Analysis of past and future temperature variability and change in Southern Ethiopia

Long-term climate variability and change analyses have been repeatedly reported from different corners of the globe since the beginning of the 19th century (e.g. IPCC, 2013; Alizadeh and Babaei, 2022; Yagaso et al., 2024). Temperature is the most widely used climate variable in climate variability and change (trend) analyses (Serdeczny et al., 2017; Zeleˇnáková et al., 2018; Kerebo et al., 2024). The mean global temperature has increased by nearly 1.1°C since 1880 and from 0.15°C–0.20°C decade⁻¹ since 1975 (IPCC, 2018; NASA, 2022). Likewise, the average global surface temperature in 2022 was 0.86°C above the average of the 20th century (13.9°C), and it was the sixth-warmest among all years between 1880 and 2022 (NOAA, 2023). NOAA also indicated 10 warmest years occurred since 2010 among which the last 9 years occurred between 2014 and 2022. The increasing global temperature has been accompanied by a rise in sea level, warming of the ocean, a strong decline in Arctic ice, widespread rises in the intensity and frequency of heatwaves and many other linked climate effects (IPCC, 2014; Alizadeh-Choobari and Najafi, 2018; Alizadeh and Lin, 2021).

Due to the consistently increasing temperature (1.5°C), the world has faced global warming (IPCC, 2018, 2021). Many regions and seasons have been warming beyond the global average, with higher average warming over the land than the ocean (IPCC, 2018, 2021). The mean precipitation responses to global warming are both positive (increase) and negative (decrease) (Feng et al., 2019; Alizadeh and Babaei, 2022). The response of precipitation to climate change is controlled by two basic mechanisms:

1. the wet-become-wetter (Held and Soden, 2006); and

2. the warmer-become-wetter (Xie et al., 2010; Zhou et al., 2019).

These two mechanisms are named the thermodynamics process for precipitation changes in which the atmospheric moisture rises with the increasing temperature (Allen and Ingram, 2002).

Beyond the two thermodynamics processes, the precipitation in regions is strongly modified by circulation changes called dynamic changes of precipitation (Dong et al., 2019; Alizadeh and Babaei, 2022). For instance, EPA (2022) showed that the global total annual precipitation has a nonsignificantly increasing trend (1.02 mm decade⁻¹) between 1901 and 2021, which is confirmed in the USA except for the Southwest part. The results of Simpson et al. (2019) indicated that the upstream jet controls the decadal disparity of winter season precipitation in Western Europe. Moreover, Neelin et al. (2013) stated that global warming is expected to cause an eastward extension of the strong Pacific jet stream, which would result in a significant precipitation rise on the California coast. Similarly, the future poleward jet swing is projected to affect the season-dependent precipitation and temperature changes, in the central USA (Zhou et al., 2022).

Precipitation in East Africa showed a significantly decreasing trend during the Spring season between 1951 and 2014 (Ongoma and Chen, 2017), and an increasing trend during September–November in the 21st century (Tierney et al., 2015). High interannual precipitation variability was noted in East Africa including Ethiopia, which caused devastating droughts and floods (Gebrechorkos et al., 2019a; Gashaw et al., 2023). Gebrehiwot and Veen (2013) and Ware et al. (2023) indicated increasing trends of annual temperatures (Tmax and Tmin) in Northern and Southern Ethiopia (1954–2020). Gebrechorkos et al. (2019b) also highlighted that the country’s maximum and minimum temperatures increased beyond 1.76 and 2°C between 2011 and 2100.

The reviewed literature indicates that the extent of temperature variability and change differs with locations, months, seasons, investigation periods and considered scenarios or representative concentration pathways (RCPs). Despite this, Southern Ethiopia lacks information about climate variability and change for the periods of 1990–2019, the 2030s (2021–2040), 2050s (2041–2060) and 2070s (2061–2080). Besides, generating detailed information about temperature variability and change contributes to achieving sustainable development goals. Hence, a better understanding of climate variability and change helps to devise feasible adaptation and mitigation measures at local and regional levels (FAO, 2017). Thus, this study was initiated to analyze the long-term past and future temperature variability and change in Southern Ethiopia.

