SPECIES IDENTIFICATION IN THE DIGITAL AGE

An investigation of the accuracy of species identification apps for mosses and lichens in forestry environments

Abstract

Species identification apps are increasingly used in biodiversity monitoring and citizen science, yet their reliability for cryptogams—mosses and lichens, is poorly understood. This study evaluated three apps, Artsorakel, Google Lens, and Seek by iNaturalist, on 30 species (15 mosses, 15 lichens), using 1,200 standardized field photographs. Identifications were assessed at family, genus, and species levels, with expert identifications as reference. The null hypothesis was that there would be no significant differences in identification accuracy between apps, organism groups, taxonomic levels or threat status. Alternative hypotheses were that: (1) the apps differ in performance; (2) accuracy varies between mosses and lichens; (3) accuracy is higher at family than genus or species level; (4) interactions exist between app, organism group, and taxonomic level; (5) red-listed species are identified less accurately than non-threatened ones. Statistical analysis (GLM with post hoc comparisons) confirmed significant effects for all factors except App × Organism. Artsorakel showed the highest overall accuracy (58.6%), particularly for non-threatened lichens (up to ~90% at family level), while Seek by iNaturalist consistently performed the worst. Species-level accuracy was low across all apps, especially for mosses and red-listed species. Results indicate that current apps, although helpful for education and broad surveys, are insufficiently reliable for conservation tasks such as ecological value assessments. App performance is strongly affected by organism group, taxonomic level, threat status, and data source quality. Locally tailored apps like Artsorakel offer advantages but cannot substitute expert field knowledge in biodiversity monitoring.