Context and purpose of the study ‐ Random sampling is often considered to be the best protocol for fruit sampling because it is assumed to produce a sample that best represents the vineyard population. However, the time and effort in collecting and processing large random samples can be cost prohibitive. When information about known field variability is available, a spatially‐explicit sampling protocol can use that information to more efficiently sample the vineyard population. A commonly used method for mapping vineyards is normalized difference vegetation index (NDVI) which can be acquired through satellite imagery or overhead flight by plane or drone. This study seeks to improve sampling efficiency by using aerial NDVI vineyard imagery to compute optimal spatially‐explicit sampling protocols that minimize both the number of locations sampled and the time required to sample, while also minimizing potential of human errors during data collection.
Material and methods ‐ NDVI imagery acquired from LANDSAT 7 was used to map spatial variability, at a resolution of 30 by 30 meter pixels, in 24 vineyards located in California’s Central Valley. Three sampling methods, each sampling twenty whole fruit clusters, were compared to determine relative efficacy: 1) Twenty pixels selected by a random number generator (RAND20); 2) Four fixed locations, representing each quadrant, near the edge of the vineyard sampling two pixels at each location (RAND4x2), and; 3) One location, determined by a novel optimization algorithm, sampling three pixels (NDVI3). The vineyards were sampled weekly between verasion and harvest to measure Brix, titratable acidity (TA), pH, and total anthocyanins.
Results – All three sampling methods were highly correlated in pair‐wise comparisons of Brix (R= 0.86 – 0.93), TA (R= 0.93 – 0.96), pH (R= 0.96 – 0.98), and anthocyanins (0.88 – 0.90). Comparing NDVI3 and RAND4x2 to RAND20, deviation from RAND20 measurements was slightly lower in NDVI3 for Brix, TA, and pH, and slightly higher for anthocyanins. These results suggest that vineyard sampling in a single row and an optimally calculated location can produce results similar to more costly random sampling.
Authors: Jim MEYERS (1), Nick DOKOOZLIAN (2), Casey RYAN (2), Cella BIONI (2), Justine VANDEN HEUVEL (1)
(1) Horticulture Section, School of Integrative Plant Science, Cornell University, Ithaca, NY 14853 and Geneva, NY 14456
(2) Viticulture, Chemistry and Enology, E&J Gallo Winery, 600 Yosemite Blvd., Modesto, CA 95354
Email: Justine@Cornell.edu
Keywords: Grapevine, Sampling, NDVI, Optimization, Spatial variability, Efficiency