Understanding and modeling the effect of temperature and daylength on soybean phenology under high-yield conditions

Temperature and photoperiod can be used to simulate soybean (Glycine max [L.] Merr) phenology because they serve as surrogates for complex biochemical pathways leading to the appearance of certain developmental stages at the whole plant level. We developed a soybean phenology model (SOYDEV) which utilizes non-linear temperature and photoperiod functions and separates floral induction and post-induction for simulating time of flowering. This model accurately simulated the dynamics of vegetative development, final node number and the occurrence of major reproductive stages such as flowering (R1), beginning pod (R3), mid-pod elongation (R3.5), beginning seed (R5), and physiological maturity (R7) in a long-term field experiment (6 years) with the semi-determinate soybean cv. NE3001, as well as in a 2-year cultivar × sowing date experiment with 12 (2004) and 8 (2005) cultivars. With cultivar-specific model calibration, root mean square errors (RMSE) of major phenological stages simulated with SOYDEV averaged 1.8 days in the long-term experiment and 3.3 days in the cultivar × sowing date experiment. Data from the cultivar × sowing date experiment were used to develop empirical relationships for estimating key cultivar-specific model parameters from published soybean maturity group ratings (0.8–4.2). Compared to using nine cultivar-specific parameters derived from the full calibration, estimation of model parameters from readily available cultivar information such as maturity group rating only slightly decreased simulation accuracy, resulting in RMSE (across stages and cultivars) values of 3.6–3.8 days. The proposed SOYDEV model may be particularly suitable for practical model applications because of its reduced need for cultivar-specific calibration. Further evaluation of the model is required under a wider range of variety maturity group ratings and environments.