Accurate leaf wetness estimation for improved grapevine disease prediction

Overview

This research initiative addresses a critical gap in precision agriculture: the lack of peer-reviewed leaf wetness estimation methods calibrated for Swiss grapevine conditions. Accurate leaf wetness data is essential for reliable plant disease prediction, yet most European grapevine farms currently rely on the default NHRH greater than 90% method without localized validation. The project aims to develop, calibrate, and compare multiple scientific methods, including NHRH, Dew Point Depression (DPD), Classification and Regression Tree (CART) analysis, and AI-based models, to identify the most accurate approach for the region. The outcome will directly support disease forecasting algorithms, helping farmers and cooperatives make informed decisions on pesticide application, crop protection, and resource management.

Technical specifications

Research approach:

• Comparison of four empirical methods for leaf wetness estimation: NHRH greater than 90%, DPD, CART analysis, and AI-based modeling
• Deployment of leaf sensors on farms to capture ground-truth leaf wetness data for validation
• Integration of weather forecast data and weather station measurements to derive the input parameters required by each method
• CART analysis extends DPD by incorporating wind speed alongside relative humidity and dew point depression
• AI-based modeling explored as a novel approach for capturing complex interactions among environmental variables
• Collaboration with fenaco, farm consultants, and farmers to cross-reference disease occurrence, severity, and pesticide application records

Key benefits for agriculture:

• More accurate disease prediction compared to the current default NHRH greater than 90% approach
• Reduced unnecessary pesticide applications through evidence-based decision-making
• Foundation for scientific precision farming practices tailored to local Swiss conditions
• Scalable methodology applicable to other crops and regions beyond grapevines

Technology readiness level

The project is at an early-to-mid research stage. While the underlying methods (NHRH, DPD, CART, AI) have shown promise in studies from the USA and New Zealand, none have been peer-reviewed or validated for Swiss, German, or French grapevine conditions. The organization has over 150 sensors installed in the field, with only a limited number currently being leaf sensors. The next phase involves installing additional leaf sensors, collecting parallel weather and forecast data, developing and calibrating each method, and benchmarking outputs against sensor-derived ground truth. The most accurate method will then be integrated into the existing disease prediction algorithm. Active partnerships with fenaco and farm consultants are already in place to support field validation and real-world deployment.