Predictive Occurrence Modeling for Three Rare Plants Within the Croatan National Forest

Abstract

Rare plants are valuable indicators of areas of biological significance, and their persistence over time serves as a measure of ecological health and good land stewardship. Knowledge of the extent of rare plant distributions and the viability of rare plant populations is valuable information for assessing the risk of extirpation, future research, and management decisions. I constructed predictive occurrence models for three rare plants, Platanthera integra, Pinguicula pumila, and Asclepias pedicellata, using the Presence-only Prediction (MaxEnt) Tool in ArcGIS Pro 2.9. Explanatory variables used to model rare plant distributions included a LiDAR-based digital elevation model (DEM) of Croatan National Forest, soil series, and natural community data. Models were trained with Elemental Occurrence data obtained from the North Carolina

Natural Heritage Program. The tool generated significant models with an area under the receiver- operator curve (AUC)>0.80 for all three species. The model for Asclepias pedicellata accurately

predicted two new populations. However, observers failed to find individuals at the predicted sites surveyed for Platanthera integra and Pinguicula pumila. This may have been due to the timing of surveys or the accuracy of the models. The results of this study demonstrate that the Presence-only Prediction (MaxEnt) Tool may prove to be a valuable predictor of undiscovered populations of rare species, especially once the environmental filters and life history traits driving the distribution patterns of a rare focal species are better understood.