Link to home

Identifying Risk Factors for European Stone Fruit Yellows from a Survey

August 2006 , Volume 96 , Number  8
Pages  890 - 899

Gaël Thébaud , Nicolas Sauvion , Joël Chadœuf , Arnaud Dufils , and Gérard Labonne

First, second, and fifth authors: Institut National de la Recherche Agronomique (INRA), UMR BGPI, CIRAD TA 41/K, Campus International de Baillarguet, 34398 Montpellier Cedex 5, France; first and third authors: INRA, Unité de Biométrie, Domaine Saint-Paul, Site Agroparc, 84914 Avignon Cedex 9, France; and fourth author: Station Expérimentale La Pugére, Chemin de la Barque, 13370 Mallemort, France

Go to article:
Accepted for publication 20 March 2006.

European stone fruit yellows (ESFY) is becoming a major economic problem for Prunus growers in Europe. The causal agent (“Candidatus Phytoplasma prunorum”) and its vector (Cacopsylla pruni) have been identified, but the present knowledge of the risk factors for this disease relies, at best, on specific experiments. To assess the relative significance of several factors correlated with ESFY incidence in the field, an exhaustive survey was performed on apricot and Japanese plum orchards in the Crau plain (France). After a preliminary multivariate exploration of the data, we used a logistic regression model to analyze and predict the cumulative number of diseased trees on the basis of a set of quantitative (age, planting density, and area of the orchard) and categorical variables (species, cultivar, and rootstock). Because of the nature of the data, we used an overdispersed binomial model and we developed a parametric bootstrap procedure based on the beta-binomial distribution to obtain confidence intervals. Our results indicated that the age, species, and cultivar of the scion were the major factors explaining the observed number of diseased trees. The planting density and the rootstocks used in the zone under study were less significant, and the area of the orchard had no effect. The residuals of the model showed that some explanatory variables had not been taken into account, because part of the remaining variability could be explained by a grower effect. The spatial distribution of the residuals suggested that one of the reasons for this grower effect was the correlation between orchards closer than 100 m, possibly caused by the flight behavior of infectious vectors.

Additional keywords: generalized linear model, Monte Carlo, over-dispersion.

The American Phytopathological Society, 2006