Assessment of Future Climate Change Scenario in Halaba District, Southern Ethiopia

Climate change is one environmental threat that poses great challenges to the future development prospects of Ethiopia. The study used the statistically downscaled daily data in 30-years intervals from the second generation of the Earth System Model (CanESM2) under two Representative Concentration Pathways (RCPs): RCP 4.5 and RCP 8.5 for three future time slices; near-term (2010-2039), mid-century (2040-2069) and end-century (2071-2099) were generated. The observed data of maximum and minimum temperature and precipitation are a good simulation with the modeled data during the calibration and validation periods using the correlation coefficient (R 2 ), the Nash-Sutcliffe efficiency (NSE), and the Root Mean Square Error (RMSE). The projected annual minimum and maximum temperatures are expected to increase by 0.091˚C, 0.517˚C, and 0.73˚C and 0.072˚C, 0.245˚C, and 0.358˚C in the 2020s, 2050s, and 2080s under the intermediate scenario, respectively. Under RCP8.5, the annual minimum and maximum temperatures are expected to increase by 0.192˚C, 0.409˚C, and 0.708˚C, 0.402˚C, 4.352˚C, and 8.750˚C in the 2020s, 2050s, and 2080s, respectively. Besides, the precipitation is expected to increase under intermediate and high emission scenarios by 1.314%, 7.643%, and 12.239%, and 1.269%, 10.316% and 26.298% in the 2020s, 2050s, and 2080s, respectively. Temperature and precipitation are projected to increase in total amounts under all-time slices and emissions pathways. In both emission scenarios, the greatest changes in maximum temperature, minimum temperature, and precipitation are predicted by the end of the century. This implies climate smart actions in development policies and activities need to consider locally downscale expected climatic changes.


Introduction
The global climate is changing, and projections suggest that the rate of change will likely increase [1]. Observed changes in the Earth's climate over the past 250 years are now widely considered to have been enhanced by anthropogenic activity [2] [3]. Climate change is happening due to natural and anthropogenic activities such as the burning of fossil fuels, industrial pollution, deforestation, volcanism, forest fires, and land use changes [4].
Climate scenarios from a global climate model (GCM) are supplied on a large scale and have a scale discrepancy for climate change effects and adaptation studies that require detailed local data. The regional outputs from a GCM are therefore "downscaled" using one of two methods: dynamic downscaling or statistical downscaling, which causes trouble [5] The Statically Downscaling Model (SDSM 4.2) was introduced by Wilby et al. (2002) and is a Windows-based decision support tool for the rapid development of single-site, ensemble scenarios of daily weather variables under current and future regional climate forcing [6]. It is a hybrid model based on multiple linear regressions (MLR) and the stochastic weather generator (SWG) [7]. The MLR represents the statistical empirical relevancy between NCEP large-scale climate variables (predictors) and local scale weather data (predictands) along with the process of screening predictors and the calibration of SDSM, which results in the production of several regression parameters. It establishes statistical relationships between the output from GCM at a large scale (i.e., predictors) and observed data from meteorological stations at a local scale (i.e., predictands) based on multiple linear regression techniques.
Statistical downscaling methodologies have several practical advantages over dynamical downscaling approaches. In situations where a low-cost, rapid assessment of highly localized climate change impacts is required, statistical downscaling (currently) represents the more promising option. The SDSM yields reliable estimates of extreme temperatures, seasonal precipitation totals, and areal and inter-site precipitation behavior [8]. The SDSM calculates statistical relationships based on multiple linear regression techniques between large-scale (the predictors) and local climate variables (the predictands). At the three future time horizons of 2010-2039 (2020s), 2040-2069 (2050s) and 2070-2099 (2080s), the emission scenarios for precipitation, maximum and minimum temperature outputs are generated [9].
Climate change is the major challenging factor in Africa. Rainfall variability is a greater challenge on the African continent, which results in variation in water availability. Greater spatial variation of precipitation (from a drop of about 25% -50% drop and rise) has been reported in East Africa. For instance, Beck and [10] have reported a decline in water availability by the 2050s. In Ethiopia, the projected mean annual temperature is found to be in the ranges of 0.9˚C -1.1˚C,  [11]. The increase in large-scale mean surface temperature by the end of the twenty-first century is estimated to be between 1.1˚C to 2.6˚C (RCP4.5) and 2.6˚C to 4.8˚C (RCP8.5) [12]. The temperatures reveal an increasing trend that tends to influence precipitation, water availability, and extremes of floods and droughts [1]. The current unpredictable conditions of the climate, such as frequent flood incidence and widespread drought, influence the country's agriculture and water resources [13]. The Halaba district is part of the main rift valley areas in Ethiopia known for their frequent drought and flood incidence. The rise in temperature and the change in precipitation affect the agricultural activities and water resources of the area. As a result, the Statistical Downscaling Model is used in this study application of a downscaling approach developed in recent years for precipitation, maximum and minimum temperature based on data from Halaba-Kulito stations for the period 1981-2018. The main aim of this study is to statistically downscale and produce future climate scenarios under different representative pathways for the future daily maximum temperature, daily minimum temperature, and precipitation for the early identification of the possible associated impacts and to set adaptation priorities in order to develop climate smart actions for development policies and activities.

