Modelling Climate Change Impact on Irrigation Water Requirement and Yield of Winter Wheat ( Triticum aestivum L.), Barley ( Hordeum vulgare L.), and Fodder Maize ( Zea mays L.) in the Semi-Arid Qazvin Plateau, Iran

: It is very important to determine the irrigation water requirement (IR) of crops for optimal irrigation scheduling under the changing climate. This study aimed to investigate the impact of climate change on the future IR and yield of three strategic crops (winter wheat, barley, fodder maize) in the semi-arid Qazvin Plateau, Iran, for the periods 2016–2040, 2041–2065, and 2066–2090. The Canadian Earth System Model (CanESM2), applying IPCC scenarios rcp2.6, rcp4.5, and rcp8.5, was used to project the monthly maximum and minimum temperatures and monthly precipitation of the region. The results indicated that the maximum and minimum temperatures will increase by 1.7 ◦ C and 1.2 ◦ C, respectively, under scenario rcp8.5 in the period 2066–2090. The precipitation will decrease (1%–13%) under all scenarios in all months of the future periods, except in August, September, and October. The IR of winter wheat and barley will increase by 38%–79% under scenarios rcp2.6 and rcp8.5 in the future periods. The increase in the IR of fodder maize will be very slight (0.7%–4.1%). The yield of winter wheat and barley will decrease by ~50%–100% under scenarios rcp2.6 and rcp8.5 in the future periods. The reduction in the yield of maize will be ~4%. Serious attention has to be paid to the water resources management of the region. The use of drought-tolerant cultivars in the region can be a good strategy to deal with the predicted future climatic conditions.


Introduction
The Earth's climate is changing at an uncharted speed [1][2][3]. This immense change in the climate is mainly due to increase in the Earth's temperature [2][3][4]. The rising temperature is projected to continue its trend in the future decades [3][4][5][6]. One of the fundamental reasons for this is the increase in the accumulation of greenhouse gases such as carbon dioxide, methane, nitrous oxide, and water vapor derived mostly from human activities [1,7]. Climate change is also a function of the volume of ozone, aerosols, and sunspots [8].
Climate change is one of the most important challenges that humans are currently facing [1,9]. Climate change has had remarkably negative effects on many countries in the world, specifically on developing countries that have inadequate resources [10]. Climate change can result in 1.
To predict the monthly maximum and minimum temperatures and monthly precipitation of the semi-arid Qazvin Plateau, Iran, until the end of the century; 2.
To analyze the impact of climate change on the future IR of three strategic crops winter wheat (Triticum aestivum L.), barley (Hordeum vulgare L.), and fodder maize (Zea mays L.) of the region; 3.
To investigate the climate change impact on the future yield of the crops.

Study Area
The area of the semi-arid Qazvin Plateau is 15821 km 2 . Its average altitude, annual precipitation, daily mean temperature, and relative humidity are 1278 m above sea level, 301 mm, 14.2 • C, and 51%, respectively. The Siberian and Mediterranean winds significantly affect and control the climate of the region. The study area consisted of six meteorological stations, namely Qazvin, Buin-Zahra, Kohin, Moalem-Kelaye, Takestan, and Avaj (Table 1).

Data Collection and Preparation
The daily maximum and minimum temperatures and daily precipitation data for a period of 30 years (1986-2015) were collected from the stations. Then, the daily maximum and minimum temperatures and daily precipitation were calculated distinctly by ArcGIS version 10 using the Thiessen polygons method.

Prediction of the Climatic Variables
The Canadian Earth System Model (CanESM2) model was used to simulate the future daily maximum and minimum temperatures and daily precipitation. In this model, the long-term time series are extracted into a one-segment text document per lattice cell. This lattice is uniform along the longitude and contains a resolution of 2.81 • . The linear regression model was used to calibrate the points (stations) against the pixels (lattice cells). For the calibration and validation of the model, 70% and 30% of the observed data were used, respectively. The Statistical Downscaling Model (SDSM) software version 5.2 was used to analyze the climatic data. SDSM is a powerful statistical technique for recognizing the impact of local climate change [34]. To make statistical connections between the variables, a few regression models were used to project the future climatic variables under the changing climate. After evaluating the regression models, the climate change scenarios were applied by the CanESM2 model for the periods 2016-2040, 2041-2065, and 2066-2090. The properties of these scenarios are indicated in Table 2. Table 2. Properties of the applied scenarios authorized by the Intergovernmental Panel on Climate Change (IPCC) [9].

