Evaluating Climate Change Impact of Rainfed Maize Production Yield in Southern Ethiopia

A developing country like Ethiopia suffers a lot from the effects of climate change due to its limited economic capability to build irrigation projects to combat climate change's impact on crop production. This study evaluates climate change's impact on rainfed maize production in the Southern part of Ethiopia. AquaCrop, developed by FAO that simulates the crop yield response to water deficit conditions, is employed to assess potential rainfed maize production in the study area with and without climate change. The Stochastic weather generators model LARS-WG of the latest version is used to simulate local-scale level climate variables based on low-resolution GCM outputs. The expected monthly percentage change of rainfall during these two-time horizons (2040 and 2060) ranges from -23.18 to 20.23% and -14.8 to 36.66 respectively. Moreover, the monthly mean of the minimum and maximum temperature are estimated to increase in the range of 1.296 0 C to 2.192 0 C and 0.98 0 C to 1.84 0 C for the first time horizon (2031-2050) and from 1.86 0 C to 3.4 0 C and 1.56 0 C to 3.18 0 C in the second time horizon (2051-2070), respectively. Maize yields are expected to increase with the range of 4.13% to 7% and 6.36% to 9.32% for the respective time horizon in the study area provided that all other parameters were kept the same. In conclusion, the study results suggest that rainfed maize yield responds positively to climate change if all field management, soil fertility, and crop variety improve were kept the same to baseline; but since there is intermodal rainfall variability among the seasons planting date should be scheduled well to combat water stress on crops. The authors believe that this study is very likely important for regional development agents (DA) and policymakers to cope up with the climate change phenomenon and take some mitigation and adaptation


Introduction
A warmer world would likely affect all aspects of our environment. Amongst all sectors (energy, agriculture, water, fisheries, livestock, etc.), agriculture is the most sensitive and vulnerable to climate change (IPCC, 1990). Climate change is now a leading issue on the environmental and socioeconomic agenda worldwide (Crossman et al., 2013). It is the primary determinant of agricultural production and hydrology balance (Adams et al., 1998). There is testimony from the scientific community of the globe that there is climate change now and 97% of climate scientists agree that it is being driven primarily by human activity (IPCC 2001b, Wang et al., 2019). Climate change, however, is expected to make agricultural development in Africa more challenging in many places (E. Blanc., 2012). Ethiopia is one of the most likely vulnerable countries in Sub-Saharan Africa facing climate change extreme events (drought and floods) as a result of climate variability and change, that would influence largely the agricultural sector (Gezie, M., 2019). The study in five regions (Amhara, Tigray, Benishangul Gumuz, Oromia, and Southern Nation Nationality and People (SNNP) of Ethiopia indicated that all the regions are negatively affected by climate change at varying levels (Baylie, M. M., & Fogarassy, C., 2021).
The Federal Democratic Republic of Ethiopia Ministry of Agricultural (2011) report on agriculture sector program plan adaptation to climate change depicted that Ethiopia agriculture would face many adverse impacts that are caused by the unpleasant appearance of climate variable change. Yet there are indications by which these impacts will continue to influence the socio-economic activities of the community at a larger scale. Deressa et al., (2008) study that used biophysical and social vulnerability indices of the Ricardian approach suggests that a decline in rainfall and an increase in temperature are more likely tends to damage Ethiopian agriculture. The second (IPCC, 1996) report predicted that tropical and subtropical regions would experience higher losses in crop production, while temperate climates might gain in productivity with climate warming. Issues of hunger and famine in Ethiopia are associated with low cereal crop production like maize as a result of poor rainfall what happens nowadays in Ethiopia once again. It is apparent, therefore; that tropical/subtropical parts of the nation wholly rely on rainfed farming. Ethiopia's food production could be greatly impacted by climate changes.
Maize is the second-largest most widely produced cereal crop next to teff in area coverage in Ethiopia and the most important commodity used to alleviate poverty in the country (CSA, 2015). Solomon et al., (2021) researched the Impact of Climate Change on Agricultural Production in Ethiopia at the national level and found that crop production will be adversely affected during the coming four decades and the severity will increase overtime period. The future prediction of the study indicates that the production of teff, maize, and sorghum will decline by 25.4, 21.8, and 25.2 percent, respectively by 2050. Timothy et al., (2019) researched the Climate change impacts on crop yields in Ethiopia and suggest that climate change will likely have only relatively small effects on average yields of maize. However, Kassaye et al., (2021) study on the Impact of climate change on the staple food crop yield in Ethiopia suggest that the yield of teff and maize will be expected to increase by 20.2 and 17.9% respectively at the end of the twenty-first century with increasing temperature and rainfall decline. Ironically, Abera et al., (2018) indicate that maize yields will decrease by up to 43 and 24% by the end of the century at Bako and Melkassa stations, respectively, while simulated maize yield in Hawassa will increase by 51%. In contrast to this, the study in the central rift valley of Ethiopia predicted that maize production will decrease on average by 20 % under climate change by 2050 (Kassie et al.,2015). The study in the Gambella region also suggests that rainfed maize production will decrease under three RCPs (2.6, 4.5, and 8.5) in the future but only it will decrease under RCP8.5 after 2040-2069 (Degife et al., 2021).
General circulation models (GCMs) are greatly supportive of the assessment of potential climate change impacts on multiple sectors on a global scale. However, a horizontal resolution of GCMs is typically between 250 and 600 km, which cannot meet the requirements of most local impact studies (Phuong et al., 2020). Hence, many dynamical and statistical downscaling methods have arisen to overcome these key disadvantages of GCMs. By creating General Circulation Models (GCM), climate conditions can be assessed for long-time scales. However, the output of these models does not have enough spatial and temporal accuracy to study the effect of climate change on agricultural and hydrological systems (Pervez and Degife et al., 2021) and none of them applied the combination of Global Circulation Models (GCMs) and AquaCrop whereby LARS-WG is used as a downscaling tool, the main objective of this study is to estimate the effect of climate change in the near future 2040 (2031-2050) and midterm 2060 (2051-2070) on maize yield production in the southern part of Ethiopia using the AquaCrop model in combination with Global circulation models. The study contains four sections: Section 1: is an introduction, section 2: Material and methods, Section 3: Results and Discussions, and Section 4: is the conclusion.

