DisplayTitle

Decision Aids for Multiple-Decision Disease Management as Affected by Weather Input Errors

First and second authors: U.S. Department of Agriculture-Agricultural Research Service (USDA-ARS), Forage Seed and Cereal Research Unit, 3450 SW Campus Way, Corvallis, OR 97331; first, second, third, and fourth authors: Oregon State University Botany and Plant Pathology, Corvallis; third author: USDA-ARS Horticultural Crops Research Unit, Corvallis, OR; and fifth author: Fox Weather LLC, Fortuna, CA.

Many disease management decision support systems (DSSs) rely, exclusively or in part, on weather inputs to calculate an indicator for disease hazard. Error in the weather inputs, typically due to forecasting, interpolation, or estimation from off-site sources, may affect model calculations and management decision recommendations. The extent to which errors in weather inputs affect the quality of the final management outcome depends on a number of aspects of the disease management context, including whether management consists of a single dichotomous decision, or of a multi-decision process extending over the cropping season(s). Decision aids for multi-decision disease management typically are based on simple or complex algorithms of weather data which may be accumulated over several days or weeks. It is difficult to quantify accuracy of multi-decision DSSs due to temporally overlapping disease events, existence of more than one solution to optimizing the outcome, opportunities to take later recourse to modify earlier decisions, and the ongoing, complex decision process in which the DSS is only one component. One approach to assessing importance of weather input errors is to conduct an error analysis in which the DSS outcome from high-quality weather data is compared with that from weather data with various levels of bias and/or variance from the original data. We illustrate this analytical approach for two types of DSS, an infection risk index for hop powdery mildew and a simulation model for grass stem rust. Further exploration of analysis methods is needed to address problems associated with assessing uncertainty in multi-decision DSSs.