Estimation of potential yield of popular cultivars using calibrated CERES-rice model in major rice growing locations in the Philippines

This study evaluated the performance of two calibrated CERES-Rice Models in simulating the effect of varying nitrogen (N) levels on the yield of popular irrigated rice cultivars at CMU, CLSU and UPLB during wet season 2014. The experimental variables were four N treatments (90,135 and 180 kg N ha-1) and three popular cultivars (PSB Rc18, NSIC Rc222 and Mestizo 19). Potential yield of the popular cultivars were estimated using the CERES-Rice Model of Decision Support System for Agrotechnology Transfer (DSSAT). Other location-specific cultivars namely: NSIC Rc238, NSIC Rc302, PSB Rc82, NSIC Rc216 and NSIC Rc298 were also estimated. Genetic coefficients were determined using DSSAT- Generalized Likelihood Uncertainty Estimation (GLUE) using 2014 experimental datasets from different sites. The calibrated models used the weather variables from the nearby weather stations at CMU, CLSU and UPLB. The DSSAT Model was validated using the generated genetic coefficients of the evaluated rice cultivars using other test experimental locations with the same season and same year. Highly acceptable nRMSEs on phenology, aboveground biomass and LAI were obtained, except for grain yield due to pest occurrence that was not simulated by the model in test cultivars grown at CMU. For UPLB and CLSU, cultivar models had moderately acceptable nRMSE in almost all parameters, except for Mestizo 19 (NSIC Rc202H) in which inconsistencies were noted on aboveground biomass, LAI and phenology. The calibrated CERES-Rice Model has predicted the phenology and aboveground biomass at anthesis and maturity, and LAI with reasonable accuracy for most of the cultivars, except for Mestizo 19. Model prediction of grain yield was better in UPLB and CLSU grown cultivars under high N applications (calibration at 135 kg N ha-1), but not with CMU due to pest infestations. Sensitivity analysis across locations revealed that PSB Rc18 has the best fitness relative to NSIC Rc222 and Mestizo 19. Further evaluation of the robustness of the generated models has to be undertaken to lessen the significant variations in phenology and grain yield of some cultivars especially the hybrid cultivars with poor predictability, through generation of more data from wet and dry seasons in various rice growing seasons.