Patterns and Teleconnection Mechanisms of Extreme Precipitation in Ethiopia during 1990–2020

Abstract

The occurrence of extreme precipitation events always leads to a mass of disasters. In this study, based on daily precipitation data from 20 meteorological stations in Ethiopia, we performed a detailed analysis of patterns and trends of ten extreme precipitation indices during 1990–2020. Our study revealed that different topographic conditions on the Ethiopian Plateau, Ethiopian savanna and Ethiopian desert resulted in great differences in patterns and trends of extreme precipitation. Notably, extreme precipitation intensity indices (Rx1day, Rx5day, SDII) and amount indices (R95pTOT) showed significant downward trends in the eastern desert (averagely −1.0 mm/year, −3.0 mm/year, −0.25 mm day⁻¹/year, −6.0 mm/year) and upward trends in the northern plateau and southern savanna (averagely 0.3 mm/year, 0.4 mm/year, 0.05 mm day⁻¹/year, 3.0 mm/year). These implied that extreme precipitation events decreased in the eastern desert and increased in the northern plateau and southern savanna during the past thirty years. Annual trends of the CDD index were upward (0.5 to 1.9 days/year) in most of Ethiopia while those of the CWD index were close to zero in most of Ethiopia, indicating that Ethiopia faced a longer duration of drought in the past thirty years. Moreover, we revealed that the local mean temperature, local mean precipitation, Southwest Asian summer monsoon and West African summer monsoon have significant impacts on the intensity, amount and duration of extreme precipitations in Ethiopia.

1. Introduction

Global warming is one of the greatest threats to human survival and seriously restricts the achievement of the United Nations Sustainable Development Goals in poor African countries [1,2]. The main factor causing global warming is the increase in global carbon emissions [3]. In March 2021, the concentration of carbon dioxide in the atmosphere reached more than 417 ppm at the Mauna Loa Observatory in Hawaii while the pre-industrial level was 278 ppm. Compared with the global air temperature during the period from 1850 to 1900, the mean air temperature during 2001–2020 rose by about 0.99 °C, and that from 2011 to 2020 rose by 1.1 °C. The past seven years (2015–2021) are on track to be the seven warmest on record [4]. Continued warming in the 21st century is projected to further intensify the global water cycle and cause risks of irregular monsoons, earlier rainfall seasons and increased flooding, leading to significantly negative impacts on many vulnerable aspects of agricultural and natural systems in many poor developing countries [5]. If the precipitation is too high, it will cause flood disasters, landslides, debris flows and other natural disasters, causing great economic losses to human production and life [6]. If the precipitation is too low, it will cause a serious shortage of water for agricultural use, land salinization and increased desertification [7]. Therefore, there is urgent need to measure evolution patterns and assess the driving mechanisms of extreme precipitations, especially for developing countries with rain-fed agriculture and weak industrial bases.

The ETCCDI has recommended a list of extreme precipitation indices to produce extremes’ estimates in a quick manner and enhance the understanding of regional extreme precipitation. Tan et al. [8] studied 14 extreme precipitation indices in the Kelantan River Basin in Malaysia in 1985–2014 and found, by using Mann–Kendall and Sen’s tests, that all indices had increased trends, except the CDD and CWD indices; notably, on a monthly scale, the Rx5day index showed a high increasing trend in January and December. Mathbout et al. [9] investigated extreme precipitation in the Eastern Mediterranean during 1961–2012 and found that the R99pTOT index had statistically significant decreasing trends for 74% of the stations while the R20mm index had increasing trends for 36% of the stations. Yang et al. [10] focused on the spatio-temporal pattern of extreme precipitation events in Canada and revealed that extreme precipitation in Canada became more severe from the mid-twentieth century, and strong links existed between extreme precipitation and ENSO/PDO/NAO, but the effects differed from region to region. Zarrin et al. [11] studied the daily precipitation records in Iran and concluded that the CWD index of 73.47% of the stations showed a decreasing trend, and the PRCPTOT index of 87.76% of the stations showed a downward trend, indicating that the drought situation in Iran is serious. Esmaeilpour et al. [12] further revealed that most trends of extreme precipitation indices in Northwestern Iran were insignificant; only the PRCPTOT and SDII indices showed significant negative and positive trends.

Africa is facing the danger of increasing drought under the global warming scenario, but extreme precipitation has not been studied sufficiently. Diatta et al. [13] studied the extreme precipitation in West Africa and revealed that extreme precipitation in the Sahel was strongly related to the Eastern Mediterranean Sea (EMS), whereas that of the Western and Northwestern Sahel was associated with AMM, MJO8, NINO.3.4, and TAPODI. Kouman et al. [14] studied the trend changes of extreme precipitation in the northeast of Côte d’Ivoire in West Africa and showed a decline in the PRCPTOT, SDII, and CWD indices, indicating a decrease in rainfall in the region. Ayugi et al. [15] studied extreme precipitation evolution over East Africa toward the end of the 21st century (2081–2100) relative to the baseline period (1995–2014) and demonstrated that SDII, R95pTOT, R20mm and PRCPTOT indices had significant changes during the OND (October to December) season compared to the MAM (March to May) season. Pinto et al. [16] studied extreme precipitation events over Southern Africa toward the end of the twenty-first century (2069–2098) relative to the reference period (1976–2005), and the Rx5day and R95pTOT indices were projected to increase significantly in the tropical and sub-tropical regions of Southern Africa and decrease in the extra-tropical region. Libanda et al. [17] investigated extreme precipitation over Zambia for 2021–2100 and showed that the CDD index had a high increase in Siavonga, Kasama and lsoka, up to the border of Zambia and Tanzania.

Due to rain-fed agriculture, the Ethiopian economy is vulnerable to precipitation since its agricultural sector, which employ 78% of the country’s labor force, heavily relies on precipitation [18]. The cumulative impact of climate-change-derived precipitation evolution has been responsible for a 33 percent yield decline in the Ethiopian economy over the last 60 years [19]. The vulnerability of precipitation evolution under global warming scenarios has urged the Ethiopian government to design strategies to mitigate the impacts of precipitation evolution [20]. In this study, we investigated the intensity, amount and duration of extreme precipitation in Ethiopia, and we explore possible driving mechanisms of extreme precipitation evolution in Ethiopia.

2. Study Area and Data

Ethiopia is located in the east of Sub-Saharan Africa and the center of the Horn of Africa [21], with a total land area of 1,103,600 square kilometers. Ethiopia is a highland landlocked country, and two-thirds of its territory is highland [22,23]. Its terrain is the highest among African countries, and it is known as the “roof of Africa”. In Western Ethiopia, many rivers flow westward along the slope, and the valley undercutting divides the highland into tableland with different sizes, flat tops and steep edges. Central Ethiopia is well preserved, with deep rivers forming canyons. Between the western and central parts of Ethiopia is the East African Great Rift Valley, which runs through Ethiopia, and the terrain to the East is relatively flat. Ethiopia has one of the most complex topographic features in Africa. Generally, Ethiopia can be divided into three regions: plateau, savanna and desert (Figure 1). The Ethiopian Plateau is located in Northwestern Ethiopia, and to its south is the Ethiopian savanna. The Ethiopian desert is located in Eastern Ethiopia and spreads from south to north.

