Abstract
Plant pathogens pose a persistent threat to grape production, causing significant economic losses if disease management strategies are not carefully planned and implemented. Simulation models are one approach to address this challenge because they provide short-term and field-scale disease prediction by incorporating the biological mechanisms of the disease process and the different phenological stages of the vines. In this study, we developed a Bayesian model to predict the probability of Plasmopara viticola infection in grapevines, considering various disease management approaches. To aid decision-making, we introduced a multi-attribute utility function that incorporated a sustainability index for each strategy. The data used in this study were derived from trials conducted during the production years 2018-2020, involving the application of five disease management strategies: conventional Integrated Pest Management (IPM), conventional organic, IPM with substantial fungicide reduction combined with host-defense inducing biostimulants, organic management with biostimulants, and the use of biostimulants only. Two scenarios were considered, one with medium pathogen pressure (Average) and another with high pathogen pressure (Severe). The results indicated that when sustainability indexes were not considered, the conventional IPM strategy provided the most effective disease management in the Average scenario. However, when sustainability indexes were included, the utility values of conventional strategies approached those of reduced fungicide strategies due to their lower environmental impact. In the Severe scenario, the application of biostimulants alone emerged as the most effective strategy. These results suggest that in situations of high disease pressure, the use of conventional strategies effectively combats the disease but at the expense of a greater environmental impact. In contrast to mechanistic-deterministic approaches recently published in the literature, the proposed Bayesian model takes into account the main sources of heterogeneity through the two group-level effects, providing accurate predictions, although precise estimates of random effects may require larger samples than usual. Moreover, the proposed Bayesian model assists the agronomist in selecting the most effective crop protection strategy while accounting for induced environmental side effects through customizable utility functions.