Application of CERES-Maize for within-season prediction of rainfed corn yields in Delaware, USA

Abstract

Listen

Translate article

The objective of this study was to determine whether the CERES-Maize crop simulation model can be used to predict rainfed corn yield up to 4 months prior to harvest. Required weather data were provided by combining the observed weather conditions up until the time of the forecast with predictions of future weather conditions. Yield forecasts were issued four times (June 1, July 1, August 1, and September 1) during each growing season (2001, 2002, and 2003). Thirty future weather scenarios were generated for each forecast date based on historical weather conditions and these forecasts were validated using the full season simulations. Not surprisingly, forecasts issued later in the growing season were more accurate than forecasts issued earlier because they incorporate more of the actual weather conditions. It appears that both the June 1 and July 1 forecasts would be of little value to decision makers because they contain too much uncertainty. However, the August 1 yield forecasts were more promising, the mean absolute error (MAE) ranged from 471 to 2407 kg ha −1 and the mean percent error (MPE) ranged from 4.8 to 46.6% and all of the September yield forecasts were extremely accurate. The September forecasts were very accurate (MAE ranged from 29 to 135 kg ha −1 and the MPE ranged from 0.5 to 1.3%) because the crop has often reached maturity by September 1. CERES-Maize would be useful to agencies that require accurate yield estimates prior to harvest. If reliable predictions of future weather conditions are available, CERES-Maize can be employed to accurately forecast yield months before harvest.