Ethiopia is dominated by a tropical savanna climate and a plateau climate; in detail, since it is located in the tropical latitudes, the lower-elevation regions of Ethiopia experience tropical savanna or desert climate, while the higher-elevation regions of Ethiopia experience temperate climate. Its mean annual air temperature is 10~27 °C. The annual precipitation in the west and northwest of Ethiopia exceeds 1900 mm. When the southeast trade wind from the Southern Indian Ocean passes through the equator and merges with the equatorial west wind from the Gulf of Guinea into a strong southwest airflow, it brings abundant precipitation into Ethiopia, especially from July to August. Combining moist air from the Indian Ocean and the altitude in Ethiopia increasing from about 800 m in the eastern regions to about 3000 m in the western regions leads to the precipitation distribution of Ethiopia being generally greater in the west than in the east. There are three seasons in Ethiopia: a minor rainy season (called Belg) from February to May, a major rainy season (called Kiremt) from June to September, and a dry season (called Bega) for the rest of the year. The Kiremt rainfall during the June–September period accounts for 50–80% of the total annual rainfall in areas with a high agricultural productivity and filling of major reservoirs [24]. In this study, the daily precipitation data for a 31-year period (1990–2020) were collected from 20 meteorological stations in Ethiopia (

Table 1. Geographical location of 20 meteorological stations in Ethiopia.

Station	Latitude	Longitude	Elevation	Station	Latitude	Longitude	Elevation
Addis	38.80	8.98	2354	Jijiga	42.72	9.37	1557
Arba	37.56	6.06	1220	Jimma	36.82	7.67	1710
Awassa	38.48	7.07	1694	Jinka	36.56	5.78	1373
Bulki	36.81	6.28	2430	Konso	37.44	5.34	1431
Butajra	38.38	8.12	2074	Mehal Meda	39.66	10.31	3084
Debre	37.74	10.33	2446	Metehara	39.92	8.87	952
Degahabur	43.56	8.23	1070	Neghele	39.27	5.33	1439
Dire	41.90	9.61	1045	Tulu	38.21	8.66	2190
Gondar	37.43	12.52	1973	Wolaita	37.75	6.82	1854
Hosana	37.86	7.57	2306	Wolkite	37.77	8.28	1884

).

3. Methodology

In order to investigate patterns, trends and mechanisms of extreme precipitation in Ethiopia during 1990–2020, we selected ten extreme indices to reveal different aspects of the extreme precipitation [25]. These indices can be naturally classified into three groups: the intensity indices include Max 1-day precipitation intensity (Rx1day), Max 5-day precipitation intensity (Rx5day) and daily precipitation intensity (SDII). The amount indices include annual total precipitation (PRCPTOT), total precipitation during very wet days (R95pTOT), and total precipitation during extremely wet days (R99pTOT). The consecutive duration indices include the number of heavy precipitation days (R10mm), the number of very heavy precipitation days (R20mm), consecutive dry days (CDD) and consecutive wet days (CWD). A detailed description of ten extreme precipitation indices can be found in

Table 2. Ten extreme precipitation indices.

Classes	Indices	Full Names	Definitions	Units
Intensity Indices	Rx1day	Max 1-day precipitation amount	Maximum 1-day precipitation amount	mm
	Rx5day	Max 5-day precipitation amount	Maximum 5-day precipitation amount	mm
	SDII	Simple daily intensity index	The ratio of annual total wetday (days with precipitation over 1 mm) precipitation to the number of wet days	mm/day
Amount Indices	R95pTOT	Very wet days	Total annual precipitation from days with daily precipitation >95th Percentile	mm
	R99pTOT	Extreme wet days	Total annual precipitation from days with daily precipitation >99th Percentile	mm
	PRCPTOT	Annual total wet-day precipitation	Total annual precipitation from days with daily precipitation ≥1 mm Percentile	mm
Duration Indices	R10mm	Number of heavy precipitation days	Annual data count when daily precipitation ≥ 10 mm	days
	R20mm	Number of very heavy precipitation days	Annual data count when daily precipitation ≥ 20 mm	days
	CDD	Consecutive dry days	Maximum number of consecutive days with daily precipitation <1 mm	days
	CWD	Consecutive wet days	Maximum number of consecutive days when precipitation ≥1 mm	days

.

The nonparametric Mann–Kendall (MK) trend test and a Sen’s slope estimator are widely used to estimate trends and their significance of time series [26,27].

The MK test is based on the statistic $S$ of a time series ${\left\{x_k\right\}}_{k=1}^n$:

$$S=\sum _{k=1}^{n-1}\sum _{j=k+1}^n\mathit{sgn}(x_j-x_k)$$

where

$$\mathit{sgn}(x_j-x_k)=\left\{\begin{array}{c}1, if x_j-x_k>0\\ 0, if x_j-x_k=0\\ -1, if x_j-x_k<0\end{array}\right.$$

For independent and randomly ordered data, $S$ follows a Gaussian distribution with mean 0 and variance $\frac{n(n-1)(2n+5)}{18}$, so the value of $S$ can reveal the significance level of the trend.

Sen’s slope estimator is used to estimate the magnitude of the trend. For a time series ${\left\{x_k\right\}}_{k=1}^n$, the slope estimator $Q_{ij}$ between $x_i$ and $x_j$ is

$$Q_{ij}=\frac{(x_i-x_j)}{i-j} \mathrm{f}\mathrm{o}\mathrm{r} i=j+1,\dots ,n, j=\mathrm{1,2},\dots ,n-1$$

Sen’s slope of the whole time series ${\left\{x_k\right\}}_{k=1}^n$ is the median value in the set $\left\{Q_{ij}\right\}$. The Sen’s slope estimator is much better than classic linear regression since it is not affected by the number of outliers and data errors.

The main drawback of the MK test is that the strong autocorrelation embedded in the time series would increase the wrong trend significance level. In order to eliminate such negative effect, Zhang et al. [28] and Wang et al. [29] proposed an iterative trend estimator through pre-whitening the time series and then estimating its trend by fitting the following model:

$$x_t=\alpha +\beta t+\phi v_{t-1}+{\epsilon }_t$$

where $v_t=\phi v_{t-1}+{\epsilon }_t$ is a red noise model and ${\epsilon }_t$ follows a white noise. The first estimate of $\phi$ is directly calculated from the time series, and an estimator of $\beta$ is obtained from the Sen’s slope estimator. After that, the estimated $\beta$ is used to remove the trend from the original series, and the newly estimated $\phi$ is used to pre-whiten the original series to obtain the new $\beta$. When the difference between $\beta$ in two consecutive iterations is sufficiently small, the resulting $\beta$ is the trend of the time series [28]. Applying the MK trend test on this prewhitened time series can obtain the significance level of the trend [29].

