Predicting Winter Wheat Heading Date: A Simple Model and Its Validation in Kansas

Abstract

Abstract Accurate prediction of winter wheat (Triticum aestivum L.) heading date is important for determining the potential incidence of diseases and abiotic stresses such as freeze or heat events. Wheat phenological modeling requires cultivar- and crop-zone-specific vernalization and photoperiod knowledge. Previous models applied in Kansas showed that the uncertainties of predicting heading date were large and could be improved. In this study, a modification to the Scientific Impact Assessment and Modeling Platform for Advanced Crop and Ecosystem Management (SIMPLACE) model was developed and implemented to improve the accuracy of winter wheat heading date estimation. The cultivar- and crop-zone-specific model parameters were calculated using a Markov chain Monte Carlo simulation. The modified models were calibrated by using the longest observation site to cover all cultivars planted in each crop zone. Model performance was then evaluated for seven winter wheat cultivars at eight experiment sites distributed across four crop zones in Kansas. Our modified model (MS) had a root-mean-square error (RMSE) between predicted and observed heading date of 4 days, which reflects an improved accuracy by 5–8 days on average compared to the Agricultural Production Systems Simulator (APSIM) or the original SIMPLACE models. There was a clear correlation between the uncertainty of the modeled heading date and the sowing date in previous models. Our modified model demonstrates that integrating nonlinear temperature response functions, temperature stress factors, and sowing date information improved prediction of the heading date in winter wheat across Kansas.