Towards Scalable Grapevine Leafroll Virus Detection in Cabernet Sauvignon Winegrapes with Airborne Imaging Spectroscopy
Abstract
Viral disease, including that caused by Grapevine Leafroll Virus Complex 3 (GLRV), causes $3 billion in damages and losses to the US grape and wine industry annually. GLRV has a long latent period (~12mo) in which vines are infectious but do not yet display visible symptoms, making it an ideal model pathosystem to evaluate both symptomatic (Sy) and asymptomatic (aSy) imaging spectroscopy-based disease detection. Imaging spectroscopy is able to detect plant disease based on the physiological and chemical changes pathogen infection induces in plants both pre- and post-symptomatically. The ability to detect disease during the latent period at scale would greatly reduce management costs, as current detection methods are entirely ground-based, labor-intensive, and expensive. Here, we use airborne imaging spectroscopy data collected in September 2020 by the NASA Airborne Visible/Infrared Imaging Spectrometer Next Generation (AVIRIS-NG) to detect GLRV in Cabernet Sauvignon grapevines in Lodi, CA. In September 2020 and 2021, industry partners scouted 835 acres of Aglianico, Cabernet Sauvignon, and Petite Sirah grapes for visible disease symptoms, and a subset was collected for confirmation molecular testing. We combined random forest with SMOTE oversampling to train a spectral model able to distinguish between non-infected and GLRV-infected grapevines. We observed robust spectral differences that allowed for differentiation between non-infected, symptomatic, and asymptomatic infected vines and combinations therein at native, 3m, and 5m resolution. Our best performing model is able to differentiate non-infected from both aSy and Sy infected grapevine at 3m resolution with 85% accuracy (0.71 kappa). At native resolution, classification accuracy is 82% (0.62 kappa). We hypothesize that these spectral differences that underlie our ability to detect disease are linked to overall plant physiology and health changes, as visible foliar symptoms were restricted to the lower canopy. The next steps for this work are to assess scalability towards spaceborne resolution (30m) and the use case potential for NASA's forthcoming hyperspectral satellite, Surface Biology and Geology.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2022
- Bibcode:
- 2022AGUFM.B22D1482R