A consistent calibration across three wheat models to simulate wheat yield and phenology in China

Calibration is essential to reduce the uncertainty caused by model parameters in applying multiple models’ ensemble. However, the method to consistently calibrate multiple crop models is still lack to date. With the three widely used crop models - CERES, CROPSIM, and NWheat - embedded in the Decision Support System for Agrotechnology Transfer (DSSAT) v.4.7, this study develops a calibration method by consistently optimizing eight genetic coefficients for the models. The study applies the method at the regional scale across seven wheat mega-environments in China. Without consistent calibration, the three models show a large variation in simulating wheat maturity date and yield components. After using regionally calibrated parameters, models’ ability and consistency in reproducing past wheat growth have been improved. Furthermore, the three-model-ensemble exhibits better performance than most of the individual models in simulating the phenology and yield. The results of this study indicate that a comparable and consistent calibration method across models can reduce the uncertainty arising from multiple models, facilitating the application of models ensemble in assessing crop growth variability in space and time.