Fig. 1
Schematic summary of the dataset. Top: Native region extents were retrieved from Kew’s Plants of the World online. Middle: Occurrence data was retrieved from the Global Biodiversity Information Facility (GBIF)24 and filtered into three different occurrence data types: raw data (blue), presence cells (grey) and thinned data (yellow). Bottom: The different occurrence data types were used in Maxent models to predict relative environmental suitability indices within native regions (i.e. range estimates). Differences between Model 0 and Model 1 to 3. Model 0 was trained to support variable selection using raw data in k-fold cross validated Maxent models (one model for each combination of feature classes, i.e. linear (L), quadratic (Q), hinge (H), product (P) and threshold (T)). The selected variables and each of the three occurrence data types were used to train a set of separate k-fold cross validated Maxent models (one model for each possible combination of feature classes, regularization multipliers and occurrence data type). The overall best performing model was selected for each species based on performance metrics.
