Simulation of within field variability of corn yield with Ceres-Maize model

Abstract

The CERES-Maize model was used to estimate the spatial variability in corn (Zea mays L.) yield for 1995 and 1996 using data measured on soil profiles located on a 30.5 m grid within a 3.9 ha field in Michigan. The model was calibrated for one grid profile for the 1995 and then used to simulate corn yield for all grid points for the 2 yrs. For the calibration for 1995, the model predicted corn yield within 2%. For 1995, the model predicted yield variability very well (r(2) = 0.85), producing similar yield maps with differences generally within +/- 300 kg ha(-1). For 1996, the model predicted low grain yields (1167 kg ha(-1)) compared with measured (8928 kg ha(-1)) because the model does not account for horizontal water movement within the landscape or water contributions from a water table. Under nonlimiting water conditions, the model performed well (average of 8717 vs. 8948 kg ha(-1)) but under-estimated the measured yield variability.