Modern apple production has moved from traditional open orchard formats to that of high-density trellised plantings. As industry has shifted, the number of trees in commercial apple orchards typically reaches 1,500 per acre, making the management of these orchard blocks on a per-tree basis impractical. The use of unmanned aerial vehicles (UAVs) in the form of portable drones offers a method for collecting image data for thousands of trees in a matter of minutes. Combining these UAVs with multispectral cameras, which collect visual data on spectra that the human eye cannot perceive, allows for the calculation of vegetation indices (VIs) such as the normalized difference vegetation index (NDVI) for individual trees. Many VIs correlate with important orchard management considerations such as tree vigor, nutrient status, and disease severity. Multiple challenges hinder the incorporation of VIs into orchard management: parsing data on a per-tree basis is challenging in high-density systems, there is no comprehensive understanding of how VIs vary across a growing season and between cultivars, and at what magnitude deviation from expected norms is indicative of a weak tree is uncertain. This experiment aimed to classify this variation in VIs by examining a large cohort of cultivars at multiple timepoints and sites. A DJI Mavic 3M equipped with a 4-channel (green, red, red edge, and near infrared) multispectral camera was used to perform flights at both commercial and research orchards in upstate New York. Orthomosaics of each site were created using Agisoft Metashape and analyzed in QGIS and R (using the FIELDimageR package). Results showed statistically significant differences in VI values between cultivars and timepoints. These differences increased in intensity later in the growing season- in September, differences in cultivar explained 40% of the observed variance in NDVI at one orchard site, compared to only 19% in July.
