Assess the impacts of climate change on the patterns of rainfall, temperature, and streamow in the Abelti Watershed of Southwestern Ethiopia

Changes in rainfall, temperature and streamow (stf) will be one of the most critical factors determining the overall impact of climate change (CC). Thus, in this study we evaluated rainfall(rf), temperature, and stf pattern under changing climate in the Abelti-Watershed (a sub-watershed of upper Omo Gibe basin), Ethiopia. The Representative Concentration Pathway (RCP) scenarios of Hadley Global Environment Model 2-Earth System (HadGEM2-ES) under Coordinated Regional Climate Downscaling Experiment (CORDEX)-Africa database selected for the near (2011-2040), med (2041-2070), and end (2071-2100) periods. Hydrologic Engineering Centers-Hydrologic Modelling System (HEC-HMS) model applied for stf projection. XL-STAT conducts average annual and seasonal rf, minimum and maximum temperature (tmin&tmax), and stf trend tests. Mean seasonal and annual rf and stf variation evaluation taken using the coecient of variation (CV). Finally, the impact of CC analysis is taken based on the baseline period. The results revealed that the climate model projection is successful for given weather stations. HEC-HMS model showed a satisfactory performance during calibration (R2=0.82) and validation (R2=0.78). The MK trend of tmin&tmax show signicantly increasing; whereas rf and stf show insignicantly decreasing except under RCP8.5. The rf and stf CV analysis indicated less, moderate, and high in the study area. And the future long year average annual rf increased by -3.6%, -1.9% and -7.7%; temperature +1.15%, +2.2% and +4.2%; and stf -2.9%, -0.05% and -8.5% under RCP2.6, RCP4.5 and RCP8.5 respectively. Thus, the decrement in rf and the increment in temperature lead to more evapotranspiration and affect the stf negatively. In conclusion, stf in the Abelti-watershed could signicantly decline with adverse consequences for water supplies, agriculture, and ecosystem health for the future. Therefore, this study may contribute to the planning and implementation of sustainable resources development and management strategies and help to mitigate the consequences of CC.

Introduction incorporating the latest information and easily adjusted to new areas. The information will contribute to future planning, managing water resources, working together towards nding lasting solutions, and will act as input for further studies. And also, it may act as a reference for future studies of cc impact on extreme events.

Materials And Methodology
The study area description Abelti-watershed is the main sub-basins of the upper Omo-Gibe River basins and characterized by a complex topography and deeply dissected steep slopes. The altitude of the watershed ranges from 3389 meters above sea level in the Shambu escarpment to an elevation of 1065 meters Abelti gage station. The area of Abeltiwatershed is 15746 km2 and which is 46.28% of the total area of the Omo-Gibe River basin. The watershed lies between 7.2 0 N & 9.5 0 N latitude and 36.5 0 E & 38.3 0 E longitude (Fig. 1). Most of the watershed from the upper part drains largely cultivated land with less permeable soils. The area falls under the eleven-land use/land cover (LU/LC) types according to Arc-GIS version 10.5. Among these, the moderately cultivated area was found to be largest, followed by bushed shrubbed grassland, intensively cultivated, disturbed high forest, dense mixed high forest, perennial crop cultivation, dense bushland, state farm, waterbody and afro-alpine heath vegetation (Fig. 2). Whereas the soil classi cation falls under eight major categories. Among these, Orthic Acrisols was found to be largest, followed by Dystric Cambisols, Pellic Vertisols, Chromic Vertisols, Orthic Luvisols, Chromic Luvisols, Lithosols, and Eutric Nitosols (Fig. 3). Also, the climate of the area is characterized by the intermediate climate of tropical sub-humid system. And rf varies from highest1800 mm to lowest 900 mm per annum, and it is monomodal. The mean monthly tmin&tmax is about 22.5 0 C and 14.5 0 C respectively.  (Collins et al., 2011) for the period 1971-2100 was obtained from the CORDEX Africa database under CMIP5. The earth system components of the HadGEM2-ES climate model compare well with observations and with other models (Jones et al., 2011). Even this model was used in several CC impact studies in Ethiopia and found consistent with other GCMs (e.g., Niguse and Aleme, 2015;Bhattacharjee and Zaitchik 2015;Urgaya 2016). The RCP is a greenhouse gas concentration (not emissions) trajectories adopted by the IPCC for its fth Assessment Report (AR5) in 2014. RCP is the latest generation of scenarios, not lie on a xed set of assumptions related to population growth, economic development, or technological change. And it is spatially explicit and provides information on a global grid at a resolution of approximately 50km's and in which concentration and land use trajectories and further local details can be considered in it. Hence, compared to the previous scenarios it has the advantages of incorporating the latest information and easily adjusted to new areas.