Ten meteorological stations located in the Wolaita Zone and the surroundings in Southern Ethiopia were used in this study (Figure 1). Seven meteorological stations among the ten are located in southern nations’ nationalities’ and peoples’ regional states (SNNPR): Arba Minch (Gamo Zone), Areka, Bilate, Boditi, Wolaita Sodo (Wolaita Zone), Hosana (Hadiya Zone) and Dilla (Gedeo Zone). The meteorological stations of Hawassa, Jimma and Chida are located in Sidama, Oromia and Southwest regions, respectively. The study area is located in the latitudinal range of 6.06–7.67°N and the longitudinal range of 36.79–38.48°E, with an altitudinal range of 1,220–2,306 m above sea level. It covers 276,152.96 km² (CSA, 2007). The monthly mean rainfall in Southern Ethiopia has a bimodal pattern (NMA, 2020) (Figure 2). The monthly rainfall has the first peak (164 mm) observed in April to May and a second peak (148 mm) observed in July to August. The mean monthly maximum temperature of the study area ranges from 24.4°C (in July) to 30.1°C (in February–March). The mean monthly minimum temperature ranges from 12.6°C (in December) to 15.1°C (in April).

Observed temperature data sets of Wolaita Zone and the surroundings were accessed from the National Meteorological Agency of Ethiopia (NMA, 2020), for the 1990–2019 period. Data quality was checked for data gaps, homogeneity and outliers. Filling the data gaps, trimming outliers and homogeneity tests are the prioritized activities in working with climate variability and trends (Hamzah et al., 2021). The missing values in the selected stations were filled using the simple arithmetic mean (Chinasho et al., 2021). The trimming of outliers (±1.5*interquartile range) and homogeneity tests (Standard Normal Homogeneity Test-SNHT, Pettitt and Buishand tests) were conducted using R-instat software, 0.7.5.

The temperature data sets of daily historical (1950–2005) and future projected (2006–2100) were accessed from the Coordinated Regional Climate Downscaling Experiment (CORDEX Africa). The future projected data sets under the two most commonly used representative concentration pathways (RCP4.5 and RCP8.5) (Muhati et al., 2018; Ismail et al., 2020) were considered in this study. The regional climate models (RCMs) have a spatial resolution of 0.44 × 0.44°. For this study, six models (Table 1) that have been reported for their outperformance in different parts of Ethiopia were considered. The selected models outperformed in the Gidabo River basin of Ethiopia (Girma et al., 2022), southwest Ethiopia (Demissie and Sime, 2021) and the Upper Blue Nile basin of Ethiopia (Takele et al., 2022).

Global Circulation (Climate) Models (GCMs) are the primary tools that provide reasonably accurate global, hemispheric and continental-scale climate information with a quite coarse spatial resolution (Trzaska and Schnarr, 2014; Lanzante et al., 2017). The resolution and accuracy of the GCMs are often insufficient to capture fluctuations in climate variables at the local level (Cannistra, 2016). It is because GCMs project atmospheric variables but not surface variables, yet the local climate is mainly affected by surface variables or features (Cannistra, 2016; Alizadeh, 2022). However, in some studies (Mekonnen and Disse, 2018; Gebrechorkos et al., 2019c), the overwhelming performances of some GCMs were noted. Generally, the GCMs have low accuracy in regional and local climate analyses (Alizadeh, 2022).

