https://orcid.org/0000-0002-6488-4402
ABSTRACT
This study examines drought patterns in Ethiopia's 12 major river basins from 1981 to 2018 using the Standardized Precipitation Index (SPI) and the Standardized Precipitation Evapotranspiration Index (SPEI). Both indices reveal historical drought episodes with slight variations, with significant differences in 1984, 2009, and 2015. Except for the Wabi-Shebelle catchment in southern Ethiopia, all river basins show an increasing trend in SPI12 and SPEI12 indices. The eastern and central regions experience more drought according to SPEI3. Seasonal correlations show that during the March–May rainy season, precipitation is negatively correlated with the Indian Ocean Dipole (IOD) index, while in the June–September season, it negatively correlates with Nino 3.4 and positively with IOD. The study also found that El Niño leads to less rainfall in the Ethiopian highlands, while La Niña results in more rainfall in the central and northern highlands but less in the south.
HIGHLIGHTS
The research assessed Ethiopia's drought conditions and their environmental implications using the SPI and SPEI, identifying significant drought periods in 1982, 1984, 1987, 1991, 1997, 2002, 2006, 2011, and 2015, with both indices showing a similar drought progression but with slight differences in severity.
The research revealed that the discrepancy between the SPI and SPEI is consistently negative. This underscores the crucial influence of temperature and evapotranspiration processes in assessing the severity of droughts.
The research evaluated the impact of multiple teleconnections on drought conditions. It was observed that distinct precipitation anomaly patterns across Ethiopia were linked to variations in Sea Surface Temperature (SST).
The results highlight the relationship of El Niño with the occurrence of drought conditions in Ethiopia, particularly its effect on cumulative drought conditions for time scales of more than 6 to 12 months. The effects of the La Niña event on drought conditions in Ethiopia are not clearly observed, as several areas experienced abnormally wet conditions during La Niña events.
INTRODUCTION
Drought is a dry period in the climate system caused by the lack of rainfall, that can occur anywhere in the world. It is a slow-onset disaster characterized by the lack of precipitation, resulting in a water shortage. Hence, its frequency and severity can produce devastating effects on agriculture, the environment, the economy, energy, and also human health (Naumann et al. 2021; Salvador et al. 2023). Unfortunately, drought has become a more recurring phenomenon in many countries of the world due to global warming, climate change, and direct human activities (IPCC 2021; Li et al. 2021). Sub-Saharan Africa is one of the driest regions in the world, and even recent findings from the Intergovernmental Panel on Climate Change (IPCC) (Trisos et al. 2022) provide ample evidence that climate change is affecting large parts of this region. Ethiopia is one of the sub-Saharan countries that has experienced many severe droughts in its history of the past six decades (a total of 30 drought events, 13 of which were severe and affected the entire country), seriously affecting a population of millions of people (Zeleza & McCann 1997; Gebrehiwot et al. 2011; Bayissa et al. 2015). Currently, the occurrence of drought in the country is increasing rapidly (once every 2–3 years) compared to the recent past (once every 10 years) (Gebrehiwot et al. 2011; Zinna & Suryabhagavan 2016). This represents a high risk, as the economy of Ethiopia is largely dependent on rain-fed agriculture, which accounts for approximately 49% of the country's GDP, with more than 80 million people dependent on agriculture (Silva & Uchida 2000, CSA 2014). Even small changes in rainfall patterns directly affect the agricultural sector, damaging crops and reducing yields. In a worst-case scenario, this could lead to a famine that kills more than a million people. The drought of 1984–1985 is still fresh in the memories of many Ethiopians, as it claimed the lives of more than 200,000 people and damaged millions of livestock (Kumar 2008).
