Use of Normalized Difference Vegetation Index Images to Optimize Vineyard Sampling Protocols

Abstract

Applying traditional statistical models to vineyards with known field variability can lead to inefficient sampling. A previously developed sampling optimization model, based on manually collected canopy data, was adapted to analyze aerial normalized difference vegetation index (NDVI) images from two Washington State Riesling vineyards for the purposes of quantifying vineyard spatial structure and computing optimal vineyard sampling protocols. A heuristic optimization algorithm was used to determine the most efficient sampling protocols needed to accurately capture canopy variability as expressed by the NDVI images, resulting in sample size reductions up to 69% and reduced distance traveled between sampling locations by over 90% compared to random sampling.