Description of the Study Area
The Halaba district is located between 7˚10'N and 7˚42'N latitude and 38˚00'E and 38˚25'E longitude in the south-eastern part of Ethiopia ( Figure 1). The total area of the district is 994.66 square kilometers, with 70% flat, 27% slope, and 3% mountainous. Altitude ranges in the area from 1554 to 2149 m.a.s.l (Ethiopia Central Statistical Agency, 2013). The area is located about 315 km south of Addis Ababa and 85 km southwest of Hawassa, the capital of the Southern Nations Nationalities and Peoples Regional State and Sidama region. The administrative center is Halaba-Kulito. According to Ethiopia's central statistics agency's population projection, the total population of Halaba district was 309,658 of which 151,545 were male and 158,113 females. Of this total figure, the rural inhabitants are 244,582 (79%) and the urban population makes up 21%.
The mean annual rainfall ranges between 544 and 1271 mm, which show large spatial variability with maximum rainfall being as large as 2.34 times the minimum rainfall. The rainfall pattern is erratic and irregular in the area. Due to the severe and heavy rains, soil erosion and flooding are very common in the low-lying areas of the study area. The mean monthly minimum and maximum temperatures are 12.8˚C and 26.8˚C. The district is considered a "woina degas", or temperate zone, having two major seasons, namely Belg (February, March, April, and May) and Kiremt (June, July, August, and September), and irrigation is practiced in some parts around the Bilate river. The livelihood strategy of the rural people of the area is predominantly mixed agriculture (crop production and animal husbandry).

Data Description
The station merged grid data for rainfall and temperature for Halaba-Kulito and an ocean carbon model (CMOC) [14].

Description of Statistical Downscaling Model (SDSM)
The Statistical Downscaling Model (SDSM), the well-recognized statistical downscaling tool which is applied widely in climate impact studies, was employed to transform the global circulation [16]. It is computationally inexpensive, able to directly incorporate the observational record of the region, etc. This software  [17]. The SDSM model contains two separate sub-models to determine the occurrence and amount of conditional meteorological variables (discrete variables), such as precipitation, and the number of unconditional variables (continuous variables), such as temperature or evaporation. Therefore, the SDSM can be classified as a conditional weather generator in which regression equations are used to estimate the parameters of daily precipitation occurrence and amount separately, making it slightly more sophisticated than a straightforward regression model [17]. The SDSM yields reliable estimates of extreme temperatures, seasonal precipitation totals, and areal and inter-site precipitation behavior. This freely available software enables the production of climate change time series at sites for which there are sufficient daily data for model calibration, as well as archived General Circulation Model (GCM) output to generate scenarios for the 21st century. The SDSM can also be used as a stochastic weather generator or to fill in gaps in meteorological data [16]. SDSM as the downscaling approach which required a proper selection of predictors established a relationship between predictors and predictand based on partial correlation coefficients. It is an important step in the downscaling process [16]. The predictors of the model were screened and selected based on the R 2 and p-values.
In order to have better prediction results, all the correlations with a p-value less than 0.05 were selected ( Table 1).
The delta method is a simple, widely used method to create scenario time series from GCM output. The method uses the delta method of the SDSM for future projections [5] [18]. The standard approach for the delta method is that the GCM-simulated difference for each calendar month (absolute difference for temperature and relative difference for precipitation) between a future period and the baseline period is determined and then this is superimposed on the historical daily temperature and precipitation series [19].
In the SDSM, a change in precipitation is obtained by: The same is true for changes in 2050s and 2080s.
The value at 2050s and 2080s is also obtained by the same formula explained in Equation (2).
 base is the mean of all ensembles (or a specific ensemble if selected) for each statistic for the baseline period. Likewise, 2020s  is the mean of all ensembles (a specific ensemble) for each statistic for period 2020s  , and so on for 2050s  and 2080s  [6].

The Representative Concentration Pathway (RCPs)
RCPs are time and space-dependent trajectories of future greenhouse gas con-

Evaluation Criteria
Due to examining the efficiency of the proposed downscale techniques, three evaluation criteria exist: correlation coefficient, Nash-Sutcliffe efficiency, and root mean square error.

Correlation Coefficient
The most familiar measure of dependence between two quantities is the Pearson product-moment correlation coefficient (R 2 ), or "Pearson's correlation coefficient," commonly called simply "the correlation coefficient." The correlation coefficient is a statistical measure of the strength of the relationship between the relative movements of your two variables. The values range between 1.0 and 1. A calculated number less than −1 or greater than 1 means an error in the measurement.
A correlation of −1 shows a perfect correlation but is negative, while a correla-  n xy x y R n x x n y y R 2 = Correlation coefficient, n = number in the given dataset, x = first variable in the context, y = second variable.

