عنوان مقاله [English]
Background and objectives: A reliable forecast of crop yield has more importance regard to the policy decision of agricultural programming. Also, the main concern of developing countries is the knowledge about the crop yield values with emphasis on the effective factors of yield. The aim of study is the yield forecasting of some crops based on the climatological, stochastic aspects. The hybrid method is the combination of two climatological and stochastic aspects.
Materials and methods: The efficiency of three aspects regard to the yield forecasting of wheat, barley, potato and maize in East and West Azerbaijan Provinces (Tabriz, Jolfa, Maragheh, Ahar, Mianeh, Urmia, Maku, Khoy, Mahabad, Maku) were evaluated in the time period 1366 to 1395. Crop yield estimation with the climatological aspect was conducted basis on the significant climatological data using the regression analysis. The determination of effective data regard to the crop yield was basis on the significant correlation coefficient of crop yield and meteorological data. The Auto Regressive Integrated Moving Average (ARIMA) model has been applied for the time series analysis. In the stochastic aspect, the time series modeling was based on the preliminary analysis, identification, estimation, diagnostics steps. The differencing method has been used for stationary of time series. In the final aspect, the meteorological data are estimated basis on the stochastic aspect and then crop yield are estimated using regression analysis.
Results: The effective meteorological data were the total sunshine hours, minimum relative humidity, maximum and minimum of temperature, mean wind speed, maximum relative humidity and mean daily temperature. The preliminary analysis of time series is checking the presence of trend and normality of time series. The time series of crop yield were normal based on the probability plot of time series. The models are passed the diagnostic step such as normality and independency of residual are introduces as the selected models: East Azerbaijan Province: wheat: ARIMA(0,1,1), barley: ARIMA(1,1,1) , potato: ARIMA(0,1,2), maize: ARIMA(3,1,2), West Azerbaijan Province wheat: ARIMA(0,1,3),maize: ARIMA(1,1,1). The comparison of average error criteria for all crops in two provinces indicated the 47.08% RMSE decreasing from regression analysis to stochastic, 21.16% from hybrid to stochastic, 49.53% MAE decreasing from regression analysis to stochastic, 30.32% from hybrid to stochastic. The results of simulation in most cases are overestimated except for maize and potato of regression analysis and hybrid which are underestimated. The average of MARE and NRMSE of three simulated methods for all crops indicated that the minimum and maximum error was related to the potato and wheat, respectively. In this case, NRMSE and MARE decreasing from wheat to potato was 44.68% and 41.66%.
Conclusion: Food supply for growing population without the sustainable agricultural development is not possible. The significant increasing trend of maximum, minimum temperature and wind speed time series can be due to the climate change which the effects of that can be observed in the water balance of Lake Urmia. According to the little research on the crop yield estimation using stochastic aspect in Iran, the time dependency modeling and trend analysis of time series improved the results. The hybrid method is highly effective when the involved parameters are accurately estimated. In the hybrid method, it must be considered a range of precision in the crop yield estimation which the mentioned range can be increased the accuracy of hybrid model. The optimization of coefficients estimation can increase the efficiency of method.