Can artificial Neural Network Systems be used for forecasting aphid flight patterns

Abstract

In Canterbury the cereal aphid (Rhopalosiphum padi) has a bimodal annual flight pattern with peaks in autumn and late spring The autumn peak coincides with the migration of aphids into autumn and winter sown wheat The timing magnitude and duration of flights vary dramatically from year to year and can have important implications for the management of this aphid in cereals Twenty years of climatic and aphid flight data were used to train a neural network to obtain a model to predict the date and size of aphid migrations in cereal crops in Canterbury New Zealand The model provided an exceptional fit to the ten years of training data but gave disappointingly poor prediction for the eight years of validation data Possible reasons for poor prediction of validation data are given and suggestions are made concerning future developments