Observed data analyses A considerable number of observations are missing in the weather and hydrological time series in the study.
Then these temporal and spatial data were screened and pre-processed by Excel spreadsheet and ARC GIS 10.5 respectively. Missed value of observed rf data was lled by Inverse Distance weighted (IDW) method (Chow et al., 1988). This method well considers spatial information than temporal correlations to interpolate (Ramos et al., 2008). A multiple linear regression method was used stf. Then the lled out of observed rf and stf data were subjected to initial quality checks such as consistencies and homogeneity by using XL-STAT software based on Pettitt's test (Pettitt's 1979). The result null hypothesis Ho and Ha according to P-value applied to detect a single change-point or continuous data. As the computed p-value is greater than the signi cance level alpha=0.05, one cannot reject the null hypothesis Ho (homogeneous) and as the computed p-value is less than the signi cance level alpha=0.05, one cannot reject the null hypothesis Ha (not homogeneous).
Climate projection data analysis Downscaled rf and tmin&tmax data of RCP2.6 (low emission scenarios), RCP4.5 (medium stabilization emission scenarios), And RCP8.5 (highest emission scenario without CC policy) (van Vurren et al., 2011) were subjected to bias correction. The bias correction procedures are used to minimize the discrepancy between observed and simulated climate variables on a daily time step. so that hydrological simulation are driven by corrected simulated climate data. Seven grid points were selected based on their proximity to the weather station in the watershed. The values of the predicted climate variables are calculated by the IDW method for each climate gage station, to predict their future values are the function of the size of the grids (Sylwia and Emilie, 2014). It relies on the observed climate and local topography features (Charron, 2014) to derive relatively ne spatial resolution of future projections.
In this study, we computed bias correction factors from the statistics of the observed and simulated variables.
For rf are determined the correction factors by matching the mean and coe cient of variation (CV) of simulated data with that of observed data (Leander and Buishand, 2007). For temperature, monthly systematic biases were calculated for the baseline period by comparing RCP outputs with the observations (Ho et al., 2012). The nonlinear correction for each daily precipitation P and a linear bias correction for temperature T was applied and transformed to a corrected p* and T*, respectively, using: The coe cient a and b are determined iteratively.
After bias correction, RCPs datasets were used for the two time periods in this study; the 1971-2000 baseline period and the 2011-2100 future period. And the impact analyses were developed for the period of near (2011-2140), med (2041-2070), and end (2071-2100) in the studied watershed under RCP2.6, RCP4.5, and RCP8.5 respectively. Then rf at a daily time scale is the single most important parameter in hydrological computer simulation models and researchers are often faced with the problem of which interpolation technique to use when determining the spatial distribution of daily rf from rain-gage positions. As the rf over a large area is not uniform, the mean depth of rf over the study area is determined by the Thiessen polygon method (Subramanya, 2009). In this method, we attempted to allow for non-uniform distribution of gages by providing a weighting factor for every seven gages in the GIS 10.5 window and the mean areal depth of rf of observed and the three RCPs were calculated.

Hydrological Model analyses
The major role hydrologic model is to generate the future stf for the projected climate. Know a day there are many models to simulate rf stf for various reigns and catchment sizes e ciently. For this study HEC-HMS model is selected because the model is applicable in many different size catchments and under considerably different conditions in the world; the same is true in Ethiopia more recently (Sintayehu, 2015). The model input parameters were prepared by a combination of different spatial analyst model tools like; Arc GIS (Zhong, 2014), Arc-Hydro Venkatesh, 2012), and Hec-GeoHMS (USACE, 2000 andMatthew andJames, 2013). Basically, by this study, the Snyder unit hydrograph method to model the transformation of excess rf into direct stf and de cit (initial loss) and constant loss method were used. The parameters have also been optimized using the optimization tools available in HEC-HMS. To get the optimum parameter values after manually calibrating the model, an automatic trial and error method was applied. The Nelder and Mead optimization method was used than the univariate method because the Nelder and Mead method use downhill simplex to evaluate all parameters simultaneously and adjust once (Feldman, 2000). Each watershed or sub-watershed responds to excess rf to form peak stf and stf volume in the outlet. rf events in each sub-basin could be different from other sub-basin events and have their unique output over the watershed. However, there are common "statistical" characteristics of the main processes, which produce the events. The three methods were used to evaluate the model performance during the calibration and validation periods, such as Nash and Sutcliffe simulation e ciency (NSE) (Nash and Sutcliffe, 1970), coe cient of determination (R2), Relative Volumetric Error (RVE) according to their respective range values (Moriasi et al., 2007).

