S05 - Session O4 - A view from above: the use of drones and image-based phenotyping in garden rose breeding

Authors: Leen Leus *, Irene Borra-Serano, Katrijn Van Laere, Peter Lootens

The huge increase of image capture and open-source tools for spatial data analysis, together with the use of drones in the field, offer a wide range of possibilities for high-throughput field phenotyping (HTFP). Therefore drones can be a useful tool for crop inventory, crop monitoring, and crop management in nurseries. Despite its importance in breeding, high-throughput phenotyping is still underused in ornamental plant breeding. Here, we evaluate the possibilities of image-based phenotyping by UAV (Unmanned Aerial Vehicle) for the selection of candidate cultivars in garden rose ( Rosa ) breeding. In garden rose breeding the most important selection traits relate to flowering and disease resistance. Ideal roses flower all summer long with an abundance of flowers, have a uniform plant architecture and are disease resistant e.g. to black spot ( Diplocarpon rosae ). These parameters were assessed in the rose selection field of a rose breeding program. The field comprised 71 genotypes with 7 to 51 plants per genotype. From June until October 2019 manual on-ground measurements and scores for plant architecture, flowering and disease resistance were correlated to data derived from RGB images obtained by UAV flights. Image analysis comprised: (1) canopy height modelling to estimate plant height and uniformity in plant architecture; (2) the quantification of flowering using the Woebbecke vegetation index; (3) the evaluation of disease resistance by the assessment of leaf drop caused by black spot through the excess green (ExG) vegetation index. Correlations between on-ground measurements and image analysis will be discussed as well as relevance of the data compared to the breeders' selection. Also best practices and recommendations to set up UAV experiments for roses and other ornamental crops in the field will be given.