Impacts of Climate Variability and Change on Sorghum Crop Yield in the Babile District of Eastern Ethiopia

Abstract

The impacts of various climatic conditions, such as temperature and rainfall variabilities, are very critical and sensitive to rain-fed crop production, particularly over the water stress arid and semi-arid regions of Ethiopia. This study was designed to evaluate the potential impact of climate variability and change on sorghum grain yield in the Babile district of eastern Ethiopia. The study was conducted based on observed and model-based simulated projected rainfall and temperature obtained from the Ethiopian Meteorological Institute and General Circulation Models (GCM) used by the Intergovernmental Panel on Climate Change (IPCC) of the Fifth Assessment Report CMIP5) and Agricultural Model Inter-comparison and Improvement Project (AgMIP). Three GCM models, namely GFDLESM2M, CanESM2, and HadGEM2-ES under RCP4.5, were considered to generate future climate projections for the near-term 21st century. Various univariate and multivariate statistical techniques were employed to compute and identify whether the impacts of climate variability and change on rain-fed sorghum crop performance were reasonably viable over the districts where grain yield is highly stable and productive under normal climate conditions. Our findings revealed that more stable and better rainfall performance from May to September, the season when sorghum crops are normally planted in the Babile district, was positively correlated, while the maximum and minimum temperatures of the season were negatively correlated with sorghum grain yield. A significant association has been detected between sorghum grain yield and its growing period rainfall, number of rainy days, and maximum and minimum temperature with multi-regression analysis. Thus, the variability of rainfall in August, June temperature, and the number of rainy days in September significantly impacted sorghum crop productivity. As a result, the multi-regression model adjusted R-squared indicated that 77% variance in annual sorghum yield performance was explained by rainfall and temperature conditions that prevailed during the crop growing period. During the past period, there was a significant increase in sorghum yields, which are projected to decline during the near term of the 21st century in the future. This revealed that declining and disturbed rainfall performance and increases in temperature are likely to reduce overall sorghum grain yield in the Babile district. We recommend that there is a need to enhance awareness for smallholder farmers on the adverse impact of climate variability and change on sorghum grain yield. In view of this, the farmers need to be geared toward employing climate-smart agriculture as a possible adaptation measure to reduce the negative impacts of climate variability and change on rain-fed crop production practices in the Babile district and other arid and semi-arid parts of eastern Ethiopia.

1. Introduction

Climate variability and change have impacted rain-fed subsistence crop production throughout the world [1]. However, the effects of climate variability on agriculture are very complex and diverse in arid and semi-arid regions [2], and rainfall and temperature variability are key climate variables that influence crop growth, development, and yield [3]. In particular, a reduction in rainfall amount and erratic seasonal characteristics of events have occurred in arid and semi-arid parts of the world [4]. In Africa, rainfall amounts are likely to decrease over most parts of sub-Saharan Africa (SSA), while rainfall variability is projected to increase [5]. These future changes in rainfall amounts and variability over Africa are expected to have high negative potential impacts on crop production, since a large proportion of the population is dependent upon rain-fed agriculture [6]. Studies by [7] also affirm that rainfall in the Eastern Africa region remains highly variable, unreliable, and likely associated with changes in the regional climate. Similarly, the long-term climatic change related to changes in rainfall patterns and variability and changes in temperature are most likely to increase the frequency of droughts and floods in Ethiopia [8,9]. The occurrence of such extreme events has been affecting the planning, performance, and management of agricultural operations and crop production in Ethiopia [10,11].

In the East and West Hararghe zones, Ethiopia has been affected by the impact of climate change-induced drought, erratic and reduced rainfall, and increasing temperature [12,13]. Climatologically, the rainfall patterns of East Hararghe zone are bimodal type, with two rainy seasons, Belg (March to May) and Kiremt (June to September), which are highly erratic in nature; there is also one dry season called Bega (October to February) [14] Based on an assessment carried out by [15], historical rainfall variability indicates that semi-arid parts of the East Hararghe zone are frequently impacted by extreme droughts and dry spell events. The purpose of this study was conducted, therefore, to quantify the state of the present and future impacts of climate variability and change on crop production at the local level. in semi-arid parts of the East Hararghe Zone’s of Babile district, where the high rainfall variability has observed and is the major cause of the agricultural problem at this study area [13,15]. The principal climatic-related challenges in semi-arid regions are rising temperatures, shifting dates of the beginning and end of the rainy seasons, prolonged and frequent dry spells, and changes in the length of the cropping season [10]. These events are due to the occurrence of climate change and variability, which are significant impacts on crop productivity [16]. Based on a CSA annual report [17], the most common crops grown in the Babile district are sorghum, maize, and groundnut. A total of 3.95 million tons of sorghum is being produced per annum, and the average yield level in Ethiopia was estimated at 2.18 t/ha less compared to the 4.54 t/ha average grain yield potential [18].

Since the sorghum crop is the most tolerant and grows in semi-arid areas under rain-fed conditions in Ethiopia, this tolerance to environmental stress makes sorghum highly suitable for semi-arid tropic (SAT) crop production systems [19]. However, the occurrences of droughts during the post-flowering period often cause premature plant death, lodging, reduction in seed size, and, ultimately, yield losses [20,21], and its productivity is significantly influenced by climate change and variability [16]. Climate variability has a significant impact on rain-fed subsistence crop production [1]. Mainly, unpredictable rainfall has a detrimental effect on crop production [22]. Similarly, the report stated by [23] also reported that sorghum phenology is inversely proportional to an increase in temperature, while grain yield is directly proportional to an increase in rainfall. More importantly, recurrent droughts frequently impact grain yield [24]. Thus, locally-based climate information is crucial for assessing and predicting the impact of climate variability and change on crop production [25]. However, there is limited information at the local level to quantify the impacts of climate change and variability on sorghum production and productivity in the study area. Because of the uncertainties in processes underpinning the changing climate, more research is needed to understand the influence of climate change and its variabilities. In this study, the impacts of climate variability and change on crop production analyzed, with adaptation options in the study area identified as important issues for minimizing climate-related risks. Therefore, the objectives of the present study are to analyze and quantify the impacts of climate variability and change on sorghum grain yield and to identify its adaptation options in the Babile district, eastern Ethiopia.

2. Materials and Methods

2.1. Description of the Study Area

The study was conducted in the Babile district, East Hararghe zone, which is located in Oromia Regional National State, eastern Ethiopia. The administrative boundaries of East Hararghe Zone are the Shebelle River on the southwest, which separates it from Bale Zone, West Hararghe Zone in the west direction, the Dire Dawa City administration on the north, and Somali Region on the northeast. The Babille district is bordered on the south and east by the Somali region, on the west by Fedis, and on the north by the Gursum districts. The Fafen River defines a portion of Babile’s eastern border. The Babile district is located between 8.35° N to 9.18° N and 42.25° E to 42.9° E at 950 to 2000 m above sea level [26]. The physical distance of the Babile district from Addis Ababa is 560 km to the east, while it is 35 km from Harar City to the east (Figure 1).

2.1.1. Climate

According to [14]’s agro-climate classification, the Babile district is classified under semi-arid areas with a length of growing period between 75 and 120 days. The rainfall distribution pattern is a bimodal type which occurs from March to May (the first rainy season) and June to September (the second rainy season). The long-term average of climate was derived from 1980–2020, at the Babile meteorological station in the Babile district. The average annual maximum temperature is 28.8 °C. Similarly, the average annual minimum temperature is 15.4 °C (

Table 1. Monthly and annual extreme temperatures (°C) profile for the Babile district (1980–2020).