According to the scientific literature, drought is categorized into meteorological, agricultural, hydrological, and socio-economic droughts (Tallaksen & Van Lanen 2004; Zhong et al. 2019), developing slowly over time and space (Lehner et al. 2001). Hence, the development of trustworthy and measurable drought indices is vital for drought monitoring and characterization (Zargar et al. 2011). By applying diverse drought indices according to their suitability for drought analysis at different timescales, it can be applied to drought monitoring. For example, the Standardized Precipitation Index (SPI) is widely used to monitor droughts as it is recommended for this purpose by the World Meteorological Organization (WMO). Its multivariate character allows for evaluating dry/wet conditions at various accumulation periods of anomalous precipitation, which can be associated with different drought types, such as SPI-1 for meteorological drought, SPI3–SPI6 for agricultural drought, and SPI6–SPI24 for hydrological drought (Tsakiris et al. 2007; Kaniewski et al. 2012, WMO 2012; Adnan et al. 2015, 2018; Tigkas et al. 2015; Mondol et al. 2016). Many drought indices have been utilized to quantify and characterize drought events, but each has its strengths and weaknesses (Park et al. 2014; Jain et al. 2015). In developing countries such as Ethiopia, precipitation is readily available only through ground-based rain gauges, so several previous studies in Ethiopia point out that the SPI index can be utilized to assess drought. For instance, Bayissa et al. (2015, 2018) used SPI for the assessment of drought conditions in the Upper Blue Nile (UBN) region, Suryabhagavan (2017) for various regions of Ethiopia, Gebrehiwot et al. (2011) for drought in the highlands of northern Ethiopia, Edossa et al. (2010) for Awash River Basin, and Gidey et al. (2018) for Raya Basin.
One of the weaknesses of SPI is to measure dryness in desert areas where evaporation is an important source of moisture (Almedeij 2014). Thus, the developed Standardized Precipitation Evapotranspiration Index (SPEI) tracks a similar approach to SPI, but it is based on a monthly climatic water balance, taking into account the role of temperature, which is the crucial strength of the SPEI over the broadly utilized drought indices (Vicente-Serrano et al. 2010; Yu et al. 2014). Recently, Mulualem & Liou (2020) predict the SPEI index in the UBN basin using an artificial neural network model incorporating multiple variables such as hydro-meteorological, climate, sea surface temperature (SST), and topographic features from 30 years of time series data. They found that the model shows a great level of similarity among observed and predicted SPEI values. The study conducted by Kourouma et al. (2022) investigated spatiotemporal climate variability of drought characteristics in 16 localities of Ethiopia in the period 1983–2020 using SPI and SPEI indices at different timescales. They concluded that the frequency, magnitude, and severity change spatially by area and the longest drought persisted for 12 months in the year 2015. Lemma et al. (2022) assessed meteorological drought in selected Ethiopian watersheds (Awash, UBN, Baro, Danakil, Omo, and Tekeze) using the Climate Hazard Group Infrared Precipitation Station (CHIRPS) dataset to calculate SPI and Effective Drought Index (EDI), confirming the usefulness of these data for this purpose. In a recent study, Stojanovic et al. (2022) also investigated the period 1980–2018 characterized by extreme drought conditions in hydro-climatological regions of Ethiopia and the role of moisture transport to precipitation from climatological sources.
The variability of drought events in Ethiopia has been hypothesized to be related to the anomaly of ocean circulations through atmospheric circulations. For example, the El Niño-Southern Oscillation (ENSO) phenomenon greatly affects Ethiopian precipitation (Camberlin 1997). The connection between El Niño events and drought in Ethiopia has been comprehensively reviewed by many researchers using a long historical rainfall record (Wolde-Georgis 1997; Comenetz & Caviedes 2002; Roop et al. 2016; Ahmed et al. 2017; Mersha & van Laerhoven 2018). A list of El Niño and La Niña events and affected regions in Ethiopia is presented in Table 1. Comenetz (2002) investigated the correspondence of the warm phase of ENSO cycles to drought events in Ethiopia, revealing that the 1997/98 El Niño caused a very abnormal rainfall distribution and the consequences were reduced crop production, outbreaks of cholera and cerebral malaria, deaths, and property damage (Babu 2009). Koo et al. (2019) divided ENSO events in Ethiopia into three periods: 2015/2016 El Niño, 2016/2017 La Niña, and neutral years (i.e. 2012/2013, 2013/2014, 2014/2015, and 2017/2018), to examine crop yield effects of climate change linked with ENSO phases at each grid element of agro-ecological zone level and revealed that the combination of El Niño and La Niña determines the final agricultural yields in Ethiopia. Some authors study ENSO effects on river basins in Ethiopia (Attia & Abulhoda 1992; Eltahir 1996; Molla et al. 2019). Some authors examine the effects of ENSO on Ethiopian river basins, showing that ENSO episodes are negatively correlated to the floods of the Blue Nile and Atbara rivers originating in Ethiopia and affecting the Nile River flow (which is the source of 85% of the Nile water) (Attia & Abulhoda 1992; Eltahir 1996; Molla et al. 2019).
