Synthesizing evidence on island biodiversity monitoring: from scattered literature to structured knowledge

Islands contribute disproportionately to global biodiversity, harboring exceptional levels of endemic species. Simultaneously, they are highly vulnerable to anthropogenic pressures, with the magnitude and speed of human impacts making islands epicenters of species extinctions and species invasions. Our knowledge of changes in island biodiversity is scattered across journals, repositories, and data formats. Moreover, knowledge about island biodiversity is often diluted in global, continental, or national syntheses, limiting its inclusion in policy and decision-making. Mobilizing and integrating this literature is essential, as islands may provide early warning systems for ecological tipping points worldwide. Here, we present a user-oriented workflow for finding, screening, and structuring biodiversity literature relevant to island ecosystem monitoring. The workflow integrates established evidence-synthesis practices following Collaboration for Environmental Evidence guidelines with AI-enabled tools to improve the scalability, efficiency, and reproducibility of the review process.A comprehensive search across Web of Science, Scopus, and OpenAlex yielded over 14,000 publications. Given the number of publications, a transformer-based binary classification model was fine-tuned using a 10% subset (1,500 articles) to rank publications by their relevance. The subset is the result of a restricted search string applied at the title and abstract level on OpenAlex. The subset was manually screened by island biodiversity experts relying on an ontology describing key concepts related to islands and Essential Biodiversity Variables (EBVs), ensuring consistency throughout the screening and coding process. Title and abstract screening retained approximately 1,100 records, and full-text assessment resulted in a final list of 700 publications, from which structured information, including island location, geological type, EBV categories covered, and protocols, was extracted.All selected publications have been transformed into machine-readable formats and deposited in Biodiversity PMC via the Biodiversity Literature Repository (BLR) community on Zenodo, enabling systematic annotation and data extraction to improve transparency and replicability. We present insights into global patterns and gaps in island biodiversity monitoring, and discuss how to make island literature FAIR (Findable, Accessible, Interoperable, and Reusable), and machine-actionable, can contribute to the Disentis Roadmap vision of liberating biodiversity knowledge from publications, thereby providing evidence for conservation planning and policy on islands and beyond.