In this study, we used the above iterative trend estimator to analyze the spatial patterns of the trends of ten extreme precipitation indices across the whole of Ethiopia. In order to further identify the associations of extreme precipitation indices with local mean temperature, local mean precipitation, Southwest Asian summer monsoon (SWASM) and West African summer monsoon (WASM), we used the difference method [30] to detrend these times series and then examined their correlations by the Pearson correlation coefficients.

4. Extreme Precipitation Evolution

We investigated patterns and trends of extreme precipitations in Ethiopia during 1990–2020 and focused on three aspects of extreme precipitations: intensity, amount and duration.

4.1. Intensity Indices

The Rx1day index measures the maximum 1-day precipitation. The mean annual Rx1day index during 1990–2020 was 42–63 mm in the whole of Ethiopia (Figure 2a). The highest value occurred in the Southwestern Ethiopian savanna, and the lowest value occurred in the Northeastern Ethiopian Plateau and the southeastern part of the Ethiopian desert. In the Ethiopian desert and Ethiopian Plateau, the range of the mean annual Rx1day index was 42–51 mm, about 10 mm lower than that of the Ethiopian savanna.

For seasonal distribution, in Ethiopian Plateau’s mean Rx1day index during the Belg season had the lowest value (23 mm), while the mean Rx1day index during the Kiremt and Bega seasons had the highest value (56 mm and 48 mm, respectively). This indicates that maximum daily precipitation in the Ethiopian Plateau rarely occurred in the Belg season (Figure 2b–d). In the Ethiopian savanna, the range of the mean Rx1day index during the Belg season was 35–51 mm, about 16 mm higher than that in the Ethiopian Plateau and 8–13 mm higher than that in the Ethiopian desert (Figure 2b). The mean Rx1day index during both the Kiremt and Bega seasons had a wide range in the Ethiopian savanna, which is consistent with the range in the whole of Ethiopia (Figure 2c,d). The mean Rx1day index during the Belg season was 35–43 mm in the southeastern Ethiopian desert, while the range of the mean Rx1day index during the Kiremt and Bega seasons was about 20 mm lower than that in the Belg season; therefore, the maximum daily precipitation in the southeastern Ethiopian desert mainly occurred in the Belg season (Figure 2b–d).

The trends of the annual Rx1day index were −1.5 to 0.4 mm/year in the whole of Ethiopia (Figure 2e,

Table 3. Trend range of Rx1day index in three regions of Ethiopia.

Regions	Ethiopian Plateau	Ethiopian Desert	Ethiopian Savanna
Annual	−1 to 0.4	−1.5 to 0.4	−1 to 0.4
Belg	0 to 0.5	−1.2 to 0.5	0 to 0.5
Kiremt	0.2 to 0.5	−1.6 to 0.5	−1 to 0
Bega	−0.7 to 0.5	−1.3 to 0.5	−1.3 to 0.5

). The highest increasing trend occurred in the southern Ethiopian savanna and the Ethiopian Plateau (0.4 mm/year). A significant decreasing trend occurred in the Ethiopian desert (−1.5 mm/year). In the Ethiopian desert, the trend variation of the Rx1day index during the Belg and Kiremt seasons was consistent with that of the annual Rx1day index: an increasing trend in the northern Ethiopian desert and a decreasing trend in the southern Ethiopian desert, while the Rx1day index during the Bega season had opposite trend features (Figure 2e–h). Comparing the Rx1day index during the Belg and Kiremt seasons in the savanna, the difference was in the northwestern savanna. The Rx1day index during the Belg season showed an increasing trend, ranging from 0 to 0.5 mm/year, but Rx1day during the Kiremt season showed a decreasing trend, ranging from −1 to 0 mm/year (Figure 2f,g).

The Rx5day index measures the maximum precipitation for five consecutive days. The mean annual Rx5day index was 61–114 mm, and the Ethiopian plateau and desert areas had the same variation ranges (61–114 mm). The mean annual Rx5day index in the Ethiopian savanna was 84–114 mm, and the lowest value was 23 mm higher than that in the other two regions. The range was 91–114 mm in the southern Ethiopian savanna, indicating that the 5-day consecutive precipitation amount in this area was relatively high (Figure 3a). This may be because the Great Rift Valley near the Omo River is low-lying and prone to updrafts, which further increase local rainfall.

For seasonal distribution, the mean Rx5day index during the Belg season was 44–67 mm in the Ethiopian Plateau, which was about 45 mm lower than the mean Rx5day index during the Kiremt season (Figure 3b,c). The mean Rx5day index during the Bega season ranged widely: 30–82 mm (Figure 3d) in the Ethiopian Plateau. In the southeastern part of the Ethiopian desert, the mean Rx5day index during the Belg, Kiremt and Bega seasons reached the lowest value of 44 mm, 24 mm and 22 mm, respectively (Figure 3b–d). In the north eastern part of the Ethiopian desert, the mean Rx5day index during the Belg season was 44–67 mm, and the mean Rx5day index was 22–56 mm during the Bega season, which was 20 mm and 40 mm lower than mean Rx5day index during the Kiremt season, indicating that this region would be drier during the Belg and Kiremt seasons (Figure 3b–d).

The trends of the annual Rx5day index were mostly downward in Ethiopia, with the largest upward trend occurring on the western edge of Ethiopia (1 mm/year) (

Table 4. Trend range of Rx5day in three regions of Ethiopia.

Regions	Ethiopian Plateau	Ethiopian Desert	Ethiopian Savanna
Annual	−1.4 to 1	−3.3 to 0.4	−1.4 to 1
Belg	−0.2 to 1	−3.1 to 1	−0.8 to −0.2
Kiremt	−0.8 to 1.2	−3.4 to 1.2	0.5 to 1.2
Bega	−1.1 to 1	−2.1 to 1.5	−2.1 to 1.5

), and the highest decreasing trend in the southeastern part of the Ethiopian desert region was −3.3 mm/year (Figure 3e). For trends of seasonal indices, the Rx5day index during the Belg and Kiremt seasons showed the fastest rising trend in the northwestern part of the Ethiopian Plateau, being 1 mm/year and 1.2 mm/year, respectively, while the Rx5day index during the Bega season showed the fastest rising trend in the southeastern desert (1.5 mm/year) (Figure 3f–h). In the Ethiopian savanna, Rx5day during the Belg season showed a significant downward trend, and the trend range was −0.8 to −0.2 mm/year (Table 4). Rx5day during the Kiremt season showed a significant upward trend in the southern savanna, and the range was 0.5–1.2 mm/year. The trends of Rx5day during the Bega season in the Afar Depression were prominently decreasing, and the range was −2.1 to −0.6 mm/year (Figure 3h), which indicated that the Afar Depression would be drier in the Bega season.