Month	Jan	Feb	Mar	Apr	May	Jun	Jul	Aug	Sep	Oct	Nov	Dec	Annual
Tx	36.6	36.6	37.5	38.8	36.0	36.4	36.6	36.5	35.6	38.6	37.6	36.6	38.8
Tn	4.5	6.0	6.0	9.6	11.5	9.0	10.0	11.0	9.0	6.5	4.0	6.0	4.0
Txmean	30.2	31.4	31.2	29.2	28.5	27.5	26.2	26.6	27.4	28.8	29.3	29.6	28.8
Tnmean	14.2	15.2	16.5	16.9	16.8	15.7	15.5	15.5	15.6	15.3	13.9	13.4	15.4
Tmean	22.2	23.3	23.8	23	22.6	21.6	20.9	21.1	21.5	22.1	21.6	21.5	22.1

Note: Tx, Tn, and T indicate extreme maximum, and extreme minimum, and Txmean, Tnmean, and Tmean indicate mean Tx, mean Tn, and annual temperature, respectively.

). The extreme maximum and minimum temperature ranges between 4°C and 38.8 °C, with extreme maximum temperatures occur in April, and in November, December, and January, respectively (Table 1). The average annual rainfall is 731 mm, while the average seasonal rainfall for March–May and June–September were 301 mm and 335 mm, respectively (

Table 2. Rainfall statistics from the Babile Meteorological station (1980–2020).

	Temporal Rainfall Distribution (mm)						Number of Rainy Days (NRD)
Month	Min	Max	Avg	SD	CV%	Cbn%	Min	Max	Avg	SD	CV%
Jan	0.0	27.3	4.2	7.8	185.7	0.6	0	7	0.9	1.8	200.0
Feb	0.0	87.6	8.9	17.5	196.6	1.2	0	7	1.5	1.9	126.7
Mar	0.0	251.7	55.8	63.0	113.0	7.6	0	19	5.4	4.8	88.9
Apr	12.2	431.3	133.1	95.5	71.8	18.2	1	23	10.6	5.6	52.8
May	0.0	252.1	112.0	75.8	67.7	15.3	0	30	12.4	8.5	68.5
Jun	0.0	276.4	76.6	59.0	77.0	10.5	0	26	9.8	7.4	75.5
Jul	0.0	191.0	64.5	49.6	76.9	8.8	0	23	8.8	6.4	72.7
Aug	5.3	241.9	92.9	45.0	48.4	12.7	2	29	12.1	6.2	51.2
Sep	0.7	286.9	100.6	65.0	64.6	13.8	0.0	27.0	12.2	5.6	45.9
Oct	0.0	334.1	61.0	68.6	112.5	8.3	0.0	25.0	6.5	6.4	98.5
Nov	0.0	68.2	12.9	18.1	140.3	1.8	0.0	10.0	2.1	2.5	119.0
Dec	0.0	104.8	8.6	21.0	244.2	1.2	0.0	5.0	0.9	1.5	166.7
MAM	57.1	656.6	301.0	142.5	47.3	41.2	9.0	56.0	28.4	14.0	49.3
JJAS	43.3	569.5	334.6	125.7	37.6	45.8	9.0	89.0	43.0	21.6	50.2
Annual	112.5	1240	731.2	232	31.7	100.0	25.0	176	83.2	41.3	49.0

Note: MAM = March–April–May; JJAS = June–July–August–September, JAS = June–August–September; NRD = number of rainy days, CV = coefficient of variation; Cbn, δ, and µ are contribution, standard deviation, and average, respectively.

).

2.1.2. Topography and Land Features

The altitude of the Babile district ranges from 950–2000 meters above sea level, with Ambelber and Sarbadin being among the highest points. A survey of the land use of the district reported in 1995/96 shows that 21.1% of the land is arable, with 17.5% under annual crops, while 3.9% is pasture, 3.7% is forest, and the remaining 71.3% is considered as built-up, degraded, or otherwise unusable [26]. Most lands in the Babile district are categorized as being in the lowland regions of Ethiopia [26].

2.2. Sources and Types of Data

Daily observations of surface air temperature and rainfall data of the district for the period from 1980–2020 were obtained from the Ethiopian Meteorological Institute (EMI). For filling the data gaps for rainfall, temperature, and solar radiation, downloaded gridded data from the NASA website (https://power.larc.nasa.gov/data, accessed on 5 October 2022-) was used. Climate model data of the IPCC fifth-generation CMIP5 was downloaded from the Agricultural Model Inter-comparison and Improvement Project (AgMIP) website through https://github.com/agmip/ClimateScenarios/Generator/tree/master/data/ (accessed on 10 October 2022). These data were used to evaluate the relationship between climate variables and sorghum yield productivity in this district. Sorghum crop grain yield at the zonal administration level of eastern Hararghe from 1995–2020 was obtained from the Central Statistical Agency of Ethiopia (CSA).

2.3. Climate Models

Daily historical and future climate projections for the near-term 21st century produced by the IPCC AR5 participating models and analyzed through the CMIP5 initiative were used in this study. At least three GCM climate models were used for climate change impact assessment based on [27] for future climate projection and impact assessment. For future projections, representative concentration pathways (RCPs; RCP4.5) were used in the near-term 21st century. The radiative forces of RCP4.5 are considered for future climate scenarios as an intermediate mitigation scenario with radiative forcing stabilizing at 4.5 Wm² by 2100 [5,28].

2.4. Statistical Downscaling of Simple Delta Approach Technique for Future Climate

The Coupled Model Intercomparison Project Phase 5 (CMIP5) global climate models (GCMs) used in this study were selected according to [29,30]. The recommendation report stated that HadGEM2 (the UK Met Office Hadley Centre Global Environment Model version two, 1.2° × 1.8° [31], CanESM2 (the Canadian Earth System Model of the Canadian Centre for Climate Modeling and Analysis second generation and is located at the University of Victoria, Victoria, British Columbia (CanESM2), 2.8° × 2.8° [32], and GFDL-ESM2M (National Oceanic and Atmospheric Administration-Global Fluid Dynamic Laboratory Princeton University Forrestal Campus) [33] models have been recommended for African climate study and, due to their long history of development and evolution, a preference for higher resolution and established performance was given [31,32].

