1. INTRODUCTION
Today, climate change has become one of important emerging issues and the biggest concern of mankind as a consequence of scientific evidence about the increasing concentration of greenhouse gases (GHG) in the atmosphere. According to the Intergovernmental Panel on Climate Change (IPCC, 2007a, 2007b, 2014a), successive scientific assessment reports proved that global average temperature would rise between 1.4 and 5.8°C by 2100 with the doubling of the CO2 concentration in the atmosphere, change in rainfall pattern from one region to another (Cubash et al., 2001). Ethiopia will be more vulnerable to projected temperatures and rainfall trends. The fifth IPCC report officially published in 2014a indicates warming in all four seasons over Ethiopia, which may result in more frequent heat waves. In terms of projected rainfall, the climate will be wetter, with more intense wet seasons and less severe droughts in different months throughout of year.
The essentiality of Ethiopian agriculture related to climate change impacts analysis and adaptation measures is self-evidenced by agriculture’s multiple roles in the country. Food security, employment, income and significant portion of GDP are drawn from agriculture. Agriculture accounts about 41 % of the GDP, 90 % of the exports, and serves as the direct source of employment and livelihood for about 85 % of the population (NMA, 2001; Declan and Lisa, 2010; CSA, 2011).
Durum wheat is grown on 70 % on total wheat areas and is part of strategic crops for contribution to food security and livelihood improvement of smallholder farmers (Kassahum et al., 2014; DZARC, 2014).
According to IPCC, unless effective adaptation strategies are carried out timely, some African countries could lose up to 50 % of yield from rain-fed agriculture by the year 2020 and access to food will be severely compromised in many African countries (IPCC, 2007a). Ethiopia cannot be an exception given its overdependence on climate driven economy. Such impacts that significantly undermine the prominent role of agriculture in food production and economic growth predominantly signify the criticality of adaptation.
Thus, there is a hypothesis that climate change might affects durum wheat yield in Central Rift Valley significantly till the year 2030s, 2050s up to 2080s mainly due to increased average temperature out of the cardinal range and rainfall fluctuation.
Therefore, climate assessment and deliberate and conscious adaptations that can cope with these evolving impacts is an immediate concern in agriculture.
Climate and crop modeling are the right and accurate tool for the projection of future climatic condition and provides necessary data to run impact models of the crops’ growth and development and develop alternate decisions for localized adaptations under climate change condition (Koocheki et al., 2001; Jones et al., 2003; Schoof et al., 2008; Mengistu and Eyale, 2011; Habtamu et al., 2012; Valizadeh et al., 2013).
Therefore, this experiment was conducted at Debre Zeit Research Center, in the Central Rift Valley of Ethiopia. The research questions addressed in this study were (i) What is the future temporal change: trend of temperature and rainfall from the year 2010 till 2099 from the base period (1980-2009) level? (ii) What is the impact of this change on the durum wheat production in Central Rift Valley? (iii) What are the possible adaptation measures to live with this change?
Objectives are (i) identify temporal changes in rainfall and temperature at different time period (2010-2039; 2040-2069; 2070-2099); (ii) to assess impact of climate change on grain yield of rainfed durum wheat and identify management options for future adaptation.
2. METHODOLOGY
2.1. Description of Study Area
The study was conducted at Ada’a District, which is located in range of 8o25’0” and 9º55’0″ North latitude and 38o45’0 and 39º10’0″ East longitude in Oromia National Regional State about 47 km southeast of the capital city of Ethiopia, Addis Ababa. Debre Zeit is the center of Ada’a district. It is characterized by moist tropical climate and experiences mainly long rainy season extending from June to September. The dry season extends from October to February. The district has a total land area of about 1610.56 km2 and divided into different agro-ecological zones. The majority of trial fields are heavy soils (Vertisol) with few pockets of light soils (Alfisols/Mollisols) (Tefera et al., 1996; WRB, 2006, CSA, 2006).
