Use of linear regression and correlation matrix in the evaluation of CERES3 (Maize)

Abstract

A historical data set (soil water content, growth, phenology and yield) consisting of six cultivars and three planting dates was used to evaluate the CERES3 crop growth model. Linear regression and correlation matrix were used to identify algorithms in the model in need of calibration. Results indicated that the model simulates yield and kernel number with low accuracy under local conditions. The number of ears per plant and water stress before and during silking were identified as factors that could explain the low accuracy.