The trend analyses
The trend analysis was carried out by using the non-parametric Mann-Kendall (MK) (Mann, 1945;Kendall, 1975) in XL-STAT software. MK trend test is a statistical test used to estimate the trend in hydro-climatic time series that has an advantage which can accommodate non-linear trends and data need not be normally distributed (Yue and Wang 2004). Also, it has low sensitivity to abrupt breaks due to inhomogeneous time series (Tabari et al. 2011 andPingale et al., 2016). To accomplish the analyses, we aggregated time series data of each climate data (rf, tmin&tmax) to areal mean. Then we checked the trend by graphically and XL-STAT statistical interpretation. As the p-value is greater or less than the signi cance level α = 0.05 at the con dence level of 99%; Ho, there is no trend; and Ha, there is trend respectively. And also, Kendall's tau or the Z-value was used to analyze either the trend is decreasing or increasing.

The mean Coe cient of Variation (CV) Analyses
The variability of mean seasonal and seasonal rf and stf were analyzed by CV and interpreted. The CV is simply the standard deviation divided by the average annual rf times by 100. According to Hare (1983), %CV is used to classify the degree of variability of rf events as less, moderate and high. When %CV<20% it is less variable, %CV from 20-30% is moderately variable, and %CV>30% is highly variable. For this study, annual, monthly, and seasonal variabilities analysis was taken for Ethiopian local season of Belg/Spring (February March, April, and May), Kiremt/Summer (June, July, August, and September), and Bega/Winter (October, November, December, and January) over Abelti Watershed.

Results And Discussions
Consistency and Homogeneity: To select representative meteorological stations, checking the homogeneity of group stations is essential. The consistency and homogeneity of rf and stf by using Pettitt's test (Pettitt, 1979) in XL-STAT software show that the observations are from the same population and homogeneous. Due to the computed p-value of each mean annual rf and four stf gage stations data is greater than the signi cance level alpha=0.05 at a 99% con dence interval, one cannot reject the null hypothesis (H0) ( Table.1) and the red horizontal broken line is shown in Fig.4 implies homogeneity of the rainfall station. And which assures that both observations are from the same population and homogeneous. The Thiessen analysis of seven weather stations revealed that ve are inside and two are out of the watershed but nearby the watershed. In which the analysis of areal mean rf is characterized by bimodal nature. And also, the general rf patterns of the watershed show very high seasonality and exhibit typical characteristics of tropical. Hence, the spatial and temporal distribution of rf governs the amount and intra-and inter-annual variability. In terms of seasons, in most parts of the watershed, rf is heavily concentrated in March, April, and May, and the second heavily rf concentration is in August, September, and October (Fig.5a). In both these seasons, 70-80% of the total annual rf occurs, and areal mean monthly (amm) rf approximately varies from the highest 200 mm to the lowest 40 mm and the average annual rf is 1404.70 mm/year.
Performance of HadGEM2-ES model: It is important to check performance the baseline or observed climate data replicate or not before use. In this study, the simulated ma rf using the HadGEM2-ES model was used and compared with the observed average annual rf of seven stations. Table.2 illustrates the performance of the climate model in the study checked by root mean square error, observation standard deviation ratio (RSR) which is a root mean square error divide by standard deviation the result is found the range between 0.19 to 0.69, correlation coe cient (R2) found between 0.89 to 1 and Nash-Sutcliffe e ciency found between 0.83 to 1, and then the result shows a good agreement and performance. The observed, uncorrected, and corrected control period of the areal mean monthly rf pattern in Fig.5a indicates there is a good agreement between them. Mean monthly rf analysis in the control period indicated that the area characterized monomodal system, high rf relatively recorded in August, and low rf recorded in January. The average monthly, tmin&tmax graphs ( Fig.5b) also show a similar pattern with observed climate datasets over the watershed.