Material and methods 2.1. Study Area
This study was conducted in the Southern part of Ethiopia in Sidama regional state at Hawassa district that covers the latitudinal area from 6 0 40 ' 0 '' N to 7 0 20 ' 0 '' N and the longitudinal area 38 0 20 ' 0'' E to 38 0 40 ' 0 '' E ( Fig.1). The main staple food crops in the zone are maize, haricot beans, kocho, and sweet potato, all produced in relatively small amounts. Chat is an income-generating crop in the higher-altitude areas of the zone, but it is not typical of the zone as a whole (USAID, 2005). There are many types of soils in Hawassa district but the soil type prevailing in the study area is Andosols (a black or dark brown soil formed from volcanic material, with an A horizon rich in organic material). The soil property of the area is characterized by the wilting point of 18.5% and field capacity of 34.8% with no restrictive soil layer and soil salinity stress (K.N. Disasa et al., 2019). Sidama region is technically falling into the borderline area between the kolla and woina dega agro-ecological zones, with altitudes in the range of 1400 -1700 meters above sea level. Average annual rainfall is in the range of 700-1200mm per year and falls during two rainy seasons, the belg (autumn) and kremt (summer) rains. Hawassa district has an annual average rainfall of 955mm with a mean annual temperature of 20 0 C. The main rainy season generally extends from June to October. In the new parallel approach radiative forcing trajectories are not associated with unique socioeconomic or emissions scenarios, and instead can result from different combinations of economic, technological, demographic, policy, and institutional futures. Four pathways lead to radiative forcing levels of RCP 8.5, RCP 6, RCP 4.5, and RCP 2.

Description of LARS-WG
Long Ashton Research Station Weather Generator is neither a predictive nor forecasting tool but is simply a means of generating time-series of synthetic weather statistically identical to the observations (Semenov, M. A. et al., 2002). Its latest version (LARS-WG Version 6.0, to date version) was used for this study to simulate future weather data (maximum and minimum temperatures, precipitations, and Solar radiation) for the near term, and mid-term in the study area. The model simulates weather data at a single site under current and future conditions (Racsko et al., 1991). It uses Minimum temperature (°C), Maximum temperature (°C), Precipitation (rainfall) (mm), and Solar radiation (MJ/m2/day) as input to generate synthetic data in daily time series. In the absence of solar radiation, the model accommodates the use of sunshine hours. LARS-WG automatically converts the sunshine hours to solar radiation using an algorithm that was described by Rietveld (1978). Table 1. Global Circulation Models used in this study

GCM Name
Name of the research center Grid Resolution RCPs 4.5 8.5

EC-EARTH
European community Earth-System Japan Agency for Marine-Earth Science & Technology The ability of LARS-WG to simulate reliable data depends on the availability of observed data. The model simulates future weather data based on as little as a single year of observed weather data. Semenov and Barrow (2002) recommend the use of daily weather data of at least between 20-30 years for better results. Weather data for long periods are significant in the way that they capture some of the less frequent events like droughts and floods. This study used daily historical observed weather data of at least 30 years from the (1981-2010) periods as baseline data.