Owing to their coarse resolution, GCM data cannot be used directly for local and regional climate change impact studies. To this end, GCMs at large spatial scales are translated to scales more representative of local environmental conditions using various downscaled techniques [34]. Thus, in this study, the daily rainfall, maximum, and minimum temperature projections were downscaled by using the AgMIP method for climate change projections from CMIP5 data using the R-Gui Software, Version 4.2.0 (R-Gui Software Version 4.2.0. Statistical Analysis of Observed and Projected Climate Data)

The statistical methods of climate data analysis were carried out based on [35]. Climate variability analysis was carried out using a statistical approach through the analysis of long-term climatic data based on [36]. Variability analysis involves the use of coefficient of variation (CV), rainfall anomaly, and other basic statistical analyses applied here. The coefficient of variation of seasonal to inter-seasonal rainfall variability was analyzed by statistical equation by employing the method of CV coefficient (coefficient of variation) adopted by [37] to express the rainfall variability. Here, the coefficient of variation of inter-seasonal to seasonal and annual rainfall variability were analyzed as follows:

$$\mathrm{C}\mathrm{V}=\left(\frac{\mathrm{S}\mathrm{D}}{\overline{\mathrm{X}}}\right)\ast 100 \mathrm{where} \overline{\mathrm{X}}=\frac{\sum _1^{\mathrm{N}}{\mathrm{X}}_{\mathrm{i}}}{\mathrm{N}}\mathrm{and}\mathrm{S}\mathrm{D}=\frac{\sqrt{\sum _1^{\mathrm{N}}({\mathrm{X}}_{\mathrm{i}}-\overline{\mathrm{X}}{)}^2}}{\mathrm{N}-1}$$

(1)

where; CV is the coefficient of variation; Xi is the rainfall of each season or year, N is the number of rainfall observations years, SD is the standard deviation and $\overline{\mathrm{x}}$ is the average of rainfall recorded in series. According to [14,37,38], CV is used to classify the degree of variability of rainfall events as less (CV < 20%), moderate (20% ≤ CV < 30%), and high variability (CV ≥ 30%), while very high CV > 40% and extremely high CV > 70% indicate extremely high inter-annual variability in rainfall.

2.4.1. Analysis of Seasonal Rainfall Contribution

The analysis of inter-seasonal to the seasonal contribution of rainfall was undertaken as follows:

$$\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{i}\mathrm{b}\mathrm{u}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}=\frac{\overline{\mathrm{X}}}{\mathrm{A}\mathrm{n}\mathrm{n}\mathrm{u}\mathrm{a}\mathrm{l}\mathrm{R}\mathrm{F}}*100\%$$

(2)

where: Contribution is an inter-seasonal or seasonal contribution for annual basis, and $\overline{\mathrm{X}}$ is the monthly or seasonal long-year average; Annual RF is the annual long-year average of rainfall of the study area.

2.4.2. Inter-Seasonal to Seasonal Rainfall Anomaly Analysis

The seasonal rainfall characteristics of each season of observed climate were analyzed using the rainfall anomaly index (RAI), first developed by [39], which constitutes the parameters stated in Equation (3). Similarly, [40] established the RAI range classification. Based on his suggestion, the seasonal rainfall performance of each rainy season was analyzed. Here, the variability is the difference between total rainfall for each season or year and the long-term mean divided by standard deviation to derive rainfall anomalies at different periods of time of each year using the following RAI:

$$\mathrm{R}\mathrm{A}\mathrm{I}=\frac{{\mathrm{X}}_{\mathrm{i}}-\stackrel{-}{\mathrm{X}}}{\mathrm{S}\mathrm{D}}$$

(3)

where: RAI is the standardized rainfall anomaly index; X_i is the seasonal rainfall of a particular year; $\stackrel{-}{\mathrm{X}}$ is the long-term average of inter-seasonal, seasonal, or annual rainfall over a period of observation; SD is the standard deviation of inter-seasonal to seasonal or annual rainfall over the period of observation.

2.4.3. Seasonal Dry Spell Probability Analysis

Daily rainfall data from the Babile meteorological station were used for the assessment of dry spell probability based on the Markov chain probability model stated by [41]. Based on [41,42], a dry spell was defined as a sequence of dry days, including days with less than a 1 mm value of rainfall. The risk of dry spells for the varying number of days following the planting period is computed using first-order Markov chain modeling, in which case analysis has to be carried out from the very beginning with the daily long year’s raw data. To provide a viable decision aid to various practitioners, different dry spell lengths (number of consecutive dry days) were examined. Accordingly, given the condition that the first of the month is a potential planting date, the probability of dry spells longer than 5, 7, 10, and 15 days was analyzed using INSTAT software.

2.4.4. Calculation of Growing Degree Days of the Sorghum Crop

Growth degree days (GDD) is a measure of accumulated heat that was calculated for the entire growing season from the planting to harvesting time period, or from May to November, respectively. It was calculated in order to analyze its impacts on sorghum crop grain yield with multiple linear regression. The pollen package in R software allows for three different versions of the GDD calculations. This function accepts the following arguments: Tmax is daily maximum temperature, Tmin is daily minimum temperature, Tbase is base temperature, Tbase is maximum base temperature, and the type of the GDD calculations suggested by [43] were used. Therefore, daily GDDs are calculated using daily maximum (Tmax) and minimum temperatures (Tmin) and a base temperature (Tbase) as follows:

$$\mathrm{G}\mathrm{D}\mathrm{D}=\sum _1^{\mathrm{N}}\left(\right(\frac{\mathrm{T}\mathrm{m}\mathrm{a}\mathrm{x}+\mathrm{T}\mathrm{m}\mathrm{i}\mathrm{n}}{2})-\mathrm{T}\mathrm{b}\mathrm{a}\mathrm{s}\mathrm{e}))$$

(4)

Tbase considered in this case is the lower base temperature as 10 °C for sorghum crop growth and development. This is because [44] stated that the optimum temperature for the growth of sorghum is (27–32) °C, and the lower base temperature is 10 °C.

2.5. Sorghum Crop Yield and Climate Variability Relationship

The impact of climate variability on crop yield has been assessed differently by using the correlation and regression methods employed as stated by [45,46]. The discrepancy in applying a standardized period between the two data sets is attributed to gaps in the production of crop data. To this extent, the period from 1995–2020, spanning 26 years of the climate variability and crop grain yield relationship was cautiously used. The data sets were subjected to bivariate correlation and cross-tabulation analysis to determine the impact of climate variability on change in sorghum grain yield. To understand the relationships between sorghum grain yield and climate variables, the Pearson product–moment correlation coefficient was used to generate statistical indices.

$$\mathrm{r}=\frac{\mathrm{n}\left({\sum }_1^{\mathrm{n}}{\mathrm{X}}_{\mathrm{i}}{\mathrm{Y}}_{\mathrm{i}}\right)-{\sum }_1^{\mathrm{n}}{\mathrm{X}}_{\mathrm{i}}{\sum }_1^{\mathrm{n}}{\mathrm{Y}}_{\mathrm{i}}}{\sqrt{(\mathrm{n}{\sum }_1^{\mathrm{n}}{\mathrm{X}}_{\mathrm{i}}^2-(}{{\sum }_1^{\mathrm{n}}{\mathrm{X}}_{\mathrm{i}})}^2)-(\mathrm{n}{\sum }_1^{\mathrm{n}}{\mathrm{Y}}_{\mathrm{i}}^2-\left({{\sum }_1^{\mathrm{n}}{\mathrm{Y}}_{\mathrm{i}})}^2\right)}$$

(5)

where r is the correlation between climate variables and sorghum yield; Y_i is sorghum grain yield; X_i is the monthly temperature or rainfall amount as well as the number of rainy days or GDD within the growing season.