HEC-HMS Model: Assessment of CC impact on water resources in a river requires a proper estimation of availability of water and that can only be achieved by hydrological modeling of the basin. In this analysis, sensitive analysis adopted for evaluating the model sensitive parameters were initial loss, constant rate, and peaking coe cient (Cp). The standard lag parameter is estimated from extracted information on the basin and drainage network. Initial loss and peak coe cient were found to be the most sensitive parameters for the simulated stream ow for all the ve sub-basins. The model has been calibrated systematically and automatically to optimize and obtain the best possible option. The model well predicts stf depth and peak values in the watershed with the optimized parameters (Table.3). The hydrograph, during the calibration and validation period, is reasonably well in simulating the stf in the study watershed ( Fig.6a and b). On the other hand, the simulated ow in these years may have a larger degree of uncertainty since rainfall was highly variable in space and time (Dile et al. 2013). The model results are reasonably accurate for high ow. The objective functions which were used to evaluate the model performance are presented in Table.3. Thus, the statistical test (Willems, 2008) of error function also justi es the validation of the model for simulation of stf hydrograph in the study watershed. The statistical parameter indicates a good model performance in terms of capturing the observed stream ow volume (RVE=-0.014% and 0.021% respectively during calibration and validation period), the pattern of stream ow hydrographs (NSE=0.82 and 0.78 during calibration and validation period respectively), and coe cient of determination (R 2 = 0.74 and 0.75 during calibration and validation period respectively).

MK Trend Test result
A very thorough analysis of the mean annual temperature, rf, and stf trend was checked by Kendall statistics by XL-STAT. This Nonparametric MK statistical test was used to detect trends in rf, tmin&tmax over the time series aggregated to areal mean. The result of the MK test gives interesting insight about annual tmin&tmax, rf and stf data for the Abelti-catchment according to XL-STAT P-values and Z-value interpretation. As the p-value is greater than the signi cance level α = 0.05; Ho, there is no trend; and as the P-value is less than the signi cance level α = 0.05; Ha, there is a trend. And also, Kendall's tau or the Z-value was used to analyze either the trend is decreasing or increasing.
Rainfall (rf): Changes in rf are variable from region to region, and may include changes in the amount, intensity, frequency, pattern, and type. Areal ma observed, baseline and future rf of seven stations over the Abeltiwatershed calculated by Thiessen method and subjected to Mk very thorough analyses in XL-STAT. MK of areal average annual rf all period under study in Table.4 result shows a mix of increasing and increasing insigni cantly. There is an insigni cant increasing trend is detected in observed, end RCP2.6 and three nonoverlapping RCP8.5 periods. While there is insigni cant decreasing trend is detected in the periods of baseline, near and med of RCP2.6 and three non-overlapping of RCP4.5. General sense the long year areal average annual future rf are decreasing trend pattern under RCP2.6 and RCP4.5 and increasing under RCP8.5. Again, the long year areal average annual rf Mk trend graph in Fig.7a shows there is a slightly decreasing trend in both RCP 2.6 and RCP 4.5 and an increasing trend in RCP 8.5. this result is closely much with the report of IPCC (2014) in which rf showed both increment and decrement. And the decreasing trend patterns in the southwestern parts of Ethiopia under the latest scenarios of RCPs near and med period supported by EPCC; Gebre and Ludwing (2015). The increment and decrement of rf in the future over the study area may lead to unexpected ooding and drought.
Based on the above results, it is of immense importance to discuss the ecological, economic, and social impacts that could result if decreasing rainfall trends continue in the future. And also, this discrepancy in the results may be mainly attributed due to climate variability and watershed characteristics. For rural farmers who have been vulnerable to drought, water stress, and the erratic nature of rainfall, appropriate adaptation strategies should have to be designed and implemented. The vulnerability of rural households might further be aggravated if extreme rainfall episodes continue in the future and consequently result in drought and surface and groundwater losses due to evaporation and overutilization.
Tmin&tmax: Recorded meteorological data analysis of tmin&tmax also indicates that increasing trends have been observed almost in all parts of the catchment. The MK test of areal average annual tmin&tmax shows an increasing trend in observed, baseline, and future three non-overlapping period RCPs scenarios. According to Z and P-value analyses result (Table.4), the average annual tmin&tmax is a statistically signi cant increasing trend in the periods of observed, near RCP2.6, near and med RCP4.5, and med and end RCP 8.5. Whereas in the other case tmin&tmax insigni cantly increasing except end period of RCP2.6. On the other hand, in the future long year, the three RCPs average annual tmin&tmax MK trend graph ( Fig.7b and c) show an increasing trend over the catchment. The tmin&tmax trend increment graphically in RCP8.5 is more than both RCP2.6 and RCP4.5 ( Fig.7b and c). In general MK statistics of areal average annual, tmin&tmax is shown an increasing trend and passé the test of signi cance of a 5% con dence level, which proved that there was obvious warming up tendency throughout the catchment. The result, in general, implies that the MK-test has shown an increasing trend for the Tmax over the area. And which proved that there was obvious warming up tendency throughout the watershed.  Fig.7c) clearly show the stf trend more related to rf in the area although land use has its in uence.