LARS-WG Calibration and validation
It is very likely important to calibrate LARS-WG before generating future scenarios that are common for most statistical downscaling models. Calibration of LARS-WG is carried out by a function on the main menu called "Site Analysis". The process is done to determine the statistical characteristics and site parameters of the observed weather data. Once LARS-WG has been calibrated, its ability to simulate future weather data in the representative study site is assessed. Validation is a process that is used to determine how well a model can simulate potential future climate variables.
The Q test function was used to determine the ability of LARS-WG to rationally estimate future climate variables. This was achieved using three statistical tests; chi-square test (X 2 ), t-test, and K-S (Kolmogorov-Smirnov) which is the output Q test function to test the performance of LARS-WG. The chi-square test will be used to determine the existence of any significant difference between the simulated and observed frequencies in the meteorological data. A t-test was used to check the existence of any reliable difference between the means of the generated and observed data sets. Additionally; a K-S test was used to decide if a sample comes from a population with a specific distribution. The Kolmogorov-Smirnov (K-S) statistic ∆ is the absolute maximum differences between observed cumulative probability P (Xm) and the theoretical cumulative probability F(Xm).
The observed cumulative probability is computed using Weibul's formula P(Xm)= - The coefficient of determination (R²) is used to determine the proportion of variance in the simulated variable that can be explained by the observed variable. The higher the value for the goodness of fit of the model.
RMSE is used to measure the difference between simulated and observed values. The lower the value the higher accuracy of the model to predict.
The Nash-Sutcliffe efficiency (NSE) is a normalized statistic that determines the relative magnitude of the residual variance between simulated and observed data. The more the value approaches 1 the high the predictive skill of the model.
where Oi is the observed data, Si is the simulated data.

Introduction to AquaCrop
AquaCrop is the FAO crop model to simulate yield response to water Steduto et al., 2009). It is a very likely friendly tool that would be used for a wide range of users and applied for the prediction of crop yield under current and future climate change scenarios. AquaCrop appears to use ground canopy cover instead of leaf area index and mostly focuses on water hereby using water productivity values normalized for atmospheric evaporative demand and of carbon dioxide concentration. This helps the model to use in diverse locations and seasons under future climate scenarios. An empirical production function is mostly suggested to use to estimate crop yield response to water to overcome the difficulty of crop responses to water deficits. Among the empirical function approaches, FAO Irrigation & Drainage Paper 33 (Doorenbos and Kassam, 1979) represented an important source to determine the yield response to water of field, vegetable, and tree crops, through Equation 5: Where Yx and Ya are the maximum and actual yield, ETx and ETa are the maximum and actual evapotranspiration, and ky is the proportionality factor between relative yield loss and relative reduction in evapotranspiration. AquaCrop develops from Doorenbos and Kassam (1979) approach by separating the Evapotranspiration (ET) into soil evaporation (E) and crop transpiration ( T ) and the final yield (Y) into biomass (B) and harvest index (HI). The separation of ET prevents the confounding effect of the non-productive consumptive use of soil evaporation water (E). This phenomenon is very likely important when the canopy does not cover the ground completely. The separation of Y into B and HI allows the distinction of the basic functional relations between environment and B from those between environment and HI. These relations are fundamentally different and their use avoids the confounding effects of water stress on B and HI. The changes described led to Equation 6 at the core of the AquaCrop growth engine: Where T the crop transpiration (mm) and WP is is the water productivity parameter (kg of biomass per m 2 and per mm of cumulated water transpired over the time period in which the biomass is produced). This step from Eq.5 to Eq.6 has a fundamental implication for the robustness of the model due to the conservative behavior of WP (Steduto et al., 2007). It is worth noticing though that both equations are different expressions of a water-driven growth engine in terms of crop modeling design (Steduto, 2003).

