Testing the suitability of thermal time models for forecasting spring wheat phenological development in western Canada

Predicting crop development stages is fundamental to many aspects of agronomy (e.g., pesticides and fertilizer applications). Temperature is the main factor affecting plant development and its impact on crop development is often measured using thermal-time. We compared different thermal-time models to identify the best model for simulating spring wheat development in western Canada. Models compared include (i) North-Dakota growing-degree-day (NDGDD), (ii) growing-degree-day base-temperature zero (GDD0), (iii) growing-degree-day base-temperature five (GDD5), (iv) beta-function (BF), and (v) modified-beta-function (MBF). We utilised agro-meteorological data collected across western Canada from 2009–2011. Results showed that accumulated heat units/daily growth rates from the different models correlated well with spring wheat phenology with R2 ≥ 0.91 and P < 0.001. However, when the developed models were used to predict time (calendar-days) from planting to anthesis for cultivar AC-Barrie, the BF and MBF models performed poorly. Average predicted times from planting to anthesis by NDGDD, GDD0, GDD5, BF, and MBF models were 63, 63, 62, 65, and 64 d, respectively; while the actual observed time was 60 d. Root-mean-square error (RMSE) for NDGDD was 4 d, 5 d for GDD0 and GDD5, and 6 d for BF and MBF. These findings suggest that simple GDD-based models performed better than more complex BF-based models.