Plant diseased models. Critical issues in development and use

Models which simulate the evolution of a plant disease during the season give important information to assess the seriousness of the situation. This activity precedes the choice of an appropriate action to be implemented for reducing the economical damage. Having effective models is a critical issue in modern agriculture, especially in Integrated Protection and Organic Farming, which rest on a set of plant disease management practices with low environmental impact. Considerable effort has gone in the study of models for the simulation of plant diseases evolution. Phenology models, population and epidemiological models have been developed for several, diffused diseases. Problems are still open, especially when the aim is that of including these models into decision support systems at use of producers and agronomists. Phenology and population models have to be developed, choosing the most promising techniques. Moreover, requirements such as that of providing justification to the user of the results computed by a model or making the user aware of the accuracy of the model results, become critical. In this paper we focus on models that address practical plant disease management issues and use mathematical techniques or Artificial Intelligence techniques (especially Machine Learning techniques). We describe relevant examples for each approach pointing out how they deal with critical issues such as adapting a model to different geographical area, or validating and maintaining the model on a long period