Scenario Properties
rcp2.6 Radiative forcing will peak at 3 W m −2 and will reach to 2.6 W m −2 by the end of 2100; CO 2 accumulation will reach 490 ppm by 2100. rcp4. 5 Radiative forcing will reach 4.5 W m −2 by 2100; CO 2 accumulation will reach 650 ppm by 2100 rcp8. 5 Radiative forcing will reach 8.5 W m −2 by 2100; CO 2 accumulation will reach 1370 ppm by 2100 To evaluate and compare the accuracy of the scenarios and to select the most efficient predictive scenario, statistical indices Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Nash-Sutcliffe coefficient (NS), and Coefficient of Determination (R 2 ) were used: Agriculture 2020, 10, 60 4 of 14 where O i and P i are the observed and simulated values, respectively, and O and P are the averaged observed and simulated values, respectively. σ o is the variance of the observed values, σ p is the variance of the simulated values, and n is the number of data. MAE and RMSE are statistical indices to evaluate the precision of variables [36]. If the MAE and RMSE values are near zero, the scenario is more efficient for predicting the variables [37]. If they are zero, it means that no prediction error exists [37]. The Nash-Sutcliffe coefficient (NS) indicates the similarity between the simulated and measured data. Its amplitude consists of the negative infinity and one, and NS = 1 means a thorough similarity or a complete performance of the scenario [38]. R 2 shows the correlation between the measured and simulated data, and its amplitude is from zero to one [39]. An R 2 value closer to zero indicates a significant correlation between the data categories [39].

Estimation of Actual Evapotranspiration (ET a )
The simulated climatic variables were the daily maximum and minimum temperatures and daily precipitation. Accordingly, the Hargreaves-Samani equation was used to estimate the reference evapotranspiration (ET 0 ) [40]: where ET 0 is the reference evapotranspiration (mm d −1 ); T mean , T max , and T min are the daily mean, maximum, and minimum temperatures ( • C); and R a is the extraterrestrial radiation (mm d −1 ). The ET 0 was multiplied by the crop coefficients to attain the actual evapotranspiration (ETa). The evapotranspiration model was calibrated and validated versus the baseline values. The R 2 (0.984) and RMSE (0.016) values indicated high similarity of the modelled evapotranspiration values to the baseline values. The crop coefficients, growth period, and duration of growth stages were achieved by the OPTIWAT software (Table 3).

Estimation of Irrigation Water Requirement (IR)
The monthly irrigation water requirement was calculated by subtracting the ET a from the effective precipitation (P e ) in each month as follows [41]: where P e is the precipitation stored in the rhizosphere and consumed by the plant. The P e was calculated by the CROPWAT software version 8.0 using the equations of SCS (Soil Conservation Service).

Estimation of Yield
The Stewart model was used to estimate the effect of irrigation water requirement changes on the yield of the crops [42]: Agriculture 2020, 10, 60 where Y a is the actual yield (ton ha −1 ), Y m is the maximum yield (ton ha −1 ), ET a is the actual evapotranspiration (mm d −1 ), ET m is the maximum evapotranspiration, and K y is the coefficient of the reaction of crop yield to water stress. The K y and Y m of crops are calculated based on field experiments. For the crops in this study, K y and Y m were attained from the experiment described in [43] on the same crops in the Qazvin Plateau ( Table 4). The Stewart model was calibrated and validated against the baseline values. The maximum yield values of the crops for the baseline period (2011) were used to attain the actual yield by the Stewart model. The attained actual yield values were compared with the actual yield values reported for the Qazvin Plateau by the Ministry of Agriculture, which were shown to have a high similarity (R 2 = 0.917, RMSE = 0.03). It is important to know that the climate change-derived CO 2 fertilization effect was not considered in the used crop yield model.

