, and Akobo Rivers. The basin is located in the wettest part of the country, with annual rainfall ranging from 1,700 to 2,500 mm during a mono-modal rainfall pattern with peak months in July and August (Figure 2(d)). The mean annual temperature ranges from 17.5 °C on highlands to around 27.5 °C on the flood plains. The total mean annual flow from the river basin is estimated to be 23.6 billion m3 (Getu Engida et al. 2021). Our study was limited to only the Baro basin due to the constraint of the recorded data, with 24,356 km2 area drainage coverage upstream at the Gambella streamflow station.
The meteorological and hydrological data for the present study have been obtained from the National Meteorology Agency (NMA) and the Ministry of Water and Energy (MoWE) of Ethiopia. Due to the poor quality of available meteorology data, only 20 meteorology stations located within or in the vicinity of the selected sub-basins (Genale and Awash) with no or minimum missing values are chosen, and those which have missing value are filled using the Thiessen polygon method. Rainfall, maximum and minimum temperature at a monthly time scale from 1985 to 2017 are used after data quality checking using outlier and homogeneity tests from neighbouring stations. Only those streamflow gauging stations which are not affected by anthropogenic influences such as dams, water obstruction, and others have been chosen for the present study. The available streamflow data were limited in duration due to missing recorded data in most of the basins and the construction of dams, even though Tekeze Dam is located upstream of the streamflow station (Emba-madre station) (Figure 1). Due to the exclusion of the anthropogenic impact on streamflow, the data are limited until the period of dam construction. In the case of the Awash and Genale basins, the streamflow gaging station (Hombole station) and Chene-masa gauging station located before Koka dam and Genale-Dawa-3 hydropower dam are selected. Awash and Baro's data with minimum missing recorded values are available from 1990 to 2006, Tekeze at Emba-Madre station has data for 1996–2007, and Genale river gaging station at Chene-mesa has an uninterrupted record for the period 1993–2005. Due to the limited availability of recorded periods of streamflow data, we extended the streamflow records to match the meteorological data period (1985–2017) by calibrating and validating a hydrological model using the overlap periods (Table S1). The model calibration and validation results indicate that Baro and Awash basins perform better with the highest R2 values, 0.79 and 0.76, and Pbias values, 2.97 and 1.93, during calibration. Tekeze and Genale sub-basins failed to capture peak flow in multiple summer seasons; however, the model performed during low flow were R2 values during calibration of 0.77 and 0.56. Detailed results of the calibration and validation processes are described in the supplementary material.
METHODS AND METHODOLOGY
This section describes the methodological framework used in this study, starting with data preprocessing. Drought analysis studies need a minimum of 30 years of data (Svoboda et al. 1987; McKee et al. 1993), but the available streamflow data for selected sub-basins are limited, so we extended the data length by simulating streamflow. The soil moisture accounting and routing (SMAR) model has been used to simulate and extend streamflow to the desired period for all four sub-basins. The National University of Ireland, Galway (Goswami & O’ Connor 2005), developed the SMAR model containing hydrological models for river simulation and flood forecasting for an early warning system. The SMAR model is a lumped conceptual rainfall-evaporation-runoff model in which its water component balance is designed based on the Nash and Sutcliffe water balance model. It performs well in other studies (Mockler et al. 2016; Dessalegn et al. 2017; Khan et al. 2018), so we selected the SMAR model to extend streamflow after calibrating and validating the model.