Nash-Sutcliffe Efficiency
Nash-Sutcliffe efficiency (NSE) is a normalized statistic that determines the relative magnitude of the residual variance compared to the measured data variance.
The Nash-Sutcliffe efficiency indicates how well the plot of observed versus si-

Root Mean Square Error
The

Performance of the SDSM Model Calibration and Validation
The future scenario construction, the observed data of maximum and minimum temperature and precipitation are correlated with the modeled data during the calibration and validation periods using the R 2 , NSE, and RMSE. during calibration and 0.78 during validation. The R 2 for precipitation is the lowest due to it being a conditional process and having a less regular distribution than the temperature distribution. Unconditional models assume a direct link between the regional-scale predictors and the local predictand. Conditional models, such as for daily precipitation amounts, depend on an intermediate variable such as the probability of precipitation occurrence [5]. The R 2 , NSE, and RMSE of the calibration and validation period are summarized in Table 2.

Maximum Temperature
The result of downscaling maximum temperature indicates that there is very good agreement between observed and simulated maximum temperature. During calibrated the observed and simulated data, for Halaba-Kilito climate station R 2 , NSE and RMSE to be 0.89, 0.89 and 0.36 respectively. Each regression models have a value very close to 1 for nine months. However, as shown in Figure 2(A), the model overestimates the maximum temperature during the months of January, April and May. In validation process, also R 2 , NSE and RMSE model computed to be 0.89, 0.91 and 0.33 respectively. This implies, as indicated on Figure  2(B), for few months (January, April and May) the model overestimates while for the other months the model showed good validation efficiency Table 2.

Minimum Temperature
The result of downscaling minimum temperature indicates that there is very  Figure 3(A), the model overestimates the minimum temperature during the months of April, November and December. In validation process, also R 2 , NSE and RMSE model computed to be 0.91, 0.79 and 0.46 respectively. This implies, as indicated on Figure 3(B), during the months of February, July and December the model overestimates while for the other months of the model showed good validation efficiency Table 2.

Precipitation
Relative to the minimum and maximum temperatures, the precipitation could not be replicated to replicate the historical (observed) data. This is due to the complicated nature of precipitation processes and their distribution in space and time. Climate model simulation of precipitation has improved over time, but is still problematic. It also stated that rainfall forecasts are more uncertain than temperature forecasts. This is because rainfall is highly variable in space, and so the relatively coarse spatial resolution of the current generation of climate models is not adequate to fully capture that variability.  Figure 4).

Future Maximum and Minimum Temperature Change Scenarios
The generated future scenarios of average annual minimum and maximum temperatures generally showed an increasing trend from the base period values. The average annual minimum temperature will increase by 0.091˚C in the 2020s, 0.52˚C in the 2050s and 0.74˚C in the 2080s using RCP 4.5. In the high emissions scenario RCP 8.5, the average annual minimum temperatures increase by 0.08˚C in the 2020s, 0.84˚C in the 2050s and 1.77˚C in the 2080s. The downscaled average annual maximum temperature scenario also indicates that it will rise by 0.07 ˚C in the 2020s, 0.24 in the 2050s and 0.34 ˚C in the 2080s using the

Future Precipitation Change Scenarios
The average annual precipitation will increase until the end of this century. In the model under the intermediate emissions (RCP4.5), the future average annual precipitation will be expected to increase by 1.3% in the 2020s and by 1.3% and 12.24% in the 2080s, respectively. Under the worst-case scenario RCP 8.5 scenario, the average annual precipitation is projected to increase by 1.27% in the 2020s, 10.32% in the 2050s, and 26.3% in the 2080s, respectively, compared to the baseline period.
Though the changes in the average annual precipitation value by the 2020s are not significant, as concluded from the results of future scenarios, the prediction of precipitation will increase in the 2050s and 2080s. In the 2080s, under the in-  [27]. Overall, the two scenarios show an increasing trend in the average annual precipitation ( Figure 6).

Conclusions
The As a result, future scenarios predict that the expected average annual temperature and precipitation will increase up to the end of the century. The changes in climatic elements up to the 2020s are not significant. However, in the 2050s and 2080s, the predictions were higher. In the worst-case scenarios, the expected average annual maximum temperature increase by 8.7˚C in the 2080s. Similarly, the increase in the average annual minimum temperature reaches 1.77˚C in the 2080s. All the downscaled projections under the two RCP scenarios show an increase in temperature with time, especially for RCP8.5. All the periods also predict an increase in average annual precipitation, especially in the worst-case scenarios, the volume of total average annual precipitation will also increase by 26.3% in the 2080s. Generally, the study area will experience an increase in average annual temperature and precipitation, which will gradually modify the ex- isting temperature and precipitation conditions resulting from urbanization and climate change. The Halaba district will continue to experience drought and flooding vulnerability. Hence, it is recommended that government planning integrate the findings of this research into its planning process, which will be informative and supportive of further investigations and practical better adaptation mechanisms in the context of a climate emergency. Further studies also need to be undertaken by adding various model ensembles in order to avoid prediction inconsistency.