Effects of correcting for geographical sampling bias on the rates of false presences and absences, and on the predicted extent of tree ferns (as a percentage of the total land area of New Zealand).

Models were fitted to two datasets (herbarium and NVS) using MaxEnt with the feature type set as “LQ”. Model predictions were based on average predictions from the 40 runs and evaluated by using the LUCAS dataset.†False presences occur when a model predicts a species as present whilst observed data indicate it is absent.‡False absences occur when a model predicts a species as absent whilst observed data indicate it is present.