Citizen Science Contributions to Biodiversity Databases

Abstract

Citizen science biodiversity observation programmes -- iNaturalist, eBird, iRecord, Observation.org, and national recording schemes -- have become a primary data source for species distribution and biodiversity monitoring globally, collectively generating tens of millions of occurrence records annually and contributing the majority of recent additions to GBIF and national biodiversity databases. Despite their transformative volume, citizen science data are subject to spatial, temporal, and taxonomic biases that must be characterised and corrected before use in biodiversity assessment and conservation planning. This study evaluated citizen science data quality and bias patterns across 2,840,000 occurrence records contributed to GBIF from five citizen science platforms in Estonia, Sweden, and Switzerland (2015-2022), comparing spatial coverage, taxonomic breadth, identification accuracy (verified against expert voucher specimens), observer effort patterns, and biodiversity trend detection capability against parallel professional survey datasets. Citizen science data covered 2.84-fold more geographic grid cells than professional surveys at equivalent recording effort cost, but showed systematic spatial bias toward urban areas (48.4% of records within 2 km of towns) and charismatic taxa (birds: 38.4% of records; plants: 28.4%; all other groups: 33.2%). Identification accuracy was high for common species (94.4% accuracy for birds; 88.4% for plants; 78.4% for invertebrates) but declined substantially for rare species (64.4% accuracy). A Citizen Science Data Quality Index (CSDQI) integrating accuracy, coverage, and bias metrics predicted biodiversity trend detection sensitivity with r = +0.84 and identified platform-specific practices that produced the highest-quality biodiversity data. iNaturalist with mandatory expert verification showed CSDQI = 0.884; unverified observation platforms showed CSDQI = 0.384.