Timeline of ENSO years and droughts in Ethiopia
| ENSO years | Droughts | Affected regions |
|---|---|---|
| 1982–1983 | 1983–1984 | All regions, particularly Tigray and Wollo |
| 1986–1987 | 1987–1988 | All regions |
| 1991–1992 | 1990–1992 | North, East, and South Ethiopia |
| 1993 | 1993–1994 | Tigray and Wollo |
| 2000 | 2000 | All regions |
| 2002–2003 | 2002–2003 | North, East, and Central Ethiopia |
| 2008–2009 | 2006 | The Southern Nations, Nationalities, and Peoples' Region (SNNPR) (Borena) |
| 2009–2010 | 2008–2009 | North, East, Central, and South Ethiopia |
| 2011 | 2010–2011 | South-central, Southeastern, and Eastern parts of Ethiopia |
| 2015–2016 | 2015–2016 | Oromia, Somali, Amhara, Afar, Tigray, SNNPR |
| ENSO years | Droughts | Affected regions |
|---|---|---|
| 1982–1983 | 1983–1984 | All regions, particularly Tigray and Wollo |
| 1986–1987 | 1987–1988 | All regions |
| 1991–1992 | 1990–1992 | North, East, and South Ethiopia |
| 1993 | 1993–1994 | Tigray and Wollo |
| 2000 | 2000 | All regions |
| 2002–2003 | 2002–2003 | North, East, and Central Ethiopia |
| 2008–2009 | 2006 | The Southern Nations, Nationalities, and Peoples' Region (SNNPR) (Borena) |
| 2009–2010 | 2008–2009 | North, East, Central, and South Ethiopia |
| 2011 | 2010–2011 | South-central, Southeastern, and Eastern parts of Ethiopia |
| 2015–2016 | 2015–2016 | Oromia, Somali, Amhara, Afar, Tigray, SNNPR |
Although several previous studies have evaluated drought in Ethiopia, the spatial and temporal variability of drought has not been established, particularly at the local and sub-regional levels. This is important because many drought studies emphasize short-term analysis (Bewket & Conway 2007), specific season analysis (Viste et al. 2013), or study the drought conditions in certain regions rather than the whole country (Bewket & Conway 2007; Edossa et al. 2010).
In this paper, we address a significant gap in the understanding of historical droughts in Ethiopia's 12 large river basins. Despite numerous studies on the impact of drought during Ethiopia's historical period and the drought situation in Ethiopia, there is a lack of comprehensive spatial and temporal estimates of these droughts. Our research fills this gap by using 38 years of meteorological variables (precipitation and temperature) to estimate these droughts.
We also extend the existing body of knowledge by studying the regional distribution of meteorological and hydrological drought. We calculate SPI and SPEI at 1–12 timescales and calculate long-term trends using the Mann–Kendall (M–K) test in all catchment areas. This analysis allows us to examine the frequency of drought events at regional and sub-regional scales and its relationship with oceanic impact (Dipole Mode Index (DMI), Oceanic Niño Index (ONI), Pacific Decadal Oscillation (PDO), and Southern Oscillation Index (SOI)) seasonally in all water bodies using the CHIRPS precipitation dataset. Furthermore, we address another research gap regarding the unclear relationship between the ENSO occurrence and its effects on different parts of the country (AKLDP 2015). Our research, focused on major river basins in Ethiopia, provides important insights to manage agricultural systems, plan water resources, and adapt ecological systems to future climate change. It also helps predict drought events and its critical parameters to improve early warning capabilities for drought management. We believe that our approach, while challenging, can significantly contribute to drought management by targeting the most vulnerable areas.
STUDY AREA AND DATA
Study area
The geographical location of Ethiopia (a) and schematic representation of the main river basins (b).
The geographical location of Ethiopia (a) and schematic representation of the main river basins (b).
Ethiopia comprises 12 major rivers, and 5 of them are transboundary rivers. The river basins are considered hydrological units defined orographically to a large extent by historical precipitation, which ultimately contributes to the volume of groundwater and river systems. The Blue Nile is the largest river in Ethiopia, which originated from Lake Tana in the northwestern part of the highlands. Next, it flows through Sudan, where, in the area of Khartoum, it joins with the White Nile (Billi 2015). The other rivers flow to the Mediterranean Sea, the Indian Ocean, or into the Endorheic basins. Because of its out-flowing rivers to the neighboring countries, Ethiopia is commonly referred to as a water tower in East Africa. The 12 river basins (Figure 1) have an estimated total flow of 122 billion metric cubes and an estimated 2.6–6.5 billion metric cubes of groundwater potential (Awulachew et al. 2007). Out of these major basins, Aysha, Danakil, Ogaden, and Mereb river basins contain no continuously flowing rivers (Ambelu 2009). Ethiopia has three drainage systems. The Western is the largest one constituting the Tekeze, Blue Nile (also called Abbay Basin), and Baro and flows into South Sudan and Sudan. The second one is the Rift Valley comprising the Awash and Omo Rivers. Omo flows into Lake Turkana on the border of Kenya. In contrast, Awash flows northeast of the Danakil depression and then dissipates into a chain of swamps and lakes at the border of Djibouti. The third consists of the rivers Shebelle and Genale originating from the mountains of the east and flowing into Somalia and the Indian Ocean.