The SDII index represents the sum of precipitation in wet days (days with precipitation over 1 mm) divided by the number of wet days. Mean annual SDII in 1990–2020 was 7.9–11.1 mm/day in Ethiopia. The lowest value occurred in the interior of the Ethiopian desert in the southeast of Ethiopia, while the highest value occurred in the eastern edge of the desert and the west of Ethiopia. The mean annual SDII index was mostly over 9.3 mm/day in the Ethiopian Plateau and the savanna (Figure 4a).

For the seasonal distribution, in the Ethiopian Plateau, mean SDII during the Belg season was 7–9.1 mm/day, which was 2.5 mm/day lower than that during the Kiremt season. This was consistent with the abundance of rainfall brought by the southwest and southeast monsoons in the Ethiopian Plateau during the Kiremt season (Figure 4b,c). In Lake Abaya, located in the south savanna, the mean SDII index during the Belg and Bega seasons was 9.9–11.3 mm/day and 9–10.5 mm/day, respectively, while mean SDII during the Kiremt season was slightly lower (7.8–9.7 mm/day) (Figure 4b–d). This indicated that the intensity of extreme precipitation in Lake Abaya Basin was not necessarily high in the major rainy season (Kiremt). In the northeastern desert, mean SDII during the Kiremt and Bega seasons showed similar spatial distributions with mean annual SDII, whereas the slight inconsistency was that mean SDII during the Bega season had the lowest value in the southeastern part of the Ethiopian desert (Figure 4a,c,d).

The SDII during the Bega season had the highest decreasing trend (−0.06 mm·day⁻¹/year) (

Table 5. Trend range of SDII index in three regions of Ethiopia.

Regions	Ethiopian Plateau	Ethiopian Desert	Ethiopian Savanna
Annual	−0.13 to 0.07	−0.28 to 0.07	−0.08 to 0.07
Belg	−0.03 to 0.11	−0.22 to 0.11	−0.13 to 0.11
Kiremt	−0.2 to 0.12	−0.32 to 0.12	−0.13 to 0.12
Bega	−0.22 to 0.1	−0.14 to 0.1	−0.06 to 0.1

), which was 0.22 mm·day⁻¹/year lower than that of the annual SDII (Figure 4e,h). In the northern Ethiopian Plateau, the SDII during the Belg and Kiremt seasons manifested a 0.09 mm·day⁻¹/year increasing trend on average, which was 0.04 mm·day⁻¹/year higher than that of the annual SDII index. In the southwestern savanna, the annual SDII index had a remarkable upward trend, and the range was 0.02–0.07 mm·day⁻¹/year (Figure 4e). Meanwhile, SDII during the Belg and Kiremt seasons in the southwestern savanna also exhibited an upward trend, and the ranges were 0.02–0.11 mm·day⁻¹/year and 0.06–0.12 mm·day⁻¹/year, respectively (Figure 4f,g). Although SDII during the Bega season showed a significant upward trend in the southwestern savanna, it showed a downward trend in the northern savanna (Figure 4h). Details are organized in Table 5.

4.2. Amount Indices

The R95pTOT index measures the annual sum of precipitation in days where daily precipitation exceeds the 95th percentile of daily precipitation. Mean annual R95pTOT was 166–304 mm in the Ethiopian Plateau and 63–304 mm in the Ethiopian desert and savanna (Figure 5a). The lowest value of mean annual R95pTOT in the Ethiopian Plateau was 166 mm, which was 93 mm higher than that in the Ethiopian desert and savanna.

For the seasonal distribution, the highest value of mean R95pTOT during the Belg, Kiremt and Bega seasons did not exceed 30 mm, and the lowest values were all found in the southeastern part of the Ethiopian desert (6 mm, 6 mm and 1 mm, respectively) (Figure 5b–d). Mean R95pTOT during the Belg and Bega seasons showed the lowest values in Northeastern Ethiopian Plateau, being 6 mm and 1 mm, respectively, while mean R95pTOT during the Kiremt season showed the highest value of 24 mm in the same region (Figure 5b–d). Compared with mean R95pTOT, during the Belg season, mean R95pTOT during the Kiremt season was 9 mm higher than that in the Afar Depression (Figure 5b,c). In most of the Ethiopian savanna, mean R95pTOT during the Kiremt season was 16–24 mm, which was 10 mm lower than the range of mean R95pTOT during the Belg season and 6 mm lower than the range of mean R95pTOT during the Bega season (Figure 5b–d).

The trends range of annual R95pTOT was −13.2 to 4.7 mm/year, but trends in most regions were concentrated from −5.5 to 4.7 mm/year (

Table 6. Trend range of R95pTOT index in three regions of Ethiopia.

Regions	Ethiopian Plateau	Ethiopian Desert	Ethiopian Savanna
Annual	−8.1 to 4.7	−5.5 to 4.7	−13.2 to 4.7
Belg	−0.4 to 0.4	−1 to 0.4	−0.4 to 0.4
Kiremt	−0.2 to 0.4	−1 to 0.4	−0.6 to 0.4
Bega	−2.3 to 0.5	−2.3 to 0.5	−0.7 to 0.5

). The significant downward trends ranging from −3 to −0.4 mm/year were observed in the central part of the Ethiopian desert and in the northern Ethiopian savanna, and the significant upward trends ranging from 2.1 to 4.7 mm/year were observed in the southwestern part of the Ethiopian savanna (Figure 5e). R95pTOT during the Belg and Kiremt seasons showed no significant change in one-third of Ethiopia, while R95pTOT during the Bega season showed no significant change in 75% of Ethiopia (Figure 5f–h). This suggests that the amount of precipitation in a considerable part of the country had not increased in three seasons. Compared with R95pTOT during the Belg and Kiremt seasons, R95pTOT during the Bega season showed an increasing trend in the southeast corner of the desert where the trend range was 0–0.5 mm/year (Figure 5f–h). In most parts of Abaya Lake in the southern savanna, R95pTOT during the Belg and Kiremt seasons showed an upward trend, while the trend range of R95pTOT during the Bega season was −0.3 to 0.1 mm/year (Figure 5f–h).

The R99pTOT index measures the annual sum of precipitation in days where daily precipitation exceeds the 99th percentile of daily precipitation. The range of the mean annual R99pTOT index was 18–84 mm, and the range for the west and northeast was 46–84 mm, which was 28–38 mm higher than that in the southeast region of Ethiopia (Figure 6a).

