Keywords

1 Introduction

Both changes and variability in the water resources have severely threatened the sustainable water resources development, particularly in the downstream part of the watershed. Therefore, determining the variability of stream runoff is essential task in water resources management and ecosystem restoration [1, 2]. The mean annual discharge provides a measure of the potential supply; intra-annual and multi-annual variation provides information which can be used to determine the probability of experiencing deficiencies under natural conditions and the amount of storage required [2,3,4]. In current situations, the hydrology of the earth is shifting with the potential to make floods and droughts more extreme [5]. There is now highly need for decision- makers to better understand the ongoing change and variation in hydroclimatic extremes in order to make preparations for the possibility of changing conditions.

The studies of changes have both a scientific and practical significance. The need to understand the impact that man is having on the ‘nature’ also another importance. The Wabi Shebele River Basin is one of frequently affected basin by hydrological extremes in Horn of Africa [6, 7]. For instance, in 1996, 1999 unexpected floods destroying homes and crops specially in three districts i.e., Kelafo, Mustahil and Burkur [8]. According to the local authorities, 34 people and 750 livestock died, with 70,000 affected by the floods in these areas. The flood of April 2005 was considered as the worst flood in past 40 years by locals, when 30,000 persons surrounded by flood waters and 6000 live stocks were washed away [7]. Ethiopian Ministry of Water, Irrigation and Electricity (MoWIE) study of Wabi Shebele River Basin master plan shows that severe hydrologic extremes in the basin especially in 1973, 1979, 1984–85 is caused by natural atmospheric variability [28]. Previous studies have focused on water resources potential assessment than trends and variabilities of hydroclimatic elements in the basin [9,10,11,12]. A few studies conducted on climate characteristics shows significant declines of annual and summer (June-September) rainfall total since 1982 in south eastern of Ethiopia, using progressive Man-kendall test on annual rainfall of Jijiga and Negele climate gauging stations [13]. Upper Wabi shebele basin at upstream of Melka Wekana is located in high rainy area, which its discharge is affected by rainfall conditions more similar to those of the high Ethiopian plateaus than to those of the downstream of Wabi Shebelle basin [9].

To understand the driving factors of changes in extreme discharges, it is important to look back to historical records and assess how variable the extreme discharges were temporally in the river basin. Also, it is useful to understand whether extreme events become more frequent or intense in the recent years. The statistics most commonly used to describe variability of river runoff are the standard deviation, and the coefficient of variation. Mann-Kendall test and Spearman tests are the most non parametric trend tests used in hydro climatic variables. However, the results of these statistical tests are often influenced by serial correlation and increase the chance of incorrectly rejecting the null hypothesis of no trend or vice versa in most hydro climatological data. To overcome this problem a rather novel approach named quantile perturbation method (QPM) [14–18] which is not dependent on the above-mentioned assumptions is utilized for analyzing temporal trend and variabilities in extreme hydroclimatic variables in this study. The study of temporal changes in extreme events identifies anomalies which can be attributed to different phenomena.

This study aims to assess temporal changes in extreme high discharges in Wabi Shebele River Basin and attribute to climate variables. Specifically, the study aims at: 1) analyzing variabilities in extreme hydro-climatic variables and 2) investigating the correlation in between extreme discharge anomalies and precipitation extreme anomalies for better communicating the science to water resource practitioners.

2 Materials and Methods

2.1 Study Area

Wabi Shebele River Basin is a transboundary basin in between Ethiopia and Republic of Somalia in horn of Africa. It originates from Bale mountain ranges of the Galama and Ahmar of Ethiopia, about 4000 m above sea level and drains portion of Somalia before draining to Indian Ocean. About 72% of the catchment (202,220 km2) is lying in Ethiopia. In this study, Wabi Shebele basin is used to represent the catchment that is lying in Ethiopia within 4o45' N to 9o45' N latitude and 38o45' E to 45o45' E longitude (Fig.1). The climate of the basin is dependent on the altitude and strong latitudinal movement of the intertropical convergence zone (ITCZ) [11, 19]. The highlands are cool and densely populated while the lowlands are arid and sparsely populated with recorded rainfall of 1487 mm and 220m respectively [2, 28]. While having the largest area coverage, the basin’s annual runoff is estimated to 3.4 BCM which is the lowest among the major river basins exist in Ethiopia [19]. The spatial variability of temperature is significant with mean, maximum and minimum value of 19.9 °C, 27.1 °C and 12.6 °C respectively. The mean annual evaporation of the basin ranges from around 1,000 mm in the north-west to 2800–3000 mm in the south-east.

