Estimating genetic coefficients of a new rice cultivar is important when a crop model is used to test its performance. Here, we estimated genetic coefficients of two rice genotypes in the Philippines, namely, inbred PSB Rc82 and hybrid Mestizo 20, for CERES-Rice using parameter estimation methods including, GENCALC (Genotype Coefficient Calculator), GLUE (Generalized Likelihood Uncertainty Estimation) and NMCGA (Noisy Monte Carlo Genetic Algorithm). Ensembling of genetic coefficients were also employed. Model calibrations were done during the 2012 wet season using observed anthesis and maturity dates and yields as calibration data. Validation was done during the dry season of that year. Calibration results suggest that genetic coefficients estimated by different methods vary and are not consistent in predicting phenology and yield accurately. One method is better in predicting phenology, while another, for yield. However, arithmetic averaging of genetic coefficients, and weighted averaging based on parameter estimation methods’ performances worked well. In this study, arithmetic averaging of model parameters during calibration produced the best predictions of phenology and yield, in both rice genotypes, and this performance persisted during validation.

当使用作物模型测试其性能时，估计新水稻品种的遗传系数很重要。在这里，我们使用包括 GENCALC（基因型系数计算器）、GLUE（广义似然不确定性估计）和 NMCGA 在内的参数估计方法估计了菲律宾两种水稻基因型的遗传系数，即自交系 PSB Rc82 和杂交 Mestizo 20，用于 CERES-Rice （嘈杂的蒙特卡洛遗传算法）。还使用了遗传系数的集合。模型校准是在 2012 年雨季使用观察到的开花期和成熟期以及产量作为校准数据进行的。验证是在当年的旱季进行的。校准结果表明，不同方法估计的遗传系数各不相同，在准确预测物候和产量方面并不一致。一种方法可以更好地预测物候，而另一种方法可以更好地预测产量。然而，遗传系数的算术平均和基于参数估计方法的加权平均效果很好。在这项研究中，校准过程中模型参数的算术平均产生了对两种水稻基因型的物候和产量的最佳预测，并且这种性能在验证过程中持续存在。