Skip to main content
CORESTA Meeting, Agronomy/Phytopathology, Montreux,1997, PT12

Fungicide, fumigant, and nematicide advisor: an expert system on the world wide web

MELTON T.A.
North Carolina State University, Dept. of Plant Pathology, Raleigh, NC, USA
Flue-cured growers in North Carolina lose at least 3-4% annually to the soilborne diseases black shank ( Phytophthora parasitica ), Granville wilt ( Pseudomonas solanacearum ), and root knot ( Meloidogyne spp.). These losses are often concentrated in areas where losses range from 25-100%. As a result, 91% of the acreage was chemically treated for one or more of these diseases in 1996. Growers continue to request individual help in selecting chemicals and rates because several chemical choices are available for each disease, some chemicals control more than one disease, chemicals vary in their efficacy, and more than one of these diseases is often present. The objective was to develop an expert system that growers, consultants, chemical dealers and county agents could readily access to help growers with their selection of fungicides, fumigants, and nematicides. An expert system is a type of artificial intelligence program that emulates the interaction the user may have with a human expert. This expert system is the first known in agriculture to operate on the world wide web using a real-time run engine (WREN, Exsys, Albuquerque, NM, USA). It was developed using rule-based developmental software (Rulebook Plus and Exysys Professional, both by Exsys) and uses 80 rules to reach one or more of 24 possible recommendations. The user is asked three to nine questions, depending on the users’ replies. For example, if the program determines during questioning that a fumigant will be recommended for Granville wilt, then no questions regarding root knot or other nematodes will be asked because the fumigant also will control root knot. Fungicide, Fumigant, and Nematicide Advisor uses backward chaining to derive its recommendations.