The methodological framework consists of two major parts: drought propagation and recovery. We use widely known drought analysis methods using the Climate Data Tool package developed by International Research Institute (IRI), Colombia University https://iri.columbia.edu/our-expertise/climate/tools/cdt/, such as SSI and SPI, for hydrological and meteorological drought to understand drought characteristics and estimate drought propagation time. The run theory is used to investigate the characteristics of droughts using the drought_feature package (Le et al. 2019) obtained from https://github.com/adrHuerta/drought_features. We used the run theory method by assigning threshold values of −0.5 and 0 in SSI and SPI values. The relationship between meteorological and hydrological drought is studied using the Pearson correlation. The relationship between meteorological and hydrological drought is studied using the Pearson correlation at a 95% confidence level. The correlation between SSI-1 and SPI at different timescales is used because the accumulation period of meteorological drought is reflected in hydrological drought due to various hydrological processes in basins. One month of runoff accumulation period (SSI-1) may be due to multiple precipitation accumulations periods (SPI-1, 3, 6, 9 and 12 months) (Oertel et al. 2018; Xu et al. 2019; Ho et al. 2021).
Drought indices estimation and characteristics
Standardised drought index
The standardised streamflow index (SSI) also follows the same procedure as SPI but uses streamflow data instead of rainfall. Monthly streamflow data are used to identify hydrological drought from 1985 to 2017; SSI over 1 and 3 months timescales are estimated; these timescales reveal monthly and seasonal water deficits over the basins.
Hydrological anomaly index
Drought events must be classified and characterised to investigate the risk and monitor their impact. Studies use dry, moderate, and wet drought classifications, but in this study, we categorised the drought events in terms of their duration, severity, and intensity using the Run theory method. The Run theory is widely used and the most effective approach for drought analysis, proposed by Yevjevich (1969). The theory answers the most crucial questions in climate hazards such as drought, ‘How long does drought persist?’, ‘severity of the drought?’ and ‘Which month the deficit started and end?’. Understanding these questions is essential in drought response measures and early warning systems. The present study utilises the run theory to detect drought events defined by duration, severity, and drought onset/offset. The drought duration is the consecutive month for which the index values are below the threshold value. Severity is the cumulative deficit of index values during the drought duration, and the intensity is estimated by dividing the severity by duration. This study set the threshold value −0.5 and 0 for the standardised drought index (SPI and SSI), respectively. The threshold value assigned for HAI follows the process mentioned in the methodology section for each sub-basin shown in Figure S5.
Copula theory and joint return period
Estimation of streamflow for drought recovery
RESULT AND DISCUSSION
This section answers the question of what persistent drought looks like in Ethiopia and the extent of its impact on water resources. Such information is crucial to monitor and mitigate the adverse impact of drought. In this section, we have presented the characteristics of meteorological and hydrological drought events over sub-basins of Ethiopia during the period from 1985 to 2017. Hydrological drought occurs when meteorological drought persists over the region; this period is explained under the propagation time section. Finally, the amount of streamflow required to recover from drought is estimated after investigating the joint return period.
Characteristics of meteorological and hydrological drought
Frequent drought over the Tekeze basin Figure 5(b) persisted before 1995 and after 2009. The years between 1996 and 2009 were relatively wettest but with mild drought during 2004/2005. The Tekeze basin experienced major drought events in 1990/1991, 2009/2010, and 2015/2016, and most of the drought events were redirected into hydrological drought except the 2009/2010 drought year (Menna et al. 2022). The study agrees with most drought years in the Tekeze basin, but it differs on the drought year 2010; however, another study in the basin agreed that the basin is frequently affected by meteorological drought but claimed it would not reflect on hydrological drought (Tareke & Awoke 2022).
The Awash basin experienced frequent short-duration meteorological droughts, as shown in Figure 5(c); late after 2010, it showed more frequently and continuously without wet years. Hydrological droughts in the sub-basin show a different pattern to meteorological droughts; instead of consecutive short months of drought, a cumulative effect of precipitation deficit will be reflected as severe hydrological drought (Figure 6(c)). Such droughts significantly impacted groundwater depletion. Since the basin is populated, the hydrologic system in the basin is highly affected by anthropogenic, which could cause such an effect on streamflow (Legesse et al. 2010).
