Predictability of CERES-Maize for Flowering Date.

The ability to accurately predict flowering date of a corn plant across genotypes and environments is useful for many applications in maize breeding, corn production and inbred characterization for seed production. Crop models have been promising tools to predict flowering date, because the model takes the photoperiod and leaf number into account to simulate the G x E interactions more dynamically than does the method based on thermal units only. However, the advantage of such a photo-thermal method in crop models has not been widely validated and applied. We calibrated CERES-Maize model using multi-environmental trials in 1997 and 2004 – 05 with 30 representative hybrids across corn relative maturity (CRM) 77 – 118. Genotype-specific parameters (commonly known as genetic coefficients) for each hybrid were fitted using the optimization method of Downhill Simplex by minimizing the root mean square error (RMSE) between observed and simulated flowering dates. We then extrapolated these parameters from genotype-specific to CRM level. Finally we validated the crop model with these generalized parameters using dataset collected at about 250 locations in North America and Europe (latitude is from 30.865 to 54.65; longitude is -121.844 ~ -72.466 and -8.658 ~ 36.008, respectively) from 2000 – 2005, covering a wide range of temperature regime. More than 2000 genotypes including over 500 commercial hybrids were used representing from 72 to 124 CRM. Planting populations and planting dates (from 3/30 to 6/9) were within the range of normal practices of local farmers. The results showed that at CRM level, CERES-Maize predicts flowering date well across different CRMs, latitudes, years and planting dates and can be used for various applications that rely on accurate flowering time prediction. The validation through this dataset also verified that the photo-thermal principle for flowering date holds steadily across genotypes and environments.