METHODOLOGY AND DATASETS
Datasets
The precipitation data from the CHIRPS (v2.0) (Funk et al. 2015) for the 1981–2018 period was utilized for the computation of drought indices. CHIRPS is a 30+ year quasi-global precipitation dataset linking satellite observations from the National Climate Forecast System version 2 and in situ precipitation observations (Dai 2013). This dataset combines satellite products and ground stations to form a gridded precipitation time series and is available at an original spatial resolution of 0.05° with a global coverage ranging from 50°S to 50°N. This dataset has been widely used for monitoring the precipitation variability and drought conditions in Ethiopia (Belay et al. 2019). Among the presently accessible satellite rainfall products, CHIRPS has better performance and is therefore suggested as a valuable substitute for the gauge precipitation dataset (Gebremedhin et al. 2021). A recent public study by Ahmed et al. (2024) evaluated the performance of ERA5 reanalysis data and CHIRPS precipitation data against ground-based measurements from 167 rain gauges in Ethiopia. These authors used point to pixel analysis for the period 1980–2018 to compare daily, monthly, seasonal, and annual precipitation data. The result indicated that over Ethiopia, CHIRPS generally outperforms ERA5, particularly in high-altitude areas, demonstrating a better capability in detecting high-intensity precipitation events. They also highlight considerable performance differences between CHIRPS and ERA5 across varying Ethiopian landscapes and climatic conditions. CHIRPS effectiveness in high-altitude regions, especially for daily precipitation estimation, emphasizes its suitability in similar geographic contexts. It has also been confirmed that CHIRPS can supplement the sparse rain gauge network and afford high spatial and temporal resolution for trend analysis (Dinku et al. 2007; Hirpa et al. 2010; Ayehu et al. 2018; Fenta et al. 2018). The CHIRPS data are downloaded from the FTP server (ftp://ftp.chg.ucsb.edu/pub/org/chg/products/CHIRPS-2.0).
The ONI and the Indian Ocean Dipole (IOD) are utilized to see how rainfall responds to climatic conditions. The National Oceanic and Atmospheric Administration's (NOAA) operational definitions of El Niño and La Niña conditions are based on the ONI. The ONI tracks the running 3-month average SST in the east-central tropical Pacific near the International Dateline, and whether they are warmer or cooler than average. The ONI is defined as the 3-month running means of SST anomalies in the Niño 3.4 region (5°N–5°S, 120°W°–170°W). The 3-month average temperature anomaly time series was retrieved from the NOAA website (https://www.cpc.ncep.noaa.gov/data/indices/oni). However, the IOD is obtained by taking the differences among the SST anomalies in the western (50°E–70°E, 10°S–10°N) and eastern (90°E–110°E, 10°S–0°N) portions of the Indian Ocean (Saji et al. 1999). The monthly IOD data are downloaded from http://www.jamstec.go.jp/frcgc/research/d1/iod/iod/dipole_mode_index.html. Furthermore, the PDO is commonly described as a long-lasting El Niño-like pattern of Pacific climate fluctuation. The PDO index values used here are from NOAA's Earth System Research Laboratory (ESRL) (http://www.esrl.noaa.gov/psd/data/correlation/pdo.data), which provides data from 1948 to the present.
