Increased uncertainty in simulated maize phenology with more frequent supra-optimal temperature under climate warming

• Six models simulated contrasting development rates in response to temperature.
• Most of models simulated maize phenology well under current climate.
• Uncertainty in simulated phenology increased with more supra-optimal temperature.
• The largest uncertainty in predicted maize phenology occurs in North China Plain.
• Model structures caused higher uncertainty than climate scenarios.

Crop phenology is related to the partitioning of assimilates to different organs, crop productivity and timing of crop management. Understanding the uncertainty in simulated crop phenology can help target future direction of model improvement and assess climate change impact more accurately. However, the uncertainty in maize phenology modelling across regions and under climate scenarios has not been properly addressed. This study investigated the uncertainty in simulated maize phenology using six widely used models (SIMCOY, MAIS, Beta, WOFOST, CERES, and APSIM). The models were firstly calibrated and validated using long-term observational data across China’s Maize Belt. The validated models were then used to simulate maize phenology changes in response to climate change. The results showed that the six models could reach acceptable precision (NRMSE < 8% for all the six models) by the calibration under current climate. However, the uncertainty between models in simulated maize phenology increased with the coefficient of fluctuation from 3.2% under the baseline to 6.3% under RCP4.5 and 7.4% under RCP8.5 in 2030s and 8.9% under RCP4.5 and 14.5% under RCP8.5 in 2080s for the simulated silking date, and from 4.2% under the baseline to 7.0% under RCP4.5 and 7.7% under RCP8.5 in 2030s and 10.2% under RCP4.5 and 16.7% under RCP8.5 in 2080s for the simulated maturity date in North China Plain. This highlights a significant knowledge gap in understanding how the key physiological processes of maize respond to changing temperature, particularly temperatures beyond the optimum. The uncertainty in predicted phenology is largest for summer maize in North China Plain, smaller for spring maize in northeast and southwest China. The increased uncertainty in North China Plain was due to more frequent supra-optimal temperatures, where different models disagree most in terms of phenology response to temperature, highlighting a key area for future model improvement. This implies that there could be a large uncertainty in simulated maize yield under future climate change in previous modelling studies conducted with a single crop growth model due to the uncertainty in simulated maize phenology.