As CC warms the atmosphere, altering the hydrologic cycle, changes to the amount, timing, form, and intensity of precipitation will continue. Another expectation is stf change patterns in watersheds are more related to rf. In some cases, trends in stf are not always consistent with changes in rf (Kiros, 2016). This may be due to the coverage of rf, the effect of human interventions, land use or land cover, soil conditions and types, topographic factor, and environmental modi cations.

Coe cient of Variation (CV)
Rainfall: Rainfall: Changes in rf are variable from region to region, and may include changes in the amount, intensity, frequency, pattern, and type. The CV is simply the standard deviation divided by the average annual rf times by 100. The overall percentage CV of areal mean seasonal and annual rf analyses was found between 10.1% and 99.3% (Table.5). This is an indication of the degree variability of rf events in the study area experiences less, moderate, and high. The average annual CV rf of observed, baseline, and the three nonoverlapping periods under three RCPs was found under 20% and it is an indication of less variation of rf annually. The percentage CV of the observed, baseline, and the three non-overlapping period future RCPs rf projection of Bega, Belg, and Kiremt seasons experiences high variability over the watershed. But in the Kiremt seasons' rf of observed, baseline and near period under RCP2.6 CV shows less and moderate variations. Generally, the watershed is more in uenced by seasonal rf variations thus for CV is greater than 30% means that high variability in the seasons. This also indicates that the watershed is vulnerable of the season to CC and vulnerable to hydro-meteorological extremes.
Stream ow: The CV analyses of mean seasonal and annual stf were found between 21.5% and 86.6 % ( Table.5).
The degree variation in the case of average annual stf during observed, baseline, and future non-overlapping periods CV is moderately varying in the watershed. Bega and Belg seasons stf CV in all studied periods show high variation (CV greater than 30%). In the Kiremt season stf CV analysis shows both moderate and high. The large no of CV in the Bega and Belg season stf similar to rf variation in those seasons over the area. The larger no of CVs indicated that the difference of stf among years at the same season was larger, which would put pressure on water management in the catchment. Generally, the study watershed is highly in uenced by seasonal variation of rf and stf, which may put pressure on natural and man-made infrastructures.
Rainfall (rf): The future RCPs projections of areal mean monthly rf show both increment and decrement in the study watershed (Fig.8a). Respectively 70% under RCP2.6 and 38% under RCP8.5 are the largest numbers of mean monthly rf increment in October and decrement in March in the near period. The mean monthly largest number increment value in the med period is up to 112% in the November month and the decrement value is up to 54% in March month under RCP8.5. Whereas in the end period respectively up to 144% and 68% are the largest increment and decrement under RCP8.5. General decrement of mean monthly rf revealed in February to May and increment in August to December in three non-overlapping periods of RCP2.6, RCP4.5, and RCP8.5. On the other hand, areal average annual and seasonal rf projection in the Table.6 reveals decrement over the watershed except for the Bega season which shows increment.
tmin&tmax: The future projections of areal ma and seasonal tmin&tmax of the study show increment in three RCPs ( Fig.8b and c). The mean monthly tmin increment respectively under RCP2.6, RCP4.5, and RCP8.5 in the near period up to 2.1oc, 1.7oc, and 2.2oc; med period 2.4oc, 2.9oc, and 3.5oc; and end period 2.1oc, 2.9oc, and 5.2oc. Whereas mean monthly tmax increment respectively under RCP2.6, RCP4.5, and RCP8.5 in the near period up to 1.5oc, 1.4oc, and 1.5oc; med period 1.7oc, 2.3oc and 3.2oc; and end period 1.4oc, 2.6oc and 4.9oc. The variation of future mean monthly tmin is magni cent compared to tmax over the watershed. The tmin&tmax variation of the end period is greater than both near and med period under the three RCPs. And also, the med period variation is more than the near period under three RCPs. On the other hand, mean seasonal and annual tmin&tmax in the Table.6 revealed there is an increment tendency over the watershed. The analyses of long-year ma temperature increments over the study are 1.3oc, 1.9oc, and 3.1oc under RCP2.6, RCP4.5, and RCP8.5 respectively. The mean seasonal and annual med period of tmin&tmax increment is more than both near and end period under RCP2.6. However, under RCP4.5 and RCP8.5 more incremental value in end than in the med period. Whereas the analyses of long year mean seasonal temperature increments respectively under RCP2.6, RCP4.5, and RCP8.5 are 1.2oc, 1.7oc, and 2.8oc in the Bega season, 1.6oc, 1.9oc, and 3.2oc in Belg season, and 1.2oc, 1.9oc, and 3.2oc in the Kiremt season. In general, average annual and seasonal tmin&tmax projections show positive increment over the study watershed but in Kiremt seasons increment less compared to the other seasons due to continuous rf season for the watershed. The projections of areal average annual tmin&tmax show increment in three RCPs projections.
