Improving the ceres-maize model ability to simulate water deficit impact on maize production and yield components
Crop models are potential tools for designing water-efficient strategies and should be tested for accurate prediction of water deficit effects on production. The objective of this research was to test and improve the CERES-Maize model (CERES-4.0) for ability to predict accurately maize biomass and grain yield components under water-limiting conditions in an environment where the model had previously given good predictions under irrigated conditions. Under a water-limited environment in northwest Spain, CERES-4.0 failed to simulate sufficiently high growth and yield; thus we evaluated aspects of model believed responsible for the poor performance. The model was tested with two evapotranspiration options [Priestley-Taylor (PT) and Penman–Monteith reference method (PFAO56, FAO no. 56 manual)] and with two values for the coefficient (KEP) that partitions evapotranspiration (ET) between crop transpiration and soil evaporation (default: KEP = 0.685; alternate: KEP = 0.500). The PT option with KEP = 0.685 underpredicted grain yield and biomass due to too early and too severe simulated water extraction. Predictions of biomass and grain yield with both PT and PFAO56 were improved when a KEP of 0.500 was used instead of the default 0.685. The PFAO56, the less water demanding of the tested crop reference (ETo) equations, gave the best predictions. In addition, a systematic underprediction of grain number and grain size with the default model in response to water deficit was observed, regardless of ET option or KEP value. Model predictions of the latter two variables was improved by replacing the default CERES-4.0 function that computes seed number per plant with the function of Edmeades and Daynard (1979).