In order to improve simulation accuracy of maize kernel number under drought stress, the study simulated canopy silking dynamic of maize under drought stress and developed the relationships among anthesis-silking interval (ASI), canopy silking percentage and maize kernel number per unit area.

Firstly, this study measured the average plant growth rate (PGR) of maize around anthesis, daily canopy silking percentage of maize, ASI, the biomass accumulation of ear after anthesis, and the kernel number per plant under different treatments of soil water content based on drought stress controlling experiment at Jinzhou Agrometeorological Experimental Station. Secondly, the parameters of maize canopy silking dynamic model were determined with experimental data. Sensitivity analysis was conducted to investigate the impact of changes in average plan growth rate (PGR_AVE) and standard deviation (PGR_SD) on simulated canopy silking percentage. Thirdly, based on the ear biomass accumulation dynamic, the canopy silking dynamic was simulated under different drought stresses before and after anthesis by considering the differences in PGR among individual plants in the canopy. The quantitative relationship between ASI and kernel setting rate (the percentage of kernel number per plant to the maximum potential kernel number per plant) was developed based on experimental data. Finally, based on simulated canopy silking percentage after anthesis, maximum potential kernel number per plant, and the kernel setting rate by canopy silking dynamic model, the maize kernel number model was developed and validated under drought stress.

Sensitivity analysis of change in canopy silking percentage in response to changes in PGR_AVE and PGR_SD showed that PGR_AVE had a greater impact on canopy silking percentage than PGR_SD. The larger the PGR_AVE and the smaller the PGR_SD, the shorter the time for the canopy reaching silking percentage of 50%. The maize canopy silking dynamic model could accurately simulate daily silking percentage after anthesis under drought stress, and the coefficient of determination (R²), the root mean square error (RMSE), and the normalized mean square error (NRMSE) between simulated and observed canopy silking percentage ranged from 0.88 to 0.98, from 4% to 12%, and from 8% to 27%, respectively. Maize kernel number model could accurately simulate the kernel number of maize per unit area under drought stress, and R², RMSE, and NRMSE between simulated and observed kernel number was 0.85, 185 kernel/m², and 10%, respectively.

By introducing the canopy silking dynamic model, the study could simulate the key phenology (silking time, anthesis-silking interval, and silking percentage) and kernel number per unit area under drought stress. The result was an important foundation for the simulation of maize yield based on canopy silking dynamic under drought stress.