Thus, the nested finer-resolution RCMs have been established to consider the local features such as land use, topography and meteorological processes (Giorgi and Bates, 1989; Alizadeh, 2022). Regional climate models have normally outperformed driving GCMs at local and regional scales (Galata et al., 2021; Alizadeh, 2022; Takele et al., 2022). However, the RCMs inherited some biases from GCMs as they are limited by the boundary condition of GCMs (Tapiador et al., 2020). Therefore, bias correction is needed to improve the accuracy of estimating the observed data sets (Chen et al., 2020; Martins et al., 2021). The climate model for the hydrological modeling tool (CMhyd) software was used to correct the biases in RCMs in this study (Rathjens et al., 2016). The CMhyd software consists of six bias correction methods: linear scaling, delta-change correction, precipitation local intensity scaling, power transformation, variance scaling and distribution mapping. From the equations available in the CMhyd software, the linear scaling (additive) bias correction method was used in this study.

Model evaluation is the process of testing the correctness of the model in estimating the actual value using standard statistical analysis (Gebrekidan et al., 2018; Bekan, 2021). Thus, the six evaluation criteria, percent bias (PBIAS) (Berhanu et al., 2016), mean absolute error (MAE), root-mean-square error (RMSE) (Wang and Lu, 2018), index of agreement (d) (Willmott, 1981), Nash–Sutcliffe measure of efficiency (NSE) (Nash and Sutcliffe, 1970) and coefficient of determination (r-square; R²) (Hahn, 1973), were used in this study. The lower MAE and RMSE indicate better model performance. In the case of PBIAS, the negative, positive and zero values indicate the model’s overestimation, underestimation and optimal performance, respectively. The low magnitude of PBIAS (positive and negative) values possibly indicates accurate model simulation. The model with higher NSE is more reliable as NSE tests the model performance between −∞ and 1. The index of agreement (d) measures the degree of model prediction error between 0 and 1 indicating no agreement at all and a perfect match, respectively.

Similarly, R² measures how well a statistical model estimates the observed values, between 0 and 1. A value for R² of 1 indicates that the model perfectly estimates the observed values, 0 indicates the model does not match and between 0 and 1 indicates the model partially matches. So, the performances of RCMs in estimating the observed data were evaluated against these criteria. The performance evaluation used the meteorological stations’ (observed) and historical (RCMs) climate data sets for the matching period (1990–2004). Hence, the observed data sets were available for 1990–2019, whereas the historical data were available for either 1949–2004/2005 or 1950–2004/2005. Furthermore, the historical data sets of some RCMs were not available for some months in 2005. Therefore, the data sets of RCMs in 2005 were not considered in the performance evaluation:

where N is the size of the data matrix (time series); Xobs[i], Xest[i] and Xobs[mean] are the ith observed, estimated and average observed values, respectively; RSS = sum of squared residuals; and TSS = total sum of squares.

The temperature data sets were analyzed annually and seasonally, for the past 30 years (1990–2019) and three future projection periods: 2030s (2021–2040), 2050s (2041–2060) and 2070s (2061–2080). The three seasons of Ethiopia are winter, spring and summer. October, November, December and January (ONDJ) are the months of winter season. The spring season consists of February, March, April and May (FMAM) whereas, June, July, August and September (JJAS) are the months of the summer season. The average temperature of the months is expressed as a seasonal temperature. The change in temperature from the baseline period (1990–2019) was indicated in degree Celsius.

Temperature variability was calculated using the coefficient of variations in percent (CV%) [equation (7)]. The trends (change) were analyzed using a nonparametric Mann–Kendall’s trend test (Mann, 1945; Kendall, 1975) and Sen’s slope [equations (8)–(13)] using R-software version 4.0.5 (R Core Team, 2021). The Mann–Kendall trend test (Z) is the most commonly used method to detect a monotonic trend in a time series climate data (Tosunoglu and Kisi, 2017). Positive and negative (Z) values refer to increasing (upward) and decreasing (downward) trends in time series data, respectively. The Sen’s slope (SS) computes the magnitude of changes in respective directions in the trends [equation (8)], which was calculated as the median from all slopes (b) and intercepts or confidence intervals [equation (13)]:

where CV% is the coefficient of variation in percent, SD is the standard deviation and X– is the mean (average) of the time series data set. The variability (CV%) rating of Gomes (1985) was used in this study: low; <10%, medium; 10%–20%, high; 20%–30% and very high; >30%:

