A Weather-Driven Model for Predicting Infections of Grapevines by Sporangia of Plasmopara viticola

Abstract

A mechanistic model was developed to predict secondary infections of Plasmopara viticola and their severity as influenced by environmental conditions; the model incorporates the processes of sporangia production and survival on downy mildew (DM) lesions, dispersal and deposition, and infection. The model was evaluated against observed data (collected in a 3-year vineyard) for its accuracy to predict periods with no sporangia (i.e., for negative prognosis) or with peaks of sporangia, so that growers can identify periods with no/low risk or high risk. The model increased the probability to correctly predict no sporangia [P(P−O−) = 0.67] by two times compared to the prior probability, with fewer than 3% of the total sporangia found in the vineyard being sampled when not predicted by the model. The model also correctly predicted peaks of sporangia, with only 1 of 40 peaks unpredicted. When evaluated for the negative prognosis of infection periods, the model showed a posterior probability for infection not to occur when not predicted P(P−O−) = 0.87 with only 9 of 108 real infections not predicted; these unpredicted infections were mild, accounting for only 4.4% of the total DM lesions observed in the vineyard. In conclusion, the model was able to identify periods in which the DM risk was nil or very low. It may, therefore, help growers avoid fungicide sprays when not needed and lengthen the interval between two sprays, i.e., it will help growers move from calendar-based to risk-based fungicide schedules for the control of P. viticola in vineyards.