Climate change is one of the greatest challenges in the 21st century and the agriculture sector is very vulnerable to this phenomenon. Since wheat is the most important cereal crop in Iran, we aim to analyze the potential impact of climatic variables (temperature and precipitation) on rainfed wheat productivity in Hamedan Province, Iran. For this purpose, generalized additive models have been used to model yields of rainfed wheat based on climatic variables during 2004–2012. Then, based on sensitivity of rainfed wheat to temperature and precipitation in this period, we predict the potential effects of climate change on rainfed wheat yield under the IPCC SRES A1FI and B1 climate change scenarios. Results suggest that yields of rainfed wheat would decrease in all Hamedan's counties primarily because of decreasing October to June precipitation and higher temperature. As a result, it is predicted that the yield of rainfed wheat in Hamedan under the A1F1 and B1 scenarios will fall by 41.3% and 20.6%, respectively, in the 2080s. In other words, according to the A1F1 scenario, in the 2080s, Hamedan Province's rainfed wheat production will decline from 1090 kg/ha to 639 kg/ha and under the B1 scenario to 865 kg/ha.
- climate change
- Hamedan Province
- rainfed wheat
Climate change is one of the greatest challenges of the 21st century and agriculture may be particularly vulnerable to climate change (IPCC 2007; Brown & Funk 2008; Khan et al. 2009; Audsley et al. 2010; Hille et al. 2012) because of its dependence on natural weather patterns and climate cycles for productivity. The impacts of climate change on agricultural production are both direct and indirect. Important direct effects will be through changes in temperatures, precipitation and radiation, more frequent extreme weather (like droughts, floods, and wind storms), atmospheric carbon dioxide concentration (Bannayan et al. 2007; Smith et al. 2007; Lobell et al. 2008), length of growing season (Saarikko & Carter 1996) and modifying evaporation, runoff, and finally soil moisture (Rosenzweig & Hillel 1998). Indirect effects will include potentially deterministic changes in diseases, pests, weeds, and sea level rise (Keane et al. 2009), the effects of which have not yet been quantified in most studies. All these factors and variables can change yield and agricultural productivity (Battisti & Naylor 2009). Although there will be gains in some crops in some regions of the world, the overall impacts of climate change on agriculture are expected to have a negative effect, threatening global food security. Predictions show that global agricultural productivity would fall by 15.9% in the 2080s if global warming continues unabated (Cline 2007) and developing countries lying in the tropical and sub-tropical regions would face catastrophic results (Morton 2007; Morison & Morecroft 2008; Mann et al. 2009).
In Iran, agriculture is the second largest sector of the economy, after the service industry, accounting for about 26 of GDP and 26% of non-oil exports. It also provides nearly 23% of the total country's employment, and 80% of food production (Dabiri et al. 2013). Nonetheless, Iran has arid or semiarid climates with long dry summers and low winter rainfall, average annual precipitation around 250 mm and high potential evapotranspiration. Such climatic characteristics associated with phenomena such as desertification, drought, water table reduction and flooding increment and vulnerability of land resources (Momeni 2003) become very vulnerable to climate change. Findings of recent studies have confirmed and documented the occurrence of climate change (IRIMO 2006a, 2006b; Tabari & Talaee 2011) and its damaging potential future effects on the Iranian agricultural sector (see Bolle 2003; Kouchaki & Nasiri 2008; Dastorani 2012; Dehghanpour et al. 2014).
Wheat is the most important cereal crop in Iran, accounting for 35% of the food grain production of the country (12 million tonnes in 2004). Rainfed wheat accounts for about 60–65% of the land area under wheat production in Iran and contributes 30–35% of wheat production in the country. The share of wheat in the daily energy supply is 47% (Ministry of Agriculture Jahad of Iran 2000) and the demand for wheat is predicted to be over 20 million tonnes in 2020 (Sharifi 2001). Rainfed agriculture is much more vulnerable to climate change (Trethowan & Pfeiffer 1999; Ludwig & Asseng 2006). For these reasons, we aimed to analyze the impacts of climate change on rainfed wheat production in the Hamedan Province as one of the main wheat producer provinces in the country.
MATERIALS AND METHODS
The study was conducted in Hamedan Province. This province covers an area of 19,543 km2 and is situated between 33 ° 59′ to 35 ° 44′ N latitude and 47 ° 47′ to 49 ° 28′ E longitude and includes eight counties, namely Hamadan, Malayer, Nahavand, Asadabad, Razan, Bahar, Kaboodarahang and Tuyserkan. Cultivable land of this province is 641,168 ha and 59% of this area – 384,003 ha – is dedicated to the cultivation of wheat: about 80% rainfed wheat and 20% irrigated wheat (Figure 1). Hamedan Province is one of the important producers of wheat in Iran; production of wheat is about 528,979 tons/year. The climate regime of this province is cold and semi-arid. Average annual temperature is 11.3°C and the mean annual precipitation is 300 mm (IRIMO 2006b), occurring mostly from March to April (Figure 2).
