Linking to images and AI-based identification tools—The only way for Flora projects to survive

Societal Impact Statement: Floras are comprehensive and authoritative catalogues of plants growing in an area of interest. They help people find and name plants, which is achieved by a combination of images, drawings, and text, rarely also maps. Like other catalogues (lexica, dictionaries, telephone books), Floras will not survive unless they move online and become portable, updatable, specimen-image-centred, and linked to AI-based identification of user-uploaded photos. The societal benefit of dynamic and interactive Floras will lie in their increased accessibility for non-expert users, with the accompanying possibility of helping more people get to know and protect local floras. Summary: Floras are taxonomic treatments of the plant species growing in an area such as a country, administrative unit, or island. Their purpose is to provide a means of naming the local plants, ideally as easily as possible. This ideal has long been constrained by the costs of printed Floras, technical language, and the naturally always-incomplete nature of Floras. These constraints can be reduced by moving Flora projects entirely online and linking them with photos of dried and living plants, and with AI-based plant identification, which will replace key-based identification. Starting from a review of major Flora projects, this review highlights Floras that are updating species treatments dynamically (instead of printing new versions) and allowing users to upload photos and thereby contribute to floristic discovery. Floras of the future will support morphological descriptions with plant photos and specimen images from open-access databases and will eventually update descriptions with features and locations extracted from both herbarium images and user-uploaded photos. The application of names, and complex nomenclatural situations, will continue to be debated, but ‘quality control’ issues should not prevent Floras from linking with AI-based identification platforms and natural language-processing models.