Eve Laroche-Pinel, Kaylah Vasquez, Madison Flasco, Monica Cooper, Marc Fuchs, and Luca Brillante*
*Department of Viticulture and Enology, California State University Fresno, 2360 E Barstow Ave, Fresno, CA, 93740 (lucabrillante@csufresno.edu)

Grapevine red blotch virus (GRBV) poses a significant threat to viticulture, leading to substantial economic losses because infected vines must be removed to prevent further spread. In this study conducted in October 2021 within a 3-ha Cabernet franc vineyard, we employed hyperspectral remote sensing with a drone-mounted camera to enhance efficiency of red blotch virus detection. The hyperspectral camera captured images in 25 spectral bands within the visible to near-infrared (VIS-NIR) domains (520 to 820 nm).

A total of 264 vines were selected randomly and sampled for PCR analysis to confirm the red blotch virus’s presence. Concurrently, images were acquired using the drone and processed through segmentation techniques to extract the vine canopy signal of the selected vines. Additionally, field experts visually inspected the vines to identify infected plants. The accuracies of both the hyperspectral images and the expert assessments were compared to the PCR results.

Six machine-learning models were trained using spectral bands as predictors. Additionally, a radiative transfer model (PROSPECT) was applied in reverse mode to predict leaf pigment concentration (anthocyanins, carotenoids, and chlorophyll) based on vine reflectance, and the output was explored as an alternative set of predictors for detecting vine infections. The overall accuracy reached 87.0% using raw spectral images and 81.4% using the PROSPECT output. The highest feature importance was attributed to the estimated anthocyanin content in leaves.

This preliminary study marks a crucial advancement toward developing an automatic system for the plant-level detection of red blotch-infected vines. Integrating hyperspectral remote sensing, PCR analysis, and machine learning techniques demonstrates promising potential for more efficient and accurate identification and management of red blotch viruses in vineyards.

Funding Support: CDFA SCBGP, CSU ARI, F3, CDFA-PDGWSS