How well documented is Australia's flora? Understanding spatial bias in vouchered plant specimens

Massive digitization of natural history collections (NHC) has opened the door for researchers to conduct inferential studies on the collection of biological diversity across space and time. The widespread use of NHCs in scientific research makes it essential to characterize potential sources of spatial bias. In this study, we assessed spatial patterns in records from the Australian Virtual Herbarium (AVH), based on >3 000 000 vouchered specimens of around 21 000 native plant species. The AVH is the main database for describing Australia's flora, and identifying its limitations is of paramount interest for the validity of conservation and environmental studies. We characterized how sampling effort is distributed across each Interim Bioregion of Australia (IBRA), then asked: (i) How complete are species inventories for each bioregion? We define completeness (C) as the ratio of observed to estimated species richness, using the Chao 1 estimator, (ii) How is sampling effort related to a commonly used Human Influence Index (HII)? and (iii) What is the probability that additional collections would result in the identification of previously unrecorded species in each bioregion? Sampling effort across bioregions is unequal, which partially reflects the collecting behaviour of naturalists in relation to species richness patterns. The density of records in bioregions ranges from 0.02–8.37 km⁻². At the bioregional scale, completeness is generally high with 79% of bioregions estimated to have records for at least 80% of their species. Completeness is partly explained by sampling effort (r = 0.43, p = 0.01), although some bioregions (e.g. Northern Kimberley and Burt Plain) have high completeness yet relatively low sampling effort. The inventory of Hampton, however, is substantially less complete than other bioregions (C = 0.66). Bioregions with high HII consistently have high completeness, while regions with low HII span the full range of completeness values. We calculated that an additional specimen collected from a bioregion has a 0.33% (Wet Tropics) to 11.7% (Arnhem Coast) probability of representing a new species for that region. Our assessment can assist with directing future systematic survey efforts by identifying bioregions where additional surveying may result in the greatest return, in terms of increasing knowledge of species richness and diversity.