Fig. 1.
figure 1

Map of study area

Full size image

2.2 Methods

Data.

Meteorological data including precipitation and temperature were collected from the National Meteorological Agency (NMA). Daily rainfall and temperature (maximum and Minimum) records for 7 stations (Table 1), with a good spatial distribution, were used to see the characteristics of precipitation and temperatures in this study. These stations are selected based on quality of data, length of record, from different watersheds and rainfall regime. The distribution of rainfall between watersheds shows variations (Fig. 3). Significant variation occurs during the rainy season (i.e. June to September). Considerable variation is also observed from April to June and from September to October.

Table 1. Annual summary of data used in the study.
Full size table

The discharge data in this study were collected from the Hydrology Department of the Ministry of Water, irrigation and Electricity (MoWIE). The measurements of river levels follow the guidelines of the World Metrological Organization (WMO) [20]. Five stations from upper, middle and lower part of the basin, relatively which have long record, one on main river and others on major tributaries of the basin were selected. The monthly database contains maximum and minimum discharges of each month in addition to the monthly average runoff, which assisted to construct the extremes time series. Main characteristics of selected discharge gauging stations are described below (Table 1 and 2).

The number of missed data in discharge is maximum in dry season (October-January) in Wabi Shebele River Basin (Table 2). This may be due to inappropriate placing of discharge gauging station. Adane, [19] reported that some of gauging stations in Wabi Shebele River Basin are installed on raised structures anchored to bridges, while others are off setted in pockets of the main course of the stream. Due to these, most of the stations has shown that the stations are not reliable to capture the low discharge situation in the basin. Therefore, such type of gauging station placing problem may be the cause for missed data especially in dry season. For existing gaps in between data, multiple regression method with adjacent discharge data and rainfall data is used to fill.

Table 2. Seasonal summary of data used in the study.
Full size table

Peak Over Threshold Selection (POT).

In this study, three highest value for each year from wet season is used to see annual and monthly hydroclimatic variability. In case of daily precipitation analysis, at least 15 extreme measurements per year is selected. To see extreme variability in precipitation and discharge at weekly aggregation level, at least 5 POT values are selected. In case of extreme discharge, similar procedures in precipitation are followed to select monthly and annual extreme discharge.

Trend Analysis.

Trends and variabilities in hydroclimatic data from year to year were determined by linear regressions and standard deviations. The statistical significancy of trends in the data was determined by the Mann-Kendall test. Both “ZMK” and “p” values were obtained to characterize trends and statistical significance. “ZMK” is the standardized Mann-Kendall (MK) statistic which follows the standard normal distribution with a mean of zero and a variance of one, and “p” is the probability value of the MK statistic also referred to the statistical significance (%). The value of near to zero regression gradient means no trend (the null hypothesis), whereas the value of regression gradient very different from zero means large trend in data (the alternative hypothesis).

Quantile Perturbation Method (QPM).

QPM is an empirical statistical analysis used to study trends and multi decadal oscillations in extremes [17]. It uses ranks of time series to detect frequency and perturbation of extreme discharge series in this paper. The perturbation is the ratio of similarly ranked data from the two series i.e. series in block length and refence series. The reference series is the long-term expected series while the other series is taken as the actual series within a particular block (sub-period). Then, for each block of years a single perturbation is calculated as the average of all perturbations above a particular threshold. Repeating the averaging over the different blocks assigns one factor to each block which eventually leads to a temporal variation of the perturbation factor. To identify periods of significant perturbations, the confidence interval is calculated and superimposed on the same plot. To calculate the confidence intervals, the values in the full time series at each site are randomly resampled to make a new series with different sequence, and the anomalies are recalculated for the resampled series based on the QPM method. The anomaly calculations are repeated 1000 times, leading to 1000 anomaly values for each block period. After ranking of the 1000 anomaly factors, the 25th and 975th values define the 95% confidence intervals for each block period. It is then graphically possible to identify periods of significant variations that the perturbation factors between the upper and lower limits of the confidence interval (the region of acceptance of the null hypothesis) are considered insignificant, whereas those outside the region of acceptance of the null hypothesis are defined as statistically significant.

