Iterative species distribution modeling results in the discovery of novel populations of a rare cold desert perennial

Abstract

Niche modeling for rare and range-restricted species can generate inaccurate predictions leading to an overestimation of a species geographic distribution. We used an iterative ensemble modeling approach and model-stratified field surveys to improve niche model formulation and better understand the ecological drivers of Ivesia webberi distribution. I. webberi is a US federally threatened herbaceous species, narrowly distributed in the western Great Basin Desert. Niche models for I. webberi were fitted using 10 replicates each of 6 modeling algorithms, while geographical projections of habitat suitability were generated using weighted ensembles of models with optimal performance. The resulting model projections were used to guide field surveys for 5 yr, generating additional spatial data, which were added to the existing dataset for subsequent modeling. Model performance across iterations was investigated and niche differences in the spatial dataset were explored. Model-guided field surveys resulted in the discovery of several new locations of I. webberi and an expansion of the species known range by 63 km. Model performance was higher in the earlier overfitted niche models. Overfitting was corrected in the final models, and predicted habitat suitability reduced from 5.98% in the 2015 model to 3.34% in the 2020 model. Findings show that I. webberi niche is associated with biotic, topographic and bioclimatic variables. Furthermore, a partial overlap was observed between environmental conditions of the initial and the new locations (Schoener’s D = 0.47), which can be decomposed into 93% of niche stability. This indicates that the majority of the newly discovered locations are within the environmental niche of the initial data.