Assessing unmanned aerial vehicle‐based imagery for breeding applications in St. Augustinegrass under drought and non‐drought conditions

Abstract

AbstractThe use of imagery collected from small unmanned aerial vehicles (UAVs) in turfgrass breeding has rapidly increased, as has the demand to develop drought‐resistant cultivars. However, prior to adopting UAVs to help guide turfgrass selection under drought stress conditions, a clear understanding of the value and predictive ability of imagery‐based turfgrass characterization is required. In St. Augustinegrass, a major warm‐season turfgrass species grown in the Southeastern United States, limited research has been published about characterizing drought stress using aerial imagery. Specifically, no efforts have compared the various vegetation indices (VIs) commonly used to evaluate vegetative health in other species and sought to identify the most useful index for phenotyping drought stress traits in St. Augustinegrass. In this study, traditional ground‐based approaches for measuring percent green cover (PGC) and normalized difference vegetation index (NDVI) were compared against their UAV‐derived counterparts as well as 13 VIs under drought and non‐drought conditions, and broad‐sense heritability (H2) was calculated. A population of 115 genotypes from a ‘‘Raleigh’’ × ‘‘Seville’’ cross were analyzed at two environmentally distinct field sites in North Carolina. At both sites, a significant relationship between ground‐based and UAV‐derived measurements for PGC and NDVI was observed before and during drought (r = 0.82 to 0.95) and suggests a clear advantage to using UAVs for phenotyping drought traits given the reduced time and labor costs compared to on‐ground efforts. Among all VIs compared, UAV‐derived NDVI (NDVI‐U) showed strong correlation with the PGC taken on the ground (r > 0.85), a similar trend over time, and a higher H2 estimate under drought conditions, suggesting that NDVI‐U has the potential to assist in the selection of St. Augustinegrass genotypes with the best phenotypic response to drought. Implementing UAV imagery‐based high‐throughput methods will allow breeders to evaluate germplasm with unbiased quantitative consistency, quickly and thoroughly, and with increased frequency—all without sacrificing the response to selection potential.