Lodi Agriculture: Leveraging NASA Earth Observations to Detect Vine Stress and Potential Viral Presence in Vineyards to Inform Management Decisions

Grapevine leafroll-associated virus complex 3 (GLRaV-3) is one of the most damaging viral grapevine diseases in the world. This virus represents a pressing concern for the Lodi Winegrape Commission (LWC), a viticultural organization based in northern California. Traditional GLRaV-3 detection methods are generally time-intensive and expensive, further complicating mitigation efforts. We partnered with the LWC, NASA Acres, and the Gold Lab at Cornell University to create and test a scalable GLRaV-3 detection model using Earth observations. We explored hyperspectral imagery from the Earth Surface Mineral Dust Source Investigation (EMIT) and multispectral imagery from the Sentinel-2 Multispectral Instrument from August and September 2022–2024. For EMIT, we ran a Principal Component Analysis (PCA) to reduce dimensionality and identify components representing maximum variance. For Sentinel-2, we calculated indices including the Normalized Difference Red Edge Index (NDRE) and Plant Senescence Reflectance Index (PSRI) to extract additional information about vegetation health. Using random forest classification models and ground-truth scouting data, we classified vineyard areas based on GLRaV-3 symptom presence. Due to training data limitations and differences in spatial resolution between sensors, we employed a binary symptomatic/unknown status classification for Sentinel-2 and a threshold-based classification (above or below 1% symptomatic rate) for EMIT. Results demonstrated that the EMIT model was more sensitive to areas of higher symptomatic vine density than the Sentinel-2 model, suggesting that hyperspectral data may be more suitable than multispectral for GLRaV-3 associated symptom mapping. This indicates that spectral resolution may be more critical than spatial resolution for this application.