Coverage and Biases in iNaturalist Records of Vascular Plants in Massachusetts

We examined the coverage and biases of iNaturalist data for three subsets of Massachusetts vascular plant species at two scales, including statewide county- and town-level data for 340 and 164 common species, respectively, and town-level data for 829 species documented in published floras of Franklin and Worcester counties. Additionally, we investigated statewide herbarium specimen data at the county and town level for an equivalent time period. iNaturalist county-level coverage was very good for a sample of common and readily identified species, which were documented in a median of 13 of the state's 14 counties. At the town level, however, iNaturalist records captured species presence in a median of just 47% of towns (of 351 total), and a median of 19% of towns (of 86 total) in Franklin and Worcester counties. iNaturalist data were more robust in areas with both higher human populations and large tracts of protected land accessible to the public. Coverage of herbarium specimens was less complete than iNaturalist coverage at both the county and town levels, except in a few areas subject to recent specimen-based floristic studies. In addition to coverage biases, iNaturalist data have numerous biases related to life history traits and environmental affiliations. These were more pronounced at the level of town than county. Species of upland habitats were more often reported than those of wetland and aquatic habitats; spore-bearing species were better represented than seed-bearers; woody species, especially trees, were better represented than herbaceous species; and forbs were better represented than graminoids. Additionally, iNaturalist data emphasized species with biotic pollination compared to those with abiotic pollination and nonnative species, especially invasive taxa, compared to native species. Our results indicate that in a populous state like Massachusetts, iNaturalist is presently a good tool for documenting county-level occurrence of common species that are easily identified from photographs. Our results also illustrate the value of iNaturalist data in documenting the arrival and spread of nonnative species. Town-level resolution is very incomplete, making such data an inadequate substitute for those generated by intensive floristic studies at that scale. Researchers and other users of iNaturalist data should be aware of the various biases and limitations in such data to avoid generating spurious patterns and drawing inappropriate conclusions.