Performance of predicted evapotranspiration and yield of rainfed wheat in the northeast Iran using gridded AgMERRA weather data

High quality of long-term daily weather data is essential for simulating crop production and its variability. However, daily weather data with adequate duration and required quality are not available in many regions. This study has evaluated the suitability of AgMERRA (The Modern-Era Retrospective Analysis for Research and Applications) weather data for simulating rainfed wheat evapotranspiration (ET_c) and yield. Daily AgMERRA were compared with corresponding observed weather data of 11 land stations across the northeast Iran, considering the different periods from 1980 to 2010. Cropwat and CSM-CERES-Wheat models were used to simulate ET_c and yield of rainfed wheat, respectively. The comparison of daily AgMERRA with observations resulted in the highest correlation (r² > 70%) and good agreement (d > 0.77 and NRMSE < 30%) between climate variables, except for daily wind speed and precipitation at all locations. However, when daily precipitation data were aggregated into 15-day periods, agreement and correlation improved. According to the monthly comparison, the largest bias between AgMERRA temperature and radiation with land observations was obtained from June to August (summer season). Results also indicated that the distribution of simulated ET_c and yield using AgMERRA was within 10% of the simulated yield using observations at 73% and 100% of locations, respectively. The degree of variation of AgMERRA-simulated ET_c and yield was very similar to the calculated coefficient of variation in simulated ET_c and yield based on observations at 73% of locations. However, simulation of ET_c and yield using AgMERRA for single years was more uncertain when compared with simulated ET_c and yield based on observations for the same year. It is concluded that AgMERRA can provide a robust estimate of long-term average ET_c and yield of wheat than the ET_c and yield of a single year in regions that there is no long-term weather data available.