Based on Tabbari et al., [21] recommendations, QPM is applied on different block length (5, 7 and 10-years) to select appropriate value of sub-period (block length) in between 5- and 15-years as preliminary analysis in the study. From preliminary analysis the block length which shows better oscillation patterns (high and low) of extreme value is selected for the whole time series variability analysis.

Pearson’s Correlation Coefficient.

Pearson’s correlation coefficient is used to see the strength of linear relationship between two variables. If the two variables are linearly related, the correlation coefficient will be near 1 or −1. The sign depends on whether the variables are positively or negatively related. The correlation coefficient related to zero (0) if there is poor relationship between the variables. Between rainfall in the Wabi Shebele basin and runoff in main and tributary rivers were determined at daily and annual time scales. The effect of seasonality in precipitation on river discharge was investigated separately by using data from the wet period from April - September (6 months).

3 Results and Discussion

The spatial and temporal variabilities in hydro-climate of Wabi Shebele River Basin is observed using data based statistical approach. Linear regression test, Mann-Kendall trend test and quantile perturbation method (QPM) are used to investigate possible trends and oscillation patterns in river discharge and climate variables. The linear trend test and Mann-Kendall trend test are performed to see trends on mean discharges, total rainfall and mean temperatures and then compare with previous results, while the QPM analysis was used to see trends and variabilities in hydroclimatic extremes due to its capacity to detect clear oscillating patterns and trends in extremes.

As observed on Figs. 2 and 3, the Wabi Shebele basin rainfall is bimodal type taking place from February-May and June – September on high land area and from March-May and September – November in lowland. In western - eastern upper and middle basin, the months from June-September (Summer) is the periods in which largest precipitation record exist, the months from march -may (Spring) is the period in which significant precipitation record exist. But in lowland around Gindhir, Gode and Degehabour, the months from March-May (Spring) and September-November (Autumn) are periods in which largest and significant rainfall records exist respectively (Fig. 2). Therefore, in highland and lowland of Wabi Shebele River Basin has two regimes of wet and dry season categories are exist. The months in which less precipitation is record exist collectively categorized as dry season in both regions. Accordingly, the highland part of the basin around Adaba, Merero, Robe (Arsi), Seru, Deder, Harar and Jijiga have dry season in between October and March, and wet season in between April and September. Whereas the lowland area of the basin around Gindhir, Degehabour, Gode have wet season (March to May; September to November) and dry season (December to February; June to August). The low-lying areas of Wabi Shebelle Basin around Degehabour, Gode, Kebridehar, and Kelafo rains from March to May is caused by moisture from the Indian Ocean, while the October to November rains is associated with the retreat of the ITCZ in a southward direction [11]. The temperature situation of sub basins shows that wet season is the season in which maximum temperature record exist specially February up to May, and the dry is the season in which minimum temperature record exist (Fig. 2).

Fig. 2.
figure 2

Mean monthly discharge distributions at selected gauging stations in record length.

Full size image

In all rivers, the discharge is recorded as maximum in wet season (July to September) season and minimum in dry season (October–March). Around 60% of the annual discharge in all rivers is from the heavy summer (June–September) rains, whereas around 19% of the annual discharge is from spring (February–May) rains.

Fig. 3.
figure 3

Mean monthly precipitation and temperatures distributions at selected gauging stations.

Full size image

3.1 The Trend Analysis

In this study linear regression test and Mann-Kendall trend test are performed to see trends and verify its significance on mean discharge, total rainfall and mean temperatures and then compare with previous results. A linear trend analysis to the annual mean runoff at all stations revealed an increasing trend at values of 1.45 Mm3, 0.58 Mm3 and 0.94 Mm3 on Wabi at Dodola bridege, Maribo and Robe respectively. The verification analysis on trend using Mann-Kendall trend test (Table 3) shows insignificant trends for the selected stations at annual and seasonal discharge levels (Table 3). In some stations significant increasing trend in discharge are observed in wet season, for example, in Wabi river at Dodola bridge and Erer river at western and eastern upper basin.