For the seasonal distribution, the highest value of mean R99pTOT during the Belg season occurred on the western edge of the Ethiopian savanna (10 mm), and the lowest value of that was distributed in the northern Ethiopian Plateau and southeastern desert areas (1 mm) (Figure 6b). The highest value of mean R99pTOT during the Kiremt season occurred in northern Ethiopia and the southeastern desert areas (6.5 mm); the lowest values occurred in the western and southeastern savanna (1.9 mm) (Figure 6c). In most parts of the Ethiopian Plateau, mean R99pTOT during the Belg season was 1–4.8 mm, which was 1.7–2.8 mm lower than that of mean R99pTOT during the Kiremt season (Figure 6b,c). Mean R99pTOT during the Belg season in the Blue Nile Basin was 1–4.8 mm, which was 1.7–4.2 mm lower than mean R99pTOT during the Bega season (Figure 6b,d). Therefore, the amount of precipitation in the Blue Nile Basin during the Belg season was less than during the Bega season.

The trend range of the annual R99pTOT index was −2.1 to 1.48 mm/year in Ethiopia, and none of the trends were significant (Figure 6e,

Table 7. Trend range of R99pTOT index in three regions of Ethiopia.

Regions	Ethiopian Plateau	Ethiopian Desert	Ethiopian Savanna
Annual	−0.56 to 1.48	−1.59 to 0.05	−2.1 to 0.46
Belg	0–0.02	0–0.02	0–0.07
Kiremt	−0.11 to 0.02	−0.11 to 0.02	−0.04 to 0.02
Bega	0	0	0

). The highest upward trend occurred on the western edge of the Ethiopian Plateau (1.48 mm/year), and the highest downward trend occurred on the western savanna (−2.1 mm/year) (Figure 6e). For trends of seasonal indices, R99pTOT during the Belg season only had an upward trend in the central savanna, and R99pTOT during the Kiremt season only had a downward trend in the northern and southeastern parts, while R99pTOT during the Bega season showed no obvious trend in the whole of Ethiopia (Figure 6f–h). This suggested that there was a slight increase in the minor rainy season, a slight decrease in the major rainy season and no change in the dry season.

The PRCPTOT index represents the annual sum of precipitation in wet days (days where precipitation is at least 1 mm). The highest value of the annual PRCPTOT index was 1566 mm (Figure 7a), and the highest values of the seasonal PRCPTOT index were 561 mm (Belg season), 925 mm (Kiremt season) and 418 mm (Bega season) (Figure 7b–d). The highest value of PRCPTOT during the Kiremt season was about twice that of PRCPTOT during the Belg and Bega seasons, since the Kiremt season is the Ethiopian monsoon season, and heavy precipitation always occurs in this season (Figure 7b–d). In the Northwestern Ethiopian Plateau, mean RPCPTOT during the Kiremt season was 421–925 mm, which was 287–730 mm higher than that of mean PRCPTOT during the Belg season and 325–721 mm higher than that of mean PRCPTOT during the Bega season (Figure 7b–d). In the southeastern savanna, compared with the range of mean PRCPTOT during the Belg season, the range of mean RPCPTOT during the Kiremt season was 91–210 mm higher (Figure 7b,c). This showed that the precipitation in the southeastern savanna was relatively low, even in the major rainy season (Kiremt). Mean PRCPTOT during the Belg and Bega seasons had the lowest value in the southeastern desert, being 134 mm and 42 mm, respectively, while mean RPCPTOT during the Kiremt season showed the highest value, being 925 mm (Figure 7b–d).

The trend range of the annual PRCPTOT index was −7.1 to 7.1 mm/year in most parts of Ethiopia, but there was a large downward trend of more than 20 mm/year in the western part of the Ethiopian savanna (Figure 7e,

Table 8. Trend range of PRCPTOT index in three regions of Ethiopia.

Regions	Ethiopian Plateau	Ethiopian Desert	Ethiopian Savanna
Annual	−21.2 to 7.1	−7.1 to 7.1	−42.5 to 7.1
Belg	−3.7 to 3	−12.6 to 3	−10.4 to 3
Kiremt	−13.4 to 6	−8.5 to 6	−27.9 to 6
Bega	−4.4 to 3.8	−1.7 to 3.8	−5.8 to 3.8

). Regardless of the trends of the annual or seasonal PRCPTOT, the Northern Ethiopian Plateau and the Southern Ethiopian savanna showed an upward trend, which indicated that precipitation in these places would increase in each season (Figure 7e–h). The trend of PRCPTOT during the Kiremt season was a downward trend near the Blue Nile River Basin in the southwest Ethiopian Plateau (Figure 7g). In this region, the highest trend was −23.1 mm/year, which was 17.2 mm higher than that of PRCPTOT (−6 mm/year) during the Belg season and 20 mm/year higher than that during the Bega season (−3 mm/year) (Figure 7f–h).

4.3. Duration Indices

The R10mm index measures the annual count of days where daily precipitation is more than 10 mm per day. The range of the mean annual R10mm index was 9–57 days in Ethiopia. For the Ethiopian desert region, the range of mean annual R10mm was 9–37 days. In the southeastern Ethiopian desert, the range of mean annual R10mm was 9–16 days. The range of mean annual R10mm in the Ethiopian Plateau was 30–57 days, which was consistent with the Ethiopian savanna. The highest mean annual R10mm was 57 days, which mainly occurred in the western part of Ethiopia (Figure 8a).

For the seasonal distribution, the range of mean R10mm during the Belg season was 4–21 days, with the highest value occurring in the southwestern part of the Ethiopian savanna and the lowest value in the Ethiopian Plateau and Ethiopian desert (Figure 8b). The range of mean R10mm during the Kiremt season was 1–35 days, which was wider than that of the other two seasons (Figure 8c). The highest value in the Kiremt season was 14 days more than that in the Belg season, occurring in the Ethiopian Plateau. The lowest value was distributed in the southeastern desert and the southern savanna (Figure 8b,c). The range of mean R10mm during the Bega season was 1–14 days, and, like the other two seasons, the southeastern desert region also had the lowest value in this season (Figure 8d).

The trend range of annual R10mm for the whole of Ethiopia was −1.53 to 0.27 days/year, with the highest decline in the Northwestern Ethiopian Plateau, southern desert and savanna. In the southeast of the Ethiopian desert, the trends of annual R10mm were upward, ranging from 0.01 to 0.27 days/year, which would help alleviate the drought in the Ethiopian desert (Figure 8e).

The trends range of R10mm during the Belg season was −0.97 to 0.43 days/year (

Table 9. Trend range of R10mm index in three regions of Ethiopia.