The southwest basin is the least drought-vulnerable region in Ethiopia because of the region's humid climate with the more extended rainy season, which recorded more than 2,000 mm per year of rainfall. However, the Baro basin underwent severe droughts in 1986/1987, 1994/1995, and 2003/2004. The hydrological drought analysis captured meteorological drought events over the sub-basin (Figure 6(d)). In recent decades, the basin has been relatively wet compared to other basins, experiencing frequent droughts (Figures 5(d) and 6(d)). The above results (Figure 5 and Figure 6) show that drought events over Tekeze, Awash, and Genale sub-basins have increased over the last two decades; Gebrechorkos et al. (2020) and Gebremeskel Haile et al. (2020) found similar results over east Africa shows an increased drought frequency associated with large-scale climate anomalies such as Indian Ocean Dipole (IOD) and El Nino-Southern Oscillation (ENSO).
Drought characteristics evaluated based on the start and end of the drought month are more crucial for water resources management. Integrating drought information into the local climate (season) data will provide input for mitigating the impacts. In this study, the selected sub-basins represent different rainfall regions of the country. Baro basins represent a mono-modal rainfall pattern (single rainy season) region. In contrast, the Awash and Genale basins represent bi-modal rainfall patterns (double rainy season) with varying peak months. Tekeze basins' rainfall patterns are categorised into bi-modal and mono-modal (Beyene et al. 2022). Due to these complex rainfall patterns over the basins, drought events occurred in four sub-basins for the last three decades, showing diverse drought onset/offset months (Figure 7(c) and 7(d)).
The onset/offset of meteorological drought over the Genale sub-basin occurs between February and September associated with the failure/gain of the primary rainfall season (March to June) over the basin Figure 7(c) and 7(d). However, the hydrological drought onset month over the basin is delayed by 3–4 months from the meteorological drought onset month Figure 7(c). The Tekeze basin is similar in drought onset months (April to October) between meteorological and hydrological drought, which show quick propagation time between droughts. Still, it shows high variation in drought offset months May to November for SPI-3 and February to July for SSI-3. The Awash sub-basin meteorological drought onset and offset months are extended for more prolonged periods from March to September than hydrological drought, which is limited to only May to August (Figure 7(c)) for drought onset and May to July for drought offset (Figure 7(d)). A mono-modal rainfall pattern (single rainy season) Baro basin shows droughts onsets and offsets from April to September for meteorological drought and May to September for hydrological droughts (Figure 7(c) and 7(d)).
Propagation time from meteorological drought to hydrological drought
The correlation coefficient between SPI and SSI over the Awash sub-basin shows the drought propagation time with the highest correlation coefficient (0.68–0.91) for 1–3 months for June to September and 6–9 months for November to April (Figure 8(c)). The result shows the minimum correlation coefficient (less than 0.23) for 1–3 months of propagation time from December to February and for greater than 1 year of propagation times for May and June. Figure 8(d) shows the highest correlation (>0.71) for more than 6 months of propagation time from meteorological to the hydrological drought between October and April. The sub-basin shows a minimum correlation coefficient for 1–3 months of meteorological drought over the basin from January to March. However, it exhibits a higher correlation coefficient during this period for the primary rainy season (June to September) over the basin.
The average drought propagation time based on drought onset months between meteorological and hydrological drought is shown in Figure 7; the Genale sub-basin shows 4 months of drought propagation on average, similar to correlation coefficient analysis (Figure 8(a)). The Tekeze basin shows an immediate drought propagation time with less than 1 month between SPI and SSI Figure 7(c). Similarly, in correlation analysis, the Tekeze basin shows a high correlation over SPI-1 (Figure 8(b)). In both results, the Awash sub-basin shows 3 months of average propagation time between meteorological and hydrological drought. The Baro sub-basin shows variation propagation time based on drought onset months and correlation analysis. In correlation analysis, drought propagation time on the sub-basin shows over 6 months, whereas the average gap between meteorological and hydrological droughts onset is only limited to 1 month (Figure 7(c)). In general, drought propagation time analysis through correlation and drought onset comparison shows more or less similar results.
