Iterative ensemble niche modeling of Ivesia webberi in the Great Basin Desert

An iterative and ensemble modeling approach was used to fit ecological niche models to better understand the ecological drivers of Ivesia webberi distribution. Geographical projections were generated from weighted average ensembles of six niche modeling algorithms (Boosted Regression Trees, Random Forests, Maximum Entropy, Artificial Neural Networks, Generalized Additive Models and Generalized Linear Models), and were used to guide field validation surveys. Alternating niche modeling and field surveys were conducted for five years (2015-2020) generating additional absence points and novel occurrences. Niche differences between the original and novel environmental conditions were estimated using tests of niche overlap, similarity, expansion, and stability, based on principal component analysis. Differences between the geographic projections of the first- and fifth-year iterative niche models were also compared. Model-guided field surveys resulted in the discovery of nine new locations of I. webberi, including two accidental discoveries, increasing the number of known locations to 32, and expanding the northern reach of the known species range by 63 km. The new locations resulted in a 7% niche expansion; however, niche similarity (p=0.10) was not significantly different. Furthermore, there was 47% overlap between the original and new locations. I. webberi niche is associated with perennial herbaceous cover, Topographic Position Index, and climatic variables. These results demonstrate the effectiveness of iterative niche modeling and model-guided field surveys, which resulted in the discovery of new locations of a rare and threatened species, expanded the known species range, and improved model performance. This can support effective conservation management for threatened species.