Trends in mean river discharge follows similar trends with rainfall exist in western and eastern upper basin sample gauging stations (Table 3 and 4). The spatial and temporal distribution of rainfall governs amount and intra and inter annual variability of discharges. This indicates that, Wabi Shebele River flow exhibits typical characteristics of tropical rainfall-dependent discharge regimes. Similar result in discharges was obtained from trend analysis on mean river discharge in between 1975 and 2015 in Wabi Shebele River Basin shows less or insignificant trends (Table 3).

Table 3. Statistical summary of trend test in mean annual discharge.
Full size table

Trend analysis on annual, seasonal and monthly rainfall indicates less or insignificant increasing trends in record length (1980–2013) at western and eastern upper Wabi Shebele Basin on Bale, Arsi and Harar highlands. But, significant decreasing trends in rainfalls is observed at middle and lowland area of the basin around Arsi robe, Gindhir and Degehabour in Wabi Sheble basin (Table 4). The wet season (April to September) rainfall has shown insignificant increasing trend in the basin. The study conducted by EPCC [22] also indicates insignificant trend of summer rainfall which range from +4 up to −5 mm/decade and substantial decreasing trends in spring rainfall during the last three decades from 1975–2007 in the basin.

Temperature trend analysis shows significant increasing trends in most of gauging stations located in upper and middle part of the basin except lower part of the basin around Degahabour which shows significant decreasing trend (Table 5). The south-eastern part of the low-lying areas of the basin around Degehabur, Gode, Kebridehare, and Kelafo receives no rainfall in July and August and has two rainy seasons [11]. The first is from March to May, and the second is from October to November. As tried to explained under Sect. 2.2, max temperature in the basin is recorded in wet season and minimum temperature is in dry season. The temperature trend difference from other part of the basin probably tied to rainfall distribution difference from others.

Table 4. Statistics summary of trend in annual rainfall.
Full size table
Table 5. Statistics summary of trend in mean temperature.
Full size table

3.2 Trends and Variabilities in Annual Extremes

Quantile perturbation method (QPM) is used to see possible trends and variabilities in extreme discharge and climatic variables. The method has a capacity to explore clear oscillating patterns and trends in hydroclimatic extremes. The method also has also the option of identifying the statistical significance of the variability observed and thus the statistical significance of the variability identified. The mean perturbation is assigned to a year which is approximately in the middle of the block. The value of confidence interval is superimposed on the same plot of extreme perturbations to identify the periods of significant variations.

From preliminary analysis conducted, 5 (five) year block of periods shows better oscillation patterns (high and low) of extreme discharges and precipitation in comparison to 7- and 10-years block lengths as shown on Fig. 4. Hence a 5-year block was used as block length of variability analysis in this study.

Fig. 4.
figure 4

Extreme perturbations at different block lengths on both discharge and precipitation.

Full size image

Extreme Discharge.

When considering extreme high discharges in annual time steps, most of perturbation values varies within confidence interval (Fig. 5). Among gauging stations increasing trend is observed on Maribo river. In early 1980s, significant negative perturbation up to −26.5%, −17.8%, −25% and −20% was observed in Maribo, Robe, Tebel and Erer watersheds respectively. Whereas, maximum significant perturbations up to 88.3%, 95.6%, 610% and 482% are observed respectively in watersheds in between 1986–1989. High oscillation pattern in annual extreme high discharges were observed in upper basin watersheds. Generally, extreme high discharge quantile perturbation shows increasing trend in Wabi Shebele River Basin from 1975 to 2015. The estimated confidence intervals are wide enough excluding the possibility of statistically significant change. Therefore, significant anomalies of extreme discharges can be due major driving factors like climate variables or catchment characteristics.

The oscillation pattern of extreme discharge quantiles changes within half a decade in both directions negatively and positively, which shows existence of external factors effects with in a given interval of years. For instance, the sea surface temperature change (SST) occurrence over the globe will occur within two to seven year interval in the form of El-Nino and La-Nino [23]. The effect this climate change indices is high over the world including our country. There is a consequence of flood and drought in Ethiopia during the occurrence year of these indices [24]. Most of positive significant anomalies of extreme discharge quantiles in Wabi Shebele River Basin occurred in moderate to very strong El-Nino years (Table 6). Contrary most of negative significant anomalies in the basin occurred in week to strong La Nino years. This indicates global climate change indices are a probable cause for changes of extreme discharge quantiles in Wabi Shebele River Basin. The impact of El Niño and La Niña are surprisingly increasing over the globe especially in developing countries like Ethiopia which their economies are dependent on regularly occurring seasonal weather conditions.

