LAUSR

Abstract Long-term biodiversity occurrence records are key to quantify long-term biodiversity patterns and trends and inform the conservation of threatened species, but they are strongly biased in terms of the species represented. This taxonomic bias, and its correlation to societal preferences, is well-identified in modern biodiversity datasets. However, it remains to be investigated, and its basis understood, in long-term occurrence datasets assembled from historical sources. Here we investigate taxonomic bias for 38 species of large terrestrial mammals using a dataset of 780 historical occurrence assembled from 16th to mid-19th century historical written sources in South Africa. We test if this bias is related to species' historical charisma, using a functional definition of non-human charisma, supported by anecdotes from the historical literature. We identify a strong taxonomic bias, with up to several order of magnitudes of difference in the likelihood of reporting between some species. Species' charisma alone explains 75% of the observed variance, the most charismatic species being largely over-reported. This is the first evidence of a positive relationship between taxonomic bias and charisma in a historical biodiversity dataset, within a homogeneous taxonomic group such as large terrestrial mammals. These results improve our understanding of the relationship between people and the large terrestrial fauna in historical times and suggest that species' charisma is a good predictor of taxonomic bias in long-term biodiversity datasets. This provides background for modern conservation by illustrating the durability of the charisma concept and of its relation with taxonomic bias, with implications for the representativeness of species in long-term conservation studies.