Components of AquaCrop
AquaCrop contains five sections (Atmosphere, Crop, Soil, Field Management, Irrigation management) which are used to calculate crop growth. Moreover, it also requires five weather data (daily minimum air temperatures, daily maximum air temperatures, daily rainfall, ETO and mean annual carbon dioxide concentration in the bulk atmosphere). psychometric constant (kPa/ o C) AquaCrop calculations are performed always at a daily time-step. However, input is not required at a daily time-step, but can also be provided at 10-daily or monthly intervals. The model itself interpolates these data to daily time steps. The only exception is the CO2 levels which should be provided at the annual time-step and are considered to be constant during the year.

Climate change in AquaCrop
In this particular study climate change is conducted only by adjusting the precipitation data file and the temperature data file under current and future climate scenarios but the impact of enhanced CO2 levels is not considered. The impact of enhanced CO2 levels is calculated by AquaCrop itself. AquaCrop uses the so-called normalized water productivity (WP*) for the simulation of aboveground biomass. The WP is normalized for the atmospheric CO2 concentration and the climate, taking into consideration the type of crop (e.g. C3 or C4). The C4 crops assimilate carbon at twice the rate of C3 crops. A C4 plant is a plant that cycle's carbon dioxide into four-carbon sugar compounds to enter into the Calvin cycle.

Calibration of AquaCrop
Calibration of AquaCrop has been achieved by comparing the actual maize production (as provided by the Department of Agriculture of southern nations, nationality, and peoples of Ethiopia regional state, Sidama zone) with the maize yield simulated by the model. Input data for various crop and soil parameters used in the model are obtained from the South Region Agricultural Research Institute and Wondo Genet Agricultural Research Center Hawassa Maize Research Sub Center of Ethiopia. By comparing the actual and simulated maize yields crop parameters (CCO, canopy development, root deepening, flowering and yield formation, canopy expansion relative to water stress, stomata closure relative to water stress, etc..), field management(soil fertility stress percentage for biomass production in respective to canopy, biomass, water productivity, and percentage soil surface covered) and soil characteristics (characteristics of soil horizon, soil surface, restrictive soil layer, and capillary rise) are adjusted through trial and error, until the closest match between recorded and simulated maize yield was achieved. Data (recorded maize yields, rainfall, minimum and maximum temperature) from Hawassa station for the year 2018 has been used to calibrate the mode l( Msowoya et al.,2014).

Validation of AquaCrop
Having calibrated AquaCrop, it was significant that the model is validated to evaluate its performance in simulating crop yields. Model validation is important to determine if the model can replicate the data, and analyze the effectiveness of model calibration and compare synthetic data with those done in previous studies. Loague and Green (1991) indicate that there are numerous statistical indicators for evaluating the performance of AquaCrop, nonetheless, Willmott (1984) argues that each of the statistical indicators has its weaknesses and strengths. To effectively evaluate the performance of the model, the use of ensemble statistical indicators is appropriate (Willmott, 1984).

Results and Discussions 3.1. Evaluation of LARS-WG
Based on the Q test function result of LARS-WG the performance of the model to generate future synthetic time series is calibrated and validated. Accordingly, the output statistical test function such as X 2 (chi-square), t-test, and K-S Kolmogorov-Smirnov) with its respective p-value is used ( Table 2). Furthermore, the coefficient of determination (R 2 ), Root Mean Square Error (RMSE), and Nash-Sutcliffe efficiency (NSE) are calculated based on the mean monthly generated and simulated climate variables ( Table 3). The significance level of 5% (0.05) which is common in statistical tests is used. The results indicate that p-values in all months except in December and January for the Chisquare test in rainfall are higher than the selected significance level of 0.05. However, the p-value for maximum and minimum temperature is higher than 0.05 for both statistical tests. This value indicates that LARS-WG is better to generate future temperatures (Max. and Min.) than rainfall (K.N. Disasa 2002) suggest that a significant difference may exist between observed and generated data if the mean climate variables do not follow the expected trend of increasing/decreasing. Therefore the model is still satisfactory to simulate future climate data.  Figure 2-3 shows the monthly mean distribution and daily maxima of climate variables (Precipitation, Maximum and Minimum temperature) in the study area respectively. The monthly mean baseline distribution of rainfall in the study area indicates that there is a bimodal rainy season with a low rainy season in March and April and a high rainy season in June to August that depicts the sowing season could be two times a year.