2.6. Sorghum Crop Grain Yield Responses to Climate Variability

This study used regression models with panel data including zonal-level sorghum crop grain yields and variations in climate variables to examine the response of yields to climate variability. This study estimated crop yields as a function of climate variables while controlling for time-invariant fixed effects, such as soil quality and other land characteristics, as suggested by [47]. As such, a multi-regression method was employed to identify the best response climate parameters to crop yield, as documented by [46]. The highest significant predictors at a 5% level of significance were conducted by multi-regression techniques using R software to evaluate the sorghum yield response to variation in monthly climate variables from the 1995–2020 data in the growing period from May to November. The relationship between sorghum grain yields with other independent climatic variables was determined by the following:

$${\mathrm{Y}}_{\mathrm{i}}=\mathrm{\mu }+{\mathrm{a}}_1{\mathrm{x}}_{\mathrm{i}1}+{\mathrm{a}}_2{\mathrm{x}}_{\mathrm{i}2}+{\mathrm{a}}_3{\mathrm{x}}_{\mathrm{i}3}+\dots {\mathrm{a}}_{\mathrm{n}}{\mathrm{x}}_{\mathrm{i}\mathrm{n}}$$

(6)

where: i = 1, 2, 3, ……n, and n is the number of observation years; µ is the average of sorghum grain yield; Y_i is the sorghum crop grain yield of each year in response to the x_i1, x_i2, x_i3, and x_in of climate parameters. These climate parameters are very useful in indicating the sorghum grain yield performance. They include the monthly rainfall amount, number of rainy days, minimum and maximum temperature, and the total heat unit value (GDD) required during crop growing period from May to November.

Many authors have commented that crop yields are affected by intra-seasonal rainfall and temperature variability [11,46,48]. Therefore, in this study, the monthly rainfall and temperature, as well as the total heat unit value, were used as explanatory variables for the entire growing season. Stepwise regression analysis was applied to select the best predictors out of the given predictors. The variation in sorghum yield was established first, and, then, how much of the year-to-year variation in sorghum grain yield was explained by year-to-year variations in monthly climate variability performance was identified based on the given fitted climate parameter indices. Finally, after the regression fit model has been developed based on the given observed climate fitted monthly climate parameters, yield simulation was estimated, and yield prediction for the near-term 21st century was carried out based on a projection of climate data obtained with three considered GCM climate models under the RCP4.5 climate scenario.

2.7. Trend and Variability of Sorghum Crop Yield Analysis

Furthermore, analyses of the coefficient of variation (CV) and annual stability (SD) of yield annual grain yield and trend of the grain yield, the trend of rainfall, and the temperature were analyzed, and their impact on sorghum grain yield was identified. Analyses of yield stability have become more important in recent years since the increased variability of climate is also associated with a decreased stability of crop yields [49]. The Mann–Kendall (MK) test [50,51] and the Theil and Sen estimator [52,53] were utilized to analyze to detect the presence of monotonic (increasing or decreasing) trends of sorghum grain yield rainfall and temperature. The Mann–Kendall statistical test has been frequently used to quantify the significance of trends in climatic parameter time series, as reported by [54]. The projected and observed trends were determined for seasonal and annual rainfall and temperature, and sorghum crop yield, respectively. Similarly, Sen’s slope estimation test is also used to compute both slopes (the linear rate of change) and intercept according to Sen’s method [53]. A positive value of Sen’s slope (Ti) and intercept (β) indicates an ‘upward trend’ (increasing values with time), while a negative value indicates a ‘downward trend’. Here, all data pairs are computed with the R software Kendall, and the trend package has been used for performing the Mann–Kendall and Sen’s slope statistical test as follows:

$$\mathrm{S}={\sum }_{\mathrm{i}=1}^{\mathrm{n}-1}{\sum }_{\mathrm{j}=\mathrm{i}+1}^{\mathrm{n}}\mathrm{s}\mathrm{g}\mathrm{n}\left({\mathrm{X}}_{\mathrm{j}}-{\mathrm{X}}_{\mathrm{i}}\right)$$

(7)

$$\mathrm{V}\mathrm{a}\mathrm{r}\left(\mathrm{S}\right)=\frac{\mathrm{n}\left(\mathrm{n}-1\right)\left(2\mathrm{n}+5\right)-\sum _{\mathrm{i}=1}^{\mathrm{p}}{\mathrm{t}}_{\mathrm{i}}\left({\mathrm{t}}_{\mathrm{i}}-1\right)\left(2{\mathrm{t}}_{\mathrm{i}}+5\right)}{18}$$

(8)

$$\mathrm{Z}\mathrm{s}=\left\{\begin{array}{c}\frac{\mathrm{S}-1}{\sqrt{\mathrm{V}\mathrm{a}\mathrm{r}\left(\mathrm{S}\right)}},\mathrm{if}\mathrm{S}>0\\ 0,\mathrm{if}\mathrm{S}=0\\ \frac{\mathrm{S}+1}{\sqrt{\mathrm{V}\mathrm{a}\mathrm{r}\left(\mathrm{S}\right)}},\mathrm{if}\mathrm{S}<0\end{array}\right.$$

(9)

where n is the length of the time series (x₁, x₂, x₃, ……, x_n), X_i and X_j are the data values in the time series i and j (j > i), respectively, and sgn(X_j − X_i) is the sign function. Positive values of Z_S indicate increasing trends, while negative Z_S values show decreasing trends. The magnitude of the trend is estimated by Sen’s slope method by [53], which proceeds by calculating the slope as a change in measurement per change in time, as follows:

$$\mathrm{Q}=\frac{{\mathrm{X}}_{\mathrm{j}}-{\mathrm{X}}_{\mathrm{i}}}{\mathrm{j}-\mathrm{i}},\mathrm{i}<\mathrm{j},$$

(10)

where X_j and X_i are the i^th and j^th data points in the time series (j > i), respectively, while Q is the slope between data points X_j and X_i, and Sen’s slope estimator is simply given by the median slope.

3. Results and Discussion

3.1. Analysis of Observed Climate Variability

Climate variability was explained in terms of the temperature and rainfall performance in the study area. Daily rainfall and maximum and minimum temperature from 1980–2020 were considered to analyze the climate variability observed in the study area. Climate variability plays a major role in crop productivity in the study area, where agriculture practice is generally rain-fed, and identifying the performance of climate variability with local climate information is crucial for assessing and predicting the impact of climate variability on crop production [25].

3.1.1. Extreme Temperature Variability Analysis

Daily maximum, minimum, and mean temperatures were analyzed from 1980–2020. The historical analysis presented within this period was used as a data set and tools were developed for the analysis of mean and extremes from daily observed data. Based on the Babile meteorological station data, the average annual Tmax for the Babile district was 28.8 °C, which generally ranges between 26.2 °C and 31.4 °C. On the other hand, the average yearly Tmin was 15.4 °C, which ranged between 13.4 °C and 16.9 °C (Table 1). The temperature generally ranges between 4 °C and 38.8 °C, with the extreme maximum temperature occurring between April and October and the extreme minimum temperature occurring between November and January (Table 1).

