Improvements in the simulation of kernel number and grain yield in CERES-Wheat

Crop simulation models can be used to quantitatively evaluate alternatives ranging from agronomic management issues to climate change, thus the need for precision and accuracy in these types of evaluation. The objective of this study was to improve the simulation of kemels/m(2) (KNO), kernel weight (KWT) and grain yield in CERES-Wheat (V 3.0, CW3), while preserving the original structure of the model. The modified model incorporates four changes: (a) KNO was determined as a function of increase in spike weight from the opening of the flag leaf sheath to 50% anthesis; (b) simulation of kernel set reduction due to high temperature; (c) a new temperature function for simulating kernel fill rate was developed; (d) the amount of stem reserves was reduced by assuming that their accumulation ceases at the beginning of the grain-filling period. The modified model (CWM) was compared against published field data of the cultivars Arapahoe and Karl 92 from 31 sowings conducted over 6 years at Lincoln and Mead, Nebraska. For both cultivars, the root mean square error (RMSE) of simulated KNO went from 4312 kernels/m(2) with CW3 to 2608 kemels/m(2) with the modified model; RMSE of simulated KWT went from 2.41 mg with CW3 to 2.78 mg with the modified model; and yield RMSE went from 1241 kg ha(-1) with the CW3 to 811 kg ha(-1) with the modified model. Consideration of reductions in kernel set due to high temperature improved simulated KNO, but was cultivar dependent. CW3 simulated KWT well because KNO was overestimated. (C) 2004 Elsevier B.V. All rights reserved.