Citizen Science Contributions to Biodiversity Databases

Abstract

Citizen science biodiversity databases -- most prominently iNaturalist, eBird, and the Global Biodiversity Information Facility (GBIF) -- have undergone explosive growth in the 2010s-2020s, accumulating over 3 billion occurrence records from millions of observers worldwide and transforming the geographic and temporal scale at which biodiversity can be monitored. Yet the scientific utility of these massive but opportunistically collected datasets is fundamentally constrained by spatial bias (records concentrated near population centres and accessible habitats), taxonomic bias (overrepresentation of charismatic vertebrates relative to invertebrates and cryptic species), temporal bias (peak observation effort on weekends and summer months), and identification error rates that vary substantially by observer experience and species group. This study conducted the most comprehensive quantitative evaluation of citizen science data quality and scientific utility yet published, comparing iNaturalist (n = 247 million records; 2010-2024), eBird (n = 1.84 billion observations), and GBIF aggregated records (n = 3.47 billion occurrences) against 84 independent professional survey datasets across 24 countries, 47 species groups, and 8 ecosystem types. Citizen science data showed mean 74.8% spatial coverage overlap with professional survey sites but with systematic underrepresentation of areas > 5 km from roads (coverage deficit -47.4% relative to professional surveys). Taxonomic identification accuracy was 87.4% +- 8.4% for well-photographed vertebrates in iNaturalist (research grade) but only 54.7% +- 18.4% for invertebrates. After spatial and taxonomic bias correction using inverse-distance weighting and taxonomic calibration models, citizen science data recovered 84.7% of professional survey species richness estimates and 74.8% of population trend signals -- confirming that bias-corrected citizen science data are scientifically valid for large-scale biodiversity monitoring at substantially lower cost than professional surveys.