Performance of nine maize phenology models in China under historical climate change conditions

Abstract

Accurate and unbiased simulation of crop phenology under various climate conditions is a necessary feature of phenology models. Nine models were evaluated for simulating the vegetative growth period (VGP) and the reproductive growth period (RGP) of maize (Zea mays L.) under historical climate variation. Seven models were based on a constant thermal/photothermal assumption (MAIS, SIMCOY, EPIC, MCWLA, WOFOST, Beta, CERES), and two models were based on a non-constant thermal/photothermal assumption (coupling response and adaptation model, RAM; average number of growing days, NGD). Phenology observations from 150 agrometeorological observation sites across China (1981–2021) were collected to evaluate model performance. Results showed that: (1) Most models simulated flowering and maturity dates well. Average RMSE of VGP was lower than that of RGP. Generally, models based on non-constant thermal/photothermal assumptions had lower RMSE than models based on constant thermal/photothermal assumptions; (2) Models having a fairly high development rate when temperature was slightly higher than base temperature (RAM, Beta, CERES, NGD, MAIS) had the lowest RMSE during RGP; (3) Simulations by some models had systematic biases. First, during VGP, standard deviations of flowering date simulations obtained from models with flexible temperature response curves across sites and years (EPIC, MCWLA, WOFOST, Beta, CERES, RAM) increased more slowly than the standard deviations of observations, while those of other models increased faster. Second, during RGP, unlike RMSE values from other models, those RMSE values obtained from RAM and NGD showed no significant correlation with the average growth period temperature. Our results suggest the importance of further investigating the impact of low temperatures on development rate during RGP in order to reduce systematic bias of models when applied under climate change conditions. Research efforts should be devoted to developing models that have flexible phenology response to temperature curves across sites and years.