Figure 1. Map for Study Area and Agro-Ecological Zones
2.2 Data set and climate model
2.2.1 Observed climate data
Daily meteorological data were obtained from Ethiopian Institute of Agricultural Research (EIAR) of Ethiopia for a period of 30 years (1980-2009). These include daily rainfall, minimum and maximum temperatures. Data were quality checked and homogeneity test was carried out (Yemenu and Chemeda, 2010).
2.2.2 Future climate data
The IPCC has defined standard greenhouse gas concentrations for the Representative Concentration Pathways (RCPs) for use in the evaluation of projected climate change in order to provide a range of possible futures for the evolution of atmospheric composition (Meinshausen et al., 2011). The scenarios show the result of different levels of emissions of greenhouse gases, from the present day to 2100, on global warming. IPCC does not indicate which policy and behavioral choices society could make that would lead to the scenarios. In all scenarios, carbon dioxide concentrations are higher in 2100 than they are today. The low-emissions scenario assumes substantial and sustained reductions in greenhouse gas emissions. The high-emissions scenario assumes continued high-emissions. The two intermediate scenarios assume some stabilization in emissions (IPCC, 2014a).
In this study, two scenarios have been used. RCP8.5 and RCP4.5. RCP8.5 is related to concentrations rise with increasing speed until the forcing (extra energy trapped in entire atmosphere) is 8.5 W/m2 in the year 2100. This is a high scenario of concentration rise while RCP4.5 represents concentrations rise with increasing speed until the forcing (extra energy trapped in entire atmosphere) is 4.5 W/m2 in the year 2100. This is a moderate scenario of concentration rise (Rosenzweig et al., 2013).
Site-specific future temperature and rainfall data were downscaled from an average ensemble of five GCMs: BNU-ESM; CCSM4; CESM1-BGC; IPSL-CM5A-MR ; MIROC-ESM, for the two Representative Concentration Pathway (RCP4.5 and RCP8.5) using the AgMIP Climate team’s methodology as outlined in the AgMIP guide for regional integrated (available for download at www.agmip.org).
2.3. Future trend analysis
Future scenarios are based on historical baseline daily weather data, with each day’s weather variables perturbed using the changes in climate model outputs for future time periods versus those same model outputs for the historical time period. The climate change scenario weather series are created for region study. The model calculates monthly changes in mean maximum temperature, minimum temperature, and precipitation by comparing future 30-year climate periods (near-term: 2010-2039; mid-century: 2040-2069; end-of century: 2070-2099) to the baseline climate period (1980-2009; use RCP 4.5 for 2006-2009 period) from the same GCM (Rosenzweig et al., 2013).
2.4. Climate change impact assessment on durum wheat production
The impact of future climate change on durum wheat production was carried out using Decision Support System for Agro Transfer (DSSAT ver. 4.5) in order to stimulate growth, development and yield of Ude and Yerer (Jones et al., 2010).
The DSSAT model is a good example of a system modelling tool. It has been used for more than 15 years for modelling crop (type and phenotype), soil, weather, and management interactions and it has also been employed to assess climate change impacts (WMO, 2010). DSSAT is a collection of independent programs that operate together, thus Crop Environment Resource Synthesis (CERES) modules which simulate for different crop as well as for wheat (Jones and Thornton 2003; Hoogenboom et al., 2004).
In this study, the CERES-Wheat model which is embedded within DSSAT version 4.5 (Hoogenboom et al., 2002) was used to simulate the phenology and yield of wheat in response to environmental and management factors. The CERES-Wheat model employs soil data which have been collected from report and published source (Hoogenboom et al., 1991; DZARC, 2008). The major soil data include soil type, slope and drainage characteristics, and chemical-physical parameters for each soil layer, such as saturated soil water content, lower limit (LL), drained upper limit (DUL), initial soil water content, relative root distribution, soil pH, bulk density and soil organic matter (Ritchie et al., 1998). Crop management data and phenological observations (Ude and Yerer) were collected from Debre Zeit Agricultural Research Center (DZARC) and daily meteorological data as an input to simulate daily leaf area index (LAI) and vegetation status parameters, biomass production, and final yield. Crop genetic coefficients included in the model relate to photoperiod sensitivity (thermal time), duration and rate of grain filling, conversion of mass to grain number and vernalization requirements (Ritchie et al., 1998; Hoogenboom et al., 2002).
