A comparison of multiple calibration and ensembling methods for estimating genetic coefficients of CERES-Rice to simulate phenology and yields

Estimating genetic coefficients of a new rice cultivar is important when crop model used to test its performance. Here, we estimated two 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) NMCGA (Noisy Monte Carlo Genetic Algorithm). Ensembling were also employed. Model calibrations done during 2012 wet season observed anthesis maturity dates yields as calibration data. Validation was dry that year. Calibration results suggest by different vary are not consistent predicting phenology yield accurately. One method better phenology, while another, yield. However, arithmetic averaging coefficients, weighted based on methods’ performances worked well. In this study, parameters produced best predictions yield, both genotypes, performance persisted validation.