Plant diversity estimates of Mediterranean islands differ among biodiversity databases

Abstract

Large-scale biodiversity databases encompass three main types of data for plants, namely single species point occurrences, cooccurrences in vegetation plots, and checklists for specific areas. Evidence shows that such data types exhibit specific biases, reporting different species assemblages at local scales. We used the Mediterranean Basin, a global biodiversity hotspot with more than 2200 islands larger than 0.01 km2, to compare island vascular plant diversity patterns emerging from occurrence data (Global Biodiversity Information Facility; GBIF), vegetation plots (European Vegetation Archive; EVA), and species checklists (Global Inventory of Flora and Traits; GIFT). We aggregated plant data at the island level and compared geographic coverage, inventory completeness, and taxonomic coverage among these data sources. The combined databases accounted for 8702 species distributed on 790 islands (35.6% of the target islands). Data availability increased from small (26.8%) over medium (75.7%) to large islands (100.0%). Spatial coverage of databases on a 30 × 30 km grid was high for GBIF (52.8%) and EVA (45.4%), and low for GIFT (21.7%). GIFT provided higher native and alien species richness values for most of the islands, whereas GBIF and EVA consistently missed a considerable fraction of the expected species richness. Taking GIFT as reference, GBIF, and to a lesser extent EVA, showed a positive bias towards perennial species and an underrepresentation of annuals. Despite their lower taxonomic coverage, GBIF and EVA data can complement our knowledge on Mediterranean islands’ plant diversity, providing data for islands lacking plant inventories. Moreover, GBIF and EVA’s large datasets can be used for investigating other levels of ecological organisation and modelling single species (GBIF) or population (EVA) trends over space and time. Finally, our results advocate for a coordinated effort to fill the knowledge gaps through data collection and digitisation, possibly integrating data collected by experts by means of citizen science initiatives.