Data set and climate model
Although radiation, soil moisture and carbon dioxide (CO2) concentration are all important variables to determine wheat productivity and assessing potential impacts of climate change, temperature and precipitation effects are more significant (Warrick 1988; Chiotti & Johnston 1995; Wang et al. 2009). For this reason, and also because of the lack of accurate information about other variables, the temperature and rainfall data has been used to analyze the effects of climate change on rainfed wheat production in Hamedan Province, for current and future years. For this purpose, climatic data series (minimum and maximum temperature and precipitation) of synoptic stations of Hamedan Province were obtained for the period 2004–2012 from the Iran Meteorological Organization. Wheat production data series used in this study cover the same period from 2004 to 2012 and include data on cultivated areas, production and yield (tonnes/ha) under rainfed conditions. The choice of these periods was justified by the availability of wheat productions statistics for the areas at those periods.
The SRES A1FI (highest future emission trajectory) and B1 (lowest future emission trajectory) climate scenarios for three time slices, namely 2020s (2010–2039), 2050s (2040– 2069) and 2080s (2070–2099), were chosen to evaluate climate change impacts (Table 1).
Generalized additive models
Many researchers have evaluated the possible impact of climate change on crop yields using mainly indirect crop simulation models (see Rosenzweig & Parry 1994; Reilly et al. 2003; Izaurralde et al. 2003; Nassiri et al. 2006; Wang et al. 2013). There are relatively few direct assessments on the impact of observed climate change based on past crop yield (see Carter & Zhang 1998; Naylor et al. 2002; Peng et al. 2004; De la Casa & Ovando 2014; Nara et al. 2014) and then prediction of the future based on fitted trends. In this study we have used the generalized additive models (GAMs) to model past climatic trends and their effects on rainfed wheat yield. Then, based on the sensitivity of rainfed wheat to temperature and precipitation in the period 2004–2012, we predicted the potential effects of climate change on rainfed wheat in Hamedan Province.
Generalized additive models (Hastie & Tibshirani 1986) extend the well-known generalized linear models (GLMs) by allowing for nonlinear covariate effects through the incorporation of additive non-parametric smooth functions. In the univariate case, assume the distribution of the response variable belongs to the exponential family and that its mean related to explanatory variables , , … through the use of a link function . GAMs are very flexible and there is no need for any assumption about the parametric form for the dependence of the response variable on variables , , … When the mean relationships are complex and cannot be easily modeled by specific linear or nonlinear functions, a GAM can be a natural choice. Due to their flexibility in model specification, GAMs have seen a variety of applications. GAMs have been widely used in environmental, biological, and ecological studies.
A GAM has the following general structure (Hastie & Tibshirani 1990): 1where are unknown smooth functions of the covariates , , … . Generalized additive models provide a flexible method for uncovering nonlinear relationships between response variable and covariates in exponential family and other likelihood based regression models. In the case study given in this paper, the iterative method known as the back-fitting algorithm was used to predict the unspecified smoothed function. The cubic spline smoother was applied to the covariates.
The details of the GAM fitting procedure are given by Hastie & Tibshirani (1990). In the present study, the response variable is Yield (tonnes/ha), and our selected covariates are total October to June precipitation (mm), mean temperature of January-February-March (JFM), April-May-June (AMJ), July-August-September (JAS) and October-November-December (OND). The October to June period is selected to calculate the precipitation because in Hamadan Province rainfed wheat cultivation commonly begins in early October with harvest in early June. In order to quantify the rainfed wheat effect of annual precipitation and temperature, precipitation and temperature are usually assumed to be linearly related to the indicators of wheat production in the literature, but the relation to the calendar time and weather variables is not assumed to be parameterized. To take into account the seasonal effect we also included the year as a covariate in the proposed model. Then the proposed model can be written as: 2where are the independent normally distributed measurement errors with mean zero and variance . As can be seen in Figure 3 (specifically in 2008), October to June precipitation is one of the main reasons for yield variation. We also studied the effect of latitude, longitude and elevation of the eight counties of Hamadan Province (Hamadan, Malayer, Nahavand, Asadabad, Razan, Bahar, Kaboodarahang, Tuyserkan) but since they are not statistically significant we have removed them from the final model.