Table 6. Summary of QPM analysis in annual extreme discharge.
Full size table

The effects of El Niño southern oscillation had a variable impact on Horn African countries, ranging from drought to floods in history [25]. El Niño refers to the large-scale ocean-atmosphere climate interaction linked to a periodic warming in sea surface temperatures across the central and east-central Equatorial Pacific and La Niña represent periods of below-average sea surface temperatures across the east-central Equatorial Pacific [26]. The humanitarian impact of La-Niña is greater when it immediately follows an El-Niño. In 1988 El Niño year’s floods affected 2.5 Million peoples, whereas the 1999 and 2011 drought years occurred following La-Niña events affected 31.5 Million and 14 Million peoples respectively in the region including Wabi Shebele basin [24].

Fig. 5.
figure 5

Extreme discharge variability with confidence interval in Wabi Shebele River Basin.

Full size image

Precipitation Extremes.

The perturbation in precipitation extreme varies within confidence interval in most of stations of the basin taken in consideration. But unique result is observed on Adaba and Robe (Arsi) stations record (Fig. 6). An increasing and decreasing trend of precipitation extremes is observed in Adaba and Robe (Arsi) stations respectively. In early 1980s significant negative anomaly up to −10.4% is observed at Adaba station, but maximum positive perturbation of +38.2% is observed at Robe station. In 2000s significant positive anomaly of +137.8% is observed at Adaba and significant negative anomaly of −39.2% at Robe (Arsi) rainfall stations are observed. In the other side most rainfall stations indicate significant positive precipitation anomalies in early 1980s specially between 1982–1983 and immediate decreasing trend in between 1984 and 1985 (Table 7 and Fig. 6). Generally, most of rainfall perturbation analysis indicates decreasing trends in middle Wabi Shebele basin.

In Ethiopia, droughts and floods are occurred frequently at every 3–5 years for last 50-years [27]. The actual power generated from Melka Wakena Dam, the hydropower plant exist in this sub basin, shows power decrement in 1991, 1996 and 2000 by 32%, 21% and 38% relative to their Preceded years [20].

In other side western and eastern upper Wabi Shebele basin indicates significant positive anomaly in precipitation extremes in 2000s. This also coincided with the resent study result [7, 19, 24, 28, 29] which indicates several devastating floods in Wabi Shebele River Basin in recent years. The destructive flood in August 2005 on the basin affected 100,000, 154 deaths and the flood occurred starting November 2008, in basin caused around 52,000 human displacement from 14 kebeles (small admirative unit), 185 villages and 164-hectare farm lands washed away [29].

Table 7. Characteristics of highest anomaly in precipitation extremes.
Full size table
Fig. 6.
figure 6

Annual precipitation extreme perturbations using 5-year block length with 95% Confidence interval at seven /7/ gauging station of Wabi Shebele River Basin.

Full size image

Temperature Extremes.

There are a few limitations in the application of the frequency perturbation method for temperature compared to rainfall. First, the independence criterion is difficult to ascertain as there are correlation in daily value similar to river discharge. For this study, the independence criterion was ignored. Instead, only a threshold for the peaks was selected. Second, temperature significance periods were not based on 95% confidence intervals because there was no reasonable distribution valid for all the periods in temperatures [17]. Therefore, temperature anomalies were then analyzed based on the long-term average perturbation.

Extreme high temperatures in Wabi Shebele River Basin perturbation analysis shows general increasing trends in the basin particularly in upper and middle of the basin but less decreasing trend in lower part of the basin (Fig. 7). Starting 1990 significant increasing trends are observed in Adaba, Kofele, Robe gauging stations record.

Fig. 7.
figure 7

Extreme high Temperature average perturbations using 5-year block length at selected gauging stations in Wabi Shebele Basin.