Stream ow: Using the capability of the HEC-HMS model for assessing the impact of CC on stf, three RCPs climate projections are simulated for further analysis. The model produced outputs, which were then analyzed to quantify the mean change of stf monthly, seasonal and annual in the study area. However, temperature and rf will affect future stf magnitude monthly, seasonally, and annually. The percentage change of the mean monthly stf result revealed both an increment and decrement (Fig.8d) Table.6. Regarding annual stf change in percentage during near period -0.2%, -12.8 and +4.4%; during med period +5.4%, -5.3% and -0.2%; during end period -12.8%, -6.4% and -8.4% respectively under RCP 2.6, RCP 4.5 and RCP 8.5 compared to the baseline period.

Discussion
The consistency and homogeneity Pettitt's test (Table.1) XL-STAT analyses of seven weather stations average annual rf and stf data in the Abelti watershed indicates ideally possess the property of consistency and homogeneity. In which we prove that the observations are from the same population and homogeneous. And the general areal average rf patterns of the watershed show seasonality and exhibit typical characteristics of tropical. Hence, the spatial and temporal distribution of rf governs the amount and intra-and inter-annual variability (Kassa, 2015). In terms of seasons, in most parts of the watershed, rf is heavily concentrated in June, Julay and August (Fig. 5a). The comparison of areal mean monthly rf of both uncorrected and observed period in the controlled period revealed that there is a large value in uncorrected period. However, it indicates a slight underestimation of simulation outputs with bias and a good linear relationship for bias-corrected simulation data. Whereas both controlled period tmin&tmax of the study revealed that there is comparable value in the observed period. These are also in line with ndings by Abdela (2013). The underestimation of the results might be due to insu cient model resolution and topography which provides mechanical uplifting and thermal forcing to air parcels (Li et al., 2016). However, there is a variation of rf during, while its performance is within the acceptable range (Table.2). This may be due to individual point regional scale predictors of rf variation, which is poorly resolved as compared to temperature (IPCC, 2014). This result reveals that the HadGEM2-ES outputs can be utilized for future projection of rf, tmin&tmax in the Abelti watershed in Ethiopia.
The calibration and validation result carried out in HEC-HMS 4.5 window for gage station in the study was satisfactory and acceptable. The graph of simulated ow caught the observed ow during calibration and validation period respectively ( Fig. 6a and b). This indicates the observed and simulated ow show that the performance of the model in simulating the base ow, rising, and recession limb of the hydrograph is fairly good.
And also, the statistical parameter indicates a good model performance in terms of capturing the observed stream ow volume (RVE), the pattern of stream ow hydrographs (NSE), and correlation coe cient (R2) ( Table.3). Based on the calibrated parameters and values the model is validated, and the performance is a little bit improved. This is also the simulated ow in both periods may have a larger degree of uncertainty since rf was highly variable in space and time (Dile et al. 2013).