The mean of S is E[S] = 0 and the Variance of S (⁠$σ2$⁠):

where p is the number of the tied groups in the data set and tj is the number of data points in the jth tied group. The statistic S is approximately normally distributed provided that the following Z-transformation is used:

where SS is the magnitude of Sen’s slope, Xj and Xi are climate data sets at time j and i, respectively for (1 ≤ i < j ≤ n).

where At = intercept, b = median slope, t = time (year), Xt = climate data at time t.

At the regional level (Southern Ethiopia), the mean annual maximum temperature (Tmax) of the past 30 years (1990–2019) was 27.3°C (Table 2). The seasonal Tmax was in the range of 25.2°C (summer) to 28.7°C (spring). The Tmax had a coefficient of variation (CV%) of 1.4%–2.4%. The mean annual minimum temperature (Tmin) of the same period (1990–2019) is 14.2°C. The seasonal Tmin ranges from 13.2°C (winter) to 14.8°C (spring season). Unexpectedly, higher Tmin was noted in the rainy seasons (spring and summer) than in the dry (winter) season. The Tmin of Southern Ethiopia in the past 30 years had a CV range of 2.5%–4.3%. Relatively, Tmin showed higher variability than Tmax in the studied area. At the station level, Tmax and Tmin showed low variability in annual and seasonal analyses ( Appendix 1 Table A1). The monthly mean Tmax of stations in Southern Ethiopia showed the same pattern in the past 30 years (Figure 3). However, some disparity was observed in the monthly mean Tmin of stations (Figure 4).

The annual mean maximum temperature (Tmax) in Southern Ethiopia has increased by 0.14°C decade⁻¹ in the past 30 years (Table 3). However, the level of warming was statistically nonsignificant. Seasonally, the study area has been warming at daytime in all seasons, by 0.06°C in winter and 0.93°C in spring. The annual mean minimum temperature (Tmin) increased by 0.36°C decade⁻¹. The level of seasonal night warming in Southern Ethiopia in the past 30 years was 0.42°C decade⁻¹ in winter, 0.3°C decade⁻¹ in spring and 0.31°C decade⁻¹ in summer (Table 3). The warming level of night temperature is statistically significant in both annual and seasonal analyses. The nights have been warming more significantly than daytimes such that its Tmin has been increasing at a rate two times greater than that noted in the Tmax. However, at local levels, Tmax showed significantly increasing trends at Bilate, Boditi, Hawassa, Hosana and Sodo stations ( Appendix 2 Table A2). Therefore, Southern Ethiopia has been warming during day and night times in the past 30 years in all seasons.

Except for HadGEM2-ES, all the outputs of RCMs underestimated the observed monthly mean maximum temperature (OBS; pink line) before bias correction between 1990 and 2004 [Figure 5(a)]. Instead, the HadGEM2-ES overestimated the observed monthly mean maximum temperature (Tmax). After applying the linear scaling (LS) additive bias correction method, the RCMs highly improved their accuracy in predicting observed Tmax data in the study area [Figure 5(b)]. This indicates that using the LS additive bias correction method to correct the Tmax data sets of six RCMs in Southern Ethiopia is reasonable. After bias correction, all six RCMs showed comparable performance in estimating the observed Tmax ( Appendix 3 Table A3). However, the best performance was noted in the MPI-ESM-LR model.

