Comparing AquaCrop and CropSyst models in simulating barley growth and yield under different water and nitrogen regimes. Does calibration year influence the performance of crop growth models?

• AquaCrop and CropSyst simulated well barley growth under different water and N regimes.
• Only one year of data could be enough for the calibration of models.
• Performance of models changed in respect to calibration/validation year and water and N regimes.
• AquaCrop seems to be easier to calibrate when limited crop growth data are available.

This work investigated the performance of AquaCrop and CropSyst in simulating barley growth under three water treatments (full irrigation, 50% irrigation and rainfed) and two nitrogen levels (high and low) with a particular attention to the influence of calibration year on the modelling results. Three years (2006–2008) of data from the experimental work carried out in Southern Italy were used. The models were calibrated for each of three years and then validated for two other years. The overall results pointed out that both models could be calibrated with data of one of any the three years and validated with all other data. Nevertheless, errors of estimate slightly changed in respect to the year of calibration and were sensitive, from one year to another, to weather conditions and different water and nitrogen regimes. The results indicated AquaCrop superior than CropSyst when the calibration was done on the basis of 2006 and 2008 data, whereas the models performed in a similar way when the calibration was done for 2007. In the case of final biomass, the relative RMSE was lower for AquaCrop (from 0.09 to 0.15) than for CropSyst (from 0.15 to 0.17). Similarly, in the case of final yield, the relative RMSE of AquaCrop was lower (from 0.11 to 0.17) than that of CropSyst (from 0.16 to 0.23). AquaCrop overestimated final biomass by 0.18 and 0.27 t ha−1 for 2006 and 2008 calibration year, respectively, and underestimated biomass by 1.02 t ha−1 when calibration was done on 2007 data. CropSyst underestimated biomass independently on the calibration year, from 0.83 to 1.26 t ha−1.