Empirical infection period models for tan spot of wheat

Infection period models for Pyrenophora tritici-repentis were evaluated for their potential use in a tan spot forecasting system. Infection periods on susceptible wheat were identified from a bioassay system that correlated t an spot incidence with crop growth stage and 24 h summaries of environmenta l data including temperature, relative humidity, wind speed, wind direction , solar radiation, precipitation and flat-plate type wetness sensor resista nce. The resulting data set was then divided for model calibration and vali dation analysis. Artificial neural networks, logistic regression, and discr iminant functions were used to classify infection periods. A neural network model had a prediction accuracy of 87% for infection periods of a validati on data set. In comparison, stepwise logistic regression correctly predicte d 69% and multivariate discriminant analysis distinguished 50% of the valid ation cases. When wetness sensor inputs were withheld from the models, both the neural network and logistic regression declined 6% in prediction accur acy. Prediction accuracy was lower when validation cases from the 1996 envi ronment were used to assess model accuracy.