In the case of minimum temperature (Tmin), before the bias correction was applied, all RCMs had low accuracy in estimating the observed values in 1990–2004. The outputs of EC-EARTH and CNRM-CM5 models underestimated and overestimated the observed Tmin (OBS; pink line), respectively [Figure 6(a)]. The rest of the RCMs showed inconsistent estimation of observed Tmin before the bias correction. On the contrary, after additive LS bias correction (Figure 6(b)], the Tmin data sets of RCMs except for HadGEM2-ES followed the same pattern and were very close to the observed values (OBS; pink line). The results of performance evaluation tests indicate that the RCMs had comparable performances in estimating the observed Tmin except for HadGEM2-ES ( Appendix 4 Table A4). The poor performance of HadGEM2-ES was indicated by the highest value of MAE, RMSE and PBIAS and the lowest value of d-index and R². In addition, the large negative value of NSE confirms the poor relationship between observed and HadGEM2-ES minimum temperature data sets. Therefore, the HadGEM2-ES model outputs were not included in the ensemble mean for projecting future temperature variability and change in Southern Ethiopia.

The annual mean maximum temperature (Tmax) variability is projected to be 0.87%–1.12% and 0.99%–1.48% CV under RCP4.5 and RCP8.5, respectively, for all projection periods (Table 4). However, the seasonal variability of Tmax is projected to be higher than that of the annual under all projection periods and RCPs. The mean annual minimum temperature (Tmin) variability of Southern Ethiopia is projected to be between 1.20% and 2.06% CV under RCP4.5 and 1.77%–2.31% CV under RCP8.5 in all projection periods (Table 4). Like Tmax, the seasonal variability of Tmin is projected to be higher than the annual ones for all projection periods and RCPs. Relatively, Tmin would have higher annual and seasonal variability than the Tmax, in the middle-of-the-road. The study area is projected to have stable temperatures that consistently increase till 2080.

The mean annual Tmax of Southern Ethiopia is projected to increase by 0.8°C–1.1°C, 1.5°C–2°C and 1.7°C–3°C in the 2030s, 2050s and 2070s, respectively, from the baseline period, under both RCPs (Table 5). The increment in the spring season was relatively higher than that noted in the annual analysis. The winter and summer seasons have had lower or equal increments with the annual ones. The mean annual Tmin of Southern Ethiopia is projected to increase by 0.8°C–1°C, 1.5°C–1.9°C and 1.6°C–3°C in the 2030s, 2050s and 2070s, respectively, from the baseline period, under both RCPs. The spring season Tmin showed a relatively higher increment than that of the annual. The results of trend tests indicate that the annual Tmax is projected to significantly increase in the 2030s under both RCPs and 2050s and 2070s under RCP8.5 (Figures 7 and 8).

Except for the spring season of 2050s and 2070s under RCP4.5, the seasonal Tmax has upward trends with a significant increase in the majority of cases (Table 6). Similarly, the annual mean Tmin of the study area is projected to increase consistently in all of the projection periods and RCPs except for the 2070s under RCP4.5 (Table 6 and Figure 8). The seasonal Tmin is projected to significantly increase in the majority of projection periods under both RCPs. Therefore, the day and night temperatures of Southern Ethiopia are projected to consistently increase in the 2030s, 2050s and 2070s, under both the modest and worst scenarios.

According to Gomes's (1985) ratings, the temperature (Tamx and Tmin) of Southern Ethiopia has had low annual and seasonal variability (<10% CV) in the past 30 years (1990–2019). The temperature of the study area has comparable variability (low) with that in the west Harerghe Zone of Ethiopia, between 1988 and 2017 (Wasihun et al., 2019). This result is also in agreement with the stable temperatures of the plateaus of East Africa between January and June (Rowhani et al., 2011). However, the relatively higher variability observed in seasons could have affected agricultural (crop and livestock) production and productivity in the study area. The temperature variability in the growing seasons of spring (FMAM) and summer (JJAS) affects crop productivity (Zerihun and Prowse, 2017). The National Meteorological Agency (NMA, 2020) indicates that the occurrence of El Nino-Southern Oscillation increases the surface temperature in Ethiopia. In line with this, Ware et al. (2023) revealed significantly increasing trends of Tmax and Tmin in Sidama Regional State in Ethiopia. However, Wang et al. (2021) affirmed that the weakening in the Walker circulation over the Indian and Pacific Ocean basins could increase rainfall and reduce the negative impacts of elevated temperature in Southern Ethiopia.