Generation of Future Climate variables
Once LARS WG is calibrated and validated future climate variables of Precipitation, Maximum and minimum temperature for two time periods (2031-2050) and (2051-2070) under two RCPs are generated. The monthly mean average of change of the climate variables for the respective RCPs and time periods is indicated in Figure 4. The seasonal climate change results indicate that the precipitation is expected to increase in September, October, and November (SON) under both RCPs in the two future time horizons relative to baseline. However, precipitation is expected to increase in 2031-2050 under both RCPs but it would tend to decrease from 2051 to 2070 for the winter (DJF) season called 'bega'. For Autumn (MAM) and winter (JJA) seasons, the precipitation is estimated to decrease for both RCPs in the two coming time horizons (Table 4). The maximum and minimum temperature is estimated to increase monthly and seasonally throughout the coming two time periods under both RCPs. The percentage of monthly precipitation tends to change in the range of -14.84% to 20.23% in 2031-2050 and -23.18% to 20.04% in 2051-2070 under the RCP 4.5 scenario. Similarly, it is predicted to change in the range of -14.8% to 26.98% in 2031-2050 and -12.83% to 36.66% in 2051-2070 under the RCP 8.5 scenario. In general, the monthly precipitation is expected to increase by 0.52% and 0.22% under the RCP4.5 in 2031-2050 and 2051-2070 respectively ( Table 5). Moreover, it is estimated to increase by 0.38% and 6.50% under the RCP 8.5 in the respective future time period. However, the future predicted precipitation indicates that there are high intermodal variabilities and the increasing trend among GCMs is non-consistence.  On the other hand, the minimum temperature is predicted to increase in the range of 1.296 0 C to 2.044 0 C and 1.  Table 5). In general Table 6 indicate the monthly average percentage change of precipitation and absolute relative change of maximum and minimum temperature under the two RCPs scenario for future two time periods. Additionally, the mean monthly climate change variables indicate that maximum and minimum temperature increase throughout the year for both RCPs in the coming two time periods but there is a seasonal shift for precipitation that depicts it would decrease in the summer ("Kiremt") and Autumn ("Belg") season of the country whereas it would likely increase the other two seasons (Figure 4).

AquaCrop Calibration
The future climate change measuring standard of the model simulation capacity is set by inputting the 2018 weather data to predict future climate effect on maize production under two RCPs (RCP4.5 and RCP8.5) and two-time periods (2040and 2060). The model grossly overestimated maize yields when compared with recorded data in the study area for the year 2018. It was therefore essential that adjustment of important and sensitive parameters in the model be carried out (Msowoya et al., 2016). By adjusting crop, management, and soil properties in the main menu of AquaCrop, an output yield close in value to the recorded yield in 2018 was derived. Before calibrating the model, first, the potential evapotranspiration (ETo) was estimated using the FAO ETo Calculator which uses the Penman-Monteith equation. Recorded data for maize production in the year 2018 for Hawassa district was averagely 7.5 tons per hectare. After varying the model parameters, the closest simulated production was 7.496 tons per hectare.

Model Validation
Once model calibration was complete, the model was then validated to determine its potential to simulate maize yields. Accordingly historically recorded maize yields for the area from 2012-2017 were used to compare with those simulated by AquaCrop to determine its potential to simulate future maize yields at two RCPS and two-time periods.  This value is considered maize production without climate change. Accordingly, the average values of future potential maize yields at two-time horizons under two RCPs were calculated similarly. Table 8 summarizes the simulated maize yields as a percentage, relative to the baseline period both RCPs and two-time periods. to 7% and 6.36% to 9.32% respectively. This result indicates that in the future two time periods maize production is yield decreased in future periods. In contrast to these studies research in Ethiopia indicates that maize yield responds positively to change in climate variables (precipitation and temperature). However, the planting date schedule should be taken to account because there is a seasonal precipitation shift in the study area. As it could be observed from the predicted result the autumn season precipitation is increasing, therefore it is recommended to shift the planting month from April which is common in Ethiopia to May.

Conclusions
Climate change is the headline of the global agenda that attracts scientific communities to deal with. It has a direct and indirect impact on the agricultural sector in general, particularly on crop yield. Therefore this study presents the effect of future climate variable changes on rainfed maize production in the southern part of Ethiopia. The study result shows that maize yield production responds positively to climate variables (precipitation, max. temp, and min. temp) change under both RCPs and two future time horizons. Nevertheless, since there is seasonal precipitation variation for both RCPs and time horizons planting dates should be considered under the future climate change scenario.
Accordingly, maize yield is predicted to increase from 4.13% to 7% and 6.36% to 9.32% for 2040 and 2060 respectively, provided that all other parameters (crop, management, and soil properties) kept the same.