3.1.2. Variations in Rainfall Amounts and Number of Rainy Days

Seasonal to inter-seasonal and annual time series of rainfall and the number of rainy days for a period of about 40 years (1980–2020) at the Babile meteorological station were used for computing extreme maximum and minimum rainfall and the number of rainy days, seasonal contribution, standard deviation, and coefficient of variation. Here, the evaluation of variability based on the coefficient of variation (CV) in rainfall amount and the number of rainy days in seasonal to inter-seasonal time periods were analyzed. The results from the coefficient of variation showed that rainfall amounts received within the seasonal (March–April–May) or Belg season and June–July–August–September or Kiremt season were highly variable (all with CV > 30) (Table 2). The observed rainfall data during all the months had a CV of above 30%, highlighting the high variability of rainfall over the study area (Table 2). Notably, the coefficient of variation in monthly rainfall amounts from the May to September time period was quite high during the months of March (CV = 113%) and June (CV = 77%) as shown in Table 2. The coefficient of variation of rainfall was the highest in the MAM season (CV = 47%) as compared to the JJAS season (CV = 38%) (Table 2). Based on [14,37,38] studies documented a CV greater than 30% value in all months and seasons, which is high to extremely high in rainfall variation (Table 2). Furthermore, the lowest CV of rainfall in consecutive months was 43%, 51%, and 46%, in July, August, and September, respectively (Table 2). Similarly, the CV of the number of rainy days during MAM, JJAS seasons, and annual time periods were 49%, 50%, and 49% respectively.

The average number of rainy days during MAM, JJAS season, and the annual time period were 28, 43, and 83 days respectively (Table 2). Due to the strengthening and establishment of main rainfall drivers during the northern hemisphere summer season, the seasonal rainfall contribution was higher in the JJAS season (46%) than in the MAM (41%). Further disaggregation of each seasonal rainfall performance on a monthly basis, April accounted for the highest monthly contribution (about 18% for MAM, with an average rainfall amount of 133 mm) (Table 2).

Furthermore, the inter-seasonal to seasonal and annual rainfall averages computed during MAM, JJAS seasons, and the annual rainfall were 301 mm, 334.6 mm, and 731.2 mm, respectively. Similarly, the study by [55] states that the annual rainfall averaged all across Ethiopia is 817 mm, but, given the complex physiography and the different seasonal and spatial influences of the prevailing air masses and winds, a large diversity is observed between various regions of Ethiopia. Here, the annual rainfall in Babile was less than the average rainfall in Ethiopia. Because of this high rainfall variability during the MAM seasonal time period and the lower average number of rainy days, it was difficult to plant sorghum crops in this time period. This result agrees with the findings of [9,56,57,58], where there was more variability in Belg rainfall than the Kiremt rainfall in most parts of Ethiopia. Moreover, the findings of [56] further show that annual and seasonal rainfall (Kiremt and Belg seasons) in Ethiopia were highly variable, with CV values ranging between 10% and 50%. Similarly, studies by [59] and [60] affirm that rainfall in the Eastern Africa region remains highly variable, unreliable, and is likely associated with changes in the regional climate. Several studies highlighted that the variability in rainfall in most East African regions is linked to large-scale climate variability, including the El Niño–Southern Oscillation (ENSO), Indian Ocean Dipole (IOD) [61], and oscillation of the inter-tropical convergence zone (ITCZ) [60].

3.1.3. Variability and Anomalies in Seasonal Rainfall

The seasonal rainfall anomaly index (RAI) was computed for the period 1980–2020. The rainfall anomaly index (RAI) classification based on [40] was used for MAM and JJAS seasons at the Babile station, as shown in Figure 2 and Figure 3. There was notable high seasonal variability and temporal anomalies in rainfall within the specified time period (Figure 2 and Figure 3). The positive RAI values represent normal to wetter rainfall conditions, while the negative values represent dry to very dry seasons, with different degrees of intensity. As shown in Figure 2, the periods that had the longest drought periods during the MAM season were 1982–1984, 1988–1992, 1998–2003, and 2011–2014. In particular, the highest negative value, with an RAI of −2.0 during the MAM season, was observed in 2011. The same year with a severe drought event was also documented in [40]. In contrast, the years with the greatest positive value of RAI during the MAM season were 1985 and 1986, with RAI of 2.5 and 2.4, respectively. Based on [40] rainfall anomaly index classification, these years were classified as very wet (Figure 2). Although it seems that the MAM rainfall anomaly has decreased, the trend line did not show any statistical significance.

Based on [40]’s drought intensity classification, the moderate drought intensity years indicated in Figure 3 during JJAS were 1987, 1991, 2000, 2005, 2011, and 2015, while the extreme droughts occurred in 2002 and 2010, which were the years with the highest negative value of RAI during JJAS, with an RAI of −2.1 and −2.0, respectively, classified as severe drought years. Here, the periods that had the longest droughts during JJAS season were from 1989–1993, from 1998–2005, from 2008–2012, and from 2015–2018 (Figure 3). Here, once the drought has occurred during JJAS, it lasts from four to seven years. On the other hand, the year of greatest positive values of RAI during JJAS were 1980, 1983, and 1988, with an RAI of 1.5, 1.9, and 1.5, respectively, being, therefore, classified as very wet (Figure 3). Although it seems that the JJAS rainfall anomaly has decreased, the trend line did not show any statitistical significance.

According to [40]’s drought intensity classification based on the rainfall anomaly index analysis from the 1980–2020 observation years, the result revealed that 41.5% were in the mild drought intensity class, 9.8% were the moderate drought intensity class, and 2.4% were a severe drought, with no extreme drought during the MAM season (

Table 3. Statistics of drought recurrence.

RAI Range	Category of the Occurrence of Drought	The Intensity of Drought in % during MAM		The Intensity of Drought in % during JJAS
−0.99 ≤ RAI < 0.0	Mild dryness	17 years	41.5%	15 years	36.6%
−1.0 ≤ RAI ≤ −1.99	Moderate dryness	4 years	9.8%	4 years	9.8%
−2.0 ≤ SPI ≤ −2.99	Severe dryness	1 year	2.4%	2 years	4.9%
≤−3.0	Extreme dryness	0	0	0	0

). During JJAS seasonal time period, 36.6% were in the mild drought intensity class, 9.8% were in the moderate drought intensity class, 4.9% were in the severe drought intensity class, and there was no extreme drought during the JJAS season (Table 3). Finally, based on [40]’s drought intensity classification, during the MAM and JJAS seasonal time periods, mild dryness to severe dryness drought intensity were observed in this study area (Table 3).

Furthermore, sorghum yield anomaly was computed from 1995–2020 and synchronized with the JJAS seasonal rainfall anomaly index (RAI) to analysis the impacts of the rainfall anomaly index on sorghum yield. The result revealed that the JJAS rainfall anomaly index showed positive values with the sorghum yield anomalies in 1996, 2019, and 2020, and negatively values with the sorghum yield in 1998, 2000, 2001, 2002, 2003, 2005, 2008, 2010, and 2015 years (Figure 4). Here, the RAI anomaly has a similar phase with a sorghum yield anomaly in most of the observed years (Figure 4).

3.2. Dry Spell Probability Analysis

The analysis of the historical occurrence of dry spells and their probability of recurrence is important for planning the time of minimum dry spell risk on the day of the year (DOY). For this study, daily observed and projected rainfall data at Babile meteorological station was used for the assessment of dry spell risk analysis based on the Markov chain probability model, as stated by [41]. The observed and projected dry spell risk analysis showed that the lowest risk of exceeding 5, 7, 10, and 15 consecutive dry days ranged between 180–260 days of the year, from July to September time of the year (Figure 5a), while the dry spell risk ranged between 100–120 in April (Figure 5a), which makes the district to start crop planting in April.

Seasonal dry spell risk analysis was made for the near-term 21st century based on climate model outputs generated from the CanESM2, GFDL-ESM2M, and HadGEM2 (CMIP5) climate projection models under RCP4.5 scenarios (Figure 5b–d). The results showed that the pattern of dry spells with specified day lengths resembled the observed events in recent years (Figure 5a) but with lower intensity in dry spell risk, from 140 to 180 days of the year as generated from the CanESM2 model output data (Figure 5b).