The low variability noted in temperature is associated with its consistently increasing trend (time series) in the past 30 years. Compared with the global average of 1975–2021 warming (0.15°C–0.2°C decade⁻¹) (NASA, 2022), daytime warming (Tmax) was in the range of the global average, whereas the night time temperature (Tmin) has been warmer in Southern Ethiopia. This result can also be supported by the findings of Gemeda (2019) who indicated an increasing trend of maximum and minimum temperature in Southern Ethiopia around Jinka between 1970 and 2015. Elevated temperature affects cellular processes responsible for plant growth and development, accelerates the phenological cycle of plants and reduces biomass and grain yield of crops (Wylie, 2008; Mathur et al., 2014; Li et al., 2019). Low biomass production and high prevalence of diseases, in response to increasing temperature, reduce livestock production (Ali et al., 2020).

The low error levels of RCMs observed after additive linear scaling (LS) bias correction confirmed that using the LS method for Tmax is acceptable. Similarly, Gebrekidan et al. (2020) indicated significant improvements in RCMs after bias correction and identical performance of LS with other bias correction methods in temperature data. Luo et al. (2018) also affirmed a substantial improvement in the accuracy levels of RCMs in predicting observed Tmax after bias corrected by LS. However, Daniel (2018) noted the lowest performance level of the additive LS bias correction method in estimating the observed Tmax. This indicates that the performance of RCMs varies with locations and considered bias correction methods, as indicated by previous studies (Lafon et al., 2013; Tong et al., 2021). Besides, the additive LS bias correction method is scientifically sound for Tmin data sets of five RCMs (except for HadGEM2-ES). In line with this, the previous studies (Gebrekidan et al., 2020; Abiy et al., 2021; Galata et al., 2021; Saranya and Vinish, 2021) showed major improvements in Tmin data sets of RCMs after additive LS bias correction. Therefore, applying the LS bias correction method on temperature data sets of RCMs except HadGEM2-ES can be suggested for Southern Ethiopia.

The low annual and seasonal temperature (Tmax and Tmin) variability projected to occur in the future years is mainly associated with its consistently increasing trend. The increasing trend of future (2030s, 2050s and 2070s) temperature (significant at a 95% confidence level in the majority of cases) is in agreement with several previous studies (e.g. Takele et al., 2022; Teshome et al., 2022). Takele et al. (2022) projected the Tmax (Tmin) to increase by 0.96°C–1.1°C (0.95°C–1.14°C) and 1.59°C–2.04°C (1.55°C–2.1°C) by 2030s and 2050s under RCP4.5 and RCP8.5 in the Upper Blue Nile Basin of Ethiopia. Likewise, the report of IPCC (2014) indicated a 0.9°C–1.1°C and 1.7°C–2.1°C increase in mean annual temperature by the 2030s and 2050s, respectively, under the mid-range emission scenario. Similarly, Teshome et al. (2022) also reported an increasing trend of Tmax and Tmin in the majority of stations in eastern Ethiopia by the 2030s and 2050s, under RCP4.5 and RCP8.5.

The projected temperature increase could cause different levels of damage to Ethiopia’s food production system. For instance, Hatfield et al. (2011) projected a 2.5%–10% decrease in crop yield due to the increasing temperature in the 21st century. Likewise, Mohammed et al. (2022) projected a maize grain yield loss of 11%–20% in the 2030s (for a 1.4°C–1.5°C increase) and 26%–29% in the 2050s (for a 1.9°C–2.5°C increase), under RCP4.5 and RCP8.5 (at Tehuledere site of Northern Ethiopia). In line with this, Abera et al. (2018) projected a decrease in maize yield up to 43% (at Bako) and 24% (at Melkassa) by the end of the century (2070–2099). Chinasho et al. (2023) also projected a 15.11% decrease in maize yield in response to climate change between 2030 and 2070 in Wolaita Zone (Southern Ethiopia) under RCP4.5 and RCP8.5. Moreover, elevated temperature reduces water bodies through evapotranspiration (Wylie, 2008). Abdella (2013) also projected that the mean annual river flow is expected to decrease by 8.6% and a reduction of peak flow (−17.47% to −30.58%) from 2015 to 2050 in the Gibe catchment (Southern Ethiopia) in response to 1.5°C increase in temperature. Therefore, proactive adaptation and mitigation measures have to be developed to reduce damages associated with increasing temperature by 2030s, 2050s and 2070s in Southern Ethiopia.

