From pet to pest? Differences in ensemble SDM predictions for an exotic reptile using both native and nonnative presence data
- Author(s): Bevan, Hannah R.;
- Jenkins, David G.;
- Campbell, Todd S.
- et al.
Published Web Locationhttps://doi.org/10.21425/F5FBG42596
As a result of the pet trade, Africa’s Nile monitor (Varanus niloticus) is now established in North America (Florida). This generalist carnivore is a potential threat to native wildlife, requiring proactive measures to effectively prevent further spread into novel regions. To determine regions at risk, we create and compare alternative ensemble species distribution models (SDMs) using a model selection approach (with 10 possible modeling algorithms grouped according to assumptions). The ensemble SDMs used presence and environmental data from both native (Africa) and nonnative (Florida) locations. The most predictive consensus SDMs for native and native + nonnative data sets (TSS = 0.87; Sensitivity = 93%; Specificity = 94%) were based on the boosted regression tree (BRT), classification tree analysis (CTA), and random forest (RF) modeling algorithms with all environmental predictor variables used. The global Nile monitor SDMs predict strong habitat suitability in tropical and subtropical regions in the Americas, the Caribbean, Madagascar, Southeast Asia, and Australia. Florida Nile monitor populations are less likely to spread into the Neotropics than if pets now in the Southwest USA are released intentionally or accidentally. Management options to avoid this spread into vulnerable regions are to actively prohibit/regulate Nile monitors as pets, enforce those restrictions, and promote exotic pet amnesty programs. The model selection approach for ensemble SDMs used here may help improve future SDM research