According to the Z-values, areal average rf of baseline, and future RCP2.6 and RCP4.5 of the watershed generally shows insigni cantly decreasing trend but RCP8.5 insigni cantly increasing trend. And based on the results (Table.4 and Fig. 7a, b, c and d) it is of immense importance to discuss the ecological, economic, and social impacts that could result if decreasing rainfall trends continue in the future. This discrepancy in the results may be mainly attributed due to climate variability and watershed characteristics. For rural farmers who have been vulnerable to drought, water stress, and the erratic nature of rainfall, appropriate adaptation strategies should have to be designed and implemented. The vulnerability of rural households might further be aggravated if extreme rainfall episodes continue in the future and consequently result in drought and surface and groundwater losses due to evaporation and overutilization. The mean annual tmin&tmax of the observed period and near period under RCP2.6, med period under RCP4.5 and med and end period RCP8.5 show a signi cant increasing trend. Whereas in the remaining period there are an insigni cant increasing trend according to Mk of p-value (Table.4). The result, in general, implies that the MK-trend test has shown an increasing trend for the tmin&tmax over the area. The trend increment of tmin shows more tmax over the watershed. In which the long-year mean annual tmin&tmax MK trend inspection; RCP8.5 trend increment greater than from both RCP4.5 and RCP2.6; and RCP4.5 greater than RCP2.6. And which proved that there was obvious warming up tendency throughout the watershed. This study in line with, Tekleab et al. (2013) reported the presence of insigni cant trends in both monthly and seasonal scale hydro-climatic variables in the majority of climate stations in the Abay basin. On the other hand, the trend test on average annual stf result revealed that insigni cantly decreasing trend in all studied period except under RCP8.5. The decreasing trend of average annual stf over study indicates there is a high tendency of temperature and evapotranspiration. And these nding is similar with nding by Negash, (2013) and Shank, (2017) in the basin of Rift-Valley Lakes basin in Ethiopia.
The future RCPs projections of areal mean monthly, seasonal and annual rf show both increment and decrement in the study watershed (Fig. 8a, Table.6). The decrement of rf in annual, Kiremt and Belg may lead the shortage of future water availability in the area, which is essential for agricultural purposes and hydropower. And the increment of Bega season rf over the study may lead to unexpected hydrological extremes. There is also a risk of a downstream cascade hydropower dam in the basin. Annual rf shows slight increment and decrement over the catchment. This result is in line with the study by Abdella (2013) in the area by using old climate model scenarios of REMO. This discrepancy in the results may be mainly attributed due to climate variability (Kassa, 2014) and watershed characteristics. And also, these study results are in good agreement with similar studies performed for the Ethiopian regions for annual and seasonal scales (e.g. Niguse and Alemu 2015;Li et al. 2016).
The increase or decrease in precipitation might be due to the combined effects of seasonal movement of the Inter-Tropical Convergence Zone (ITCZ), local land surface heating, and topographic forcing (Li et al. 2016).
The mean monthly tmax&tmin of three RCPs projection for future non-overlapping periods revealed in Fig. 8b and c expected positive increment in the watershed. The observed period average monthly tmax analysis revealed a positive increment between +0.16 ˚C to +1.2 ˚C in the study watershed. The average annual and seasonal tmax deviation stated in Table.6 shows a positive increment in three RCPs projections over three nonoverlapping periods in the study area. When we inspect the gure the mean monthly tmin&tmax positive increment in RCP2.6 is greater than in RCP4.5 and RCP8.5. In a general sense average monthly tmin&tmax of the study revealed a positive increment in three RCPs in the three-non-overlapping periods. When we come to the seasonal case in the study the Belg and Kiremt season tmin&tmax increment is more than the other seasons in three RCPs under three non-overlapping periods in the area. Annually air temperature is expected to increase in all months in both RCP scenarios. Increasing air temperature was however much higher in RCP8.5 and was associated with the fact that RCP8.5 is a higher greenhouse gas emission scenario with a higher degree of global warming. And also, the maximum amount of temperature change from RCP8.5 is since RCP8.5 produces more greenhouse gas as compared to RCP4.5 and RCP2.6, which is medium and low in greenhouse gas production (Riahi et al., 2011). In general, average annual and seasonal tmin&tmax projections show positive increment over the study watershed but in Kiremt seasons increment less compared to the other seasons due to continuous rf season for the watershed. The projections of areal average annual tmin&tmax show increment in three RCPs projections. According to the result close agreement with the study by Abdella (2013) in the area by using old climate scenarios of REMO. In the 2080s of RCP 2.6, RCP 4.5, and RCP 8.5, the change in monthly mean tmin&tmax magni cent compared to near and 2050`s for the catchment. The result generally shows clear agreement with the indication of EPCC (2015) over southwestern parts of Ethiopia future the area experiences rise in tmin&tmax.
The watershed is more in uenced by seasonal rf variations thus for CV is greater than 30% means that high variability in the seasons. This also indicates that the watershed is vulnerable of the season to CC and vulnerable to hydro-meteorological extremes. The larger no of CVs indicated that the difference of stf among years at the same season was larger, which would put pressure on water management in the catchment. Generally, the study watershed is highly in uenced by seasonal variation of rf and stf, which may put pressure on natural and manmade infrastructures.