Estimation of threshold values and joint return period
|Sub-basin .||Copula .||AIC .||RMSE .||N-Sc .|
|Sub-basin .||Copula .||AIC .||RMSE .||N-Sc .|
Assessment of the amount of streamflow required for drought recovery
The four hydrological drought events for each sub-basin were selected mainly within the prediction band (Figure 11) and considered to represent different drought years from each return period (2, 5, 10, and 25 years). The estimation was carried out using Equation (14); for example, for 6 months (October/2005) to March/2006 of drought in the Genale sub-basin, the stream flow declined below the threshold with values −190, −289, −304, −208, −116, −69.9, and 69 Million cubic meters (MCM). The first 2 months (October and November/2005) are the developing stage of the drought. The following month (December/2005) is maximum drought intensity; after that, the river started to gain flow, which means it is recovering stage. The amount required during the first month of recovery is the difference between January/2006 and December/2005 ((−2.08108m3)–(−3.04108m3)), which gives as 0.96108m3 and consecutive 2 months the river required 0.91108 and 0.45108m3. The recovery amount is estimated in the final drought month by subtracting the basin's threshold value from the subsequent non-drought month (6.9107 − 3.94107m3) gives 0.29108m3. The total amount required for the recovery of the sub-basin is the summation of all streamflow required during hydrological drought months (0.96108, 0.91108, 0.47108, and 0.29108m3), which gives 2.63108m3. The amount of streamflow required was estimated similarly for each sub-basin, namely Genale, Tekeze, Awash, and Baro sub-basins for 2,5,10, and 25 return periods hydrological drought events.
Awash sub-basin requires 2.13108 m3 flow to recover from 3 months drought (July/1990–September/1990). It may be noted that the amount of flow required increases for longer drought duration. Other selected events are 10, 12, and 17 months of drought duration for drought years July/2015–April/2016, June/1995–May/1996, and February/2002–June/2003; these require streamflow replenishment of 7.23108, 6.14108, and 7.92108 m3, respectively, for drought recovery. The most prolonged hydrological drought over the Baro sub-basin was 14 months (May/1986–June/1987) and 12 months (October/1991–March/1992), which required 1.809 and 2.5109m3 to recover. The other selected drought events of 6 months (October/2008–March/2009) and 9 months (September/2010–May/2011) required 8.07108 and 1.016109m3 of streamflow for recovery from drought.
The frequent and devastating meteorological drought events and subsequent hydrological drought over the last three decades in Ethiopian sub-basins underscore the necessity of undertaking more significant research to implement a drought resilience policy. Many studies have recently been carried out over Ethiopia and East Africa. However, most of these studies focus on meteorological, agricultural, and socioeconomic aspects (Bayissa et al. 2017; Degefie et al. 2019; Kourouma et al. 2022), but the understanding of hydrological drought is crucial to managing water during a drought. The main problem with carrying out hydrological studies in Ethiopia is the lack of historically recorded streamflow data. In this study, we overcome this problem for selected four sub-basins by extending streamflow data after calibrating and validating a hydrological model. The standardised and anomaly index methods prove to be capable of drought analysis (Hänsel et al. 2016; Guo et al. 2020). The results (Figures 5 and 6) show that drought events over Tekeze, Awash, and Genale sub-basins have increased over the last two decades (Gebrechorkos et al. 2020; Gebremeskel Haile et al. 2020). Similar results over East Africa showed an increased drought frequency associated with large-scale climate anomalies such as IOD and ENSO.