3.3. Rainfall Trend and Variability Analysis over Babile District

Seasonal to inter-seasonal rainfall amount trend and variability in the observed time period (1980–2010) was extended, with future time period rainfall projected in the near-term 21st century using three GCM climate models: CanESM2, HadGEM2 and GFDL-ESM2M model outputs under RCP4.5 were analyzed. The results revealed that Kiremt seasonal rainfall trend statistically decreases during the observation and in the near-term 21st century (

Table 4. Mann–Kendall results of rainfall trend and variability from 1980–2040 at Babile station.

	CanESM2 (1980–2040)				HadGEM2 (1980–2040)				GFDLESM2M (1980–2040)
Time Series	MKZ	Sen’s Slope	p-Value	CV%	MKZ	Sen’s Slope	p-Value	CV%	MKZ	Sen’s Slope	p-Value	CV%
Jan	1.03	0.0	0.30	201	0.41	0.0	0.68	238	1.06	0.0	0.29	202
Feb	−0.45	0.0	0.65	186	−0.44	0.0	0.66	185	0.39	0.0	0.69	177
Mar	1.11	0.26	0.27	112	1.34	0.39	0.18	112	0.62	0.09	0.53	114
Apr	−1.02	−0.56	0.31	72	0.63	0.35	0.53	69	−1.45	−0.65	0.15	72
May	−1.31	−0.59	0.19	71	−1.17	−0.54	0.24	70	−0.39	−0.2	0.69	69
Jun	−1.47	−0.41	0.14	80	−3.35	−0.8 *	0.01	91	−0.37	−0.13	0.71	76
Jul	1.07	0.28	0.28	73	−0.16	−0.05	0.87	74	0.40	0.15	0.68	73
Aug	−4.43	−1.16 *	0.00	54	−3.48	−0.88 *	0.00	50	−4.54	−1.19 *	0.00	55
Sep	−1.64	−0.57	0.10	63	−3.52	−1.02 *	0.00	70	−1.64	−0.57	0.10	62
Oct	0.72	0.24	0.47	121	0.06	0.01	0.95	112	0.11	0.01	0.91	113
Nov	2.33	0.16 *	0.02	158	0.88	0.0	0.38	156	1.21	0.0	0.23	153
Dec	−0.01	0.0	0.99	244	0.18	0.0	0.86	234	0.35	0.0	0.73	225
MAM	−0.87	−0.94	0.39	50	0.43	0.51	0.67	49	−0.81	−0.66	0.42	49
JJAS	−2.53	−1.98 *	0.01	40	−4.17	−3.29*	0.00	44	−2.49	−1.96 *	0.01	39
Annual	−0.79	−1.47	0.43	33	−1.44	−2.39	0.15	33	−1.34	−2.20	0.18	32

* Indicates the trend is statistically significant at 95% significant level.

). Similarly, Belg seasonal rainfall trend also shown a decreasing trend, even though it is not statistically significant (Table 4). The result agrees with a previous study carried out in East and West Hararghe by [12]. While Kiremt seasonal rainfall variability analysis from 1980–2040 with CanESM2, HadGEM2 and GFDL-ESM2M model output showed 40%, 44%, and 39%, respectively, Belg seasonal rainfall variability with those respective model outputs showed 50%, 49%, and 49%, respectively (Table 4). Similarly, the annual rainfall variability under the RCP4.5 climate scenario with the CanESM2, HadGEM2, and GFDL-ESM2M (CMIP5) model outputs were 33%, 33%, and 32%, respectively (Table 4). Based on [37] rainfall variability analysis, inter-seasonal to seasonal rainfall variability, as well as the annual rainfall variability of the observed time period to the future time period in the near-21st century at Babile station, shows extremely high and highly variable (Table 4).

3.4. Temperature Trend and Variability Analysis over Babile District

Seasonal to inter-seasonal maximum and minimum temperature amount trend and variability in the observed time period (1980–2010) and future time period temperature projected in the near-21st century by three GCM climate models’ (CanESM2, HadGEM2, and GFDL-ESM2M) output under RCP4.5 were analyzed. The results revealed that both seasonal to inter-seasonal maximum and minimum temperature trends were shown to be statistically increasing from 1980–2040 under the RCP4.5 climate scenario (

Table 5. Maximum temperature trend and variability from 1980–2040 at Babile station.

	CanESM2 (1980–2040)				HadGEM2 (1980–2040)				GFDLESM2M (1980–2040)
Time Series	MKZ	Sen’s Slope	p-Value	CV%	MKZ	Sen’s Slope	p-Value	CV%	MKZ	Sen’s Slope	p-Value	CV%
Jan	4.48	0.04 *	0.00	4.2	5.25	0.05 *	0.00	4.6	4.79	0.043 *	0.00	4.3
Feb	5.09	0.07 *	0.00	6.3	4.91	0.06 *	0.00	6.1	3.92	0.045 *	0.00	5.6
Mar	4.51	0.06 *	0.00	6.8	4.26	0.056 *	0.00	6.6	4.12	0.053 *	0.00	6.5
Apr	3.39	0.05 *	0.00	6.9	3.28	0.045 *	0.00	6.9	3.13	0.043 *	0.00	6.9
May	3.73	0.05 *	0.00	6.8	3.96	0.056 *	0.00	6.9	3.51	0.047 *	0.00	6.7
Jun	2.57	0.03 *	0.01	6.2	2.42	0.031 *	0.02	6.1	1.99	0.025 *	0.05	6.0
Jul	2.18	0.02 *	0.03	4.9	2.18	0.021 *	0.03	4.9	3.20	0.033 *	0.00	5.2
Aug	2.47	0.02 *	0.01	4.5	2.47	0.022 *	0.01	4.5	4.88	0.054 *	0.00	5.4
Sep	3.06	0.03 *	0.00	5.0	4.60	0.052 *	0.00	5.7	2.37	0.022 *	0.02	4.9
Oct	3.64	0.04 *	0.00	6.3	3.73	0.041 *	0.00	6.3	3.19	0.033 *	0.00	6.2
Nov	2.76	0.03 *	0.01	5.5	4.02	0.042 *	0.00	5.8	2.21	0.019 *	0.03	5.4
Dec	4.14	0.04 *	0.00	5.3	3.93	0.035 *	0.00	5.3	3.93	0.035 *	0.00	5.3
MAM	4.77	0.06 *	0.00	5.6	4.73	0.056 *	0.00	5.6	4.51	0.051 *	0.00	5.4
JJAS	3.33	0.03 *	0.00	3.9	3.70	0.031 *	0.00	4.0	3.89	0.033 *	0.00	4.1
Annual	6.19	0.04 *	0.00	3.4	6.50	0.043 *	0.00	3.5	5.93	0.039 *	0.00	3.3

* Indicates the trend is statistically significant at a 95% significant level.

and

Table 6. Minimum temperature trend and variability from 1980–2040 at Babile station.

