Logistic Regression Modeling of Dollar Spot Epidemics Using Weather Variables as Inputs.

Dollar spot, caused by Sclerotinia homoeocarpa, is the most damaging disease of cool-season turfgrasses throughout the United States.  A reliable dollar spot prediction model would be useful for making management decisions for high-value turfgrasses.  Logistic regression was used to develop a model that input weather variables to predict probability of the occurrence of dollar spot on creeping bentgrass putting greens and fairways at sites in Oklahoma and Wisconsin.  Numbers of disease foci were determined daily in plots receiving no fungicide or treated with fungicide.  Various on-site weather variables were recorded hourly.  Weather data were transformed to 2-, 3-, 4-, and 5-day moving averages.  Weather data and class variables (season and fungicide application) were used as independent variables and disease data as dependent variables in logistic regression analysis to identify best fitting models.  Models using 5-day moving averages were better than models using other moving averages.  Relative humidity was the only highly significant (P=0.0006) weather variable.  The best models also included season and fungicide application class variables.  Temperature variables were not significant (P=0.60), but minimum thresholds for disease symptom appearance were set at 14 °C based on field and controlled environment chamber studies.  The temperature threshold was used to determine when to implement the predictive model.  Model validation was initiated in 2009 using a predicted probability, action threshold of 10%.  Fungicide sprays were applied when disease reached the threshold.  Initial validation experiments indicate that the model satisfactorily predicts onset and increase of dollar spot disease.