Methodology
The SPI and SPEI
The SPI is the most commonly used indicator worldwide designed by McKee et al. (1993) for identifying meteorological droughts. The SPI calculates precipitation anomalies at a given position based on a comparison of observed precipitation amounts for the period of interest (e.g. 3, 6, 12, 24 months) with the long-term historical precipitation record for that period. In general, the historical record is fitted to a Gamma probability distribution, which is then back-transformed to a normal distribution such that the mean of the SPI values becomes zero and the variance of one. Conceptually, SPI is equivalent to the Z-score, which is a commonly used technique in statistics. Detailed formulation and more information on the SPI can be found in Mckee et al. (1993) and WMO (2012). We have to take into consideration that the SPI shows the anomalies of the observed total precipitation for any given location and accumulation period of interest. Thus, the magnitude of the departure from the average value is a probabilistic measure of the severity of wet (positive values) or dry (negative values) conditions. Since SPI can be calculated over different precipitation accumulation periods (typically ranging from 1 to 24 months), the resulting various SPI indicators allow for estimating different potential impacts of drought. In conventional and operational applications, SPI1–SPI3 indicate immediate consequences such as reduced soil moisture, snowpack, and flow in smaller creeks; SPI3–SPI12 are used as indicators for reduced stream flow and reservoir storage, while SPI12–SPI24 are used as indicators for reduced reservoir and groundwater recharge. To get a full picture of the potential impacts of drought, the SPI should be calculated and compared at different timescales, and also a comparison with other drought indicators is also needed. One of the disadvantages of SPI is that it takes into account only precipitation and can fail to measure the role of temperature and other meteorological variables like the wind on the evaporation from the surface and, therefore, their role in the occurrence and severity of drought (Vicente-Serrano et al. 2010). For this reason, the SPEI is also used in this study.
The SPEI utilized the same method as the SPI, but its sensitivity to evapotranspiration permitted it to assess the water supply and demand relationships at various timescales (Yu et al. 2014). Its multi-scalar characteristics also permit the assessment of the temporal evolution of the climatic water balance, which facilitates the attribution of drought occurrence and severity at different timescales (Beguería et al. 2010). Thus, potential evapotranspiration (PET) is a crucial water loss indicator to consider for investigating drought conditions in the hydrological cycle in the context of global warming (Mukherjee et al. 2018). Vicente-Serrano et al. (2010) described the method to calculate the SPEI, which takes as input data the difference between precipitation and PET. For calculating PET, there are various methodologies, such as the Thornthwaite (Thornthwaite 1948), the Hargreaves (Hargreaves 1975), the modified Hargreaves (Ravazzani et al. 2012), and the Penman–Monteith FAO56 methods (Allen et al. 2006). Based on SPI/SPEI indices, a classification of drought into seven categories ranging from an extremely dry to an extremely wet class is given in Table 2. A drought episode starts when the SPI/SPEI value is below zero, followed by a value of −1 or less, and ends when the SPI/SPEI returns to a positive value (Mckee et al. 1993). The severity of each episode is calculated by summing all SPI/SPEI values over the months of the episode.
Drought category based on SPI/SPEI values (Mckee et al. 1993)
| SPI/SPEI values | Drought category |
|---|---|
 | Extremely wet |
 | Severely wet |
 | Moderately wet |
 | Near normal |
 | Moderately dry |
 | Severely dry |
 | Extremely dry |
| SPI/SPEI values | Drought category |
|---|---|
| | Extremely wet |
| | Severely wet |
| | Moderately wet |
| | Near normal |
| | Moderately dry |
| | Severely dry |
| | Extremely dry |
M–K trend test
The M–K trend test (Mann 1945) is a non-parametric method widely used to detect significant trends in time series data. It identifies the presence of a monotonic tendency in a chronological series of a variable, with the null hypothesis stating that there is no monotonic trend. The test does not make assumptions about the underlying distribution of the data, and its rank-based measure is not influenced by extreme values. The primary output of this method is the Kendall Tau, which measures the monotonicity of the slope. In addition, it provides the Sen's slope, estimating the overall slope of the time series. This makes the M–K trend test a reliable and robust tool for time series analysis. In this study, statistical significance trend analysis was done by using the M–K trend test, while the magnitude of trend was determined by the non-parametric Sen's slope estimator.
RESULTS AND DISCUSSION
Drought episode frequency based on SPI and SPEI
Temporal evolution of SPI and SPEI over 1-, 3-, 6-, and 12-month timescales for the whole of Ethiopia. Yellow, red, and violet colors represent different drought categorizations.
Temporal evolution of SPI and SPEI over 1-, 3-, 6-, and 12-month timescales for the whole of Ethiopia. Yellow, red, and violet colors represent different drought categorizations.
Temporal evolution of the SPEI and SPI values on different timescales (1–12 months) for the 1981–2018 period for the whole of Ethiopia.
