P-233 - An assessment of environmental factors influencing grape downy mildew incidence with a random forest approach

An oomycete, Plasmopara viticola, can infect both the leaves and fruit of grapevines (Vitis spp.), potentially leading up to 75% crop loss. We periodically deployed grapevine cuttings (Vitis vinifera, 'Chardonnay') grown in greenhouses to monitor grape downy mildew development in field conditions. These cuttings were placed at four to five random locations near experimental vineyards in Winchester, VA, USA, during the 2022-2024 period, and they remained in place for two to seven days. We returned the vines to greenhouses, kept them for two weeks, and then visually estimated disease incidence on ten leaves per vine. The deployments were repeated six to ten weeks per year. To explore the influence of environmental conditions on mean disease incidence, we employed a random forest machine learning approach (R ver. 4.3, package 'randomForest'). Hourly ambient weather data was obtained through OpenWeatherMap.org during deployment and seven days before the deployment date. We run it with 80% training and 20% test data with the mean incidence thresholds set to 10% to 40% in 10% increments. Preliminary results showed model accuracy around 0.80 to 0.81 and sensitivity ranging from 0.90 to 0.99. The results highlighted the effects of humidity, dew point, and temperature factors five to six days before the deployment date, wind factors four days before the date, and barometric pressure and temperature factors during the deployment period. We plan to investigate these influences further to gain deeper insights into the conditions conducive to grape downy mildew infection.