Empirical versus mechanistic models for primary infections of Plasmopara viticola

Plasmopara viticola, the causal agent of grapevine downy mildew, requires fungicide treatments to avoid severe yield losses, so epidemiological models have been elaborated to better manage fungicide scheduling. Primary oosporic inoculum plays a key role in epidemic development, and some of these models have been elaborated following an empirical approach to define quantitative relationships between occurrences of primary infections and weather factors influencing them. Recently, a mechanistic dynamic model was elaborated which accounts for the biological effects of weather on the different stages of the primary infection chain, from the progressive breaking of dormancy in the overwintering oospore population to infection establishment. In this work, three widely used models (3-10 rule, EPI and DMCast) were compared with this model (named UCSC model) on the basis of their theoretical approaches and their accuracy was tested against real data collected over a 9-year period in Sardinia (Italy). The two empirical models, 3-10 and EPI, seem to be too over-simplified to simulate correctly a complex biological phenomenon such as the sexual stage of P. viticola they lack consideration of some key stages of the infection chain and simultaneously simulate various biological events, each of which are influenced by specific weather conditions. EPI, in particular, needs specific calibrations and ongoing processes of adaptation that make its use in disease warning difficult. DMCast overcomes these problems because of its mechanistic approach; nevertheless, the empirical method used to model oospore maturation requires validations and, as in the present case, modifications before practical use. The UCSC model, due to its fully mechanistic approach, does not present this problem and provides accurate dynamic simulations of the sexual stage of P. viticola, with a high degree of detail and requiring neither calibrations nor corrections.