The drought duration and severity of meteorological and hydrological droughts show a linear relationship over selected sub-basins (Figure 7). The longer the duration, the higher the severity of the drought. The drought years 2010/2011 and 2015/2016 were severe and more prolonged drought events over Tekeze, Genale, and Awash sub-basins, lasting more than 10 months. Band et al. (2022) studied drought severity over East Africa during the 2010/2011 drought years, which agreed with our results. The 2010/2011 drought covered most of the country and persisted for more than a year, including four sub-basins investigated in this study. Awange et al. (2016) studied the duration of drought events over the Horn of Africa using remote sensing data, which shows the drought duration over Ethiopia much extended than the results obtained in the present study. Still, the result coincided with our results in which drought years (1987/1988, 1991/1992, 2002/2003, and 2010–2012) had the highest duration. Figures 6(d) and 7(c) show that the onset and offset of droughts are associated with the amount of precipitation during the primary rainy season (June to September). The findings of Zeleke et al. (2022) support our results for the Tekeze basin, except for the Genale sub-basin, which has multiple drought onset and offset months due to its two rainy seasons taking place, primary rainy season from March to June with peak rainfall during April and the second rainy season from February to May (Beyene et al. 2022).
The drought propagation analysis helps understand the impact of drought on the streamflows, including when it starts. Quantifying the flow deficit during a drought event is essential in mitigating drought impact. The propagation time analysis (Figure 8) showed the month when the precipitation deficit was reflected on the streamflow using the Pearson correlation coefficient between SSI-1 and SPI over different timescales. In this study, the propagation time among the sub-basins, the Genale sub-basin, showed the highest correlation for 3–6 months, whereas the Tekeze and Awash sub-basins start within the first months at different months of the year. Gu et al. (2020) and Li et al. (2020) explained that basin characteristics and other factors influence basin propagation time. In this study, we observed the variation of drought propagation time among sub-basin in drought duration months.
The Archimedean copula method is applicable for investigating dependence between multiple variables, such as meteorological and hydrologically extremes studies, flood analysis (Reddy & Ganguli 2012; Tahroudi et al. 2021), and drought analysis (Wu et al. 2020; Menna et al. 2022). Using two copula functions (Clayton and Gumble), this study estimated the joint return period between drought duration and severity. The selected four drought events from each return period are used to estimate the streamflow required to recover from the drought (Figure 11). Drought duration and the required streamflow show a linear relationship in most basins. However, during some drought events, even if the recovery stage is shorter due to the prolonged developing stage of the drought, the amount of streamflow required will be higher than the others.
The persistent drought in Ethiopia has significantly impacted various sectors, including agriculture and water supply. This study aimed to investigate the impact of meteorological and hydrological droughts on four sub-basins in Ethiopia. The study's findings revealed that the Genale sub-basin in southeastern Ethiopia experienced the most frequent drought events in the last two decades. The Tekeze and Awash basins also showed significant drought events after the 2009/2010 drought. In contrast, the Baro sub-basin experienced fewer drought events, but severe droughts were observed in 1986/1987, 1994/1995, and 2003/2004. The hydrological drought persisted longer than the meteorological drought, and the onset and offset months of droughts depended on local weather conditions. In particular, the onset of drought months mainly coincided with the failure primary rainfall season (Kiremt season) for bi-modal sub-basins.
The study also investigated the propagation time from meteorological to hydrological droughts over sub-basins, which varied depending on the catchment characteristics of the basin. High topography areas of the basin experience shorter months of meteorological drought reflected in streamflows. The amount of flow required for the stream to recover from prolonged drought exhibited a linear relationship with drought duration. The study found that longer drought duration needed the highest streamflow for most sub-basins to recover from the drought. So the most recognised drought events over the East African region during 2010/2011 and 2015/2016 show that longer drought duration required the highest streamflow for most sub-basins to recover. The main limitation of this study is attributed to the constraints posed by the observed meteorological and streamflow data. The study only investigated upper sub-basins and had access to a limited period of observed data. Additionally, the study did not account for the impact of artificial water abstraction from the river during drought events. Since drought propagation time is influenced by topography, future studies should focus on including downstream basins to fully understand the propagation time of the basins. The study's results could help design drought monitoring and management strategies in Ethiopia to mitigate the negative impact of droughts on various sectors of the economy.
DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.
CONFLICT OF INTEREST
The authors declare there is no conflict.