Regions	Ethiopian Plateau	Ethiopian Desert	Ethiopian Savanna
Annual	−0.76 to 0.27	−0.76 to 0.27	−1.53 to 0.27
Belg	−0.57 to 0.43	−0.57 to 0.43	−0.97 to 0.43
Kiremt	−1.01 to 0.21	−0.48 to 0.21	−1.01 to 0.21
Bega	−0.13 to 0.08	−0.08 to 0.13	−0.17 to 0.13

). A significant decreasing trend occurred in the southwestern savanna, being from −0.37 to −0.17 days/year (Figure 8f). During the Kiremt season, the northern Ethiopian Plateau and the southern savanna showed an upward trend, with the maximum upward trend reaching 0.21 days/year, which decreased by 0.22 days/year compared to the Belg season (Figure 8f,h). In the Bega season, the variation range of the trends value was smaller than that in the previous two seasons, ranging from −0.17 to 0.13 days/year, indicating that the duration of precipitation did not increase or decrease significantly in the Bega season, and the duration of rainfall changed mostly in the Belg and Kiremt seasons (Figure 8f–h).

The R20mm index measures the annual count of days where daily precipitation is more than 20 mm per day. Mean annual R20mm was 3.8–22.7 days, with the highest value occurring in Western Ethiopia and the lowest value in Southern Ethiopia. The Ethiopian Plateau’s values were 9.2 to 22.7 days, with the lowest value being 5.4 days higher than the lowest value in Ethiopia, indicating a higher precipitation duration in the Ethiopian Plateau region (Figure 9a).

Mean R20mm during the Belg season was 1.2–9.1 days, with the highest value occurring in the western savanna and on the northeastern edge of the Ethiopian desert (Figure 9b). Mean R20mm during the Belg season in the Blue Nile Basin of the Ethiopian Plateau was 1.2–4.6 days, while mean R20mm during the Kiremt season in the Blue Nile Basin was 12.1–14.1 days, which was about 10 days higher than that in the Belg season, indicating that the precipitation in the Blue Nile Basin had a longer duration in the Kiremt season (Figure 9b,c). Mean R20mm during the Bega season was 0.2–5.6 days, with the lowest value occurring in the southeast of the Ethiopian desert area (Figure 9d). Although mean R20mm during the Bega season was 0.2–2.5 days for most of the Ethiopian desert region, the range was 2.5–4.1 days in the northeastern fringe of the Ethiopian desert, which may be due to winds across the Red Sea bringing precipitation (Figure 9d).

The trend range of the annual R20mm index was −0.92 to 0.21 days/year, with large declines in the Western Ethiopian Plateau (−0.92 to −0.27 days/year) (Figure 9e,

Table 10. Trend range of R20mm index in three regions of Ethiopia.

Regions	Ethiopian Plateau	Ethiopian Desert	Ethiopian Savanna
Annual	−0.92 to 0.21	−0.27 to 0.21	−0.6 to 0.21
Belg	−0.13 to 0.02	−0.13 to 0.13	−0.24 to 0.08
Kiremt	−0.59 to 0.13	−0.18 to 0.13	−0.28 to 0.13
Bega	−0.07 to 0.06	−0.03 to 0.06	−0.07 to 0.06

). The trend range of R20mm during the Belg season was −0.24 to 0.13 days/year, and there were no significant trends in most regions during the Belg season (Figure 9f). The trend range of R20mm during the Kiremt season was −0.59 to 0.13 days/year, and the trend distribution of R20mm during the Kiremt season was basically similar to that of annual R20mm (Figure 9e,g). In the Blue Nile Basin, located on the Ethiopian Plateau, the trends of R20mm were mainly downward in the Bega season, ranging from −0.07 to −0.01 days/year (Figure 9h).

The CDD index measures the maximum number of consecutive days with a daily precipitation of less than 1 mm. Mean annual CDD was 36–134 days, with the lowest values occurring in the southwestern savanna and northeast of the Ethiopian desert, and the highest values occurring in the southeast of the desert (Figure 10a).

Mean CDD during the Belg season was 18–66 days and was similar to mean annual CDD in distribution, but the difference was that mean CDD during the Belg season was relatively high in the northwest region of the Ethiopian Plateau (Figure 10a,b). The range of mean CDD during the Kiremt season was 6–75 days. Except for the northern part of the Ethiopian Plateau, mean CDD during the Kiremt season was relatively low, ranging from 6 to 36 days, indicating that the Ethiopian Plateau was relatively wet during the Kiremt season (Figure 10c). Mean CDD during the Bega season was 28–96 days (Figure 10d). In the southeastern desert region, mean CDD during the Bega season was 86–96 days, and combined with the range of mean CDD during the Belg season, which was 59–66 days, the range of CDD during the Kiremt season was 65–75 days, indicating that the number of consecutive days without rain would increase in three seasons, and the areas would become more arid (Figure 10b–d).

The trend range of annual CDD was 0.27–2.62 days/year (

Table 11. Trend range of CDD index in three regions of Ethiopia.

Regions	Ethiopian Plateau	Ethiopian Desert	Ethiopian Savanna
Annual	0.27 to 1.95	0.27 to 2.62	0.27 to 2.62
Belg	−0.5 to 1.2	−0.5 to 1.2	−0.5 to 1.2
Kiremt	−0.94 to 0.18	−0.94 to 0.18	−0.94 to 0.18
Bega	−0.37 to 1.08	−0.37 to 3	−0.11 to 3

), indicating that the number of consecutive days without rain was increasing throughout Ethiopia. In addition, in the southern part of the Ethiopian Plateau, the southwestern part of the Ethiopian savanna and the eastern part of the Ethiopian desert, the trends of annual CDD were significantly upward (Figure 10e).

The trend range of CDD during the Belg season was −0.5 to 1.2 days/year. In the southeastern Ethiopian Plateau and the southern Ethiopian savanna, the trends were declining with a range of −0.5 to −0.01 days/year (Figure 10f). The trend range of CDD during the Kiremt season was −0.94 to 0.18 days/year, and the trends were rising near the Afar Depression, ranging from 0.02 to 0.18 days/year (Figure 10g). The trend range of CDD during the Bega season was −0.37 to 3 days/year, with a decline in the northern Ethiopian Plateau, where the highest decline reached −0.37 days/year (Figure 10h).

The CWD index measures the maximum number of consecutive days with a daily precipitation of at least 1 mm. Mean annual CWD was 3–21 days, with higher values in the northern Ethiopian Plateau. The range was 14 to 21 days in the Northern Ethiopian Plateau, which was 5–10 days higher than that in the Southern Ethiopian savanna (Figure 11a).

Mean CWD during the Belg season was 3–9 days and was lower in the Ethiopian Plateau and most of the Ethiopian desert, ranging from 3 to 5 days (Figure 11b). Mean CWD during the Kiremt season varied from 2 to 21 days, and the mean in the Ethiopian Plateau was relatively high in this season (Figure 11c). Mean CWD during the Bega season was 2–7 days, and the maximum value decreased by 2 days compared to the Belg season and 14 days compared to the Kiremt season (Figure 11b–d).

