Predictive Modeling of Grapevine Red Blotch Disease Using Multi-Temporal Remote Sensing and Spatial Epidemiology

Keywords: Red Blotch Disease, Remote Sensing, Multi-Agent System, Epidemiology, Viticulture

Abstract: Grapevine red blotch virus (GRBV) causes significant economic losses in viticulture, necessitating early detection and prediction to mitigate its spread. This study develops a predictive model for 2024 GRBV incidence using multi-temporal remote sensing and spatial epidemiological data collected prior to August 2024. We integrate hyperspectral imaging, spatial autocorrelation metrics, and host susceptibility factors within an automated machine learning framework. Our approach employs iterative feature engineering and addresses class imbalance, achieving a final F1-score of 0.97. Results demonstrate the critical importance of historical infection patterns, neighborhood effects, and vegetation health metrics, aligning with vector-mediated dispersal dynamics. The model highlights both the promise and limitations of remote sensing for pre-symptomatic detection, particularly its reliance on prior-year data. This work contributes an operational, data-driven framework for GRBV forecasting, with implications for precision viticulture and broader plant disease management. Future efforts should incorporate vector population dynamics and validate the approach across diverse environments.

Submission Number: 239