Vouchered biodiversity specimens are of paramount importance in systematics, and evolutionary biology because they can be validated through tangible materials. However, they are being replaced by mass-production of observation-based records, which are seldom validated by experts. As this gap continues to widen, the link from biodiversity data to natural history collections becomes unclear and underrepresented. The need to assess sampling gaps and biases in these disparate biodiversity sources has also been raised in theoretical and applied ecology. In this study, we disentangled areas of mismatches and congruencies in sampling biases between observations and vouchered specimens in the plant occurrence records of the world.
Results/Conclusions
We revealed unequal representation of sampling biases spanning variable temporal, spatial and phylogenetic breadths. Shared biases included about 20% overlap in hotspots of sampling density around coastlines of the United States. Unique biases for vouchered specimens include massive gaps in southern United States, great temporal sampling depths and phylogenetic clustering – the tendency of closely related species to be collected similarly than expected randomly. Species richness and abundance hotspots were seen in the Central Americas for vouchered specimens while these hotspots were seen focused around Northern Europe for observation-only occurrence records. The sampling of observation records was more phylogenetically dispersed, and spatially under sampled in most of central and northwestern United States. Our results challenge users of biodiversity occurrence records to adequately account for these shared and unique biases while making efforts to ensure random sampling methods in their own work to preserve accuracy and validity in global biodiversity data.