A Logistic Regression Model for Predicting Risk of White Mold Incidence on Dry Bean in North Dakota

Abstract

White mold, caused by Sclerotinia sclerotiorum, is the most important disease affecting dry bean production in North Dakota. This disease currently is managed mainly through fungicides applied during the flowering stage. A disease-forecasting model was developed to help growers with their decision to apply these fungicides. The model was built using weather variables collected during eight consecutive half-month periods between 1 May and 31 August 2003 to 2005 and white mold incidence data obtained from 150 fields. The model was produced using logistic regression analysis, and includes total rainfall, average minimum temperature, and number of rainy days in the first half of June, July, and August, respectively, as predictors and explained 85% of the variability. The model was validated using an independent disease data set collected from 100 fields during the same years. The model exhibited high true positive ratio (0.79) and very high accuracy (0.91) between observed and predicted probabilities of white mold incidence. Results from this study suggest that in-season macro-weather variables could be used to predict the risk of white mold, which in-turn could help growers make better-informed decisions on whether or not to apply fungicides for white mold control.