Degree-day models for 43 pests of ornamental plants: Comparing the accuracy of phenological predictions based on a generalized model, optimized models, and calendar days

Growing Degree-Day models are valuable tools for predicting pest activity and accurately timing pest management tactics. Though helpful, in practice there are logistical challenges associated with the use of degree-day models. Degree-day models are typically optimized for a particular pest species by evaluating combinations of starting dates and base temperatures to identify the one with the greatest predictive power based on minimizing the mean square error. However, when confronted with a large complex of pests, such as the case in ornamental landscapes and nurseries where biodiversity is high, the use of different models for each pest is not logistically feasible. Consequently, a generalized model based on a single starting date and base temperature is often used, which presumably compromises predictive accuracy in favor of computational simplicity. For example, the Ohio State University Degree-Day and Phenology Website (http://www.oardc.ohio-state.edu/gdd) employs a model that uses January 1 and 50°F to predict the phenology of nearly 50 key arthropod pests of woody ornamental plants. The objective of this study was to quantify the magnitude of error in the generalized model relative to optimized models developed individually for each of the key pests, to determine the degree to which generalized models have utility for timing pest management decisions. A generalized and optimized model was developed for 43 species based on 5 years of phenological and degree-day data. In the sixth year, the accuracy of phenological predictions of the generalized model was compared to predictions based on optimized models and calendar days.