2026/05/06 by Iyesha Madushani Wanigasinghe, Diane L. Haughland, Lysandra A. Pyle +2 · 2 voices
Environmental Science · Psychology · #Animal and Plant Science Education #Rangeland and Wildlife Management #Species Distribution and Climate Change
paper · doi:10.1093/aobpla/plag020
openalex publication_date 2026/05/06 · openalex created_date 2026/05/09 · openalex updated_date 2026/07/30
Abstract While mobile apps are becoming more accurate at identifying vascular plants, it is unclear whether accuracy is maintained when doing plot-based surveys, where image acquisition is limited to a small pool of individuals that often lack ideal identification features. We evaluated two free plant identification apps, Flora Incognita and iNaturalist, using 5291 field images of 119 species from 54 plots of 0.8 m2, from graminoid-dominated grassland and forested sites. We examined the top-1 identification (ID) accuracy as a function of botanical (growth form and reproductive state) and photographic (image quality and background) parameters. We used score fusion to test whether combining multiple images of the same individual (perspective combination) improved ID accuracy. Top-1 ID accuracy was consistently higher in Flora Incognita (79.2% overall) as compared to 66.5% in iNaturalist. No groups reached the reference expert accuracy of plot-based surveys (90%–95%). Images featuring reproductive structures significantly improved ID accuracy, particularly for graminoids (overall +8.9% in Flora Incognita, +17.1% in iNaturalist), and high-quality images increased accuracy significantly for graminoids and herbs in iNaturalist. Combining multiple images improved the overall ID accuracy of vascular plants in iNaturalist by 6% and 1.7% in Flora Incognita, with significant improvement for herbs and shrubs in iNaturalist. We provide recommendations for ID success in the field for app users as well as app developers to optimize image acquisition for species identification in surveying and monitoring efforts.