Temporal evolution of the SPEI and SPI values on different timescales (1–12 months) for the 1981–2018 period for the whole of Ethiopia.
Temporal evolution of the SPEI and SPI (Supplementary Material Figure S1) values on different timescales (1–12 months) for the 1981–2018 period for each river basin.
Temporal evolution of the SPEI and SPI (Supplementary Material Figure S1) values on different timescales (1–12 months) for the 1981–2018 period for each river basin.
Drought episode frequency
Annual frequency of drought based on the SPI3 and SPEI3 counts for the period 1981–2018.
Annual frequency of drought based on the SPI3 and SPEI3 counts for the period 1981–2018.
Monthly drought episode frequency from monthly data using SPEI3 (a) and SPEI12 (b).
Monthly drought episode frequency from monthly data using SPEI3 (a) and SPEI12 (b).
At a 12-month timescale, a hydrological drought event is more reflected (Figure 6(b)). Surprisingly, the monthly drought episode frequency did not show significant variations from that of SPEI3. It is slightly noticeable that drought happens more often in the Awash, Abbay, and Wabi-Shebele basins. Drought analysis studies in the country revealed that extreme hydrological drought occurs in the upper and middle Awash basin (Edossa et al. 2010) and the UBN basin (Bayissa et al. 2018).
Trend analysis
Time series of SPEI12 (light blue) and SPI12 (red line) and their corresponding trend lines for the main river basins of Ethiopia during 1981–2018.
Time series of SPEI12 (light blue) and SPI12 (red line) and their corresponding trend lines for the main river basins of Ethiopia during 1981–2018.
The highest and lowest values of SPI12 and SPEI12 provide a synopsis of hydrological dry and wet years. Major hydrological drought episode frequencies are identified in 1982, 2001, and 2011 with the most extended hydrological drought episode frequency occurring in the 1980s and early 2005 (El Kenawy et al. 2016). SPI and SPEI values from all the basins were, in general, highly correlated, specifying that either of the indices is suitable for drought depiction. The trends of SPEI12 and SPI12 for the Abbay, Aysha, Baro, Genale, Omo, Rift Valley, and Tekeze showed a positive trend. However, Wabi-Shebelle showed a decreasing trend. The difference between the two indices indicates the role of temperature in moderating the dryness trend.
and 
for the 3-month timescale. A similar increase in the magnitude of the slope is observed for the SPI index. We can conclude that the wetting and drying magnitude of the main river basins of Ethiopia is significantly higher when computed at longer timescales. A spatial differentiation of drought index trends is observed between the basins occupying the eastern and western half of Ethiopia, suggesting an evolution toward opposite hydro-climatic conditions and characterized by a longitudinal differentiation. 
The Sen's slope (×1,000) of the SPEI (top) and the SPI (bottom) at temporal scales of 1, 3, 6, and 12 months. Red asterisks indicate statistically significant trends at a 95% confidence level (period 1981–2018).
The Sen's slope (×1,000) of the SPEI (top) and the SPI (bottom) at temporal scales of 1, 3, 6, and 12 months. Red asterisks indicate statistically significant trends at a 95% confidence level (period 1981–2018).
In Table 3, it is noted that differences among the SPEI and SPI trends were negative for the majority of the basins except for Baro, Omo, and Rift Valley for all analyzed timescales. For the analysis of drought conditions in the above basins, if we have only considered the SPI trend, we can only observe the wetting trend and favorable conditions. This is because the effect of temperature was omitted in the computation of SPI. A moderate trend than the SPI was identified by SPEI, which has considered the effect of temperature and thereby affecting the water balance of a region.