The trend range of annual CWD was −0.25 to 0.1 days/year (Figure 11,

Table 12. Trend range of CWD index in three regions of Ethiopia.

Regions	Ethiopian Plateau	Ethiopian Desert	Ethiopian Savanna
Annual	−0.25 to 0.03	−0.11 to 0.1	−0.18 to 0.03
Belg	0.01 to 0.07	−0.14 to 0.04	−0.11 to 0.04
Kiremt	−0.24 to 0.17	−0.11 to 0.23	−0.24 to 0.17
Bega	−0.05 to 0.08	−0.07 to 0.08	−0.07 to 0.08

), and both the increasing and decreasing trends in Ethiopia were not significant. The trends were downward in the northwestern part of the Ethiopian Plateau and upward in the Afar Depression in the northeastern desert region (Figure 11e). The trend range of CWD during the Belg season was from −0.14 to 0.07 days/year (Figure 11f). Compared with the downward trends of annual CWD, the northwest of the Ethiopian Plateau showed an upward trend of CWD during the Belg season, but the upward trends were not significant (Figure 11e,f). The trend range of CWD during the Kiremt season was −0.24 to 0.23 days/year, and the trends rose near the northern Afar Depression, ranging from 0.03 to 0.23 days/year (Figure 11g). The trends of CWD during the Bega season ranged from −0.1 to 0.08 days/year, and there was no obvious change in trends in more than half of the regions (Figure 11h). The maximum of the upward trends was 0.08 days/year, which appeared in the Northern Ethiopian Plateau, the Southwestern Ethiopian savanna, and the Southeastern Ethiopian desert.

5. Teleconnection Mechanisms

In this section, we explore links and derive mechanisms of extreme precipitations in Ethiopia during 1990–2020.

5.1. Link with Mean Annual Temperature

We examined the link between mean annual temperature and the annual extreme precipitation indices (Figure 12). The intensity indices of extreme precipitation include the Rx1day, Rx5day and SDII indices. These three indices were negatively correlated with mean annual temperature in the southern savanna, and the correlations reached −0.54, −0.68 and −0.61, respectively (Figure 12a–c). In the southern Ethiopian Plateau, the correlations with the Rx1day and SDII indices were significant and moderately positive, being 0.13–0.29 and 0.11–0.47, respectively (Figure 12a,c). In the southeastern desert, the correlations with SDII were 0.47–0.65, which were about 0.18 higher than those with the Rx1day and Rx5day indices (Figure 12c).

For the amount indices of extreme precipitation, the correlations with R95pTOT were 0.12–0.14 positive, and they were 0.24–0.37 positive with PRCPTOT, while they were −0.37 to −0.02 negative with R99pTOT in the Northern Ethiopia Plateau and savanna (Figure 12d–f). In the southern savanna, three indices had a significant negative correlation: among them, the correlation with R95pTOT reached −0.6, which was 0.11 higher than that with R99pTOT, and was 0.05 higher than that with PRCPTOT (Figure 12d–f). The R95pTOT and PRCPTOT indices showed positive correlations in most of the Ethiopian desert, while R99pTOT showed negative correlations, ranging from −0.49 to −0.2 (Figure 12d–f).

For the duration indices of extreme precipitation, the correlations with the R10mm, R20mm and CWD indices were significantly negative in the central part of the Ethiopian savanna, while the correlations with CDD were significantly positive in similar places (Figure 12g–j). In the northeastern desert, the R10mm and R20mm indices were positive, while CDD showed significantly negative correlations, ranging from −0.81 to −0.64. In the western savanna, only CWD had negative correlations, but they were not significant (Figure 12g–j).

5.2. Link with Mean Annual Precipitation

We examined the link between mean annual precipitation and the annual extreme precipitation indices (Figure 13). For the extreme precipitation intensity indices, the correlations with the Rx1day and Rx5day indices were significantly positive, except for the southwestern Ethiopian savanna, while the correlations with SDII were significantly positive, and the range was 0.55–0.88 in Ethiopia (Figure 13a–c). The correlations with Rx1day were strong in the northern Ethiopian Plateau and desert, and the range was 0.62–0.72, which was approximately 0.3 higher than that in the southern savanna (Figure 13a). The correlations with Rx5day were strong in the Western Ethiopian Plateau and savanna, and the range was 0.67–0.75, which was 0.2 higher than that in the southern savanna (Figure 13b). SDII had the smallest correlation in the western savanna, and the range was 0.55–0.6, which was about 0.28 lower than that in the Ethiopian Plateau (Figure 13c).

For the amount indices of extreme precipitation, the correlations with the R95pTOT, R99pTOT and PRCPTOT indices were significantly positive in Ethiopia (Figure 13d–f). The correlations with R95pTOT were strong in the western Ethiopian Plateau and the southern desert, and the range was 0.78–0.83, which was approximately 0.3 higher than that in the southwestern savanna (Figure 13d). The correlations with R99pTOT were strong in the Ethiopian Plateau and the northern desert, and the range was 0.5–0.63, which was 0.35 higher than that in the southern savanna and desert (Figure 13e).

For the duration indices of extreme precipitation, the correlations with the R10mm and R20mm indices were significantly positive in the whole of Ethiopia, and those with CWD were negative in a small region of Ethiopia, while the correlations with CDD were negative in most of Ethiopia (Figure 13g–j). The correlations with R10mm were strong in the southwestern savanna, and the range was 0.9–0.97, and correlations with R20mm were strong in the southern savanna, and the range was 0.84–0.95, which was about 0.1 higher than that in the Ethiopian Plateau (Figure 13h). The correlations with CDD were strongly negative in the Ethiopian Plateau and reached −0.81, while the positive correlations occurred in the southern savanna and the edge of the desert (Figure 13i). CWD had negative correlations in the northern and western Ethiopian Plateau and desert, but they were weak, being −0.11(Figure 13j).

5.3. Link with Southwest Asian Summer Monsoon Index (SWASMI)

We examined the link between Southwest Asian Summer Monsoon Index (SWASMI) and annual extreme precipitation indices (Figure 14). For the intensity indices of extreme precipitation, the correlations with the Rx1day and SDII indices were significantly positive and exceeded 0.5 in the Northeastern Ethiopian desert, while the positive correlations were not prominent with Rx5day in this region (Figure 14a–c). In the Southeastern Ethiopian desert, the correlations with three indices were negative and about −0.3. In the Western Ethiopian Plateau and desert, the correlations with the Rx5day and SDII indices were negative while the correlations with Rx1day were positive (Figure 14a–c).

For the amount indices of extreme precipitation, the correlations with R95pTOT were highly positive in the Northeastern Ethiopian Plateau, and the value was 0.49–0.62 while the correlations with other extreme precipitation amount indices were negative in the same position (Figure 14d–f). Furthermore, the correlations with three indices were all significantly positive in central Ethiopia, and the magnitude of correlations was about 0.3 (Figure 14d–f). In the Western Ethiopian desert, R95pTOT was negative, and the range was −0.34 to −0.2, which was about 0.08 higher than that with R99pTOT (Figure 14d,e).