	CanESM2 (1980–2040)				HadGEM2 (1980–2040)				GFDLESM2M (1980–2040)
Time Series	MKZ	Sen’s Slope	p-Value	CV%	MKZ	Sen’s Slope	p-Value	CV%	MKZ	Sen’s Slope	p-Value	CV%
Jan	1.14	0.02	0.25	18	0.73	0.01	0.46	19	0.73	0.01	0.46	19
Feb	1.71	0.03	0.08	15	1.05	0.02	0.28	16	0.00	0.0	0.99	16
Mar	3.97	0.07 *	0.00	14	1.34	0.02	0.18	14	0.89	0.01	0.37	15
Apr	4.89	0.05 *	0.00	9	3.26	0.03 *	0.00	8	3.77	0.03 *	0.00	9
May	4.91	0.04 *	0.00	7	4.51	0.03 *	0.00	7	4.31	0.03 *	0.00	7
Jun	3.86	0.03 *	0.00	8	2.76	0.02 *	0.01	7	1.69	0.01 *	0.09	7
Jul	3.26	0.02 *	0.00	8	2.49	0.02 *	0.01	8	2.73	0.02 *	0.01	8
Aug	4.15	0.03 *	0.00	5	3.84	0.02 *	0.00	5	5.02	0.03 *	0.00	6
Sep	5.45	0.04 *	0.00	7	5.20	0.04 *	0.00	6	2.34	0.02 *	0.02	6
Oct	3.85	0.06 *	0.00	13	2.69	0.03 *	0.01	12	1.86	0.02	0.06	12
Nov	3.12	0.04 *	0.00	14	1.90	0.02 *	0.05	14	1.07	0.01	0.28	15
Dec	2.37	0.04 *	0.02	15	1.95	0.03 *	0.05	15	1.37	0.02	0.17	15
MAM	4.81	0.05 *	0.00	9	2.74	0.02 *	0.01	8	2.55	0.02 *	0.01	8
JJAS	4.82	0.03 *	0.00	6	4.1	0.03 *	0.00	6	3.60	0.02 *	0.00	5
Annual	3.39	0.03 *	0.00	8	2.30	0.02 *	0.02	8	1.64	0.01	0.10	8

* Indicates the trend is statistically significant at a 95% significant level.

). This result agrees with the study carried out by [62] in East and West Hararghe, where the minimum temperature shows an increasing trend and is projected to increase by 0.34 °C for 2030 and 0.41 °C for 2050 under RCP4.5, while the maximum temperature is projected to change by 0.02 °C for 2030 and 0.54 °C for 2050 under RCP4.5. Similarly, in the study conducted by [63] using a multi-model data set, the mean annual temperature is likely to rise significantly when compared with the 1961–1990 level by a maximum of 1.1 °C by 2030, and 2.1 °C by 2050. Furthermore, the minimum temperature variability in certain months (January, February, and March, and October, November, and December) of observed and GCM model output in the specified range of year shows more variability than the rest of months (Table 6), while maximum temperature variability is shown to be less variable in January, February, and March, and October, November, and December than during the other months (Table 5).

3.5. Sorghum Crop Production

Sorghum productivity in the Babile district was analyzed for the period from 1995–2020. The analyzed results revealed that the minimum, maximum, and mean of sorghum grain yield were 0.5, 2.26, and 1.62 t/ha, respectively, while the standard deviation was 4.1, with the annual coefficient of variation of grain yield being 25.3% (

Table 7. Summary statistics of climate data set response with sorghum grain yield (1995–2020).

Variables	Mean	Std. dev	CV	Min	Max	Corre	t-Value	p-Value	R2
Grain yields (t/ha)	1.62	4.1	25.3	0.5	2.26
(GDDs) May–November	2594.4	181.0	7.0	2170.2	3039.1	0.2	0.473	0.064	0.092
May RF (mm)	106.7	76.9	72.1	0	255.6	0.23	1.154	0.260	0.052
June RF (mm)	56	42.7	76.3	0	173.4	0.12	0.567	0.575	0.015
July RF (mm)	75.2	42.8	56.9	5.8	190.8	0.27	1.365	0.185	0.072
August RF (mm)	82.6	45.1	54.6	5.3	201.8	0.34	1.761	0.051 *	0.115
September RF(mm)	79.4	54.2	68.3	0.7	233.5	0.39	2.102	0.046 *	0.156
JJAS RF (mm)	292.1	106.3	36.4	43.3	505.6	0.26	1.322	0.199	0.068
NRD in Sept (days)	11	7.0	63.6	0	26	0.31	1.595	0.063	0.096
MayTmax (°C)	28	2.0	7.0	25.7	32.5	−0.37	−1.923	0.066	0.134
JunTmax (°C)	27.7	1.7	6.1	24.5	33.0	−0.35	−1.813	0.082	0.121
SeptTmax (°C)	27.8	1.6	5.8	25.3	32.2	−0.24	−1.217	0.235	0.058
JunTmin (°C)	15.7	0.7	4.5	12.9	16.5	−0.39	−2.045	0.05 *	0.148
AugTmin (°C)	15.6	0.6	3.9	14.4	16.4	−0.24	−1.221	0.233	0.059
SeptTmin (°C)	15.8	1.0	6.4	12.2	17.3	−0.27	−1.377	0.181	0.073
OctTmin(°C)	15.2	1.2	7.7	12.3	17.0	−0.35	−1.832	0.079	0.123

* Indicates the correlation between July and September rainfall amounts and sorghum grain yield is statistically significant at a 95% significant level.

). From the previous study, the average national sorghum grain yield level in Ethiopia was estimated at 2.18 t/ha less as compared to the 4.54 t/ha average grain yield potential [18]. Hence, according to our analysis, the average sorghum crop yield for the Babile district is far less than the national level.

3.6. Correlation and Regression of Sorghum versus its Growing Period Climate Variables

In this study, we employed climate and crop data for the period 1995–2020. Such a length of data is supported by [64], which dictated the accumulation of crop yield, and climate data for a period of 15 or more years are useful in expressing seasonal climate variability in relation to crop yield. Thus, the Pearson product–moment correlation coefficient analysis was carried out for considering sorghum crop yield and climatic variables in terms of month-based maximum and minimum temperature, monthly rainfall amount, and number of rainy days during the growing period, extending from May to November to find out how sensitive sorghum crops were to those variables. The computed correlation values between sorghum grain yield and some climate variables (Table 7) are statistically significant at a 95% significant level.

The results revealed that sorghum grain yield is weakly positively associated with monthly rainfall (Table 7), whereas it has a negative relationship with monthly maximum and minimum temperature during the crop growing period (Table 7). Here, the results of sorghum grain yield variability evaluated by inter-seasonal climate variability separately showed that changes in sorghum grain yield variability were influenced 15.6% differently with a correlation value of (+0.39) and 11.5% differently with a correlation value of (+0.34) due to September and August rainfall variability, respectively, whereas there was a 14.8% difference with a correlation value of (−0.39) and a 13.4% difference with a correlation value of (−0.37) due to the June minimum and May maximum temperature variability, respectively (Table 7). Similarly, this study agrees with the study carried out by [23], which also documented that sorghum crop yield has a negative relationship with temperature, while grain yield has a positive relationship with monthly rainfall.

