Typifying late-season weather for maize yield prediction

Abstract

Long-term weather data were used to generate typical weather records. By substituting typical weather data for actual late-season weather data it was possible to compute yield predictions using the model CERES-Maize, and compare these with predictions based on full season actual data. Yield predictions made at anthesis deviated by an average of 31,5% from those obtained using actual weather data, while predictions made 30 days later reduced the mean percentage prediction error to 1,8%.