The globe has been warming due to the increasing temperature; as a result, many adaptation and mitigation measures have been suggested globally and nationally (IPCC, 2021). FAO (2017) indicates that the level of vulnerability to the impacts of climate change varies with geographic location, economy and demography, the adaptation measures need to be local. The detailed information on temperature variability and change in the past and future helps to understand the associated negative impacts on agriculture, hydrology, biodiversity, environment and human well-being, among others. The projected climate pattern of the future helps the country to devise proactive adaptation and mitigation measures for the associated damages at different levels from local to national level. This could improve the resilience of farmers as well as the country to climate change impacts. This in turn contributes to the achievement of sustainable development goals (e.g. no poverty, zero hunger and climate action). This is because the agriculture sector in Ethiopia accounts for 80% of employment, 33% of the gross domestic product (GDP) and 76% of exports (EPRSS, 2023). Therefore, the findings from this study can be used by farmers (when interpreted by local governments), government and nongovernment organizations, policymakers of Ethiopia, initiatives working on climate change adaptation and mitigation in Africa and international scholars for further investigations.

The results were presented in several national research workshops/symposiums and panel discussions. The experts who attended the presentations from the Ethiopian Institute of Agricultural Research and different academic institutions are expected to work on drought-tolerant crops and livestock varieties. Likewise, the information will be disseminated to local farmers through media (i.e. television and radio) and Telecom companies in regional languages. The damages associated with the increasing temperature can also be lowered by setting consistent climate information services through the joint implementation by the Ministry of Agriculture, the Ethiopian Agricultural Transformation Institute, the National Disaster Risk Management Commission and Telecom companies in Ethiopia. Through the myriad dissemination methods including social media and paper presentations, policymakers can get valuable information to consider during policy development in Ethiopia.

In the past 30 years, Tmax and Tmin of Southern Ethiopia have had low variability with a significantly increasing trend in most cases. The LS additive bias correction method improved the performance of temperature data sets of all RCMs except for HadGEM2-ES. The Tmax and Tmin will increase in all future projection periods and RCPs. The annual mean Tmax is projected to significantly increase by the 2030s under the two RCPs and by the 2050s and 2070s under RCP8.5. Except for the spring season of the 2050s and 2070s, under RCP4.5, the seasonal Tmax has a significantly increasing trend in most cases. The mean annual Tmin is projected to significantly increase in all projection periods and RCPs except for the 2070s under RCP4.5. The seasonal Tmin is also projected to increase consistently in most cases. Based on the results, it can be concluded that the Tmax and Tmin have low variability in the past and future. The increasing temperature could have been affecting agricultural production systems in Southern Ethiopia. Therefore, the projected temperature information can be disseminated to agriculture experts, development agents, health workers and farmers by the Government of Ethiopia through media (i.e. television and radio) and Telecom companies. Moreover, further empirical investigations on the impacts of increasing temperature on crop and livestock production, hydrology and biodiversity can be suggested for devising proactive adaptation and mitigation measures in the study area. This would improve the farmers’ resilience to the negative impacts of projected temperature variability and change in the study area.

The authors of this article are grateful to the Africa Center of Excellence for Climate Smart Agriculture and Biodiversity Conservation, Haramaya University (World Bank) and Wolaita Sodo University, Ethiopia, for financial support.

Table A1 

Table A2 

Table A3 

Table A4