Evaluation of Model and Scenarios
All statistical indicators showed that the mean daily maximum and minimum temperatures and daily precipitation values had a relatively high fit to the observed values (Table 5). However, scenarios rcp2.6 and rcp8.5 represented a better fit to the observed values than scenario rcp4.5. Therefore, scenarios rcp2.6 and rcp8.5, the most optimistic and least optimistic ones, respectively, were used to estimate the evapotranspiration, irrigation water requirement, and yield of the crops.

Prediction of Temperature
The simulation results showed that the maximum temperature would rise in the first eight months and decrease in the last four months for all periods and under all scenarios, in comparison with the observed period ( Figure 1). The highest increase in the maximum temperature was attributed to the period 2066-2090 and the rcp8.5 scenario, in which the maximum temperature increased by 1.7 • C. The lowest increase in the maximum temperature belonged to the period 2016-2040 of the rcp2.6 scenario, with an increase of 0.3 • C.

Prediction of Temperature
The simulation results showed that the maximum temperature would rise in the first eight months and decrease in the last four months for all periods and under all scenarios, in comparison with the observed period ( Figure 1). The highest increase in the maximum temperature was attributed to the period 2066-2090 and the rcp8.5 scenario, in which the maximum temperature increased by 1.7℃. The lowest increase in the maximum temperature belonged to the period 2016-2040 of the rcp2.6 scenario, with an increase of 0.3℃.

Prediction of Precipitation
All scenarios projected that the monthly precipitation will decrease in all months of all periods, except in August, September, and October, compared to the observed period ( Figure 3). The highest decrease in the monthly precipitation was related to the scenario rcp2.6 (13% decrease) in the period 2016-2040 in comparison with the observed period. The lowest decrease in monthly precipitation belonged to the scenario rcp4.5 (1% decrease) in the period 2066-2090 compared to the observed period. Investigating the month by month precipitation indicated that the precipitation will increase in August, September, and

Prediction of Precipitation
All scenarios projected that the monthly precipitation will decrease in all months of all periods, except in August, September, and October, compared to the observed period ( Figure 3). The highest decrease in the monthly precipitation was related to the scenario rcp2.6 (13% decrease) in the period 2016-2040 in comparison with the observed period. The lowest decrease in monthly precipitation belonged to the scenario rcp4.5 (1% decrease) in the period 2066-2090 compared to the observed period. Investigating the month by month precipitation indicated that the precipitation will increase in August, September, and October under all scenarios and periods, with September in 2066-2090 having the highest amount. Except for these months, the precipitation will decrease in all months of all scenarios and periods, with June in 2066-2090 having the greatest decrease.

Estimation of Irrigation Water Requirement (IR)
The results demonstrated that the irrigation water requirement (IR) of winter wheat will increase in all periods under both scenarios ( Figure 4). Meanwhile, the IR will further increase as the end of the century approaches. The monthly and total change percentage of the IR is also shown in Table 6. An IR change between 38% and 79% was observed under both scenarios in all periods compared to the baseline period.

Estimation of Irrigation Water Requirement (IR)
The results demonstrated that the irrigation water requirement (IR) of winter wheat will increase in all periods under both scenarios ( Figure 4). Meanwhile, the IR will further increase as the end of the century approaches. The monthly and total change percentage of the IR is also shown in Table 6. An IR change between 38% and 79% was observed under both scenarios in all periods compared to the baseline period.