Full size image

3.3 Trends and Variabilities in Seasonal Extremes

The QPM approach is applied to extreme value at seasonal aggregation level. Three maximum events are selected from each season to see temporal variability in extreme seasonal flood events. Two seasons, i.e. dry (October–March) and Wet (April– September), are used to see extreme quantiles in this study.

Extreme Discharge.

QPM analysis conducted on seasonal extreme discharges indicates significant positive anomalies in dry season streamflow of upper Wabi shabele basin at five years interval in average i.e., 1977, 1982, 1987, 1997 and 2008. These years are directly coincided with weak to very strong El-Nino years.

Table 8. Characteristics of highest anomaly in seasonal discharges.
Full size table

For Wet season quantile perturbation analysis also shows similar dry season perturbation oscillation patterns. There are both negative and positive significant anomalies occurred in upper Wabi shebele river discharge extremes at half a decade interval. Accordingly, the years 1976, 1982, 1986, 1995 and 2010 are the yeas in which positive significant anomalies observed in wet season. Whereas the years like 1978–80, 1984–85, 1990 and 2001–02 are the years in which significant decrement of extreme discharge observed in upper Wabi Shebele River Basin. Similar to dry season, wet season positive significant anomalies are separately observed in moderate to very strong scale El-Nino years. Whereas negatively significant anomalies are observed in La-Nina years.

Precipitation Extremes.

Except the years like 1988 and 1997, most of the precipitation extreme over Wabi Shebele River Basin changes within confidence interval. The 1988 the rainfall across the basin shows significant positive anomalies in the basin during dry season. It is also evident that, there is increasing trend in extreme precipitation on Adaba station rainfall. The 1988 was the year in which high flood disasters occurred in the horn of Africa as a whole [24]. Similar results are reported by MoWR [20] on upper Wabi Shebele basin around Adaba. In early 1990s and late 2000s significant increasing trend is observed in dry season.

In wet season (April- September), the extreme precipitation anomalies obeyed two phases in Wabi Shebele River Basin, decreasing trend in between 1980–1990 and less increasing trend in between 1990 to 2013. Early 1980s and 1990s are years in which significant positive and negative anomalies are observed respectively in most of stations (Table 9).

Table 9. Characteristics of highest anomaly in seasonal extreme precipitation
Full size table

Temperature Extremes.

QPM analysis on seasonal extreme high temperatures indicates increasing trend in both dry and wet season (Table 10). The size of positive trends decreases from western to eastern upper basin and indicates general decreasing trend in downstream of Wabi Shebele River Basin around Degehabour and Gode area. There is strong arguments in literatures that, with more water vapor and heat in the atmosphere, results for high storms [30,31,32].

Table 10. Characteristics of highest anomaly in seasonal extreme high temperatures.
Full size table

3.4 Effect of Aggregation Level on Detected Anomalies

The effects of time aggregation on variability analysis is investigated based on QPM analysis on daily, monthly and annual anomalies in the study area. In extreme discharge perturbation analysis, anomalies at daily aggregation level is less than both monthly and annual perturbation (Fig. 8). Whereas, the precipitation extreme perturbation at daily aggregation level are higher than both monthly and annual aggregation level (Fig. 9) which coincided with the study result of Ntegeka and Willems [17]. This indicates that, the extreme discharge perturbation conducted by QPM at different aggregation level gives completely opposite result with the study result of Ntegeka and Willems [17] in which the perturbations of smaller aggregation level are greater than larger level. It is known that the first day discharge has significant effect on the next day or lack of independency in daily discharge may be the cause for less anomaly in small time aggregation level.

Fig. 8.
figure 8

Comparison between extreme discharge perturbations at different aggregation levels

Full size image
Fig. 9.
figure 9

Comparison between precipitation perturbations at different aggregation levels.

Full size image

3.5 Correlation Analysis

Intergovernmental Panel on Climate Change (IPCC) [26] indicate that climate change “has detectably influenced” several of the water-related variables that contribute to floods, such as rainfall and snowmelt [26]. In this study the rainfall variability in Wabi shebele basin was taken as one of the major driving factors for flood change in the basin. To confirm this hypothesis, Pearson’s correlation analysis is conducted on extreme anomalies of discharge and precipitation in different watersheds of the basin (Table 11). The statistical significance of correlation value is checked using student test at 5% significance error.

