Implementation of a regional wheat disease forecasting system

A forecasting system for tan spot, Stagonospora leaf blotch and Fusarium head blight (scab) of wheat was deployed in North Dakota and Minnesota in 1999. Tan spot and Stagonospora leaf blotch infection periods were predicted by backpropagation neural network models using environmental data from 17 automated weather stations. A generalized regression neural network predicted wetness duration as an additional input. A disease control advisory was based on cumulative infection periods and a scouting protocol between stem elongation and milk stages of growth. Burkard cyclonic flow volumetric samplers were placed near the weather stations in fields with wheat stubble on the soil surface. Air samples were collected three times a week and examined under a microscope for fungal spores of Gibberella zeae, the cause of Fusarium head blight. A heuristic forecast was based on spore level, wetness duration, and proximity to an inoculum source. Forecasts were provided to growers and consultants via the Internet and a toll-free telephone message. Challenges were faced in fund raising, physically accessing sampling sites, timely sample processing, sampling inefficiency, programming the interfaces between environmental data, prediction model, and web page, and updating the web pages. More than 1400 distinct hosts were served by the Internet web site during the summer of 1999. An improved forecasting system will be operational during the 2000 growing season.