Understanding the impact of co-occurring diseases on aerial hyperspectral detection: A UAS case study with grapevine leafroll virus and downy mildew

Grapevine disease prevention and management challenges sustainable viticulture globally. Comprehensive vineyard monitoring is time- and labor-intensive, especially in regions where multiple pathogens regularly cause co-occurring infections. We used a visible-to-near-infrared (VNIR) imaging spectrometer mounted on an uncrewed aerial system (UAS) to monitor two New York State vineyards affected by grapevine downy mildew (GDM) and Grapevine Leafroll-associated Virus 3 (GLRaV-3). Four UAS datasets were collected during the 2024 growing season, accompanied by ground scouting, physiological measurements, and PCR diagnostics for a subset of vines. Partial least squares regression (PLSR) models detected GLRaV-3 symptoms with symptomatic-class F1 scores of 0.69–0.72 in August, and models differentiating GDM-only from GLRaV-3-only symptoms achieved accuracies of 0.55–0.82 after down-sampling. To test whether co-occurring GDM impacted accurate GLRaV-3 detection, we compared PLSR recall for GLRaV-3-only vines with recall for vines showing both GLRaV-3 and GDM symptoms. Across three datasets, recall differences (co-occurring minus GLRaV-3-only) were +0.4, +9.0, and −4.6 percentage points and all 95% confidence intervals included zero, suggesting that co-occurring GDM did not consistently reduce GLRaV-3 symptom classification accuracy. This study provides a direct, field-based evaluation of aerial imaging spectroscopy to detect co-occurring grapevine diseases in commercial vineyards. Our results suggest that UAS VNIR data can be informative amidst the biological complexity of ongoing operational agricultural systems. Early-season disease monitoring remains challenging, and larger representative studies are needed to better quantify co-occurrence effects on disease identification via aerial imaging spectroscopy.