Record error and range contraction, real and imagined, in the restricted shrub Banksia hookeriana in south-western Australia

Banksia hookeriana Meissn. (Proteaceae) is a fire‐killed shrub endemic to the northern sandplains of south‐western Australia that could be described as endangered based on its small geographical range (< 5000 km²) and area of occupancy (～500 km²). Impacts on the species’ geographical range by land clearance for farming and mining, and by altered fire regime, were investigated using three lines of evidence: records of herbarium collections, a comprehensive field survey of extant populations, and air photo and satellite images revealing the recent history of land clearance and fires. These show that the species’ range has contracted by up to 40% in area and 26% latitudinally through the loss of outlier and range limit populations since 1960. In addition, 22% of remaining native shrubland on the Eneabba sandplain has been lost over this period through clearing for farming and mining, representing further habitat loss for B. hookeriana. Detailed investigation of B. hookeriana herbarium collections (n = 46) revealed important errors that artificially affected the geographical range of the species and emphasized the importance of close examination of all data captured from collection records. Recorded locations occurred hundreds of kilometres outside the current geographical range of the species in areas with different climate and substrate. Incorrect species identification of herbarium specimens further extended the apparent geographical range of the species. On the other hand, credible records indicated the loss of the species from localities at the limits of its range. Overall, a disconcertingly high proportion of records contained errors that may be difficult to identify without close examination of the original collections and detailed ground‐truthing. Were these records to be used to model climate envelopes, identify potential habitat where the species might occur, or might migrate to either as pests or under climate change scenarios, or to analyse evolutionary or ecological theory (for example) — as is now becoming commonplace — large errors may ensue.