The difference between SPEI (1, 3, 6, and 12) and SPI (1, 3, 6, and 12)
| All basins | SPEI1 − SPI1 | SPEI3 − SPI3 | SPEI6 − SPI6 | SPEI12 − SPI12 |
|---|---|---|---|---|
| Abbay | −0.003 | −0.046 | −0.032 | −0.012 |
| Awash | − 0.119 | − 0.082 | − 0.107 | − 0.189 |
| Aysha | − 0.3037 | − 0.225 | − 0.223 | − 0.268 |
| Baro | 0.005 | 0.010 | 0.085 | 0.049 |
| Danakil | − 0.109 | − 0.143 | − 0.209 | − 0.1521 |
| Genale | −0.020 | 0.050 | 0.028 | 0.006 |
| Mereb | − 0.161 | 0.005 | 0.031 | − 0.114 |
| Ogaden | − 0.393 | −0.151 | −0.013 | 0.044 |
| Omo | 0.015 | 0.017 | 0.081 | 0.078 |
| Rift -Valley | 0.006 | 0.021 | 0.059 | 0.012 |
| Tekeze | − 0.126 | −0.060 | −0.013 | − 0.106 |
| Wabi-Shebelle | −0.064 | −0.018 | −0.052 | − 0.147 |
| All basins | SPEI1 − SPI1 | SPEI3 − SPI3 | SPEI6 − SPI6 | SPEI12 − SPI12 |
|---|---|---|---|---|
| Abbay | −0.003 | −0.046 | −0.032 | −0.012 |
| Awash | − 0.119 | − 0.082 | − 0.107 | − 0.189 |
| Aysha | − 0.3037 | − 0.225 | − 0.223 | − 0.268 |
| Baro | 0.005 | 0.010 | 0.085 | 0.049 |
| Danakil | − 0.109 | − 0.143 | − 0.209 | − 0.1521 |
| Genale | −0.020 | 0.050 | 0.028 | 0.006 |
| Mereb | − 0.161 | 0.005 | 0.031 | − 0.114 |
| Ogaden | − 0.393 | −0.151 | −0.013 | 0.044 |
| Omo | 0.015 | 0.017 | 0.081 | 0.078 |
| Rift -Valley | 0.006 | 0.021 | 0.059 | 0.012 |
| Tekeze | − 0.126 | −0.060 | −0.013 | − 0.106 |
| Wabi-Shebelle | −0.064 | −0.018 | −0.052 | − 0.147 |
The statistically significant trends at a 95% confidence level were represented with bold numbers.
Teleconnections
resolution and climate indices such as IOD, Nino 3.4, and PDO, which are well known to impact Ethiopian precipitation, are performed. The seasonal correlations map (only statistically significant values for p < 0.05) for the main rainfall seasons of March, April, May (MAM), June, July, August, September (JJAS), and October, November, December (OND) is presented in Figure 9. This figure shows the complex spatial variations of the seasonal teleconnections over Ethiopia. The MAM rainfall season, which is the primary precipitation season in southern Ethiopia, and the secondary rainfall season over central and northeastern Ethiopia are identified with a band of positive correlation with PDO extending from central Ethiopia to the northeast, while negative correlations are restricted over the Baro river basin in the west of the country. The correlation pattern with the IOD is highlighted because it is very similar but reversed in sign. Meanwhile, weak positive correlations with Nino 3.4 were found in the central region of the country. However, for JJAS, which is the primary precipitation season over most parts of Ethiopia, we found areas of negative correlations with Nino 3.4 to the center and northwest in the UBN and Tekeze, in agreement with previous findings (e.g. Diro et al. 2011; Degefu et al. 2017). In addition, areas with positive correlations are observed in the southeast over the Genale-Dawa, Wabi-Shebelle, and Ogaden. In JJAS, rainfall also displays a statistically significant positive relationship with IOD over the basins of the northeast of the country. Similarly, JJAS showed weak positive correlations with PDO but to the southeast. A correlation analysis for the OND season in southern Ethiopia (Kassahun 1987) displayed less teleconnection with all of the climate indices. So, it is rather striking that the OND shows a statistically significant positive correlation with the PDO in the south of the country, although the area that shows this relationship is visually small. In general, the correlation analysis between SST indices and gridded Ethiopian precipitation depicted in Figure 9 shows that SST fluctuations according to modes of climate variability in different forms for the seasonal and spatial pattern of precipitations over Ethiopia. Thus, for an enhanced and better-equipped forecasting of rains as well as other drought indicators, it is essential first to detect regional patterns and separate the country into uniform areas regarding the impacts of the SST. 
Correlation between SST and seasonal precipitation over Ethiopia during 1981–2018. Both SST and precipitation data are averaged over the seasonal periods: MAM, JJAS, and OND.
Correlation between SST and seasonal precipitation over Ethiopia during 1981–2018. Both SST and precipitation data are averaged over the seasonal periods: MAM, JJAS, and OND.