For the duration indices of extreme precipitation, the correlations with the R0mm and R20mm indices were significantly positive in central Ethiopia, and the ranges were 0.42–0.53 and 0.41–0.52, respectively (Figure 14g,h). In the Southern Ethiopian desert, the correlations with CDD were highly negative, and the range was −0.5 to −0.37, while the correlations with CWD were positive, and the range was 0.37 to 0.48 (Figure 14i,j). In the Western Ethiopian savanna, the correlations with the R10mm and R20mm indices were negative, and the magnitude of the correlations was also similar, ranging from −0.23 to −0.12 (Figure 14g,h).

5.4. Link with West African Summer Monsoon Index (WASMI)

We examined the link between the West African summer monsoon index (WASMI) and the annual extreme precipitation indices (Figure 15). For the intensity indices of extreme precipitation, the correlations with the Rx1day and Rx5day indices were significantly positive, and the highest value reached 0.45 and 0.5, respectively (Figure 15a,c). In the Northwestern Ethiopian Plateau, the correlations with the Rx1day and Rx5day indices were negative while the correlations with SDII were positive (Figure 15a–c).

For the amount indices of extreme precipitation, the correlations with R95pTOT were significantly positive in the Northeastern Ethiopian desert, which were about 0.09 higher than those with R95pTOT and 0.15 higher than that those with PRCPTOT (Figure 15d–f). In Western and Southeastern Ethiopia, the correlations with three indices were negative, and the values were about −0.4 (Figure 15d–f).

For the duration indices of extreme precipitation, the correlations with R10mm were not significant in the whole of Ethiopia, and the range was −0.31 to 0.33 (Figure 15g). The correlations with R20mm were significantly positive in the Ethiopian Plateau and negative in the Southeastern Ethiopian savanna and desert (Figure 15h). Although the correlations with R10mm were not significant, the spatial correlation distribution was similar to that with R20mm (Figure 15g,h). The correlations with CDD were significantly negative in the central and Southern Ethiopian savanna, and the range was −0.49 to −0.16 (Figure 15i), and the correlations with CWD were significantly negative in the Southeastern Ethiopia desert (Figure 15j).

6. Conclusions

Extreme precipitation is an important part of extreme climate change. It is of great significance to study the distribution, trend and mechanism of extreme precipitation in Ethiopia in order to mitigate climate change. Therefore, we analyzed in detail patterns and trends of extreme precipitation through 10 widely used extreme precipitation indices. These indices can reflect different aspects of extreme precipitation, including extreme precipitation intensity, extreme precipitation amount and extreme precipitation duration. Three indices (Rx1day, Rx5day, SDII) reflect extreme precipitation intensity, three indices (R95pTOT, R99pTOT, PRCPTOT) reflect extreme precipitation amount and four indices (R10mm, R20mm, CWD, CDD) reflect extreme precipitation duration. Our analysis found that extreme precipitation evolution in the Ethiopian Plateau, Ethiopian savanna and Ethiopian desert are as follows:

➢

Ethiopian Plateau: the Rx1day, Rx5day, SDII and R95pTOT indices had increasing trends in the northeast, and the trends were 0.2–0.4 mm/year, 0.4–1 mm/year, 0.02–0.07 mm/year and 2–4.7 mm/year, respectively; the trends in the southwest decreased slightly, and the trends were −0.4 to −0.1 mm/year, −0.8 to −0.2 mm/year, −0.2 to −0.08 mm/year and −8.1 to –0.4 mm/year, respectively; R99pTOT had a slightly decreasing trend in the east, and the trend was −0.56 to −0.05 mm/year; PRCPTOT and CDD had no trend; R10mm and R20mm had slightly increasing trends, and the trends were 0.01–0.27 days/year and 0.05–0.21days/year, respectively. CWD had a slightly increasing trend in the southeast (−0.25 to −0.04 days/year) and a decreasing trend in the northwest (0.03–0.1 days/year).

➢

Ethiopian savanna: Rx1day had an increasing trend in the south, and the trend was 0.2–0.4 mm/year; Rx5day and R10mm had significantly decreasing trends in the central savanna, and the trends were −0.8 to −0.2 mm/year and −0.5 to −0.25 days/year, respectively; SDII and R95pTOT had significantly increasing trends in the southeast, and the trends were 0.02–0.07 mm/year and 2–4.7 mm/year, respectively; R99pTOT and PRCPTOT had increasing trends in the southeast, and the trends were 0.46–0.97 and 0–7.1 mm/year, respectively; R20mm had a significantly decreasing trend in the north, and the trend was −0.27 to −0.11 days/year; CDD had a significantly increasing trend in the southwest, and the trend was 1.28–2.28 days/year; CWD had no trend in most areas.

➢

Ethiopian desert: the Rx1day and SDII indices had increasing trends in the north, and the trends were 0.2–0.4 mm/year and 0.02–0.07 mm/year, respectively; they also had decreasing trends in the south, and the trends were −1.5 to −0.4 mm/year and −0.28 to −0.03 mm/year; the Rx5day and R95pTOT indices had decreasing trends, and the trends were −3.3 to −0.2 mm/year and −3 to −0.4 mm/year; CDD showed a significantly increasing trend (1.95–2.62 days/year) in the southeast. R99pTOT, PRCPTOT, R10mm, R20mm and CWD had no trend in the central desert.

Due to poor research conditions and the unstable political environment in East Africa, almost all extreme precipitation studies are based on modeling data (e.g., 15, 17). Our study is the first investigation on extreme precipitation in Ethiopia conducted directly through observed precipitation data from 20 meteorological stations in Ethiopia. Our study revealed that different topographic conditions in the Ethiopian Plateau, Ethiopian savanna and Ethiopian desert resulted in great differences in the patterns and trends of different extreme precipitations. Generally, extreme precipitation intensity indices (Rx1day, Rx5day, SDII) and amount indices (R95pTOT) showed significant downward trends in the Eastern Ethiopian desert and upward trends in the Northern Ethiopian Plateau and Southern Ethiopian savanna. These implied that extreme precipitation events decreased in the eastern desert and increased in the northern plateau and southern savanna during the past thirty years. In terms of the duration indices, annual trends of CDD were upward in the whole of Ethiopia while those of CWD were slight, indicating that Ethiopia faced a longer duration of drought in the past thirty years, and it is expected that Ethiopia would be dryer in the near future. Moreover, we revealed that the local mean temperature, local mean precipitation, Southwest Asian summer monsoon and West African summer monsoon have significant impacts on intensity, amount and duration of extreme precipitations in Ethiopia.