Predicting Key Grapevine Phenological Stages in Eastern Canada: Insights From 15 Years of Field Observational Data

The study aimed to develop predictive models for key grapevine phenological stages in northern climates using 15 years of field observations. The research focused on major cultivars grown in eastern Canada (Frontenac, Marquette, Seyval, and Vidal), where viticulture is expanding. Key phenological stages were grouped into six categories, and their distribution as a function of weather‐based explanatory variables was modeled using multinomial logistic regression. Temperature‐ and humidity‐related variables were evaluated based on the percent correct classification of the six categories. There were the categories: 1 (BBCH 5–9); 2 (BBCH 51–53); 3 (BBCH 57–61); 4 (BBCH 71–76); 5 (BBCH 81–83); and 6 (BBCH 88–91). Overall, the best explanatory variables were the degree days in either Tb = 0°C or Tb = 10°C. Adding humidity‐related variables did not significantly improve model accuracy. The final models resulted in a percent correct classification of all six phenological stage categories of 90.1%, 91.4%, 89.2%, 88.2%, 87.4%, and 88.4% for the Frontenac, Marquette, Seyval, and Vidal cultivars; other cultivars; and all cultivars, respectively. The resulting models provide probability curves for each phenological stage category, allowing for prediction of stage onset and duration. We explored a different approach to modeling phenological stages in our study. Our goal was to develop single models that can predict when multiple stages are reached. These models are intended for integration into decision support tools for managing grapevine diseases, optimizing canopy management, and harvest timing.