RESULTS AND DISCUSSION
Fitting the proposed model
For the proposed model fitting we used the mgcv package in R statistical software which provides functions for generalized additive modeling and generalized additive mixed modeling. Figures 4 and 5 depict the effect of covariates on the yield and describe the diagnostic plots for the fitted model which indicate that the proposed model is a reasonable model with normally distributed residuals. Based on the residual diagnostics (Q_Q plot, histogram and scatter plots of residuals against the fitted value reported in the paper), residuals of the fitted model did not significantly violate the normality, equal variance, and independence assumptions. The seasonal relation of the year and yields is illustrated very well by the model. Most of the nonparametric terms in the fitted model were significant. The selection of predictors and the decision on their entry to or exclusion from the model was based on the generalized cross-validation (GCV) score and Akaike Information Criterion (AIC). Note that 67.4% of variations of yield are well described by the fitted model. From the fitted model we concluded that there is a positive correlation between precipitation in October to June and mean temperature in JFM. The same result is true for yield. However, with increasing OND mean temperature the wheat yield decreases.
Analyzing climate change impacts on wheat production
As mentioned previously, Iran and especially its agricultural sector is facing problems arising from climate change. The results of much research on the impacts of climate change in Iran (see Kehl 2009; Abbaspour et al. 2009; Amiri & Eslamian 2010) have shown that, during the last half century, in many parts of the country, higher mean temperatures and a reduction in total annual precipitation have been observed. Accordingly, in this section we analyze the effects of climate change on rainfed wheat productivity based on the SRES A1FI (highest future emission trajectory) and B1 (lowest future emission trajectory) climate change scenarios and according to Hamedan rainfed wheat sensitivity to temperature and precipitation described above for the period 2004–2012. Figure 6 indicates that, in each scenario, the yield increases with increasing precipitation. However as we can see overall yield decreases in the 2020, 2050, and 2080s. Results based on the proposed model, presented in Figure 6 and summarized in Table 2, indicate that the yield of rainfed wheat in Hamedan Province will fall by 41.3% and 20.6% in the 2080s under the A1F1 and B1 scenarios, respectively. As can be seen in Table 2 and Figure 7, Hamadan County will have the minimum reduction and Asadabad County will experience the sharpest decline in rainfed wheat production in the future. Figure 8 indicates that the annual yield will increase with increasing mean temperature in JFM, but there is an inverse relationship in AMJ as illustrated by Figure 9; however the effect of mean temperature in OND is not statistically significant. The JAS effect is not of direct interest. Figures 8, 9, 10 and 11 show the partial effects of seasonal temperature covariates on yield.
Climate change is now recognized as one of the most challenging and complex problems facing the globe. Literature review shows that climate change is global; likewise its impacts; but the most adverse effects will be felt mainly by developing countries. The agriculture sector, especially in terms of rainfed crop yields, may be particularly vulnerable to this phenomenon because of the great dependence on natural weather patterns and climate cycles for productivity.
Since wheat is the most important cereal crop in Iran in terms of nutrition and food security, the present study was designed to determine the impact of variation in temperature and rainfall on rainfed wheat in Hamedan Province. In order to quantify the effect of annual precipitation and temperature on rainfed wheat, precipitation and temperature are usually assumed to be linearly related to the indicators of wheat production in the literature, but the relation to the calendar time and weather variables is not assumed to be parameterized. For this purpose, the generalized additive model has been used to analyze past climatic trends and their effects on rainfed wheat yield.
Then, based on sensitivity of rainfed wheat to temperature and precipitation in the period 2004–2012, we predict the potential effects of climate change on rainfed wheat yield in Hamedan Province. To take into account the seasonal effect, we also included the year as a covariate in the proposed model. We also studied the effect of latitude, longitude and elevation of the eight counties of Hamadan Province but since they are not statistically significant we have removed them from the final model. For the proposed model fitting, we used the mgcv package in R statistical software which provides functions for generalized additive modeling and generalized additive mixed modeling.
The results derived from the analysis by R statistical software indicate that the proposed model is a reasonable model with normally distributed residuals. Approximately 67.4% of the variation in yield is well described by the fitted model. Finally, based on the SRES A1FI (highest future emission trajectory) and B1 (lowest future emission trajectory) climate change scenarios we analyzed the potential impacts of climate change on rainfed wheat in Hamedan Province. Results suggest that yields of rainfed wheat would decrease in all Hamedan's counties primarily because of decreasing October to June precipitation (rainfed wheat-growing seasons) and higher temperature under the climate change scenarios. Future research needs to take into account the all-important variables such as radiation, soil moisture and carbon dioxide (CO2) concentration to determine wheat productivity and assess potential impacts of climate change in Hamedan Province.
The authors would like to thank the two referees and the editor whose comments have been very useful in improving the manuscript.
- First received 6 April 2015.
- Accepted in revised form 3 July 2015.
- © IWA Publishing 2016