Estimation of Irrigation Water Requirement (IR)
The results demonstrated that the irrigation water requirement (IR) of winter wheat will increase in all periods under both scenarios ( Figure 4). Meanwhile, the IR will further increase as the end of the century approaches. The monthly and total change percentage of the IR is also shown in Table 6. An IR change between 38% and 79% was observed under both scenarios in all periods compared to the baseline period.    The IR of barley was also enhanced in all periods under both scenarios, with a further increase toward the end of the century ( Figure 5). As in the case of winter wheat, an IR change between 38% and 79% was observed for barley under both scenarios in all periods compared to the baseline period ( Table 7).
Agriculture 2020, 10, 60 9 of 16 The IR of barley was also enhanced in all periods under both scenarios, with a further increase toward the end of the century ( Figure 5). As in the case of winter wheat, an IR change between 38% and 79% was observed for barley under both scenarios in all periods compared to the baseline period (Table 7).    For fodder maize, the IR increased in May, June, and July, and decreased in August and September in comparison with the baseline period ( Figure 6). In total, the IR decreased by 0.2% and 1.4% in the periods 2016-2040 (under rcp8.5 scenario) and 2066-2090 (under rcp2.6 scenario), respectively (Table 8), while it increased in the remaining periods. For fodder maize, the IR increased in May, June, and July, and decreased in August and September in comparison with the baseline period ( Figure 6). In total, the IR decreased by 0.2% and 1.4% in the periods 2016-2040 (under rcp8.5 scenario) and 2066-2090 (under rcp2.6 scenario), respectively (Table 8), while it increased in the remaining periods.

Estimation of Yield
Except in the two periods 2066-2090 under scenario rcp2.6 and 2016-2040 under scenario rcp8.5 for fodder maize, the crops yield decreased in all periods under both scenarios (Table 9). Meanwhile, the yield reduction became more severe as the end of the century approached. The actual yield of the crops is shown in Table 10. Moreover, the total yield of the crops in the Qazvin Plateau under both scenarios for the future periods is presented in Table 11.

Estimation of Yield
Except in the two periods 2066-2090 under scenario rcp2.6 and 2016-2040 under scenario rcp8.5 for fodder maize, the crops yield decreased in all periods under both scenarios (Table 9). Meanwhile, the yield reduction became more severe as the end of the century approached. The actual yield of the crops is shown in Table 10. Moreover, the total yield of the crops in the Qazvin Plateau under both scenarios for the future periods is presented in Table 11.

Evaluation of Model and Scenarios
The CanESM2 model seems to be appropriate for the prediction of the temperature and precipitation of the semi-arid Qazvin Plateau. The scenarios rcp2.6 and rcp8.5 were shown to be the most efficient ones for the prediction of the minimum and maximum temperatures and precipitation. In a study of the Qazvin Plateau using the CanESM2 model, rcp2.6 was found to be the best scenario among the others [5], which is in agreement with the findings of this study. Moreover, the authors of [44,45] showed that CanESM2 model can efficiently predict the temperature and precipitation in the multiple sites with diverse climatic conditions. Evaluating the results of the present study and the other studies presents the usefulness of the CanESM2 model for the prediction of the future temperature and precipitation. However, performing more studies with different climatic models will help us in selecting the most appropriate model for the future research. Furthermore, the model and methods used in the present study can be applied for studies in other regions throughout the world to check whether compatible results are derived. It is also necessary to note that the present model and methods are easy to use and inexpensive. This means that they can be exploited by researchers in regions with less accessibility to scientific facilities. Nonetheless, such models and methods demand a huge data set for being ran efficiently, which might be difficult to attain.

Prediction of temperature
The maximum temperature was projected to increase in all months of the periods but to decrease in September, October, November, and December under all scenarios in all periods. The same trend was predicted for the minimum temperature. Such predicted increasing trends in the future temperature have been also reported by other studies [3,5,6]. Shifts in the seasons would, however, be the reason for the prospective temperature decrease in September, October, November, and December. Since the Qazvin Plateau is severely impacted by Siberian winds, it is possible that the winds will lead to shifts in the seasons of the area [5]. Furthermore, the highest maximum and minimum temperature will relate to the scenario rcp8.5 and period 2066-2090. The results of the present study are in agreement with the results attained by [44]. The highest temperature was predicted by scenario rcp8.5. This seems acceptable due to the governing physical rules to simulate the continuing increase in the radiative forcing and CO 2 accumulations until the end of the century.