The increase in temperature accelerates the evapotranspiration process which further in uences the rf amount and ultimately contributes to the modi cation of seasonal stf. The change of mean monthly, seasonal and annual stf results showed that the future increment in both directions over the watershed (Table.6 and Fig. 8d).
Regarding the study, positive and negative direction increments in local seasonal stream ow in the watershed under three RCPs. The three non-overlapping periods' future Bega seasons stf projection result shows an increment over the watershed compared to baseline under three RCPs. Whereas the Belg and Kiremt seasons stf projection show decrement under three RCPs scenarios of non-overlapping period and which corresponds to rf projection in the area. Overall, in the Bega season, a bigger percentage change increment is recorded, which may lead to unexpected extreme oods. Again, the bigger negative percentage stf value in the Belg season will lead to the shortage of water demand for agriculture and hydropower management in the basin. Hence the future monthly and seasonal great variation of stf leads a signi cant effect to domestic use and irrigation activity in the area as well as hydropower projects downstream. These projected effects of possible future climate change would signi cantly affect many hydrologic systems, which in turn affect the water availability and runoff, and the ow in rivers. The maximum reduction of ow can be anticipated in the Belg season as compared to the other seasons. This may be attributed to the increase in average temperature and potential-evapotranspiration. The projected precipitation under RCP4.5 and RCP8.5 scenarios shows that precipitation may signi cantly decrease in the near-term and mid-term periods, respectively. This decline of stream ow may cause an acute shortage of irrigation water owing to increasing demands for fresh water in the future periods.

Conclusions
The assessment of the CC impact over hydro-climatic variables and trends represents an important issue for resources management. The observational records and climate projections provide abundant evidence that hydro-climatic variables and trends are dramatically changing and have the potential to be strongly impacted by CC. Impact studies with hydrological models on the effects of CC are important as they can indicate how the hydrological processes are likely to be affected and how strong they are going to be affected in the future in the area. This is especially relevant for policymakers, those charged with the responsibility of selecting appropriate adaptation measures. HEC-HMS model predicted the measured stf provided satisfactory in both calibration and validation according to performance evaluation ( Increasing air temperature was however much higher in RCP8.5 and was associated with the fact that RCP8.5 is a higher greenhouse gas emission scenario with a higher degree of global warming. In general, in the Abelti watershed mean areal monthly, seasonal and annual stf has shown both increment and decrement in three RCPs scenarios. Increment and decrement of monthly, seasonal and annual stf under three RCPs but its projection change increment of 2080`s greater than in near and med. However, more increment and decrement mean stf percentage change recorded in the seasons of Bega and Belg in three projections. Hence, the increase and decrease in flow volume in the basin may have a signi cant contribution to the sustainability of existed and undergoing water development projects. This study found that the Abelti-catchment is very sensitive to climatic change in terms of temperature, rf, and stf. For this study results, are highly much because from rf and stf analysis for future period increase under RCP 2.5 and 4.5 for the med period. The study output is relevant because supported by the two investigators Chaemiso et.al, (2016) and Abdella and Teshome (2013) in the study area and by EPCC (2015). An increase in the magnitude of extreme rf in the study area may have a signi cant negative effect on the sustainability of existed agriculture practice (UNICEF, 2015), the life of the community, and hydropower development projects downstream. Moreover, it will lead to unexpected ood hazards and drought on community settlers. However, a precaution of mitigation and preparedness measures ought to be taken for possible ooding in the oodplain area of the river basin and downstream.

Recommendations
Knowing the monthly, seasonal, and annual simulated long-term average annual rf, temperature, and stf components in the future is useful to recommend better alternative and complementary action. Keeping in view, the threats to the survival of the Abelti watershed, the following steps are recommended are sought.
The models and model outputs used in this study possessed a certain level of uncertainty. The model simulation considered land-use changes remain constant for the future time horizons and single GCM emission scenarios although it is not true in the actual case. Hence, the results of this research should be taken carefully and be considered as an indicative prediction of the future and further researches should be extended by considering the future land-use changes, other climatic variables and different GCM outputs.
The status of the study area environment in recent years has been known to be seriously adjusted by man's technology coupled with the glaring climatic change which together greatly increase the risk of high extreme events. The study points to upcoming water resource problems so that mitigation and adaptation strategies can be made ahead. The study will also assist policymakers in decision-making on irrigation development in the watershed.