GrapeMilDeWS: A Formally Designed Integrated Pest Management Decision Process Against Grapevine Powdery and Downy Mildews

Introduction

Since 2001, The INRA¹ santé végétale (plant health) laboratory has undertaken the design of pest management “agronomical decision rules” in viticulture. Based on observation, scientific knowledge and operational expertise, these decision rules aim to come as close to “Integrated Pest Management” (IPM) (Boller, Avilla, Gendrier, Jörg, & Malavolta, 1998; Kogan, 1998) as possible and to allow for significant reductions in the number of pesticide applications.

After a few years of experimenting with single pest designs, the team chose to focus on downy mildew (Plasmopara viticola) and powdery mildew (Erysiphe necator). In France, these two diseases represent 80% of the treatments applied on grapevine (ASK, 2000) and the practice of coupling treatments is particularly widespread with these diseases.

The first aim of the work presented here is therefore to move from a one-plot/one-pest approach to a more pragmatic approach that pairs treatments against two diseases, as is common among vine growers. Furthermore, the team considered grapevine mildews to be a good case study for the design of multiple target (i.e. integrated) pest management strategies.

Besides a demonstrative case study, a desired outcome of the research would be to transfer an operational Decision Support System (DSS) at the farm scale that would help to reduce significantly the number of fungicide applications and yet would guarantee that the production targets (both qualitative and quantitative) are reached.

Before implementing a DSS however, the phytopathologists started by designing a prescriptive crop protection decision strategy to support the scientific evaluation of the various innovations that they blended together. This strategy involves a process, beginning at bud break in spring through to harvest. The decisions taken in a given period are influenced by previous decisions as well as the phenological development of the plot and the evolution of the crop’s sensitivity to each pathogen. Designing a multi-target strategy is quite complex. In particular, response priorities between the two diseases evolve during the season, which makes decision rules difficult to write out. In order to carry out this program, computer scientists were brought into the team to help formalize, specify and evaluate the design. The model that follows is the result of this collaboration. This mathematically formal model representation had to be both understandable by phytopathologists other than its designers and suited for computer simulations. Indeed computer simulation is considered to be very helpful in the design and testing of new cropping systems (Aubry et al., 1997; Cros, Duru, Garcia, & Martin-Clouaire, 2004; Martin-Clouaire & Rellier, 2003; Sebillote, 1987). Moreover, a formalized model of the strategy should be easy to implement into a DSS once fully validated.

It must be noted that current knowledge of the vineyard pathosystem makes it difficult to simulate epidemics and damages at the plot scale and thus even more difficult to compute optimal crop protection strategies (see review of optimization techniques, in Dent, 1995). These techniques involve linear programming or dynamic programming (e.g. Feldman & Curry, 1982; Shoemaker, 1984), but these theoretical achievements are difficult to implement in practice. For instance, dynamic programming has been largely abandoned in recent years as building a realistic model for the pathosystem often requires a large number of variables which makes the problem quickly intractable.