Table 1 Cultivar coefficients used with the CSM-CERES Wheat model.
Varieties P1V P1D P5 G1 G2 G3 PHINT
Ude 5.8 45.22 740.7 40 50 1 60
Yerer 5.1 84.5 700.8 35 45 1 60
P1V= Days, Optimum vernalizing temperature, required for vernalization, P1D= Photoperiod response (% reduction in rate/10h drop in pp); P5= Grain filling (excluding lag) phase duration (ºC.d); G1= Kernel number per unit canopy weight at anthesis (#/g); G2= Standard kernel size under optimum condition (mg); G3= Standard non stressed mature tiller wt (incl grain) (g dwt); PHINT= Interval between successive tip appearance (ºC.d); CSM: Crop Simulation Modeling.
Ude(CD95294-2Y) and Yerer (CD94026-4Y) are semi-dwarfs currently registered in National Trial Varieties (NTV) of Ethiopia ( Gerba et al., 2012; MoA, 2010/2011). Moreover, these varieties are most popular for farmers, high yielders currently, good quality of grain, varieties that are planted every year and have used as baseline for comparison, and varieties that have past all stages of selection and which have historical data for at least 5 years in the study region (DZARC, 2014). The difference between Ude and Yerer is mostly their adaptation conditions where Ude performs well in optimal condition while Yerer in waterlogging conditions (DZARC, 2014). For the model calibration, the treatments with the recommended fertilizer rates, i.e. 60 kg/ha urea and 100 kg/ha Di-ammonium Phosphate, thus 45 kg N/ha and 20 kg P/ha for each variety (MoA, 2010/11).
The crop management data include these two varieties, planting date, planting density and fertilizer (application dates and rates) (Jones et al., 2003).
An assessment of climate change impact on the yield of the wheat varieties was performed. The CERES-Wheat model were calibrated using GLUE (Generalized Likelihood Uncertainty Estimation), which is used to estimate automatically genotype-specific coefficients for the DSSAT crop models (James et al., 2011).
2.4.1. Crop Model evaluation and validation
The performance of the model was validated using an independent crop data from years that were not used for model calibration. For Ude variety, calibration was done using five years and validation using three years, while for Yerer variety, calibration has been done using six years and validation with three years. Model performance was assessed through root mean square error (RMSE) and index of agreement or d-statistic. The Root Mean Square Error (RMSE) and normalized Root Mean Square Error (RMSEn) (Equations (1) and (2)) were computed to measure the coincidence between measured and simulated values. The comparison has been done with simulated mean values of days to heading, days to maturity and grain yield (kg/ha) with measured ones. The value of RMSE approaching to zero indicates the goodness of fit between the simulated and observed values. The RMSEn was computed using the following equations: (1)
where n= number of observations, Pi= predicted value for the ith measurement and Oi= observed value for the ith measurement.
The RMSEn was also computed as follows:
(2)
where RMSE= root mean square error and = the overall mean of observed values.
RMSEn (%) gives a measure of the relative difference of simulated versus observed data (Valizadeh et al., 2013).
On the other hand, d-statistic provides a single index of model performance that encompasses bias and variability and is a better indicator of 1:1 prediction than R2. The closer the index value is to unity, the better the agreement between the two variables that are being compared and vice versa (Willmott et al., 1985). The d-statistic was computed as:
,0<d<1 (3)
where n: number of observations, Pi= predicted value for the ith measurement, Oi= observed value for the ith measurement, = the overall mean of observed values, P’i = Pi- ; O’i = Oi- .
Moreover, linear regression was applied between simulations and observations to evaluate model performance and correlation coefficient (R2) for each simulation (Loague and Green,1991).