Cooperative Fish and Wildlife Research Units Program: Mississippi
Education, Research and Technical Assistance for Managing Our Natural Resources

Ramirez-Reyes, C., M. Nazeri, G. Street, D. Todd Jones-Farrand, F.J. Vilella, and K.O. Evans. 2021. Embracing ensemble species distribution models to inform at-risk species status assessments. Journal of Fish and Wildlife Management 12(1):98-111.


Effective conservation planning requires reliable information on the geographic distribution of species, which is often incomplete due to limited availability of presence data. Species distribution models (SDMs) and associated tools have proliferated in the past decades and have proven valuable in evaluating suitability of habitat for species. However, practitioners have yet to fully adopt the potential of SDMs to inform surveys and other conservation efforts for information-limited species. Instead, most efforts rely on expert knowledge and other traditional methods to locate extant populations. In particular, the Species Status Assessment (SSA) framework of the United States Fish and Wildlife Service would benefit from incorporating SDM approaches to facilitate conservation decisions. Here, we describe an SDM approach for at-risk species that would benefit SSAs and similar species conservation efforts. We applied four modeling techniques (generalized additive, maximum entropy, generalized boosted, and weighted ensemble) to recent monitoring data for three at-risk species proposed for listing under the U.S. Endangered Species Act (Papaipema eryngii, Scutellaria ocmulgee, Balduina atropurpurea) in the Southeastern U.S. The ensemble models reduced uncertainty caused by differences among modeling techniques and improved the predictive accuracy of fitted models. Incorporating an ensemble modeling approach into the SSA framework would benefit monitoring efforts and provide more robust status assessments for at-risk species. We emphasize the importance of producing SDM in close collaboration among the stakeholders involved in use of model outputs.