Prediction of Precipitation
It was predicted that the precipitation will decrease under all scenarios in all months of the periods but not in August, September, and October. The precipitation was shown to increase in the three mentioned months, with the greatest increase being in September of the period 2066-2090. The scenario-based projected declines in the precipitation compared to the observed values can overall be explained by the prospective elevation of the radiative forcing and CO 2 accumulations by the end of the century. One of the proofs for the projected increase in the precipitation in August, September, and October might be the shift in seasons. Another important reason for the enhancement of the precipitation, specifically in the period 2066-2090, might be the decline in the thickness of the rivers ice and increase in the discharge from May to June in Siberia [46]. This event can be led by redundant warming around Siberia in the period 2066-2090. The authors of [47] predicted that the mean temperature would increase in Siberia by the end of the century. The Qazvin Plateau is severely affected by the Siberian winds. Thus, the highest precipitation projections for the period 2066-2090 might be describable.

Estimation of Irrigation Water Requirement (IR)
Our calculated irrigation water requirement (IR) was a function of actual evapotranspiration and effective precipitation. The evapotranspiration in this study was partly derived from the maximum and minimum temperatures. Therefore, the higher IR for winter wheat and barley by the approach of the end of the century was reasonable, since the maximum and minimum temperatures of the attributed months were predicted to increase and the precipitation was predicted to decrease by approaching the end of the century. Moreover, the slight decrease in the IR of fodder maize in August and September of all periods follows the same regulation. It was predicted that the maximum and minimum temperatures will decrease in September and the precipitation will increase in August and September of each period. Evapotranspiration in arid and semi-arid areas is expected to increase, which can increase the requirement for agricultural irrigation [16]. The irrigation water requirement was shown to be enhanced by an increase in temperature and evapotranspiration [48].

Estimation of Yield
The crop yield was predicted to decrease under both scenarios in all periods, except for fodder maize in periods 2066-2090 (rcp2.6) and 2016-2040 (rcp8.5). Overall, the projected reduction in the crop yield can be linked to the predicted increase in the maximum and minimum temperatures, as well as the decrease in the precipitation. Meanwhile, the predicted slight increase in the yield of fodder maize in the two mentioned periods might partly be attributed to the increase in the precipitation and to the decrease in the maximum and minimum temperatures in September and October. Precipitation and temperature are two important factors affecting crop evapotranspiration, which can indirectly affect the yield. High and abnormal evapotranspiration rates derived from climate warming can impose serious water stress on a crop, which might lead to a significant reduction in the yield. In agreement with this study, it was forecasted that the low amount of precipitation and high temperature will decrease the yield of winter wheat in the Qazvin Plateau as the end of the century approaches [28]. The dryland cereals yield in three semi-arid provinces of Iran was also negatively affected by the low amounts of precipitation [49]. Scenario rcp8.5 (1370 ppm CO 2 ) was shown to have a more negative effect on the crops yield compared to scenario rcp2.6 (490 ppm CO 2 ). The crop yield model used in this study did not consider the future CO 2 fertilization effect on the yield. Otherwise, the CO 2 fertilization may have possibly alleviated the negative impact of water stress on the yield. Especially in C 4 plants like maize, which are capable of assimilating higher amounts of atmospheric CO 2 , the CO 2 fertilization effect can significantly compensate reduction in the yield. It would be interesting and important to consider the CO 2 fertilization effect in future studies using more advanced crop models.

Conclusions
The present study investigated the impact of climate change on the three strategic crops of the semi-arid Qazvin Plateau, Iran. The coupling of a modern climatic model with the classical irrigation water requirement and yield models was shown to be successful and efficient. The CanESM2 climatic model was appropriate for the prediction of the temperature and precipitation of the semi-arid Qazvin Plateau. Moreover, scenarios rcp2.6 and rcp8.5 were the most efficient ones for the prediction of the temperature and precipitation of the region. The models of the present research can be used to study climate change impacts on agroecosystems of other semi-arid regions of the globe.
In general, the irrigation water requirement of winter wheat, barley, and fodder maize will increase as the end of the century approaches. This will lead to noticeable reductions in the yield and can endanger the food security in the region as well as the country. Therefore, significant attention has to be paid to the water resources management of the Qazvin Plateau. In addition, the use of drought-tolerant cultivars can be a good strategy to deal with the predicted future climatic conditions.