The result indicates that oscillation pattern of both variables, i.e. extreme discharge and precipitation anomalies are similar even though the statistical correlation value shows a little bit small. From Table 8, it is evident that in dry season a better positive correlation value in between extreme discharge and precipitation up to 62%, 20%, 30%, 53% and 65% are observed at Maribo, Wabi at Dodola, Robe, Tebel and Erer watersheds respectively. Further in Wet season, low negative correlation value up to −42%, −42% and −9% are in Maribo, Wabi at Dodola and Robe watersheds respectively. Whereas positive wet season correlation value up to 31% was observed in Erer watershed. In general, statistical correlation using Pearson’s correlation coefficient method shows, less than 50% correlation value in average in between extreme discharge and precipitation correlation. This implies that for change in extreme discharge in Wabi Shebele River Basin, there is other driving factors in addition to climate elements. There was study result by MoWR, [20], in upper Wabi Shebele River Basin (upstream of Melka wakena) strong flood only occurs when soil has been previously saturated with water after the first rainy phase. This is because of not very intense rainfall conditions and considerable perviousness of soil in sub basin.

Table 11. Annual and seasonal correlation coefficient between discharges and rainfall.
Full size table

4 Conclusions

Temporal variabilities in hydro climate variables were examined for Wabi Shebele River Basin in between 1975–2015. The preliminary trend detection done on discharges at five sample rivers using linear trend regression test which is verified by Mann-Kendall significance test indicates less/insignificant increasing trends on mean annual discharge. Rain fall trend analysis conducted on total rainfall of different stations in the basin indicates significant decreasing trend in middle and lower part of the basin. The western and eastern highland area of the basin indicates less increasing trends in rainfall. The trend analysis conducted on mean temperatures indicates significant increasing trend in Wabi Shebele River Basin.

The temporal trend and variability analysis done on extreme high discharge using QPM approach indicates significant increasing trend in Maribo river from western upper basin and Erer river discharge from eastern upper basin. The other most extreme discharge anomalies vary within confidence interval. The dry and wet seasons anomalies mostly follow similar oscillation pattern with annual extreme discharge anomaly. Most of positive significant anomalies of extreme discharge quantiles in Wabi Shebele River Basin occurred in moderate to very strong El-Nino years. Contrary most of negative significant anomalies in this sub basin occurred in week to strong La Nino years. Therefore, we can conclude that global climate change indices may be the other influential factors for increments and decrements of extreme discharge quantiles in Wabi Shebele River Basin in addition to other driving factors in the sub basin. The precipitation extreme perturbations analysis shows increasing trend at all aggregation levels in western and eastern upper basin and decreasing trend in middle in between 1980 and 2018. QPM analysis on extreme high temperature indicates significant increasing trend with high oscillation pattern. The study conducted by EPCC [22] reveals similar result that substantial increasing trend in rainfall over south eastern highlands of Ethiopia and warming trend of about 0.4 °C/decade in the region which strengthen the current study.

The QPM analysis conducted to explore the effect of time aggregation level on extreme quantile perturbations indicates that, the perturbations for the annual aggregation level are generally greater than both monthly and daily perturbations in case of extreme discharge, which completely contradict to the result obtained in case of precipitation, in which extreme perturbation at daily aggregation level is greater than both monthly and annual perturbation. This indicates, the extreme discharge perturbation conducted by QPM at different aggregation level on extreme discharge gives completely opposite result with the study result of Ntegeka and Willems [17] done on precipitation extreme, which is the perturbations of smaller aggregation level are greater than larger level. The correlation analysis done between extreme discharge and precipitation indicates strong correlation of dry season extreme discharge and precipitation anomaly up to 65% in some watersheds, even though it falls below 50% in other sample watersheds.

Wabi Shebele River Basin is largest basin in Ethiopia with ungagged hydrologic characteristics. It is hard to characterize the whole basin extreme hydroclimatic variabilities with only five sub basins which most of it confined at upper highland area of the basin, while the basin has more than 11 sub basins with larger than 500 km2 catchment area. Therefore, extension of river discharges and precipitations in ungagged sub basin using robust model and analysis can explore better information on trends, temporal variabilities, correlations and periodicities in extreme hydroclimate variabilities of the basin.