Whole-crop Silage Corn Yield Prediction Using Neural Network

Abstract

Although whole-crop silage corn is one of the major forage crops in Japan, the yield is presently stagnating. To improve the yield, the potential yield under each cultivation condition should be predicted and the cultivation methods improved. In this study, we developed a neural network model that predicts the dry matter yield in the yellow-ripe stage of three corn varieties, using the experimental data of each variety and meteorological data obtained from the automated meteorological data acquisition system observation points near each experimental site. The correlation coefficient between the estimated and measured dry matter yield using this model was 0.556 and the root mean square error was 460.2 kg/10 a. The developed model was also used to create a “Takanestar” yield prediction map covering the Tohoku to Kyushu regions, and the regional differences were compared.