Predicting phenological development in winter wheat

Abstract

Accurate prediction of phenological development is important in the winter wheat Triticum aestivum agroecosystem. From a practical perspective, applications of pesticides and fertil- izers are carried out at specific phenological stages. In crop-simulation modeling, the prediction of yield components (kernel number and kernel weight) and wheat-grain yield relies on accurate pre- diction of phenology. In this study, a nonlinear multiplicative model by Wang & Engel (WE) for pre- dicting phenological development in differing winter wheat cultivars was evaluated using data from a 3 yr field experiment. In the vegetative phase (emergence to anthesis) the daily development rate (r) was simulated based on the product of a maximum development rate (Rmax) in the vegetative phase, a temperature response function (ƒ(T)), a photoperiod response function (ƒ(P)), and a vernal- ization response function (ƒ(V)). ƒ(T) was a nonlinear function of the 3 cardinal temperatures for phe- nological development (minimum, Tmin, optimum, Topt, and maximum, Tmax). ƒ(P) was an exponential function of the actual and critical photoperiods and a sensitivity parameter unique to each cultivar. ƒ(V) was calculated using ƒ(T) based on the cardinal temperatures for vernalization (Tmin,vn, Topt,vn, and Tmax,vn). In the reproductive phase, r was simulated based on the product of Rmax for the repro- ductive phase and ƒ(T). Predictions from this nonlinear model were compared to predictions from the phenology submodel of CERES-Wheat V3.0 (CW3). The nonlinear model performed very well for predicting phenological development in the 3 winter wheat cultivars, the mean root mean square error (RMSE) ranged from 2.9 to 4.1 d from booting to maturity. For the CW3 model, the mean RMSE ranged from 4.8 to 5.9 d for the same phenological stages. The WE model predicted double ridge with a mean RMSE of 7.3 d. Both models predicted terminal spikelet with a mean RMSE ranging from 6.2 to 7.1 d. The WE model was generally a better predictor of phenology between booting and maturity than the CW3 model.