Quantitative relationships between Erisyphe necator airborne inoculum density above grape canopy and weather variables.
Carisse, O., Van der Heyden, H., and Morissette-Thomas, V. (2015). "Quantitative relationships between Erisyphe necator airborne inoculum density above grape canopy and weather variables.", Acta Horticulturae (ISHS), 1068, pp. 179-188.
Production of airborne conidia is a major component of grape powdery mildew (GPM) epidemics. Conidia of Erisyphe necator are produced on lesions induced by ascospores, from overwintered cleistothecia or conidia from previous infections. A practical approach to using aerial conidium concentration (ACC), as a risk factor for GPM, would be to estimate inoculum density based on the relationship between weather variables and ACC. In an attempt to describe the relationships between weather variables and airborne inoculum of E. necator, within a grape canopy, airborne conidia were monitored continuously during the growing season, in a vineyard planted with susceptible cultivars, from 2000 to 2002. Airborne inoculum concentration was quantified as the daily number of conidia per m3 of air. A total of 54 weather variables were generated from measurements of ambient weather data. Fundamentally, weather variables are highly correlated, which causes multicollinearity, and the quantity of airborne inoculum, at any given time, is influenced by weather conditions, on the current day and for some period of time prior to the current day. Daily ACC and the corresponding weather variables are time series, characterized by temporal autocorrelation; hence, data were analysed using time series. Only few weather variables allowed prediction of ACC with a lag of 4 to 7 days: daily mean temperature, daily hours of temperature between 25 and 30°C, hours of temperature below 20°C, mean temperature when relative humidity is greater than 85%, mean relative humidity when temperature is below 20°C, and mean vapour pressure deficit when temperature is below 20°C. Polynomial distributed lag regression models, based on 3 of these variables, predicted 65% of the variation in ACC.