The correlation between seasonal rainfall total and sorghum grain yield was only 0.26, which is lower than the monthly rainfall time period. The implications of these findings are also supported by the long-known fact that seasonal rainfall totals are weakly correlated as compared to intra-seasonal rainfall variability (Table 7). As result, sorghum crop yield was positively correlated with monthly rainfall and the number of rainy days, while it was negatively correlated with monthly maximum and minimum temperature in the study area (Table 7). In Botswana, a comparable study found a 0.53 correlation coefficient between sorghum yield and seasonal rainfall [65]. However, the seasonal rainfall correlation coefficient computed by [66] in Ethiopia over west Oromia was 0.68, and the maximum and minimum temperature correlated with sorghum grain yield were −0.55 and 0.23, respectively, which indicated that rainfall and minimum temperature were positively associated with sorghum grain yield, while there was a negative relationship with the maximum temperature of its growing period. However, in our study area, both monthly maximum and minimum temperature during the crop growing period have negative relationships with sorghum grain yield (Table 7).

3.6.1. Analysis of Sorghum Grain Yield Response to Climate Variability

Sorghum grain yield response to inter-seasonal climate variability was examined by using climate and crop data for the period (1995–2020) in this study. Figure 6 and Figure 7 demonstrate the relationship between sorghum grain yield and month-based climate variables during the growing season separately. Even though the goodness-of-fit of the regression equations is statistically not significant at alpha 0.05, the results revealed that sorghum grain yield is weakly associated with monthly rainfall (Figure 6) and inversely proportional to monthly maximum and minimum temperature (Figure 7). As indicated in both Figure 6 and Figure 7, each climate variable did not explain the variation in sorghum yield performance separately. However, there is a tendency for sorghum crop yield to increment with rainfall while declining with an increase in temperature. This study agrees with the study carried out by [23], which also documented that sorghum crop yield is inversely proportional to the change in temperature, while grain yield is directly proportional to the change in rainfall. The adjusted R-squared statistic of multiple regression revealed that the overall effect of month-based rainfall and temperature conditions during the growing period was 76.9% (

Table 8. The goodness-of-fit test for the constructed multiple linear regression model.

Parameter	Estimate	Std. Error	t-Value	Pr (>|t|)
(Intercept)	7.6585998	1.3591759	5.635	1.63 × 10−5 *
JunTmin	−0.3606583	0.0608823	−5.924	8.56 × 10−6 *
GDDMN	0.0014246	0.0002712	5.253	3.85 × 10−5 *
JunTmax	−0.1712791	0.0261689	−6.545	2.23 × 10−6 *
AugRF	0.0037350	0.0008755	4.266	0.000377 *
Sept NRD	0.0276674	0.0075862	3.647	0.001603 *
R-squared (R2)	0.7686			1.015 × 10−6 *
F-statistics	17.61 *

* indicates the multiple regression model that constructed using temperature, number of rainy das and rainfall amount has high predictive skills, with goodness-of-fit being highly statistically significant at a 99% significant level. (, 0.05 ‘.’, 0.1 ‘ ’).

), which indicated that five variables explained 76.9% of the variation in sorghum grain yield. In contrast, 23.1% of the sorghum grain yield variation remained unknown, and could be required to explain sorghum grain yield variability in the study area.

3.6.2. Sorghum Yield Prediction

The multi-regression model based on climate variability was constructed as follows: sorghum yield in t/ha = 7.66 − 0.361JunTmin + 0.0014GDDMN − 0.171JunTmax + 0.00374AugRF + 0.028SepNRD. In this multiple linear regression equation, the projected climate data from three GCM model outputs, namely GFDLESM2M(I), CanESM2(D), and HadGEM2-ES(K) (CMIP5) under the RCP 4.5 climate scenario, were used separately to make the sorghum grain yield prediction for the near-term 21st century (2020–2040). However, the line graph of the observed years from 1995–2020 of sorghum grain yield variability shows upward and downwards trends. The trend tested by the Mandel–Kendall method showed a Z value of 3.42 and a sense slope of 0.0363 with a p-value of 0.00063, which showed an increasing trend is statistically significant at an alpha less than 0.05, but in the prediction time period from 2020–2040 it was shown to be decreasing relative to observation trend. A comparable study in Ethiopia showed an increased trend of rain-fed sorghum production from 1993–2019 [67].

According to the result shown in Table 4, decrease in rainfall amount in August, September, or JJAS during cropping season in the near-term 21st century leads to rainwater deficits for crop growth and development, whereas warming over the study area as indicated in Table 5 and Table 6 could also have more negative impacts on grain yield. Based on [68]’s method, an increase in temperature leads to an increase in the rate of water loss through evapotranspiration from field crops. This leads to rainwater deficit, which has a negative impact on crop production. Thus, the result of sorghum grain yield predicted by the multi-regression equation in the coming years of the 21st century show a decrease relative to observed grain yield (Figure 8). This result agrees with previous studies, which have already revealed that a combination of increased rainfall variability and increasing ambient air temperatures will cause a significant decline in the yields of major crops [48]. Furthermore, the sorghum crop productivity is significantly influenced by climate change and variability in India, as documented by [22]. Similar studies carried out in sub-Saharan African showed evidence of negative climate variability impact on crop yield for major staple cereal food crops, such as maize, sorghum, and millet [69,70,71].

4. Conclusions

The impacts of climate change and variability on sorghum grain yield was performed using the rainfall anomaly index, correlation, and multi-regression models. Monthly rainfall amounts and the number of rainy days during the growing season were both positively correlated with sorghum grain yield, whereas maximum and minimum temperatures were negatively correlated with sorghum grain yield. In addition to this, the rainfall variability analysis during the JJAS and MAM seasonal and dry spell risk analysis showed a highly variable and high risk of dry spell occurrence. Even though farmers are using the May–November months as the sorghum crop growing period, a relatively lower dry spell risk was identified from July to September, making this the optimum crop growing period. However, the reality of the occurrence of climate variability characteristics varies from year to year, which forces farmers to frequently plant outside of this optimum planting window. Such rainfall variability and the high risk of dry spells play a major role in crop productivity in the study area, where agricultural practice is generally rain-fed.

Based on generated future GCM climate models’ data output during the Kiremt season, the monthly rainfall in the near-term 21st century showed a decreasing trend, while the maximum and minimum temperatures showed an increasing trend. From the climatological point of view, there exists a relationship between sorghum grain yield and monthly climate parameters. Similarly, rainfall and the number of rainy days have positive relationships with sorghum grain yield, but the maximum and minimum temperature is negatively associated with sorghum grain yield. The result revealed that an increase in temperature and rainfall variability in the near-term 21st century induced increased water loss though evapotranspiration, which results in more impacts on sorghum crop yield. The result revealed that an increase in temperature and rainfall variability in the near-term 21st century induced am increase in water loss though evapotranspiration, which impacts on sorghum crop grain yield reduction. As a result, the predicted sorghum grain yield in the future will decline. Similarly, the multi-regression model that used future-generated climate data as the input for near-term 21st century results showed a decreasing trend in the study area.

Therefore, this finding showed that sorghum productivity varied and changed over time as a result of climate variability and change, implying the necessity for developing an adaptation strategy to this impact. This implies that the need for adjustments, such as on-farm rainwater management and utilization, are prime issues for the sustainability of rain-fed agriculture. Thus, the use of soil moisture conservation agronomic practices, manure application, and high dry spell resistance and short cycle crops are highly recommended to overcome this impact of climate variability and change. Despite the fact that the climate system is complicated and has many variables, the current study only considered sorghum grain yield with monthly rainfall and temperature-related characteristics. As a result, future research should add other non-climate parameters for more in-depth investigations in this study area.