Spatial and temporal sampling biases and spatial scale affect species distribution models and applicability to conservation management

We used Maxent to model the distribution of Cypripedium acaule in North Carolina using records from 1) publicly available databases (GBIF and iNaturalist) and 2) herbaria. We compared distribution models made with the different sets of occurrence records to evaluate how spatial and temporal biases in records affect model results.

The data provided here include the original iNaturalist dataset (prior to cleaning as described in our methods) and the code for the evaluation of models based on ground-truthed populations. We cannot provide the original herbaria dataset because location records are kept in confidence due to poaching concerns.