Often the relationship between ENSO and droughts in Ethiopia, particularly during the El Niño phase, has been studied (Gebrekirstos et al. 2008; Zeleke et al. 2017; Gobie & Miheretu 2021). They found that droughts are significantly correlated with ENSO. One such common approach to analyzing ENSO teleconnections in the region is through composites of El Niño and La Niña. Though correlation can capture linear relationships, since it does not handle non-linear relationships, composite analysis is widely used to understand physical mechanisms. There are various steps required to form composites of any given phenomenon. First is necessary to choose a means to define events for compositing. In a previous study on ENSO composite analysis, generally, a positive basis was utilized for El Niño events, and a negative basis was utilized for La Niña events (Karoly 1989). This study has used the definition of events described by the ONI to assign the events and nonevents. Especially, El Niño or La Niña events are compared with non-ENSO events. After the events are chosen, they are averaged. These averaged events are subtracted from the average of the nonevents to form the composite.
Composite maps of precipitation in JJAS season during warm and cold ENSO years.
Composite maps of precipitation in JJAS season during warm and cold ENSO years.
Pearson correlation between DMI, ONI, PDO, SOI, and SPEI at 1–12 timescales. Black lines specify a statistically significant correlation coefficient at p < 0.05.
Pearson correlation between DMI, ONI, PDO, SOI, and SPEI at 1–12 timescales. Black lines specify a statistically significant correlation coefficient at p < 0.05.
CONCLUSIONS
The primary aim of this research was to study and evaluate existing drought conditions and investigate their usefulness in assessing the impact of drought conditions on the environment of Ethiopia. The SPI and SPEI were applied to analyze drought development over the study period. The analysis indicated that the main periods of drought occurred all over the country in 1982, 1984, 1987, 1991, 1997, 2002, 2006, 2011, and 2015. Similar temporal evolution of drought is observed from the SPI and SPEI series. However, the temporal drought analysis shows that there is a small difference in magnitude among the droughts depicted by the precipitation-based SPI and temperature-based SPEI as their difference (SPEI − SPI) was always negative, signifying the significant role of temperature and evapotranspiration processes on modulating the severity of drought.
Furthermore, this study assessed the role of various teleconnections on drought based on a composite analysis. Initially, we found different patterns of precipitation anomalies over Ethiopia associated with SST variations characterized through the ENSO index in the Atlantic Ocean, with the IDO index in the Indian Ocean, and with the Pacific Decadal Oscillation in the Pacific Ocean. The results highlight the relationship of El Niño with the occurrence of drought conditions in Ethiopia, highlighting its effect on cumulative drought conditions for timescales of more than 6–12 months. The precipitation anomalies for the JJAS period showed the impact of El Niño on the precipitation reduction in central highlands and favorable conditions during La Niña. This could be explained by the northeastward evolution of rainfall anomalies developing during El Niño. Interestingly, mild drought conditions are seen during MAM and OND with SPEI12. The effects of the La Niña event on drought conditions in Ethiopia are not clearly observed, as several areas in the northern, central, and eastern regions experienced abnormally wet conditions during La Niña events, especially in JJAS. The result showed that the rainfall over Ethiopia during the summer monsoon (JJAS) is negatively correlated with SST indices, and as such the study of the links between ENSO and drought occurrences, lagged correlations between ENSO and regional drought indicators such as rainfall can provide critical information for water resources and seasonal forecasts.
ACKNOWLEDGEMENTS
Milica Stojanovic acknowledges the grant no. ED481B-2021/134 from the Xunta of Galicia (regional government), and Rogert Sorí the grant RYC2021-034044-I funded by Ministerio de Ciencia, Innovación y Universidades, Spain (MICIU/AEI/10.13039/501100011033) and the European Union Next Generation EU/PRTR. Both are supported by the Xunta de Galicia under Project ED431C2021/44 (Programa de Consolidación e Estructuración de Unidades de Investigación Competitivas (Grupos de Referencia Competitiva) and the Consellería de Cultura, Educación e Universidade) and by the European Union `ERDF A way of making Europe'. This work has also been possible thanks to the computing resources and technical support provided by CESGA (Centro de Supercomputación de Galicia) and RES (Red Española de Supercomputación).
DATA AVAILABILITY STATEMENT
All relevant data are available from an online repository or repositories. The Nino sea surface temperature and PDO time-series values were obtained from the US National Center for Atmospheric Research (https://climatedataguide.ucar.edu/climate-data/nino-sst-indices-nino-12-3-34-4-oni-and-tni). The SOI values were acquired from the National Oceanic and Atmospheric Administration (https://www.ncdc.noaa.gov). The IOD index was obtained from the Japan Agency for Marine-Earth Science and Technology (https://www.jamstec.go.jp/frsgc/research/d1/iod/).
CONFLICT OF INTEREST
The authors declare there is no conflict.
